Productivity, Health and the Quality of Buildings FALL 2014
M. Yvonne Hidle CARNEGIE MELLON UNIVERSITY | VIVIAN LOFTNESS
BIDS™ Revised Case Study Template M. Yvonne Hidle
December 2014
TABLE OF CONTENTS Introduction 1.1 Case Study List 1.2 Executive Summary 1.3 Baseline Numbers
Case Study 1: Acoustic Privacy = Individual Productivity The Effect of Speech and Speech Intelligibility on Task Performance 2.1 Power Point Slide 2.2 BIDS Evaluation Form 2.3 BIDS Summary 2.4 Cost Resources 2.5 Paper
Case Study 2: Acoustic Privacy = Individual Productivity Cognitive Performance During Irrelevant Speech: Effects of Speech Intelligibility and Office-Task Characteristics 3.1 Power Point Slide 3.2 BIDS Evaluation Form 3.3 BIDS Summary 3.4 Cost Resources 3.5 Paper
Case Study 3: Treadmill Workstations = Health + Reduced Absenteeism Treadmill Desks: A 1-Year Prospective Trial 4.1 Power Point Slide 4.2 BIDS Evaluation Form 4.3 BIDS Summary 4.4 Paper 4.5 Supporting Paper Case Study 4: Exercise at Work = Health + Reduced Absenteeism The Effects of the Nintendo Wii Exercise Program on Chronic Work-Related Low Back Pain in Industrial Workers 5.1 Power Point Slide 5.2 BIDS Evaluation Form 5.3 BIDS Summary 5.4 Paper 5.5 Supporting Paper
BIDS™ Revised Case Study Template M. Yvonne Hidle
December 2014
CASE STUDY LIST Acoustic Privacy = Individual Productivity The Effect of Speech and Speech Intelligibility on Task Performance Haapakangas, A., Haka, M., Keskinen, E., & Hongisto, V. (2008). The Effect of Speech and Speech Intelligibility on Task Performance â&#x20AC;&#x201C; An Experimental Laboratory Study. ICBEN, 1068-091. Retrieved September 23, 2014. In a 2007 controlled lab experiment studying speech intelligibility with 37 participants at the Finnish Institute of Occupational Health, Haapakangas et al identifies an increase in individual productivity of 9.5% for serial recall (p < 0.05) and 8.5% for complex memory (p < 0.001) when the Speech Transmission Index was kept between 0.10 (representative of a closed office) and 0.35 (representative of a closed office with an open door or acoustically private open office) compared to 0.65 (representative of a poorly designed open office).
Acoustic Privacy = Individual Productivity Cognitive Performance During Irrelevant Speech: Effects of Speech Intelligibility and Office-Task Characteristics Jahncke, H., Hongisto, V., & Virjonen, P. (2013). Cognitive Performance During Irrelevant Speech: Effects of Speech Intelligibility and Office-Task Characteristics. Applied Acoustics, 74(3), 307-316. Retrieved October 22, 2014. In a 2013 controlled lab experiment with 42 participants at the University of Gavle, Jahncke et al identifies an increase in individual productivity of up to 5% for word memory tasks (p < 0.0001), 6% for information searching tasks (p = 0.07), and 3% for math tasks (p < 0.05) when the Speech Transmission Index (STI) is kept below 0.34 in an open-plan office through the use of speech masking noise (between 40 and 45 dB), high cubicle walls (4.7 feet or greater), and sound absorption panels on the cubicle walls. Treadmill Workstations = Health + Reduced Absenteeism Treadmill Desks: A 1-Year Prospective Trial Koepp, G., Manohar, C., Mccrady-Spitzer, S., Ben-Ner, A., Hamann, D., Runge, C., & Levine, J. (2013). Treadmill Desks: A 1Year Prospective Trial. Obesity, 705-711. Retrieved November 5, 2014. In a 1-year intervention study with 36 participants at an office for the Educational Credit Management Corporation in Minnesota, Koepp et al identifies a 2.6% decrease in overall weight (p =0.04), a 5% decrease in waist circumference (p < 0.001), a 9% increase in High-Density Lipoproteins (p < 0.005), and a 37.3% increase in daily physical activity among obese individuals who use a treadmill workstation compared to their baseline condition. In a 2005 analysis of the National Health Interview Survey (NHIS) and the Medical Expenditure Panel Survey (MEPS) with 45,756 respondents, Finkelstein et al identifies up to a 40% decrease in absenteeism for men (p < 0.05) and up to a 58.5% decrease in absenteeism for women (p < 0.05) with the reduction of obesity from a baseline of 5 days/year for men and 8.2 days/year for women.
Exercise at Work = Health + Reduced Absenteeism The Effects of the Nintendo Wii Exercise Program on Chronic Work-Related Low Back Pain in Industrial Workers Park, J., Lee, S., & Ko, D. (2013). The Effects of the Nintendo Wii Exercise Program on Chronic Work-Related Low Back Pain in Industrial Workers. Journal of Physical Therapy Science, 985-988. In an 8-week intervention study with 24 workers at a K Tire Factory, Park et al identifies a 13% decrease in Lower Back Pain (LBP) (p < 0.05), a 6.11% increase in back strength (p < 0.05) and a 15.8% increase in self-reported emotional wellbeing (p < 0.05) among factory workers who participated in a 30 minute Nintendo Wii Fitness program 3 times a week compared to their baseline. In a 2011 LBP clinical study that analyzed claims data from 40 self-insured employers and 211,551 patients in the United States. Ivanova et al identifies a 234.4% decrease in employer costs for employees without LBP compared to employees with LBP (p < 0.0001), at roughly $1,856 per employee.
BIDS™ Revised Case Study Template M. Yvonne Hidle
December 2014
EXECUTIVE SUMMARY BIDS Case Studies This document focuses on acoustics and physical activity in the work environment and their impact on the health and productivity of employees. Acoustics is a key factor in the design of offices that is often overlooked, especially with the mass adoption of open plan offices in corporate America. In fact, “about 70 percent of U.S. employees now work in open offices” according to the International 1 Management Facility Association. Bad acoustics, in particular high speech intelligibility, have been found to reduce worker 2 3 productivity significantly as can be seen in both of the studies related to acoustics within this document. Architects should be cognizant of acoustics and speech intelligibility when designing offices, in particular open plan schemes. Studies by Haapakangas et al. and Jahncke et al. suggest that employers either move toward closed offices or provide more acoustic treatments in open plan offices, such as higher cubicle partitions, partition sound insulation, and sound masking. Lack of physical activity is another issue that many American workers face today. 80% of jobs within the United States today 4 are “sedentary” or “light”; compared to a little over 50% of jobs in 1960. As one New York Times article put it: “The findings also put pressure on employers to step up workplace heath initiatives and pay more attention to 4 physical activity at work.” This increase in sedentary work is potentially a key factor in the obesity epidemic. As of 2011-2012 approximately 35.1% of 5 adults over the age of 20 in the United States are obese and 69% are overweight. Obesity can lead to a number of health ailments which have been shown to decrease worker productivity and increase health-related absenteeism. One study within this document, conducted by Koepp et al. supports the effectiveness of treadmill workstations to reduce worker obesity. Architects and designers should space plan with worker activity in mind; in fact, the study by Koepp et al. suggests that offices be laid out in a manner that encourages walking to shared facilities such as trash bins and printers. In addition, employers should consider active workstations such as treadmill and cycling desks to help reduce worker obesity, improve worker health and emotional wellbeing. This is particularly important for populations with lower incomes and lower level of educational 6 attainment as their rates of inactivity and obesity tend to be higher than those in higher income brackets although the 7 correlation has partially broken in more recent years. In addition to space planning measures and active workstations, employers should also consider setting aside work time to allow employees to exercise during the workday. In fact, the 4th case study in this document focuses on industrial workers with Lower Back Pain (LBP) and the addition of exercise into their daily work schedule. Park et al. suggests the use of the Nintendo Wii Fitness Program to alleviate Lower Back Pain and improve worker strength and balance. These studies indicate the importance and benefits of integrating facilities and policies for physical activity for the health of employees when designing offices and other work environments.
1
Wong, V. (2013, July 1). Ending the Tyranny of the Open-Plan Office. Retrieved December 7, 2014, from http://www.businessweek.com/articles/2013-07-01/ending-the-tyranny-of-the-open-plan-office 2 Haapakangas, A., Haka, M., Keskinen, E., & Hongisto, V. (2008). The Effect of Speech and Speech Intelligibility on Task Performance – An Experimental Laboratory Study. ICBEN, 1068-091. Retrieved September 23, 2014. 3 Jahncke, H., Hongisto, V., & Virjonen, P. (2013). Cognitive Performance During Irrelevant Speech: Effects of Speech Intelligibility and Office-Task Characteristics. Applied Acoustics, 74(3), 307-316. Retrieved October 22, 2014. 4 Trends over 5 Decades in U.S. Occupation-Related Physical Activity and Their Associations with Obesity. (n.d.). Retrieved December 8, 2014, from http://www.plosone.org/article/info:doi/10.1371/journal.pone.0019657 5 Obesity and Overweight, Data for the U.S. Retrieved December 8, 2014, from http://www.cdc.gov/nchs/fastats/obesityoverweight.htm 6 Physical Inactivity. (n.d.). Retrieved December 8, 2014, from http://www.americashealthrankings.org/all/sedentary 7 Adult Obesity Facts. (2014, September 9). Retrieved December 8, 2014, from http://www.cdc.gov/obesity/data/adult.html
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
December 2014
BASELINE NUMBERS New Cost Data
Item Sound Masking
Cost $1.00
SQFT
Used in case study? Case study 2
Higher Cubicle Screens
$200.00
Panel
Case study 2
Better Sound Absorption Panels for Cubicles
$58.00
Panel
Case study 2
Drywall
$4.70
SQFT
Case study 1
Paint
$0.87
SQFT
Case study 1
Carpet Tile
$4.24
SQFT
Case study 1
Acoustic Tile
$3.82
SQFT
Case study 1
Large Hydronic Heating System
$7.67
SQFT
Case study 1
Wet Pipe Sprinkler System
$2.85
SQFT
Case study 1
Florescent Lighting
$7.52
SQFT
Case study 1
Wall Switches
$2.51
SQFT
Case study 1
Receptacles
$3.90
SQFT
Case study 1
Data and Voice System
$3.25
SQFT
Case study 1
Notes Numbers from Florida Sound Masking. For more information see cost resources in case study 2. Numbers from OBEX Cubicle & Desk Extenders. $200 per panel. For more information see cost resources in case study 2. Numbers from Sound Away. UltraQuiet Acoustic Cotton Panel 3#, 4" x 2' x 4', 8 square feet. $58 per panel. For more information see cost resources in case study 2. Numbers from RSMeans Online 2014. 5/8" F.R. drywall, 3 5/8" Metal Stud Framing @16" O.C., 5/8" F.R. drywall. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. Painting, interior on plaster and drywall, brushwork, primer & 1 coat. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. Nylon, fusion bonded, 18"x18", or 24"x24" 24oz. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. 5/8" fiberglass board 24"x48" tee grid suspended support. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. 149,000 S.F., 2,400 K.W. 8 floors. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. Light Hazard, steel black, sch. 40, 50,000 s.f. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. Flourescent Fixtures 23 fixtures per 1600 S.F. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. 10 per 1000 S.F. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. 20 per 1000 S.F., 2.4 watts per S.F. For more information see cost resources in case study 1. Numbers from RSMeans Online 2014. 2 data/voice outlets per 1000 s.f. x5 for 10 data/voice outlets per 1k sf. For more information see cost
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
December 2014 resources in case study 1.
Cubicle
$20,000.00
8-Group Cubicle
Case study 1
Treadmill Workstation
$3,000.00
Workstation
Case study 3
Nintendo Wii Fit Bundle
$449.99
Console Bundle
Case study 4
Television with 50" Screen
$549.99
Unit
Case study 4
Indirect health savings associated with LBP
$1,856.00
Employee
Case study 4
Absenteeism obese employees
3.4 Extra days
Employee
Case study 3
Unisource Mirage 8 L-shaped cubicles with complete electrical wiring @ $20,000. For more information see cost resources in case study 1. Steelcase treadmill. For more information see summary for case study 3. Bundle includes console, games, controls, and accessories. For more information see cost resources in case study 4. For more information see cost resources in case study 4. Ivanova, Birnbaum, Schiller, Kantor, Johnstone, & Swindle, 2011. For more information see the supporting paper in case study 4. Fiebelkorn, Finkelstein, & Wang, 2005. For more information see the supporting paper in case study 3.
Case Study 1 THE EFFECT OF SPEECH AND SPEECH INTELLIGIBILITY ON TASK PERFORMANCE
M. Yvonne Hidle CARNEGIE MELLON UNIVERSITY | VIVIAN LOFTNESS
Acoustic Privacy = Individual Productivity Participant Scores in Serial Recall and Complex Working Memory Tasks
Haapakangas et al 2008
First cost increase: $5,085 / employee Annual productivity savings: $770 / employee ROI: 15%
90 80 70 60 50 40
Serial Recall Task (assumed to translate to reading)
76.7
82.6
83.2
STI 0.65:
STI 0.35:
STI 0.10:
56.4 STI 0.10:
0
55.9
10
STI 0.35:
20
51.5
30
STI 0.65:
Task Scores
In a 2007 controlled lab experiment studying speech intelligibility with 37 participants at the Finnish Institute of Occupational Health, Haapakangas et al identifies an increase in individual productivity of 9.5% for serial recall (p < 0.05) and 8.5% for complex memory (p < 0.001) when the Speech Transmission Index was kept between 0.10 (representative of a closed office) and 0.35 (representative of a closed office with an open door or acoustically private open office) compared to 0.65 (representative of a poorly designed open office).
Complex Working Memory Task (assumed to translate to data entry/calculation and designing/engineering/programming)
STI 0.65 (poorly designed open office) STI 0.35 (closed office with open door or well-designed open office) STI 0.10 (closed office with closed door)
Reference: Haapakangas, A., Miia, H., Esko, K., & Valtteri, H. (2008). Effect of speech intelligibility on task performance - an experimental laboratory study. ICBEN .
F14-Productivity, Health and the Quality of Buildings M. Yvonne Hidle Overview: 1. Article citation: Venetjoki, N., A. Kaarlela-Tuomaala, E. Keskinen, and V. Hongisto. "The Effect of Speech and Speech Intelligibility on Task Performance." Ergonomics (2006): 1068-091. ICBEN. Web. 23 Sept. 2014. <http://www.icben.org/2008/PDFs/Haapakangas_et_al_Laboratory_study.pdf>. 2. Study objective: To understand the effect of sound disturbances, in particular speech with varying intelligibility, on cognitive performance in three different office environments: cellular closed office, well-designed open office, and unsatisfactory open office. 3. What type of study is it? Simulation study Building case study Cross sectional (multiple building) study Before and after study
x
Lab experiment Field experiment Meta-analysis (multiple studies) Other: _____________________________
4. When did the study occur? (eg., month(s)/year) December 2007. 5. Where did the study occur? (eg., “Xerox headquarters, Boston”, “office buildings in Paris”) The study does not specifically mention the location, although it does state that the study was conducted in a laboratory that was setup to replicate an office-like environment. At the top of the study, under the list of authors, Finnish Institute of Occupational Health Indoor Environmental Laboratory, is mentioned. 6. How many people?
37
7. How many buildings?
1
Characteristics: Students (no additional information) Laboratory setting. 8 workstations/cubicles. Participants were positioned in the 4 middle workstations. Speakers producing speech noise were placed in the outer 4 workstations. Masking sound was produced by Characteristics: loudspeakers hidden in the ceiling.
8. What design strategies were compared in the study? Varying STI (Speech Transmission Index) levels, in particular STIs that corresponded with the three different types of office environments: cellular closed office, well-designed open office, and unsatisfactory open office. The STI levels given per office environment are as follows: Cellular closed office = STI 0.10 (“speech intelligibility is extremely low”) Well-designed open office = STI 0.35 (“speech intelligibility is reasonably low”) Unsatisfactory open office = STI 0.65 (“speech intelligibility is nearly perfect”) 9. What design strategy does the study support? The study concludes that office environments with an STI of 0.10 or 0.35 are better for tasks that require serial recall and complex working memory. It recommends the use of sound separation and noise masking between open office cubicles.
BIDS™ Study Evaluation Form
48-729 Fall 2014
Study method: 10. How were the design strategies measured? The design strategies were measured using five different tasks that the participants were asked to perform: 1) A serial recall task 2) A complex working memory task 3) A proofreading task 4) A visual short-term memory task 5) A reading comprehension task Conclusions were mainly drawn from the first three tasks. The last two tasks were “included as filler tasks to increase the length of test sessions.” 11. How was the benefit measured? Benefits were measured through the results of the aforementioned five tasks. In the serial recall task, participants subjected to an STI of 0.10 scored 9.5% better than participants subjected to an STI of 0.65. Performance between STI 0.10 and STI 0.35 did not differ much (0.89%). In the complex working memory task, participants subjected to an STI of 0.10 scored 0.73% better than participants subjected to an STI of 0.35 and 8.5% better than participants subjected to an STI of 0.65. In the proofreading task, no significant score difference was found between the three STI scenarios. This is believed to be the case because in proofreading, unlike in memory, you have the option to go back and re-read a section. The study states that with proofreading “the disturbing effects of intelligible background speech may be compensated by changes in strategies and enhanced effort.” It was found, however, that participants subjected to the higher STI of 0.65 rated the proofreading task to be harder than participants subjected to the STI of 0.10 and 0.35. The last two task results were not mentioned in the paper. 12. Other measurements: Subjective ratings were also taken for speech condition and subjective disturbance of different sounds for all three STI scenarios. Each participant took a questionnaire after each speech condition (STI 0.10, STI 0.35, and STI 0.65). The questionnaire used a 5-point Likert scale. Speech conditions were found to be more favorable and pleasant in the scenarios with a lower STI. Speech was also found to be the most distracting sound in all the scenarios; it surpassed both background hum and computer tapping. 13. How was the relationship between the design strategy and benefit determined? x
Simulation ANOVA Linear regression
Logistic regression/odds ratio Other statistical method: ______________________________ Observed only/no analytical method
14. How were confounding factors addressed? Confounding factors were avoided in the following ways: 1) By applying all three STI scenarios to the same individual for all participants in the study. 2) A practice session for the participants. 3) “The order of speech conditions was counterbalanced across subjects, as was the presentation of different speech samples and test versions in different speech conditions.” 4) All participant groups were tested at the same time of day, between 8:15 am and 12:30 pm.
BIDS™ Study Evaluation Form
48-729 Fall 2014
Study results 15. What is the quantitative benefit of the design action? The study found that reducing the STI in office environments from 0.65 to between 0.1 and 0.35 increases worker productivity by as much as 9.5% for serial recall tasks and 8.5% for complex working memory tasks. When relating this to the office environment, it means that closed cellular offices and well-designed open office plans with appropriate acoustic features are better for worker productivity than poorly-designed open office plans. The study lobbies the use of masking noise and lowering speech intelligibility in open office plans. 16. List other significant relationships found: The study concluded that STI is “a good predictor of both performance loss and subjective disturbance.” In addition, the study found that speech conditions were considered to be more favorable when the STI was lower. Furthermore, speech was the most distracting sound in all STI scenarios (more distracting than background hums and computer tapping). 17. List important non-significant factors: Lighting (410 lx in all scenarios) Temperature (23 C in all scenarios) Sound level (48 dBA in all scenarios) Input air rate (64 1/s in all scenarios) CO2 (700 ppm in all scenarios) Time of day (between 8:15 am and 12:30 pm in all scenarios) Masking noise generated by speakers in the ceiling (present for all scenarios, shown to not disturb participants according to subjective ratings from the questionnaire). These factors might determine the baseline. 18. What are the limitations of the study? 1) Open office spaces and rooms are reverberant. Materials, layout, and room size vary. 2) Speaker’s direction can be constantly changing (in the study, speakers were stationary the entire time for every scenario). 19. Costs provided in the study: None.
BIDS™ Study Evaluation Form
48-729 Fall 2014
BIDS™ Revised Case Study Template M. Yvonne Hidle
October 2014
Privacy – Haapakangas et al 2008 Acoustic Privacy = Individual Productivity In a 2007 controlled lab experiment studying speech intelligibility with 37 participants at the Finnish Institute of Occupational Health, Haapakangas et al identifies an increase in individual productivity of 9.5% for serial recall (p < 0.05) and 8.5% for complex memory (p < 0.001) when the Speech Transmission Index was kept between 0.10 (representative of a closed office) and 0.35 (representative of a closed office with an open door or acoustically private open office) compared to 0.65 (representative of a poorly designed open office). Overview In a 2007 controlled lab experiment with 37 student participants at the Finnish Institute of Occupational Health, Haapakangas et al investigated the relationship between speech intelligibility in the office environment and individual productivity. The Speech Transmission Index (STI) for three different office environments was explored: closed office with a closed door, closed office with an open door (or alternatively, a well-designed acoustically private open office), and a poorly designed open office. Each scenario was given an STI: 0.10, 0.35, and 0.65 respectively. 8 cubicle work stations were setup in the laboratory to best 2 simulate an office environment. Study conditions such as lighting, temperature, sound level, input air rate, CO and time of day were the same for all STI experiments. The 37 participants were divided into groups of 3-4; all participants were subjected to each STI scenario. Participants in each group were seated at the middle work stations and speakers that transmitted speech noise were situated in the outer work stations. Speakers emitting sound masking noise were hidden in the ceiling for all STI scenarios. Demographics such as education level, occupational level, age, race and gender were not noted in the study.
Image 1: Experimental setup (Haapakangas, Miia, Esko, & Valtteri, 2008) Study Method Each participant was asked to complete 5 tasks for all 3 STI scenarios: a serial recall task (computer, digits 1 to 9), a complex working memory task (computer, calculations), a proofreading task (pen and paper), a visual short-term memory task (not specified), and a reading comprehension task (not specified). Results from the study were based entirely on the first three tasks; the latter two were included as fillers and their results were not published in the study. At the end of each STI scenario, participants were asked to fill out a questionnaire with a 5-point Likert scale. This questionnaire was used to get subjective ratings on the background noise (level of distraction and sound pleasantness), difficulty of the tasks, and self-rated work efficiency. Study Results The study found that participants subjected to the 0.10 STI scored 9.5% better on the serial recall task than those subjected to the 0.65 STI condition (p < 0.05). A negligible difference was found between the 0.10 STI and 0.35 STI conditions (0.89%). In the complex working memory task, participants in the 0.10 STI condition scored 8.5% better than those in the 0.65 STI condition (p < 0.001) and a negligible 0.73% better than those in 0.35 STI condition. For the final measured task, proofreading, the study found that there were no significant score differences between the three different STI scenarios. However, they did find that participants rated the proofreading task harder in the 0.65 STI scenarios than in the 0.10 or 0.35 STI scenarios (p < 0.05). Haapakangas et al hypothesized that the scores did not differ significantly in proofreading because the participants had the ability to re-read and go over the material if the speech noise distracted them. This differed from the serial recall and complex working memory tasks because the individual must work at the computer’s pace rather than their own. In addition, the questionnaire found that participants “self-rated work efficiency declined with ascending STI values (p < 0.05).” (Haapakangas, Miia, Esko, & Valtteri, 2008) Sound pleasantness in the environment was rated higher as the STI values declined (p < 0.001).
BIDS™ Revised Case Study Template M. Yvonne Hidle
October 2014
Lastly, speech noise was considered the most disturbing noise when compared to the background hum of masking sound and computer tapping; participants rated the speech noise to be the most disturbing at the highest STI level (p < 0.001).
Participant Scores in Serial Recall and Complex Working Memory Tasks 90 80 70
Serial Recall Task (assumed to translate to reading)
83.2 STI 0.10:
82.6
0
STI 0.35:
10
76.7
20
STI 0.65:
STI 0.10 (closed office with closed door) 56.4
30
STI 0.10:
STI 0.35 (closed office with open door or well-designed open office)
55.9
40
STI 0.35:
STI 0.65 (poorly designed open office)
51.5
50
STI 0.65:
Task Scores
60
Complex Working Memory Task (assumed to translate to data entry/calculation and designing/engineering/programming)
BIDS Life Cycle Assumptions The 2014 BIDS office baseline assumes a 100,000 square foot building, 500 employees with an average salary of $45,000. RSMeans Online estimates a first cost increase of $29.18/SF ($7,586.80 per employee) for the construction of individual closed offices for every employee with union labor using 3 5/8” metal studs @ 16” O.C. with painted 5/8” gypsum board on both sides, a door, acoustic tile ceiling, new carpeting, new ductwork, new electrical outlets, and 30,000 square feet added to the building to allow for 120SF offices per employee. (RSMeans) When you subtract out the cost of existing cubicles, which Cost Owl estimates to be around $12.50/SF, the revised first cost increase becomes $19.56/SF ($5,085.60 per employee). (Smith, 2014) Serial recall and complex working memory tasks are assumed to translate into office work involving data entry/calculation (4% office time), designing/engineering/programming (9% office time), and reading (6% office time). In a baseline building, a 9.5% increase in productivity for data entry/calculations, a 8.5% increase in productivity for designing/engineering/programming and reading tasks, due to individual closed offices, yields annual savings of $771.75/employee, resulting in an ROI of 15.18%. First cost increase per square foot: Baseline first cost increase: First cost increase per employee: Annual productivity savings per employee:
$19.56 $2,542,800 $5,085.60 $771.75
Baseline annual savings: NPV of the savings: ® EVA : ROI:
$385,875 $2,934,965.25 $392,165.25 15.18%
(see appendix for breakdown) ($19.56 x 130,000sf) ($2,542,800 / 500 employees) ($45,000 x 9.5% improvement x 10% of time + $45,000 x 8.5% improvement x 9% of time) ($771.75 x 500 employees) ($771.75 x 500 x 7.606) (NPV of the savings - FCI) (Savings per employee /FCI per employee)
Limitations One major limitation of the study is the size, only 37 participants were included in the experiment. Any future studies should have more participants. In addition, open office spaces and rooms have varying degrees of reverberance, depending on materials, layout, and room sizes. Future studies should be in the field, to take into account different materials and layouts in multiple office buildings. Lastly, a speaker’s direction can be constantly changing in the real-world office environment. In the
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
October 2014
study, speakers were stationary the entire time for each STI scenario. Future studies should find a way to make this more realistic by having the source of the speech move occasionally. References Haapakangas, A., Miia, H., Esko, K., & Valtteri, H. (2008). Effect of speech intelligibility on task performance - an experimental laboratory study. ICBEN . RSMeans. (n.d.). RSMeansOnline. Retrieved September 26, 2014, from RSMeansOnline: Construction cost estimates made easy: http://www.rsmeansonline.com/ Smith, A. (2014, 10 6). How Much Does a Call Center Cubicle Cost? Retrieved 10 6, 2014, from Cost Owl: http://www.costowl.com/b2b/cubicle-call-center-cost.html Appendix Finalized Calculations Total Cost Total cost of constructing a closed office for every employee
$3,793,200
Original Building Area BIDS Baseline
100,000
SQFT
Baseline Cost Total cost for existing cubicle infrastructure
1,250,000
Revised Building Area To allow for a 10x12 closed office for every employee
130,000
SQFT
Revised Cost Total Cost - Baseline Cost
$2,543,200
Total Employees BIDS Baseline
500
Cost per SQFT Revised Cost / Total Building Area
$19.56
Closed office wall area 10x12 office; 8 foot ceiling
352
SQFT
Cost per Employee Revised Cost / Total Employees
$5,086.40
Total office wall area (wall area per office) x (number of employees)
176000
SQFT
Closed office floor/ceiling area 10x12 office; 8 foot ceiling
120
SQFT
Total office floor/ceiling area (floor/ceiling area per office) x (number of employees)
60000
SQFT
Per SQFT
C10101265060
Total Cost
$827,200.00
5/8" F.R. drywall, 3 5/8" Metal Stud Framing @16" O.C., 5/8" F.R. drywall
Numbers may not match above exactly due to rounding.
Interior Construction Partitions
$4.70
Wall area Interior Finishes Paint
$0.87
Per SQFT
C30102300060
Wall area Carpet tile
$4.24
Total cost Per SQFT
$153,120.00 C30204100060
Floor area Acoustic tile
$3.82
Total Cost Per SQFT
$254,400.00 C30204100060
Total Cost
$229,200.00
Per SQFT
D30201041520
Total Cost
$460,200.00
Per SQFT
D40104100640
Total Cost
$171,000.00
Ceiling area HVAC Large Hydronic Heating System Using this to get the assembly cost for the ductwork system. Ceiling area Fire Protection Wet Pipe Sprinkler System Ceiling area
$7.67
$2.85
Painting, interior on plaster and drywall, brushwork, primer & 1 coat Nylon, fusion bonded, 18"x18", or 24"x24" 24oz 5/8" fiberglass board 24"x48" tee grid suspended support
149,000 S.F., 2,400 K.W. 8 floors
Light Hazard, steel black, sch. 40, 50,000 s.f.
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
October 2014
Electrical Florescent Lighting
$7.52
Per SQFT
D50202080640
Ceiling area Wall Switches
$2.51
Total Cost Per SQFT
$451,200.00 D50201300400
Wall area Receptacles
$3.90
Total Cost Per SQFT
$441,760.00 D50201100680
Wall area Data and Voice System
$3.25
Total Cost Per SQFT
$234,000.00 D50309200102
Total Cost
$571,120.00
Wall area
Baseline cost (cost of existing cubicle infrastructure) Partitions $12.50 Per SQFT Total Cost
Cost Owl $1,250,000.00
Flourescent Fixtures 23 fixtures per 1600 S.F. 10 per 1000 S.F. 20 per 1000 S.F., 2.4 watts per S.F. 2 data/voice outlets per 1000 s.f. x5 for 10 data/voice outlets per 1k sf
Unisource Mirage 8 L-shaped cubicles with complete electrical wiring @ $20,000
Cost of cubicle
(500 employees) / (8 cubicle group)
(62.5 cubicle groups) * ($20,000 per group)
($1,250,000.00 for cubicles) / (100,000 sqft)
$20,000.00
62.5
$1,250,000.00
$12.50
cubicle groups
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
Effect of speech intelligibility on task performance an experimental laboratory study Annu Haapakangas1*, Miia Haka2, Esko Keskinen2, Valtteri Hongisto1 1 Finnish Institute of Occupational Health, Indoor Environment Laboratory, 20520 Turku, Finland 2 University of Turku, Department of Psychology, 20014 Turku, Finland *corresponding author: e-mail: annu.haapakangas@ttl.fi
ABSTRACT The aim of this study was to examine the effects of speech varying in intelligibility on cognitive performance and subjective perceptions of sound environment disturbance. 37 subjects performed a series of tasks in three conditions in which speech transmission indexes were 0.10, 0.35 and 0.65. These correspond to cellular office, well-designed open office and unsatisfactory open office, respectively. The experiment was conducted in an office laboratory in which the direction of the speech source varied. The sound environments were presented at 48 dBA. Performance deteriorated in condition 0.65 compared to the other two conditions in a serial recall task (p < .05) and in a complex working memory task (p < .001). Proofreading performance did not differ in different conditions but the task was experienced as easier in the condition with lowest STI value than in the other two (p < .05). Questionnaire measures showed consistent, statistically significant differences between all three situations: the higher the STI value was, the more it was experienced to disturb performance and draw attention away from the task (p < .001). Similarly, the ease of habituation and the pleasantness of sound were rated higher as the STI value declined (p < .001). Self-rated work efficiency declined with ascending STI values (p < .05). Continuous background noise was not experienced to disturb performance in any condition. This supports the use of continuous masking sound in minimizing speech intelligibility, i.e. increasing speech privacy in open offices. INTRODUCTION Office surveys have shown that speech is experienced as the most distracting noise source in offices (e.g., Haapakangas et al. 2008). The problem is particularly great in open offices. Earlier laboratory experiments have shown that the disruptive effects of speech are not produced by the level of speech but speech intelligibility (Colle 1980; Ellermeier & Hellbr端ck 1998; Venetjoki et al. 2006; Schlittmeier et al. 2008). According to Hongisto et al. (2007), speech intelligibility in open offices varies greatly but it can be lowered by proper acoustical design. Research is needed on the effect of speech intelligibility on task performance in order to encourage investments in acoustic improvements. However, there are only a few published studies that have studied the effect of speech on cognitive performance with varying levels of speech intelligibility. Speech intelligibility can be evaluated with Speech Transmission Index (STI). For example, STI value 0.50 expresses roughly that 50 percent of syllables are correctly heard. STI can be easily determined between office workstations using acoustical measurements. The first experiment that used STI as a descriptor of irrelevant speech was published by Venetjoki et al. (2006). It was found that proofreading performance deteriorated significantly in STI 0.80 compared to STI 0.00. Later, similar experiments have been conducted by Schlittmeier et al. (2008).
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
The aim of this experiment was to show how the level of speech intelligibility affects cognitive performance. Speech was expected to deteriorate performance more in higher levels of speech intelligibility. The study focused on the STI range from 0.100.65 because the acoustic conditions of offices are typically in this area. The study also aimed to validate the model of Hongisto et al. (2008) that predicts the deterioration of performance as a function of STI. METHODS Subjects and test setup A repeated measures design with three speech conditions was used. Altogether 37 students took part in the experiment in December 2007. The three speech conditions were STI 0.10, STI 0.35 and STI 0.65. A group of four participants was tested at a time. Experiments were conducted between 8.15 a.m. and 12.30 p.m. The experiment included a practice session and three test sessions lasting for about 55 minutes each. Each situation was followed by a 5-minute break. The order of speech conditions was counterbalanced across subjects, as was the presentation of different speech samples and test versions in different speech conditions. Subjects performed five tasks: a serial recall task, a complex working memory task, a proofreading task, a visual short-term memory task and a reading comprehension tasks. The latter two were mainly included as filler tasks to increase the length of test sessions and were being piloted for future experiments. The serial recall task was conducted following classic procedures in which digits from 1 to 9 are presented on a computer screen in random order. Subjects' task is to recall the digits in the same order. The number of digits recalled in correct serial positions is measured. The complex working memory task was modified from the operation span task developed by Turner and Engle (1989). In the task, subjects had to state whether simple arithmetic calculations, presented on a computer screen, were true or false. Each calculation was followed by a presentation of a word that the subject had to memorize. At the end of each equation-word pair list, subjects were asked to recall the presented words in the same order. The number of words to be remembered increased from 3 to 8. The total number of correctly recalled words was measured. The proofreading task was the same as in our earlier study (Venetjoki et al. 2006). It was a pen-and-paper task in which subjects looked for mistakes in the text. Half of the errors were spelling errors whereas half required semantic processing. Each speech condition was followed by a questionnaire that assessed subjective perceptions of the sound environment. A 5-point Likert scale was used in most questions. Disturbance of other environmental factors was included in the questionnaire that was presented after the last sound condition. Background information was gathered with a separate questionnaire before practice session. Speech conditions The investigations were carried out in an office laboratory (Figure 1). Three speech conditions were used: â&#x20AC;˘
STI=0.10 - corresponds to a private office room with the door closed. Speech intelligibility is extremely low.
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
•
STI=0.35 - corresponds to a very good open office or a private office room with opened doors. Speech intelligibility is reasonably low.
•
STI=0.65 - corresponds to a typical open office without adequate acoustic design. Speech intelligibility is nearly perfect.
STI values were obtained by changing the relative sound pressure levels of speech and masking as shown in Figure 2. However, the total sound pressure level was constant in all speech conditions, LA,eq=48 dB. Special effort was made to create an office-like environment. All factors of indoor environment were monitored during all experiments. Typical values are given in Figure 1.
Indoor environment: Lighting: 410 lx 23 °C Temperature: Sound level: 48 dBA Input air rate: 63 l/s CO2: 700 ppm
Figure 1: The office laboratory (30 m2) contains 8 workstations. Four subjects were tested simultaneously and they sat in the middle. Speech was produced from four other workstations in random sequences. Masking sound was produced from four loudspeakers hidden above the suspended ceiling. Sound pressure level [dB]
STI
70
48
0.65
60
38 48
50
48 46 44
0.35
40 30
48 48
0.10
20
39
Masking (48 dBA)
10
Speech (48 dBA)
30
Octave band [Hz]
8000
4000
2000
1000
500
250
125
63
0
40 50 Sound level [dBA] Altogether Masking Speech
Figure 2: The spectrum of speech and masking was constant in each sound condition. The relative Aweighted sound pressure levels in three speech conditions 0.65, 0.35 and 0.10.
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
RESULTS Data was analyzed with SPSS 16.0 program using variance analysis of repeated measures. Performance measures Task performance was affected by speech intelligibility in the serial recall task (p < .05) and in the complex working memory task (p < .001). In the serial recall task, the contrast comparisons showed a significant decline in the percentage of correctly recalled digits in STI 0.65 compared to other two speech conditions (Table 1). The performance did not differ between STI 0.35 and STI 0.10. Table 1: Serial recall task. The table shows the percentages of digits recalled in correct serial position in different speech conditions. Speech condition STI 0.10 STI 0.35 STI 0.65
Mean 56,4 55,9 51,5
SD 13,7 14,0 14,6
A similar pattern emerged for the complex working memory task (Table 2). The number of correctly recalled words did not differ between STI values 0.10 and 0.35 but performance deteriorated significantly between STI 0.35 and STI 0.65 (p < .01). Proofreading task was not affected by different levels of speech intelligibility but the task was experienced as easier in STI 0.10 than in the other speech conditions (p < .01). Table 2: Complex working memory task. Percentages of correctly recalled words in different speech conditions are shown. Speech condition STI 0.10 STI 0.35 STI 0.65
Mean 83,2 82,6 76,7
SD 9,7 8,8 12,2
Subjective ratings Analyses of questionnaire measures showed consistent, statistically significant differences between all three speech conditions. The results for perceptions of speech conditions are shown in Figure 3. The higher the STI value was, the more the sound condition was experienced to disturb performance and draw attention away from the task (p < .001). Similarly, speech condition was perceived more pleasant and easier to habituate to as the STI value declined (p < .001). Self-rated efficiency also declined with ascending STI values with a significant deterioration in STI 0.65 in comparison to the other two speech conditions (p < .01).
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
5 Habituation to SC 4 Pleasantness of SC 3 Disturbance of SC 2 Capture of attention from SC 1 0,00
0,20
0,40
0,60
0,80
1,00
Speech Transmission Index STI Figure 3: Subjective ratings of the qualities of speech conditions (SC). The figure shows average values on a scale from 1 (not at all) to 5 (very much).
The subjective disturbance of different sound sources is shown in Figure 4. Speech was the most distracting sound in all speech conditions. The disturbance of speech increased with growing STI-values (p < .001). The perceived disturbance of the sound level followed the pattern of disturbance from speech, increasing with ascending speech intelligibility (p < .001) although the actual sound pressure level was the same in all speech conditions. It must be noted that, in speech condition STI 0.10, the sound pressure level of speech was 9 dB below the sound pressure level of masking (Figure 2). Yet, speech was experienced as the most disturbing sound source even in this condition and the difference between the disturbance of speech and masking noise was statistically significant (p < .01). The disturbance of masking noise did not differ in different speech conditions.
5 Speech
4 Overall sound level
3 Computer tapping
2 Background hum (masking noise)
1 0,00
0,20
0,40
0,60
0,80
1,00
Speech Transmission Index STI Figure 4: Distraction caused by different noise sources on a scale from 1 (not at all) to 5 (very much). Mean values for each condition are shown.
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
DISCUSSION The present study demonstrated the deteriorating effects of intelligible speech on cognitive performance. As in other similar studies (e.g., Schlittmeier et al. 2008), serial recall deteriorated in the speech condition in which speech was most intelligible. The performance in a complex working memory task followed a similar pattern but the degree of deterioration due to intelligible speech was slightly bigger than in the serial recall task. The working memory task taps into both the storage and the processing functions of working memory whereas serial recall only requires short-time storage of information. Such tasks are generally better predictors of complex cognitive abilities than simple digit recall tests (Daneman & Merikle 1996) and complex working memory tasks may therefore have more relevance when the purpose is to find measures that capture cognitive processes essential in real office work. Speech transmission index proved to be a good predictor of both performance loss and subjective disturbance. It is important to note that performance in the situation corresponding to private room conditions (STI 0.10) did not differ from the speech condition corresponding to well-designed open office conditions (STI 0.35). The drop in performance occurred between 'good open office' condition (STI 0.35) and 'poor open office' condition (STI 0.65). This is in line with Hongisto's (2005) model that predicts that performance starts to deteriorate when STI exceeds 0.30. Unlike performance tests, the subjective disturbance ratings showed consistent differences between all three speech conditions. Schlittmeier et al. (2008) who observed similar differences between objective and subjective measures, have suggested that the experienced disturbance might cause participants to invest more effort in performing tasks. This could amplify differences in subjective perceptions while reducing differences in performance between different conditions. It is likely that noise affects performance via several routes. Disturbance of cognitive mechanisms, such as working memory, represents a direct link between environmental conditions and performance effects. It is possible that subjective perceptions of disturbance predict long-term effects of noise that may indirectly influence performance, e.g. motivational effects and stress. However, this question cannot be answered in laboratory settings alone. Speech intelligibility did not affect all tasks in the same way. Contrary to expectations, proofreading performance was not affected by different levels of speech intelligibility. A few possible explanations exist for this. In our earlier experiment (Venetjoki et al. 2006), proofreading performance differed between STI values 0.00 and 0.80 but a narrower range of STI values was used in the present experiment. Proofreading may be less sensitive to changes in speech intelligibility than short-term and working memory tasks because it allows more flexible use of different strategies. For example, in proofreading, one can compensate temporary disruptions of attention by stopping and going back in the text. In serial recall and working memory tasks, the pace of information processing is to a large degree set by the test program. Thus, the disturbing effect of intelligible background speech may be compensated in proofreading by changes in strategies and enhanced effort. The finding that the task was experienced as more difficult with higher STI-values supports this possibility. Another explanation concerns the method of presenting speech samples. In our previous experiment (Venetjoki et al. 2006), speech was produced from loudspeakers that were in front of the subject and visible. In the present experiment, open office conditions were better simulated by producing speech from adjacent workstations and varying the direction of speech source randomly. Even with equal STI values, the
Performance: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, Foxwoods, CT
listening experience differs depending on the speech production method. This may complicate the comparison of different experiments. Other studies have typically produced speech monaurally or binaurally using headphones. In real open offices, rooms are reverberant, the speaker's direction is changing, listening is binaural and the speech enters, not directly, but via room reflections to the listener. These issues should be taken into account when speech production is planned for experiments on speech intelligibility. It would also be important to investigate whether the way of presenting speech confounds the relation between STI and performance. Finally, the results have practical relevance for acoustic design of offices. It is suggested that more appropriate conditions for individual work performance can be created in open offices by lowering speech intelligibility, e.g. by using masking noise. The subjective disturbance ratings showed that masking noise was not experienced to disturb participants at all. ACKNOWLEDGEMENTS This study was carried out as a part of a national research programme MAKSI (Perceived and Modelled Indoor Environment) funded by Tekes, National Technology Agency, Finnish Institute of Occupational Health and several participating companies. The authors thank Jukka Hyönä, Johanna Kaakinen, Virpi Kalakoski, Andreas Liebl and Sabine Schlittmeier for co-operation in designing tests for the experiment. REFERENCES Colle HA (1980). Auditory encoding in visual short-term recall: effects of noise intensity and spatial location. J Verbal Learn Verbal Behav 19: 722-735. Daneman M, Merikle PM (1996). Working memory and language comprehension: A meta-analysis. Psychon Bull Rev 3: 422433. Ellermeier W, Hellbrück J (1998). Is level irrelevant in irrelevant speech? Effects of loudness, signal-to-noise ratio, and binaural unmasking. J Exp Psychol: HPP 24: 1406-1414. Haapakangas A, Helenius R, Keskinen E, Hongisto V (2008). Perceived acoustic environment, work performance and wellbeing - survey results from Finnish offices. In: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008, July 21-25, Mashantucket, Connecticut, USA. Hongisto V (2005). A model predicting the effect of speech of varying intelligibility on work performance. Indoor Air 15: 458468. Hongisto V, Virjonen P, Keränen J (2007). Determination of acoustic conditions in open offices and suggestions for acoustic classification. In: 19th International Congress on Acoustics, Madrid, Spain, Sept 2-7, 2007. Hongisto V, Haapakangas A, Haka M (2008). Task performance and speech intelligibility - a model to promote noise control actions in open offices. In: 9th International Congress on Noise as a Public Health Problem (ICBEN) 2008 July 21-25, Mashantucket, Connecticut, USA. Pejtersen J, Allermann L, Kristensen TS, Poulsen OM (2006). Indoor climate, psychosocial work environment and symptoms in open-plan offices. Indoor Air 16: 392-401. Schlittmeier SJ, Hellbrück J, Thaden R, Vorländer M (2008). The impact of background speech varying in intelligibility: Effects on cognitive performance and perceived disturbance. Ergonomics 51: 719-736. Turner ML, Engle RW (1989). Is working memory task dependent? J Memory Lang 28: 127-154. Venetjoki N, Kaarlela-Tuomaala A, Keskinen E, Hongisto V (2006). The effect of speech and speech intelligibility on task performance. Ergonomics 49: 1068–1091.
Case Study 2 COGNITIVE PERFORMANCE DURING IRRELEVANT SPEECH: EFFECTS OF SPEECH INTELLIGIBILITY AND OFFICE-TASK CHARACTERISTICS
M. Yvonne Hidle CARNEGIE MELLON UNIVERSITY | VIVIAN LOFTNESS
Treadmill Workstations = Health + Reduced Absenteeism
First cost increase:1 Annual absentee savings:2 ROI:
$3000 / employee $600 / employee 20%
1 Koepp, G., Manohar, C., Mccrady-Spitzer, S., Ben-Ner, A., Hamann, D., Runge, C., & Levine, J. (2013). Treadmill desks: A 1-year prospective trial. Obesity, 705-711. Retrieved November 5, 2014, from http://onlinelibrary.wiley.com/doi/10.1002/oby.20121/pdf 2 Finkelstein, E., Fiebelkorn, I., & Wang, G. (2005). The Costs of Obesity Among Full-time Employees. American Journal of Health Promotion, 45-51.
12 Months (p < 0.001): 4205
25.41%
6 Months (p < 0.001): 4460
1000
Baseline: 3353
2000
12 Months (p < 0.02): 5041
3000
6 Months (p < 0.01): 4960
In a 2005 analysis of the National Health Interview Survey (NHIS) and the Medical Expenditure Panel Survey (MEPS) with 45,756 respondents, Finkelstein et al identifies up to a 40% decrease in absenteeism for men (p < 0.05) and up to a 58.5% decrease in absenteeism for women (p < 0.05) with the reduction of obesity from a baseline of 5 days/year for men and 8.2 days/year for women.
4000
37.28%
5000
Baseline: 3672
In a 1-year intervention study with 36 participants at an office for the Educational Credit Management Corporation in Minnesota, Koepp et al identifies a 2.6% decrease in overall weight (p =0.04), a 5% decrease in waist circumference (p < 0.001), a 9% increase in High-Density Lipoproteins (p < 0.005), and a 37.3% increase in daily physical activity among obese individuals who use a treadmill workstation compared to their baseline condition.
6000
Activity Units per Day
Health â&#x20AC;&#x201C; Koepp et al 2013
Daily Physical Activity (AU/Day)
(office)1
0 Obese Participants All participants Participant Groups by Body Type
F14-Productivity, Health and the Quality of Buildings M. Yvonne Hidle Overview: 1. Article citation: Koepp, G., Manohar, C., Mccrady-Spitzer, S., Ben-Ner, A., Hamann, D., Runge, C., & Levine, J. (2013). Treadmill desks: A 1-year prospective trial. Obesity, 705-711. Retrieved November 5, 2014, from http://onlinelibrary.wiley.com/doi/10.1002/oby.20121/pdf 1.1 Supporting article citation: Finkelstein, E., Fiebelkorn, I., & Wang, G. (2005). The Costs of Obesity Among Full-time Employees. American Journal of Health Promotion, 45-51. 2. Study objective: This study seeks to understand the effects of treadmill desks on the daily activity, health and weight of sedentary employees. 3. What type of study is it? Simulation study Building case study Cross sectional (multiple building) study Before and after study
x
Lab experiment Field experiment Meta-analysis (multiple studies) Other: _____________________________
4. When did the study occur? (eg., month(s)/year) 1-year intervention study; year not specified in the paper. 5. Where did the study occur? (eg., “Xerox headquarters, Boston”, “office buildings in Paris”) The intervention study took place in an office building for the “Educational Credit Management Corporation (ECMC), a financial services corporation; in Oakdale, Minnesota.” (Koepp, et al., 2013) 6. How many people? 36 People characteristics: All employees held jobs that were described as sedentary: “a job that requires sitting at a desk/table for the majority of the work day.” (Koepp, et al., 2013) The participants volunteered for the trial and were not monetarily compensated. The study consisted of 25 women and 11 men. 10 participants were considered lean (BMI < 25 kg/m2), 15 were considered overweight (25 < BMI < 30 kg/m2), and 11 were considered obese (BMI > 30 kg/m2).
BIDS™ Study Evaluation Form
48-729 Fall 2014
7. How many buildings? 1 Building characteristics: Financial services building. 8. What design strategies were compared in the study? Treadmill workstations versus a standard office desk. 9. What design strategy does the study support? The study supports the use of treadmill workstations for sedentary employees. The study also suggests that physical activity be increased by improving access to health clubs and redesigning the office work place in a way that encourages employees to walk to spaces such as meeting rooms, printers, and trash cans. Study method: 10. How were the design strategies measured? All participants were required to wear belt accelerometers that measured their daily physical activity during waking hours every day for the entire year. All participants were also required to have: their weight taken, their height measured, their body composition measured (through air-displacement plethysmography), have venous blood withdrawn, have their blood pressure measured, have their energy expenditure while resting, sitting, and walking measured (using indirect calorimetry) and take surveys regarding their work performance. Daily physical activity measurements taken with the accelerometer allowed researchers to gather data on the amount of time participants were sedentary and active. This data enabled them to measure how many minutes a participant walked at work and outside of work for the entire year. Height and body measurements were taken to keep track of participant weight, BMI, and waist circumference. Venous blood was withdrawn so the researchers could analyze glucose, HDL (high density lipoprotein), LDL (low density lipoprotein), total cholesterol, triglycerides, thyroid stimulating hormone (TSH), and hemoglobin A1C. Energy expenditure measurements were taken while resting, sitting, and walking through indirect calorimetry. Participants were required to fast, not undertake vigorous activity and not consume caffeine or alcohol prior to this test. Subjects were measured at rest (lying down) for 30 minutes, sitting for 15 minutes, and walking at 1, 2, and 3 mph for 15 minutes. 11. How was the benefit measured? The study does not specifically measure a benefit. It does mention that the cost of the treadmill workstation could be offset by the cost of obesity related diabetes which costs on average $10,000 per person per year. The study states that increased daily activity through the use of a treadmill workstation results in weight loss, waist circumference loss, and increased HDL. This daily activity and its benefits overtime would decrease obesity amongst office workers. A supporting study was used to measure benefits of decreasing obesity in the office environment. 12. Other measurements: Workplace performance surveys were used to address â&#x20AC;&#x153;four aspects of workplace performance: overall performance, quality of work, quantity of work, and quality of interactions with coworkers.â&#x20AC;? (Koepp, et al., 2013) The surveys took approximately 4 minutes to complete and were administered every Wednesday. Participant supervisors were also required to complete a similar survey to assess the work of the participants. More detailed surveys were given every 3 months, these surveys took approximately 20 to 30 minutes to complete.
BIDSâ&#x201E;˘ Study Evaluation Form
48-729 Fall 2014
13. How was the relationship between the design strategy and benefit determined? Simulation Logistic regression/odds ratio x ANOVA Other statistical method: ______________________________ Linear regression Observed only/no analytical method 14. How were confounding factors addressed? Confounding factors were addressed with the following: 1) All participants underwent the same physical tests and measurements. All participants also took the surveys. 2) All participants were subjected to the same environmental conditions within the office (lighting, heating, CO2 etc.) 3) All participants held a job within the company defined as sedentary. A control group would have been useful to address additional confounding factors. Study results 15. What is the quantitative benefit of the design action? The study found that obese participants who had the treadmill workstation experienced the best results with a 2.6% decrease in overall weight (p =0.04), a 5% decrease in waist circumference (p < 0.001), a 9% increase in HighDensity Lipoproteins (p < 0.005), and a 37.3% increase in daily physical activity. Participants classified as lean lost the least amount of weight, 0.81% of their baseline weight (p = 0.04). Lean participant did show an increase in daily physical activity which initially peaked at 20.9% for the first 6 months (p < 0.05) and then dropped to 2.4% over the course of the year (p < 0.07). When factoring in lean participants, the group as a whole had a modest weight loss of only 1.3% (p < 0.05). The High-Density Lipoproteins increase was the same for the entire group. 16. List other significant relationships found: Treadmill workstation use does not appear to affect performance in the long term according to self-reported surveys from participants and surveys from the participants’ supervisors. There is an adaptation period of approximately 3 months, after which performance returns to the baseline. 17. List important non-significant factors: Participant age, gender and education level. 18. What are the limitations of the study? The study recognizes several limitations: 1) The treadmills workstations replaced existing employee workstations. For this reason treadmill workstations were scattered throughout the office, with treadmill workstation users often sitting next to traditional desk users. The study suggests grouping treadmill workstation users to encourage the formation of a “microcommunity.” 2) The study also suggests that future studies work more with enhancing behavioral strategies. The study specifically states that offices should be reorganized to encourage employees to walk to meeting rooms, printers, and trash cans. The study also mentions that health club measurements should be encouraged. The study does not, however, address diet and the behavior strategies associated with a healthier diet. 3) The study cited study duration as being too short. Future studies should be longer than one year. 4) The study also recommends more participants. 5) The study states that a control group is needed for the entire intervention. This particular study only had a 2 week baseline and no control group (a group of participants with no treadmill workstations).
BIDS™ Study Evaluation Form
48-729 Fall 2014
19. Costs provided in the study: It costs approximately $3000 to $4000 for the Steelcase treadmill workstation used in the study. Obesity related diabetes which costs on average $10,000 per person per year.
BIDSâ&#x201E;˘ Study Evaluation Form
48-729 Fall 2014
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
Health – Koepp et al 2013 Treadmill Workstations = Health In a 1-year intervention study with 36 participants at an office for the Educational Credit Management Corporation in Minnesota, Koepp et al identifies a 2.6% decrease in overall weight (p =0.04), a 5% decrease in waist circumference (p < 0.001), a 9% increase in High-Density Lipoproteins (p < 0.005), and a 37.3% increase in daily physical activity among obese individuals who use a treadmill workstation compared to their baseline condition. In a 2005 analysis of the National Health Interview Survey (NHIS) and the Medical Expenditure Panel Survey (MEPS) with 45,756 respondents, Finkelstein et al identifies up to a 40% decrease in absenteeism for men (p < 0.05) and up to a 58.5% decrease in absenteeism for women (p < 0.05) with the reduction of obesity from a baseline of 5 days/year for men and 8.2 days/year for women. Overview In a 1-year intervention study with 36 participants at an office for the Educational Credit Management Corporation in Minnesota, Koepp et al observed how a treadmill workstation affected the health, daily activity, and weight of sedentary employees. All participants volunteered for the trial and were not monetarily compensated. All participants had sedentary job descriptions, 25 participants were women and 11 participants were men. 10 participants were considered lean (BMI < 25 kg/m2), 15 were considered overweight (25 < BMI < 30 kg/m2), and 11 were considered obese (BMI > 30 kg/m2). Participants were observed during a 2-week baseline period with no treadmill workstation. During this period of time, subjects were weighed and their height measured. In addition they underwent the following: daily physical activity measurements, energy expenditure measurements, body composition measurements and venous blood tests. Participants were also required to take short survey-based assessments of their performance. After the 2-week period, the participants were divided into 2 groups. The first group, composed of 13 employees, got treadmill workstations immediately. The second group, composed of 23 employees, received treadmill workstations 6 months after the first group. The treadmill workstations replaced the employee’s original desk and could be used as a treadmill, a standing desk, or a sitting desk. Employees were allowed to walk, stand, or sit at their desks and were not required to meet a daily quota of walking time. The treadmill desks could go up to 3 miles per hour. During the 1 year trial, all participants were asked to take body composition measurements, be weighed and have their height measured, wear accelerometers, have venous blood withdrawn, have their blood pressure measured, and to take surveys regarding their work performance. All participants were kept thermally comfortable in an environment between 68 and 74 degrees Fahrenheit. Study Method All participants were required to wear belt accelerometers that measured their daily physical activity during waking hours every day for the entire year. All participants were also required to have: their weight taken, their height measured, their body composition measured (through air-displacement plethysmography), have venous blood withdrawn, have their blood pressure measured, have their energy expenditure while resting, sitting, and walking measured (using indirect calorimetry) and take surveys regarding their work performance. Daily physical activity measurements taken with the accelerometer allowed researchers to gather data on the amount of time participants were sedentary and active. This data enabled them to measure how many minutes a participant walked at work and outside of work for the entire year. Height and body measurements were taken to keep track of participant weight, BMI, and waist circumference. Venous blood was withdrawn so the researchers could analyze glucose, HDL (high density lipoprotein), LDL (low density lipoprotein), total cholesterol, triglycerides, thyroid stimulating hormone (TSH), and hemoglobin A1C. Energy expenditure measurements were taken while resting, sitting, and walking through indirect calorimetry. Participants were required to fast, not undertake vigorous activity and not consume caffeine or alcohol prior to this test. Subjects were measured at rest (lying down) for 30 minutes, sitting for 15 minutes, and walking at 1, 2, and 3 mph for 15 minutes. Workplace performance surveys were used to address “four aspects of workplace performance: overall performance, quality of work, quantity of work, and quality of interactions with coworkers.” (Koepp, et al., 2013) The surveys took approximately 4 minutes to complete and were administered every Wednesday. Participant supervisors were also required to complete a similar survey to assess the work of the participants. More detailed surveys were given every 3 months, these surveys took approximately 20 to 30 minutes to complete.
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
Study Results The study found that obese participants who had the treadmill workstation experienced the best results with a 2.6% decrease in overall weight (p =0.04), a 5% decrease in waist circumference (p < 0.001), a 9% increase in High-Density Lipoproteins (p < 0.005), and a 37.3% increase in daily physical activity. Participants classified as lean lost the least amount of weight, 0.81% of their baseline weight (p = 0.04). Lean participants did show an increase in daily physical activity which initially peaked at 20.9% for the first 6 months (p < 0.05) and then dropped to 2.4% over the course of the year (p < 0.07). When factoring in lean participants, the group as a whole had a modest weight loss of only 1.3% (p < 0.05). The High-Density Lipoproteins increase was the same for the entire group. Daily walking increased for all the participants, from a baseline of 70 ± 25 min/day to 128 ± 62 min/day after 6 months (p < 0.001) and 109 ± 62 min/day after 12 months (p < 0.001). “No significant whole group changes were observed in triglycerides, glucose, LDL, TSH, and total cholesterol. Similarly, energy expenditure and energy efficiency did not change for the group as a whole over the 1-year intervention.” (Koepp, et al., 2013)
Image 1: Treadmill Desk (Koepp, et al., 2013)
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
BIDS Life Cycle Assumptions The 2014 BIDS office baseline assumes a 100,000 square foot building, 500 employees with an average salary of $45,000. The first cost increase is $15/SF ($3000 per employee) for the purchase and installation of a treadmill workstation. Absenteeism was calculated using the numbers given in a supporting study by Finkelstein et al. Normal weight women are absent from work 3.4 days/year and normal weight men are absent from work 3 days/year. The average of the two genders was calculated to be 3.2 days/year (baseline). Obese (grade-III) women are absent from work 8.2 days/year and obese (grade-III) men are absent from work 5 days/year. The average of the two genders was calculated to be 6.6 days/year. (Fiebelkorn, Finkelstein, & Wang, 2005) In a baseline building, a 49.25% decrease in absenteeism due to daily physical activity among employees using a treadmill workstation, yields annual savings of $597.66/employee, resulting in an ROI of 19.9%. First cost increase per square foot: Baseline first cost increase: First cost increase per employee: Annual productivity savings per employee: Baseline annual savings: NPV of the savings: ® EVA : ROI:
$15.00 $1,500,000 $3000.00 $597.66 $298,830 $2,272,900.98 $772,900.98 19.9%
($3000 per employee x 500 employees / 100,000 sf) ($15.00 x 100,000sf) ($1,500,000 / 500 employees) $45,000 x (3.4 days not missed / 256 workdays) ($597.66 x 500 employees) ($597.66 x 500 x 7.606) (NPV of the savings - FCI) (Savings per employee / FCI per employee)
Limitations The study recognizes several limitations: 1) The treadmills workstations replaced existing employee workstations. For this reason treadmill workstations were scattered throughout the office, with treadmill workstation users often sitting next to traditional desk users. The study suggests grouping treadmill workstation users to encourage the formation of a “micro-community.” 2) The study also suggests that future studies work more with enhancing behavioral strategies. The study specifically states that offices should be reorganized to encourage employees to walk to meeting rooms, printers, and trash cans. The study also mentions that health club measurements should be encouraged. The study does not, however, address diet and the behavior strategies associated with a healthier diet. 3) The study mentioned that the intervention duration was too short. Future studies should be longer than one year. 4) The study also recommends more participants. 5) The study states that a control group is needed for the entire intervention. This particular study only had a 2 week baseline and no control group (a group of participants with no treadmill workstations). References Fiebelkorn, B. I., Finkelstein, P. M., & Wang, P. G. (2005). The Costs of Obesity Among Full-time Employees. American Journal of Health Promotion , 45-51. Koepp, G. A., Manohar, C. U., McCrady-Spitzer, S. K., Ben-Ner, A., Hamann, D. J., Runge, C. F., et al. (2013). Treadmill Desks: A 1-Year Prospective Trial. Obesity , 705-711. Appendix None
Obesity
Original Article CLINICAL TRIALS: BEHAVIOR, PHARMACOTHERAPY, DEVICES, SURGERY
Treadmill Desks: A 1-Year Prospective Trial Gabriel A. Koepp1, Chinmay U. Manohar1, Shelly K. McCrady-Spitzer1, Avner Ben-Ner2, Darla J. Hamann3, Carlisle F. Runge4 and James A. Levine1
Objective: Sedentariness is associated with weight gain and obesity. A treadmill desk is the combination of a standing desk and a treadmill that allow employees to work while walking at low speed. Design and Methods: The hypothesis was that a 1-year intervention with treadmill desks is associated with an increase in employee daily physical activity (summation of all activity per minute) and a decrease in daily sedentary time (zero activity). Employees (n ¼ 36; 25 women, 11 men) with sedentary jobs (87 6 27 kg, BMI 29 6 7 kg/m2, n ¼ 10 Lean BMI < 25 kg/m2, n ¼ 15 Overweight 25 < BMI < 30 kg/m2, n ¼ 11 Obese BMI > 30 kg/m2) volunteered to have their traditional desk replaced with a treadmill desk to promote physical activity for 1 year. Results: Daily physical activity (using accelerometers), work performance, body composition, and blood variables were measured at Baseline and 6 and 12 months after the treadmill desk intervention. Subjects who used the treadmill desk increased daily physical activity from baseline 3,353 6 1,802 activity units (AU)/day to, at 6 months, 4,460 6 2,376 AU/day (P < 0.001), and at 12 months, 4,205 6 2,238 AU/day (P < 0.001). Access to the treadmill desks was associated with significant decreases in daily sedentary time (zero activity) from at baseline 1,020 6 75 min/day to, at 6 months, 929 6 84 min/day (P < 0.001), and at 12 months, 978 6 95 min/day (P < 0.001). For the whole group, weight loss averaged 1.4 6 3.3 kg (P < 0.05). Weight loss for obese subjects was 2.3 6 3.5 kg (P < 0.03). Access to the treadmill desks was associated with increased daily physical activity compared to traditional chair-based desks; their deployment was not associated with altered performance. For the 36 participants, fat mass did not change significantly, however, those who lost weight (n ¼ 22) lost 3.4 6 5.4 kg (P < 0.001) of fat mass. Weight loss was greatest in people with obesity. Conclusions: Access to treadmill desks may improve the health of office workers without affecting work performance. Obesity (2013) 21, 705-711. doi:10.1002/oby.20121
Introduction Obesity represents the combined effect of obesogenic environments and individual biological susceptibility; causative elements include sedentariness and food eaten in excess of need. Although obesity treatment has focused on individual solutions, less effort has been expended in environmental reengineering. This may be because environmental reengineering is multifaceted and involves architecture, furniture design, workflow innovation, and expense. Evidence suggests that sedentariness is associated with a myriad of health complications, including heart disease, metabolic syndrome,
and cancer, and is associated with premature mortality (1,2). These conditions are associated with higher insurance costs and losses in workplace performance (3). As a result, many corporations and their respective insurance providers are exploring relationships with health clubs and gyms that provide access, discounts, and monetary incentives for employees (4). However, such programs are not ubiquitous and poorly accessed by people, especially with obesity (5). The modern workplace fosters sedentariness (6) and people with obesity tend to sit more than lean individuals (7). Individuals move less at work than they do in leisure time (8). Moreover, there are benefits of intermittent bouts of walking to break-up sitting time (e.g., blood glucose) (9). Therefore, workplace reengineering to reverse sedentariness may be beneficial to employee health.
1
Endocrine Research Unit, Mayo Clinic, Rochester, Minnesota, USA. Correspondence: James A. Levine (Levine.james@mayo.edu) 2 The Carlson School of Management, University of Minnesota, Minneapolis, Minnesota, USA 3 Department of Public Administration, The University of Texas, Arlington, Texas, USA 4 Department of Applied Economics, University of Minnesota, Minneapolis, Minnesota, USA Funding agencies: This study was supported by grants DK56650, DK63226, DK66270, DK50456 (Minnesota Obesity Center), and RR-0585 from the US Public Health Service and by the Mayo Foundation and by a grant to the Mayo Foundation from Mr. R Stuart and ECMC. Disclosure: Dr. Levine prior to 2008 received consulting/royalty fees from Steelcase. He currently has no financial or legal agreement of any kind with Steelcase Inc. The following authors received grant support: Koepp G, Manohar C, McCrady-Spitzer S, Levine J, and Ben-Ner A. The following authors received Consulting Fees: Carlisle F. Runge C and Flint-Paulson D. Received: 25 January 2012 Accepted: 10 September 2012 Published online 6 November 2012. doi:10.1002/oby.20121
www.obesityjournal.org
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
705
Obesity
Treadmill Desks and Physical Activity Koepp et al.
One approach to workplace reengineering is to reexamine the modus operandi of the work desk. A treadmill desk (10) is a desk that enables an employee to walk on a treadmill while continuing normal work activities. Commercial units are available that include speedrestricted treadmills and a hydraulic desk top with an integrated treadmill control panel. Preliminary research examining treadmill desks (11) demonstrates that people can conduct normal office tasks (e.g., computer use and telephone calls) while walking on a treadmill desk at about 1 mph. In so doing, energy expenditure increases by 100–150 kcal/h and people, even with obesity, can tolerate this intervention for several hours per day (12). Access to treadmill desks might be useful to help people lose weight because many people work at desks for many hours each day. We were interested in whether changing a person’s desk—without any specific behavioral intervention—could increase daily physical activity, decrease sedentariness, and improve body weight.
Subjects and Methods
FIGURE 1 Study design of 36 office workers in a 1-year prospective trial of a treadmill desks intervention.
Subjects The 1-year study was conducted at Educational Credit Management Corporation (ECMC), a financial services corporation; in Oakdale, MN. Thirty-six office workers with sedentary job descriptions (‘‘a job that requires sitting at a desk/table for the majority of the work day’’) were recruited. Any employee with a sedentary job at ECMC could volunteer for the trial; however, subjects were excluded if their personal physician advised them against the study, they were unable to walk at 3 mph for 30 min, or were pregnant. The company, ECMC, agreed to allow subjects to complete the 1-year treadmill desk intervention irrespective of potential negative workplace performance.
Protocol We conducted a 1-year prospective trial to evaluate access to treadmill desks dedicated to a single employee in their workspace. Employees were not relocated due to the installation of the treadmill desk. We addressed the hypothesis that a 1-year intervention with access to treadmill desks was associated with increased employee daily physical activity, defined as the summation of low-, moderate-, and highintensity physical activity (Clinicaltrials.gov; NCT01461382). Our secondary hypothesis was that a 1-year intervention with access to treadmill desks was associated with decreased employee daily sedentary time defined as time spent accumulating zero physical activity (as measured with an accelerometer). The Mayo Clinic Institutional Review Board approved the study and subjects provided informed written consent. All subject measurements were recorded in a private room.
Baseline period. A 2-week baseline period preceded the treadmill desk intervention. Over the two baseline weeks, the 36 subjects worked normally (seated) at their usual desk. Subjects underwent a 14-day measurement daily physical activity, duplicate measurements of body composition, assessment of workplace performance, and venous blood determination of glucose, insulin, Hemoglobin A1C, and lipids. Measurement techniques are detailed below.
12-Month treadmill desk intervention. A treadmill desk replaced each of the 36 participants’ pre-existing desks for 12 months. One treadmill desk was dedicated to each individual. The employee’s original (traditional) desk was removed. As noted above, personnel were not located. The treadmill desk we used (Steelcase, Grand Rapids, MI) was a desk that can be elevated or lowered using a hydraulic
706
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
motor at the press of a button (Figure 1). The treadmill beneath the desk ran silently up to a maximum speed of 2 mph. The unit cost was $3,000–4,000 and was compliant with office safety standards. The treadmill desk, by design, did not compel the owner to walk because it could be lowered for chair use at the press of the button. The baseline measurements of daily physical activity, daily sedentary time (minutes with no activity), body composition, venous blood, and workplace performance outcomes were repeated at 6 and 12 months of the intervention. The techniques are described below.
Measurement techniques. All measurements and study activities took place in a research laboratory that we built inside the ECMC facility. It was staffed daily by a member of the study team. The subjects were weighed on a calibrated scale (model 644, Seca Corporation, Hanover, MD). Height was measured using a standiometer (model 242, Seca Corporation). Body composition was measured using air-displacement plethysmo graphy (Life Measurement Incorporated, Concord, CA) (13). Daily physical activity was monitored during waking hours, 7 days a week, throughout the 1-year intervention using a belt-worn accelerometer (Actical; Respironics, Philips, Eindhoven, The Netherlands). To account for nonwear time, data were excluded for missed whole and/or partial days. The download process took about 1–2 min and does not interfere with work time. The actical used a lithium coin cell battery CR2025 and were changed approximately every 6 months by loosening four small screws on the back of the device. The battery changing process required about 2 min per device. Venous blood was drawn a by nursing staff for measurements of glucose, HDL, LDLcalculated, total cholesterol, triglycerides, thyroid stimulating hormone (TSH), and hemoglobin A1C. Duplicate resting blood pressure (after 30 min of rest) was measured using an automated cuff (HEM-711ACN, Omron, Bannockburn, IL). Lying (resting), sitting, and walking energy expenditure were measured using indirect calorimetry (7). Subjects fasted for 10 h, had
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS: BEHAVIOR, PHARMACOTHERAPY, DEVICES, SURGERY
not undertaken vigorous activity for 6 h, had not consumed caffeine for >10 h nor alcohol for >12 h prior to the start of the measurements. Subjects were in thermal comfort (68–74 F). Subjects were allowed to drink water at room temperature and were encouraged to empty their bowel and bladder before the measurements. Energy expenditure was measured using indirect calorimetry while subjects were at lying rest (30 min), sitting in their office chair reading (15 min), and walking at 1, 2, and 3 mph each for 15 min. Workplace performance was assessed by employees and their supervisors using validated surveys (14). The surveys addressed four aspects of workplace performance: overall performance, quality of work, quantity of work, and quality of interactions with coworkers. A weekly survey was administered on Wednesdays to all employees. The survey required 4 min to complete. Supervisors completed similarly structured weekly surveys that focused on their employees’ workplace performance. A more detailed survey was administered every 3 months and took 20–30 min to complete (15,16). The weekly surveys focused on the week prior to the survey; the quarterly surveys asked employees and their supervisors about workplace performance during the prior 4 months.
Data and statistical analysis Accelerations were gathered at 32 Hz, summated, time stamped, and stored each minute on the device internal memory. Data from the device were then downloaded once a week by the study team using a standard laptop. All values are provided as mean 6 SD. ANOVA and post hoc paired t-tests were used to compare changes over time. Where indicated linear regression analysis was used. Statistical significance was defined as P < 0.05.
Results Thirty-six subjects were recruited for the 1 year treadmill desk intervention; 25 women, 11 men; 42 6 9.9 years, 87 6 27 kg, and BMI 29 6 7 kg/m2 (Table 1). Ten subjects (9F:1M) were lean (BMI < 25 kg/m2), 15 overweight (11F:1M) (25 < BMI < 30 kg/m2), and 11 had obesity (5F:6M) (BMI > 30 kg/m2). Four additional subjects were initially recruited but did not complete the study; one because of the diagnosis of inflammatory bowel disease, one because of pregnancy, one person developed connective disease requiring high dose steroid use, and one because the person left the company. The data from these four participants are not included in the analyses. For the 36 subjects in general, it took <5 min to acclimate to the treadmill desks. There were no injuries or reportable events associated with use of the desks. The subjects tolerated the treadmill-based system well over the 1-year intervention. To address our primary hypothesis, we compared daily physical activity at baseline and at 6 and 12 months. Subjects showed increased daily physical activity with the treadmill desk intervention. At baseline, the 36 subjects averaged 3,353 6 1,802 activity units (AU)/ day, at 6 months, 4,460 6 2,376 AU/day (P < 0.001), and at 12 months, 4,205 6 2,238 AU/day (P < 0.001) (Figure 2A). Daily sedentary time (zero activity) was 1,020 6 75 min/day at baseline and decreased with the treadmill desk intervention; after 6 months, daily sedentary time decreased by 91 6 66 min/day to 929 6 84 min/
www.obesityjournal.org
TABLE 1 Anthropometric, body composition, and metabolic variables at baseline and after 6 and 12 months in 36 subjects enrolled in the treadmill desk intervention
Weight (kg) Body fat (%) Fat mass (kg) Fat free mass (kg) Glucose (mg/dl) Hemoglobin A1c (%) Total cholesterol (mg/dl) Triglycerides (mg/dl) HDL (mg/dl) LDL (mg/dl) Waist Circ. (cm) Systolic BP (mm Hg) Diastolic BP (mm Hg)
Baseline
6 Months
12 Months
86.9 31.4 28.4 59.3 93.8 5.3 193 133 55 112 95 132 89
85.6 31.8 28.1 57.6 95.6 5.5 192 124 55 112 92 128 83
85.5 30.5 27.0 58.5 92.7 5.5 193 121 59 107 91 129 84
6 6 6 6 6 6 6 6 6 6 6 6 6
26.6 7.9 15.0 15.3 23.0 0.9 32 88 20 31 19 13 18
6 6 6 6 6 6 6 6 6 6 6 6 6
26.4** 8.2 13.9 15.4* 36.3 1.0*** 34 77 20 29 19*** 14* 8
6 6 6 6 6 6 6 6 6 6 6 6 6
25.8* 8.6 14.4 14.9 11.1 0.6 28 57 23** 30 18*** 13* 9
Data are shown as mean 6 SD (comparisons to baseline). *P < 0.05. **P < 0.01. ***P < 0.001.
day; P < 0.001. At 12 months, the decrease in daily sedentary time was by 43 6 67 min/day to 978 6 95 min/day; P < 0.001 (Figure 2B). The increase in daily physical activity at 6 and 12 months occurred predominantly during the work day. We were able to access accelerometer data for treadmill walking at work; at baseline, walking was 70 6 25 min/day; at 6 months, 128 6 62 min/day (P < 0.001); and at 12 months, 109 6 62 min/day (P < 0.001). Desk use itself, accounted for approximately 63% of the increased activity at 6 months and 90% of the increased activity at 12 months. The treadmill desk intervention was not associated with altered workplace performance. Employee self-rated workplace performance assessments (Table 2) varied more than assessments by supervisors, but overall, the results are similar. There were no significant changes in employee workplace performance either as self-assessed or assessed by supervisors. The relationship between time spent working on the treadmill desk and several performance measures suggested that there was a minor loss in workplace performance for the first 3–5 months. However, by the end of the 1-year intervention, workplace performance exceeded baseline. Interestingly, there was significant interindividual consistency in daily physical activity. There was a significant positive relationship for the 36 employees between daily physical activity at baseline and at 6 months (r2 ¼ 0.58; P < 0.001) and at 12 months (r2 ¼ 0.52; P < 0.001) of the intervention. Thus, interindividual variance was maintained even with the intervention. Although this was not an a priori hypothesis, we examined the effect of season post hoc. There were 36 employees in the 1 year treadmill desk intervention. Thirteen employees started the treadmill desk intervention in May of 2008 and ended the intervention in May 2009. Twenty-three other employees started the 1 year treadmill
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
707
Obesity
Treadmill Desks and Physical Activity Koepp et al.
TABLE 2 Employee rated performance and supervisor rated performance at baseline, 6 and 12 months of the treadmill desk intervention
Phase I Employees Baseline Employee rated performance Overall 7.71 6 Quality 3.56 6 Quantity 3.40 6 Interaction 3.54 6 Supervisor rated performance Overall 6.99 6 Quality 3.37 6 Quantity 3.23 6 Interaction 3.45 6
Phase II Employees
D6 months
D12 months
Baseline
D6 months
D12 months
0.96 0.39 0.41 0.50
0.35 0.16 0.13 0.12
6 6 6 6
0.19 0.07 0.08 0.10
0.21 0.02 0.10 0.02
6 6 6 6
0.20 0.09 0.09 0.08
7.32 3.53 3.37 3.41
6 6 6 6
1.42 0.46 0.38 0.40
0.25 0.17 0.05 0.13
6 6 6 6
0.16 0.07 0.06 0.07
0.05 0.08 0.15 0.05
6 6 6 6
0.18 0.08 0.07 0.07
1.21 0.48 0.39 0.52
0.42 0.06 0.06 0.08
6 6 6 6
0.20 0.09 0.12 0.10
0.73 0.07 0.07 0.12
6 6 6 6
0.15 0.07 0.11 0.16
7.54 3.70 3.64 3.40
6 6 6 6
1.21 0.48 0.42 0.41
0.42 0.15 0.11 0.02
6 6 6 6
0.16 0.09 0.11 0.09
0.37 0.04 0.15 0.10
6 6 6 6
0.30 0.07 0.11 0.09
Industry standard surveys were used to address four aspects of workplace performance (14–16): overall performance, quality of work, quantity of work and, quality of interactions with coworkers. A weekly survey was administered on Wednesdays to all employees. Supervisors completed similarly structured weekly surveys that focused on their employees’ workplace performance. Negative values represent a decline in performance.
desk intervention, 6 months later (starting in November of 2008 and finishing in November of 2009). Therefore, there was a 6-month period (May 2008 through November 2008) where 23 employees had entered the study but received no intervention. During the 6 months of surveillance, the 23 subjects showed no significant change in body weight from 6 months to baseline (82.8 6 18.2 at 6 months versus 83.5 6 19.4 kg at baseline). Subjects starting in May showed higher baseline daily physical activity than subjects starting in November (4,252 6 2,022 and 2,827 6 1,400 AU/day, P < 0.02). However, there were no group by time interaction (i.e., the pattern of changes within group were similar between groups). Weight loss occurred with the treadmill desk intervention, but for the group as a whole, it was modest. For the 36 participants, fat
mass did not change significantly, however, those who lost weight (n ¼ 22) lost 3.4 6 5.4 kg (P < 0.001) of fat mass. At baseline, subjects weighed 86.3 6 26.5 kg and after 12 months of intervention this decreased to 85.1 6 25.6 kg (P < 0.05) (Table 1). There was variability in weight change over the 12-month intervention; weight change ranged from 9 kg to þ4 kg. High Responders to the intervention were defined as having increased daily physical activity by 100 AU/day at Month 12. High responders lost more weight than non-responders (4 6 4 kg weight loss versus 1 6 3 kg weight loss; P < 0.05). For the 22 subjects that lost weight, body fat decreased (P < 0.001). Body Composition data (17) (Table 1) that demonstrated fat loss at 12 months were 3.4 6 5.4 kg. With the treadmill desk intervention, waist circumference significantly decreased for all subjects; 95 6 19 cm versus 91 6 18 cm (P <
TABLE 3 Energy expenditure (kcal/h) and energy efficiency (kcal/kg/h) measured using indirect calorimetry at baseline and after 6 and 12 months in 36 subjects enrolled in the treadmill desk intervention
Baseline RMR (kcal/h) 1 MPH (kcal/h) 2 MPH (kcal/h) 3 MPH (kcal/h) RMR (kcal/kg/h) Sitting (kcal/kg/h) 1 MPH (kcal/kg/h) 2 MPH (kcal/kg/h) 3 MPH (kcal/kg/h) FIGURE 2 Treadmill desk. Steelcase Walkstation in an Educational Credit Management Corporation employee cubical.
708
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
70 176 241 313 0.83 1.02 2.06 2.82 3.66
6 6 6 6 6 6 6 6 6
20 53 70 100 0.19 0.23 0.37 0.46 0.49
6 Months
12 Months
71 173 239 318 0.83 0.97 2.03 2.81 3.76
73 166 236 315 0.86 0.97 1.94 2.75 3.63
6 6 6 6 6 6 6 6 6
24 63 79 111 0.18 0.20 0.37 0.42 0.55
6 6 6 6 6 6 6 6 6
25 58 82 126 0.14 0.18 0.31 0.38 0.62
Data are shown as mean 6 SD.
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS: BEHAVIOR, PHARMACOTHERAPY, DEVICES, SURGERY
0.001). In subjects with obesity, waist circumference decreased from 101 6 26 to 96 6 25 cm (P < 0.001). Employees with obesity lost more weight than lean subjects (2.3 6 3.5 kg weight loss versus 0.7 6 2.3 kg weight loss; P ¼ 0.04). Employees with obesity significantly increased daily physical activity from baseline (3,672 6 1,959 AU/day) to 6 months (4,960 6 2,666 AU/day, P < 0.01) and continued to increase activity through 12 months (5,041 6 2,383 AU/day, P < 0.02). Lean subjects significantly increased daily physical activity; from, at baseline, 3,897 6 2,187 AU/day to, at 6 months, 4,713 6 2,229 AU/day (P < 0.05). However, this increase was not sustained; at 12 months daily physical activity was 3,990 6 2,513 AU/day (for decline, P < 0.07). For the entire group, HDL increased over the 1-year intervention from baseline, 55 6 20 mg/dl to 60 6 23 mg/dl (P < 0.005). No significant whole group changes were observed in triglycerides, glucose, LDL, TSH, and total cholesterol. Similarly, energy expenditure sand energy efficiency did not change for the group as a whole (Table 3) over the 1-year intervention.
Discussion Traditional chair-based desks were replaced with treadmill desks in 36 employees. We addressed the hypothesis that a 1-year intervention with access to treadmill desks was associated with increased employee daily physical activity. Our secondary hypothesis was that a 1-year intervention with access to treadmill desks was associated with decreased employee sedentariness. We found that with access to treadmill desks, employees’ daily physical activity increased over the course of a year and sedentariness decreased. There was no significant change in workplace performance. Weight loss occurred, especially in subjects with obesity. Access to treadmill desks may help promote daily physical activity in chair-based office workers. High income countries are battling sedentariness (18), overweight, obesity, and its health and fiscal consequences (19). The challenge of obesity extends into middle and low-income countries as well (20). Sedentariness is associated not only with obesity (1) but also with heart disease, diabetes, cerebrovascular diseases, cancer, and mood (18). One variable that determines sedentariness is job-type (21). This is especially important because most jobs in high income countries are screen based and sedentary (22). In addition, most people work for most of their waking hours and workdays are associated with more sedentary time than leisure days (8). With respect to energy expenditure and obesity, active jobs can lead to energy expenditure of 2,000 kcal per day more than sedentary jobs (21). Furthermore, obese people are sedentary for approximately 2 h/day more than their lean counterparts (7). The goal of the study was to examine the impact of transforming a traditional workspace (e.g., desk, chair, and computer) into an activity-promoting workspace (treadmill desk). We targeted a sedentary, computer-based, office environment because most Americans work in this manner (22,23). This type of work environment increases both screen time and seated time (24). One approach to increase office physical activity levels is through behavioral modification and improving access to health clubs (25–27). Another approach is to redesign the employees’ physical workspace in such a way that
www.obesityjournal.org
physical activity is environmentally promoted (28). For instance, meeting rooms, printers, and trash cans can be placed distant from where employees work. Another approach for reversing workplace sedentariness might be to replace desk chairs with treadmill desks. In pilot studies, we found that people can conduct daily work tasks such as computer use and telephone calls while walking on a treadmill (11). However, these were pilot studies. Here, we evaluated access to treadmill desks in a prospective 1-year trial using objective outcome measures and workplace performance assessments. We found that access to treadmill desk use was associated with increased daily activity, decreased sedentariness, and unaltered workplace performance. Although in this study, body weight decreased significantly, the average amount of weight loss was modest. This suggests that further efforts are needed to maximize the weight-associated health and cost benefits of active work. We suspect that a degree of compensatory eating occurred; maybe neurological mechanisms drove counter-regulatory responses to preserve body weight (29). Broadly, our observations are compatible with agricultural work studies which demonstrate that during peak harvest times, where physical activity increases, food intake increases as well (30). Thus, unless food intake is exogenously restricted with access to treadmill desks, weight loss cannot be maximized. Noting the degree of responsive compensatory eating that occurred with access to treadmill desks, we feel that future treadmill desk programs may incorporate programs of caloric restriction if the goal is to maximize weight loss (31). The implementation of treadmill desks was associated with increases in daily physical activity throughout the year. Importantly, the initial 6-month effect was greater than the 12-month effect and suggests that the effect of the environmental change began to wane within a year, as occurs with other interventions. We also noted that the impact of access to treadmill desks on daily physical activity occurred regardless of seasonal effect although season impacted baseline activity (Spring activity levels were greater than Fall). Seasonal variance in daily physical activity is known to occur (32); the observation that the intervention impacted employees working indoors, regardless of seasonality, is noteworthy. The interindividual consistency of daily activity we found to be intriguing. This is not the first observation of this phenomenon (8). Beyond self-selecting for more sedentary jobs (33) which cannot be the case here, it may be that people with obesity have less ‘‘appetite’’ for activity not because of habitus but because of biological drives to enhance sedentariness and potentially conserve metabolic fuel (3). While workplace performance was not affected significantly, the data suggested that there was an early adaptation response in the first 3 months, but thereafter employees adapted to the new work environment without impaired workplace performance. This is akin environmental interventions (34). Overall, with access to treadmill desks, workplace performance was unaffected. Obese employees benefited from access to treadmill desks, significantly more than lean people with respect to weight loss. This may contrast with other workplace interventions where obese people tend not to use free or subsidized gym memberships as much as lean people (35). With the degree of weight loss we found, the impact of access to treadmill desks may not seem to be cost effective. However, if a $3,000 desk helps prevent a person from developing diabetes, the
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
709
Obesity
Treadmill Desks and Physical Activity Koepp et al.
costs saved from diabetes approximate, $10,000/person/year (36). Thus, deploying such approaches to targeted individuals may prove cost effective. We recognize several limitations to this study. First, the treadmill desks were scattered throughout the workplace so that a treadmill desk employee sat next to a traditional seated desk user. Thus, we did not develop or support a ‘‘micro-community’’ of mutually supportive treadmill desk users; which might have increased daily activity further (37). Second, future interventions might benefit by embracing behavioral strategies to affect a sustained intervention (38). Third, this study was relatively short in duration. Thus, we feel that the next step is to link active work environments to sustainable behavioral programs (39), test these, and then proceed to longer term trials. Fourth, there were relatively few subjects in this study; this is important especially if we are suggesting that the walk-atwork approach may be widely applicable. Fifth, there was no control group matched for the entire intervention; for instance the Afternoon Dip we report (Figure 3) before the intervention began could have been anticipatory. It is, however, difficult to have a none-intervention control in a corporate setting over a year. Although we acknowledge this, we feel our conclusions are valid because the study was adequately powered to address our primary hypothesis and we were diligent to include a representative work population.
FIGURE 4 Hour-by-hour daily physical activity (AU/h) for 36 employees in a treadmill desk intervention at baseline, 6 months, and 12 months after treadmill desk intervention.
In conclusion, access to treadmill desks can help improve daily physical activity and reduce daily sedentary time.O
Acknowledgments We thank Mr. R. Boyle, Ms. C. Dubbs, and the employees of E.C.M.C. C 2012 The Obesity Society V
References
FIGURE 3 Daily physical activity (AU/day) (A) and daily sedentary time (min/day) (B) for 36 employees at baseline, 6 months, and 12 months after treadmill desk intervention. Comparisons to baseline, *P < 0.01, **P < 0.001. Data are shown as mean 6 standard error.
710
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
1. Hamilton MT, Hamilton DG, Zderic TW. Exercise physiology versus inactivity physiology: an essential concept for understanding lipoprotein lipase regulation. Exerc Sport Sci Rev 2004;32:161-166. 2. Duncan GE. Exercise, fitness, and cardiovascular disease risk in type 2 diabetes and the metabolic syndrome. Curr Diab Rep 2006;6:29-35. 3. Brown H, Roberts J. Exercising choice: the economic determinants of physical activity behaviour of an employed population. Soc Sci Med 2011;73:383-390. 4. Manning J. Healthcare Informatics 100. Company profile: Epic Systems Corp. Private club. Healthc Inform 1999;16:86. 5. Chaput JP, Tremblay A. Obesity and physical inactivity: the relevance of reconsidering the notion of sedentariness. Obes Facts 2009;2:249-254. 6. Belkic K, Nedic O. Workplace stressors and lifestyle-related cancer risk factors among female physicians: assessment using the occupational stress index. J Occup Health 2007;49:61-71. 7. Levine JA, Lanningham-Foster LM, McCrady SK, et al. Interindividual variation in posture allocation: possible role in human obesity. Science 2005;307:584-586. 8. McCrady SK, Levine JA. Sedentariness at work: how much do we really sit? Obesity (Silver Spring) 2009;17:2103-2105. 9. Dunstan DW, Kingwell BA, Larsen R, et al. Breaking up prolonged sitting reduces postprandial glucose and insulin responses. Diabetes Care 2012;35:976-983. 10. John D, Bassett D, Thompson D, Fairbrother J, Baldwin D. Effect of using a treadmill workstation on performance of simulated office work tasks. J Phys Act Health 2009;6:617-624. 11. Fidler JL, MacCarty RL, Swensen SJ, et al. Feasibility of using a walking workstation during CT image interpretation. J Am Coll Radiol 2008;5:1130-1136. 12. Thompson WG, Foster RC, Eide DS, Levine JA. Feasibility of a walking workstation to increase daily walking. Br J Sports Med 2008;42:225-228; discussion 8. 13. Miyatake N, Takenami S, Kawasaki Y, Kunihashi Y, Nishikawa H, Numata T. Clinical evaluation of body fat percentage in 11,833 Japanese measured by air displacement plethysmograph. Intern Med 2005;44:702-705. 14. Pronk NP, Martinson B, Kessler RC, Beck AL, Simon GE, Wang P. The association between work performance and physical activity, cardiorespiratory fitness, and obesity. J Occup Environ Med 2004;46:19-25. 15. Ben-Ner A, McCall BP, Stephane M, Wang H. Identity and in-group/out-group differentiation in work and giving behaviors: experimental evidence. J Econ Behav Org 2009;72:153-170. 16. Ben-Ner A, Putterman L. Trust, communication and contracts: an experiment. J Econ Behav Org 2009;70:106-121.
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS: BEHAVIOR, PHARMACOTHERAPY, DEVICES, SURGERY 17. Fields DA, Goran MI, McCrory MA. Body-composition assessment via air-displacement plethysmography in adults and children: a review. Am J Clin Nutr 2002;75: 453-467. 18. Dunstan DW, Barr EL, Healy GN, et al. Television viewing time and mortality: the Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Circulation 2010;121: 384-391. 19. Finkelstein EA, Trogdon JG, Cohen JW, Dietz W. Annual medical spending attributable to obesity: payer-and service-specific estimates. Health Aff (Millwood) 2009; 28:w822-831. 20. James WP. The epidemiology of obesity: the size of the problem. J Intern Med 2008;263:336-352. 21. Black AE, Coward WA, Cole TJ, Prentice AM. Human energy expenditure in affluent societies: an analysis of 574 doubly-labelled water measurements. Eur J Clin Nutr 1996;50:72-92. 22. Seghers J, Rutten C. Clustering of multiple lifestyle behaviours and its relationship with weight status and cardiorespiratory fitness in a sample of Flemish 11- to 12-year-olds. Public Health Nutr 2010;13:1838-1846. 23. National Telecommunications & Information Administration. http://www.ntia.doc.gov/. Accessed August 2011. 24. Thompson JL, Allen P, Cunningham-Sabo L, Yazzie DA, Curtis M, Davis SM. Environmental, policy, and cultural factors related to physical activity in sedentary American Indian women. Women Health 2002;36:59-74. 25. Brown WJ, Bauman AE, Owen N. Stand up, sit down, keep moving: turning circles in physical activity research? Br J Sports Med 2009;43:86-88. 26. Brownell KD. The LEARN Program for Weight Management 2000. Dallas, Texas: American Health Publishing Company; 2000. 27. Brownell KD, Stunkard AJ, Albaum JM. Evaluation and modification of exercise patterns in the natural environment. Am J Psychiatry 1980;137: 1540-1545.
www.obesityjournal.org
28. Zenk SN, Schulz AJ, Matthews SA, et al. Activity space environment and dietary and physical activity behaviors: a pilot study. Health Place 2011;17:1150-1161. 29. Leibel RL, Rosenbaum M, Hirsch J. Changes in energy expenditure resulting from altered body weight. N Engl J Med 1995;332:621-628. 30. Coward WA. Contributions of the doubly labeled water method to studies of energy balance in the Third World. Am J Clin Nutr 1998;68:962S-969S. 31. Levine JA, Vander Weg MW, Klesges RC. Increasing non-exercise activity thermogenesis: a NEAT way to increase energy expenditure in your patients. Obes Manag 2006;1:146-151. 32. Mitsui T, Barajima T, Kanachi M, Shimaoka K. Daily walking activity among male office workers in a rural town in northern Japan. J Physiol Anthropol 2010;29: 43-46. 33. Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71-80. 34. Lin CJ, Feng WY, Chao CJ, Tseng FY. Effects of VDT workstation lighting conditions on operator visual workload. Ind Health 2008;46:105-111. 35. Boriani F, Bocchiotti MA, Guiot C. ‘‘Self-sustainable’’ gym clubs tackling obesity: ‘‘exercise’’ for ‘‘energy’’. Exerc Sport Sci Rev 2008;36:212. 36. American Diabetes Association. http://www.diabetes.org/diabetes-basics/diabetesstatistics/. Accessed 1 August, 2011. 37. Paffenbarger RS, Jr., Hyde RT, Wing AL, Lee IM, Jung DL, Kampert JB. The association of changes in physical-activity level and other lifestyle characteristics with mortality among men. N Engl J Med 1993;328:538-545. 38. Wing RR, Greeno CG. Behavioural and psychosocial aspects of obesity and its treatment. Baillieres Clin Endocrinol Metab 1994;8:689-703. 39. Blair D, Buskirk ER. Habitual daily energy expenditure and activity levels of lean and adult-onset and child-onset obese women. Am J Clin Nutr 1987;45:540-550.
Obesity | VOLUME 21 | NUMBER 4 | APRIL 2013
711
NOTICE WARNING CONCERNING COPYRIGHT RESTRICTIONS
The copyright law of the United States [Title 17, United States Code] governs the making of photocopies or other reproductions of copyrighted material. Under certain conditions specified in the law, libraries and archives are authorized to furnish a photocopy or other reproduction. One of these specified conditions is that the reproduction is not to be used for any purpose other than private study, scholarship, or research. If a user makes a request for, or later uses, a photocopy or reproduction for purposes in excess of "fair use," that use may be liable for copyright infringement. This institution reserves the right to refuse to accept a copying order if, in its judgement, fullfillment of the order would involve violation of copyright law. No further reproduction and distribution of this copy is permitted by transmission or any other means.
Case Study 3 TREADMILL DESKS: A 1-YEAR PROSPECTIVE TRIAL
M. Yvonne Hidle CARNEGIE MELLON UNIVERSITY | VIVIAN LOFTNESS
Acoustic Privacy = Individual Productivity Task Performance with Different STI Conditions
Jahncke et al 2013 (office) 75
Math (p < 0.05): Assumed to translate into office work involving data calculations (4% office time) and designing, engineering and programming (9% office time)
70
65
Percentage of Correct Answers (%)
In a 2013 controlled lab experiment with 42 participants at the University of Gavle, Jahncke et al identifies an increase in individual productivity of up to 5% for word memory tasks (p < 0.0001), 6% for information searching tasks (p = 0.07), and 3% for math tasks (p < 0.05) when the Speech Transmission Index (STI) is kept below 0.34 in an open-plan office through the use of pink speech masking noise (between 40 and 45 dB), high cubicle walls (4.7 feet or greater), and sound absorption panels on the cubicle walls.
Phonemic Fluency: data considered insignificant (p > 0.05) 60
55
Information Search (p = 0.07): Assumed to translate into office work involving researching (9% office time)
50
Semantic Fluency: data considered insignificant (p > 0.05)
45
40 Word Memory (p < 0.0001): Assumed to translate into office work involving e-mail processing (11% office time) and reading (6% office time)
35
First cost increase: $1170 / employee Annual productivity savings: $800 / employee ROI: 68%
30 0.71
0.34
0.23
0.16
0.08
0.00
Speech Transmission Index (STI)
Reference: Jahncke, H., Hongisto, V., & Virjonen, P. (2013). Cognitive performance during irrelevant speech: Effects of speech intelligibility and office-task characteristics. Applied Acoustics, 74(3), 307-316.
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
Privacy – Jahncke et al 2013 Acoustic Privacy = Individual Productivity In a 2013 controlled lab experiment with 42 participants at the University of Gavle, Jahncke et al identifies an increase in individual productivity of up to 5% for word memory tasks (p < 0.0001), 6% for information searching tasks (p = 0.07), and 3% for math tasks (p < 0.05) when the Speech Transmission Index (STI) is kept below 0.34 in an open-plan office through the use of speech masking noise (between 40 and 45 dB), high cubicle walls (4.7 feet or greater), and sound absorption panels on the cubicle walls. Overview In a 2013 controlled lab experiment with 42 participants at the University of Gavle, Jahncke et al investigated the relationship between speech intelligibility and cognitive performance in an open-plan office environment. 6 Speech Transmission Index (STI) values were explored: 0.00, 0.08, 0.16, 0.23, 0.34, and 0.71. Where 0.00 was considered “silent” and 0.71 was considered “perfect sentence intelligibility.” 4 foot wide cubicle workstations with 4.7 foot high screens were used to simulate an open-office environment. Study conditions such as lighting, temperature, room background noise levels, and headphone noise levels were kept the same for all STI experiments. The 42 participants were divided into groups of 7 to 8. All participants were subjected to the same STI levels (0.00 through 0.71) and asked to perform the same tasks. All were compensated with 500 SEK (~$69 U.S. dollars) at the end of the experiment. Speech noise was played on headphones that all participants were required to wear. The speech noise consisted of 240 four word sentences, half of which were spoken by a female actress and the other half by a male actor, all words were in Swedish. 18 participants were female and the mean age was 24. Participants were students recruited from the University of Galve. Study Method Each participant was asked to complete 5 computer tasks in all 6 STI scenarios (for a total of 30 tasks): word memory task, information search task, math task, and word fluency task (two parts – semantic and phonemic). For the word memory task, participants were visually presented words on a computer screen; these words were organized into 8 lists with 10 words in each list. Lists were viewed one word at a time; 1000ms each word, and a 1000ms break between words. After being presented all words in a list, the participants were asked to recall all 10 words in the list in free order. The task took 6 minutes to perform. For the information search task, participants were asked to view a table of objects on the computer screen. Each table consisted of 20 rows and 7 columns. Each row had 1 object and each column described 1 characteristic of that object. “Participants were asked to find the object that met a set of criteria, by using 3 of the 7 columns.” (Jahncke, Hongisto, & Virjonen, 2013) There were 12 questions, with 1 minute allotted per question. The task took 12 minutes to perform. For the math task, participants were asked to solve basic addition problems. These problems consisted of adding 2 triple digit numbers. Problems were presented on the computer screen. There were 25 problems total, each problem had an 18 second time limit. After the 18 seconds were up, the computer would move to the next question. This task took 7 minutes to perform. For the word fluency task, the semantic part, participants were presented with a semantic category and then told to list as many words related to that category as possible in 60 seconds. Participants were asked to do this 2 times. The first part of this task took 2 minutes to perform. For the phonemic fluency part, participants were presented with a letter and instructed to list as many words as possible that start with that letter in 60 seconds. Participants were not allowed to use variants of the same word. Participants were asked to do this 2 times. The second part of this task took 2 minutes to perform. Each participant underwent a practice period for all tasks that was 5 minutes long and done in silence. Tasks and the STI condition were presented in a random order by the computer. The total experiment time was 3 hours and 15 minutes. Participants were given a 15 minute break in another room after half of the experiment time had elapsed. Study Results The study found that participants subjected to the 0.00 STI condition scored about 5% better in word memory tasks (p < 0.0001), 6% better in information searching tasks (p = 0.07), and 3% better (p < 0.05) in math tasks than participants subjected to the 0.71 STI condition. No significant score difference was found in the fluency task (semantic and phonemic) scores for all acoustic conditions.
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
November 2014
Task Performance with Different STI Conditions 75 Math (p < 0.05): Assumed to translate into office work involving data calculations (4% office time) and designing, engineering and programming (9% office time)
70
Percentage of Correct Answers (%)
65
Phonemic Fluency: data considered insignificant (p > 0.05)
60
55 Information Search (p = 0.07): Assumed to translate into office work involving researching (9% office time)
50
45 Semantic Fluency: data considered insignificant (p > 0.05)
40
35
30 0.71
0.34
0.23
0.16
0.08
0.00
Speech Transmission Index (STI)
Word Memory (p < 0.0001): Assumed to translate into office work involving email processing (11% office time) and reading (6% office time)
BIDS Life Cycle Assumptions The 2014 BIDS office baseline assumes a 100,000 square foot building, 500 employees with an average salary of $45,000. The first cost increase is $5.87/SF ($1,174.00 per employee) for the installation of a sound masking system (Miller, 2003), cubicle height extenders (OBEX), and additional cubicle sound insulation (Sound Away). The word memory task was assumed to translate into office work involving e-mail processing (11% office time) and reading (6% office time). The information search task was assumed to translate into office work involving researching (9% office time). And the math task was assumed to translate into office work involving data calculations (4% office time) and designing/engineering/programming (9% office time). The two word fluency tasks were not added to the calculation as the study found the relationship between the STI level and performance to be insignificant (p > 0.05). In a baseline building, a 5% increase in productivity for e-mail processing and reading, a 6% increase in productivity for researching, and a 3% increase in productivity for data calculations and designing/engineering/programming, due to improved acoustic conditions in an open-plan office, yields annual savings of $801.00/employee, resulting in an ROI of 68.2%. First cost increase per square foot: Baseline first cost increase: First cost increase per employee: Annual productivity savings per employee:
$5.87 $587,000 $1,174.00 $801.00
Baseline annual savings: NPV of the savings: ÂŽ EVA : ROI:
$400,500 $3,046,203 $2,459,203 68.2%
(see appendix for breakdown) ($5.87 x 100,000sf) ($587,000 / 500 employees) ($45,000 x 5% improvement x 17% of time + $45,000 x 6% improvement x 9% of time + $45,000 x 3% x 13% of time) ($801 x 500 employees) ($801 x 500 x 7.606) (NPV of the savings - FCI) (Savings per employee / FCI per employee)
Limitations One of the major limitations for the study was the lack of STI experiments between 0.34 and 0.71. Because the researchers expected the strongest change to occur between STI 0.00 and 0.34, they only had one outlying measure at 0.71. Future studies should have STI points between 0.34 and 1.
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
November 2014
There was some overlap between STI conditions 0.16 and 0.23, leading to situations where the two STI conditions were not sufficiently different from one another. The study recommends that all STI levels be more than 0.10 points apart. Headphones may not be fully representative of an actual open-plan office environment. Future studies should take into account the acoustic environment of real offices: reverberation, materials, layout, and direction and movement of speakers. Lastly, the study mentions that the content of the speech noise was very simple (four word sentences). If the sentences were more complex and/or the sentences were longer, the results from the study may be different. References Florida Sound Masking. (2008). How much does sound masking cost? Retrieved October 25, 2014, from Florida Sound Masking: http://www.floridasoundmasking.com/faq/sound-masking-faqs/75-how-much-does-sound-masking-cost Jahncke, H., Hongisto, V., & Virjonen, P. (2013). Cognitive performance during irrelevant speech: effects of speech intelligibility and office-task characteristics. Applied Acoustics , 307-316. OBEX. (n.d.). Cubicle Panel Extenders. Retrieved October 25, 2014, from OBEX: Office Panel Extenders: http://www.panelextenders.com/ Sound Away. (n.d.). UltraQuiet Acoustic Cotton Panels. Retrieved October 25, 2014, from Sound Away: http://www.soundaway.com/Acoustic_Cotton_Panels_s/91.htm Appendix Finalized Calculations Total Cost Total cost for acoustic improvements
$587,000
Building Area BIDS Baseline
100,000
Cost per SQFT Total Cost / Total Building Area
$5.87
Total Employees BIDS Baseline
500
Cost per Employee
$1,174.00
SQFT
Total Cost / Total Employees
Interior Construction Sound Masking
$2.00
Per SQFT
$3.00
Total Cost Per SQFT
$200,000.00
Cubicle Wall Extenders
Total cost Per SQFT
$300,000.00
$0.87
Total Cost
$87,000.00
Numbers from Florida Sound Masking
Numbers from OBEX Cubicle & Desk Extenders
($200 per panel x 3 panels per desk x 500 employees) /100,000 SQFT
Sound Absorption Panels for Cubicle Walls
Numbers from Sound Away. UltraQuiet Acoustic Cotton Panel 3#, 4" x 2' x 4', 8 square feet
($58 per panel x 3 panels per desk x 500 employees) /100,000 SQFT
F14-Productivity, Health and the Quality of Buildings M. Yvonne Hidle Overview: 1. Article citation: Jahncke, H., Hongisto, V., & Virjonen, P. (2013). Cognitive performance during irrelevant speech: Effects of speech intelligibility and office-task characteristics. Applied Acoustics, 74(3), 307-316. Retrieved October 22, 2014, from http://www.sciencedirect.com/science/article/pii/S0003682X12002629 2. Study objective: This study seeks to understand the effect of irrelevant speech on cognitive tasks in an open-plan office environment. 3. What type of study is it? Simulation study Building case study Cross sectional (multiple building) study Before and after study
x
Lab experiment Field experiment Meta-analysis (multiple studies) Other: _____________________________
4. When did the study occur? (eg., month(s)/year) Does not state 5. Where did the study occur? (eg., “Xerox headquarters, Boston”, “office buildings in Paris”) The study was conducted at a 60 square meter (~646 square foot) office laboratory at the University of Gavle. 6. How many people? 47 initially until 5 participants had to be removed. Data from 42 participants was taken. People characteristics: All participants had normal hearing and vision. They were recruited from the University of Gavle. 18 of the participants were female, the mean age was 24. Participants were awarded 500 SEK (~$69 U.S. dollars) at the end of the study. 7. How many buildings? 1 Building characteristics: The study was conducted at a 60 square meter (~646 square foot) office laboratory at the University of Gavle. The room was setup to resemble an open-plan office. It had artificial windows with artificial outdoor lighting. Workstations were separated with a 1.43 meter (4.7 foot) tall screen. Each cubicle workstation was approximately 1.24 meters (4 feet) wide. Ambient conditions were kept the same for all STI scenarios and participants: • The room temperature was kept at around 21°C (69.8°F). The CO2 was always below 1000 ppm. • “Vertical luminance was set to 480-520 lx on the height of the seating in each workstation.” (Jahncke, Hongisto, & Virjonen, 2013) • Background noise levels were kept at approximately 30dB. • Headphone sound level was kept at 33 dB.
BIDS™ Study Evaluation Form
48-729 Fall 2014
8. What design strategies were compared in the study? The study compared performance at certain tasks with varying STI (Speech Transmission Index) levels: 0.00, 0.08, 0.16, 0.23, 0.34, and 0.71. The 0.00 STI is referred to as the “silent condition” and 0.71 STI is referred to as the “perfect sentence intelligibility” condition. 9. What design strategy does the study support? The study recommends that STI levels be kept below 0.50 between the speaker and the listener when employees are doing “cognitive work that requires concentration and high short-term memory load.” (Jahncke, Hongisto, & Virjonen, 2013) Study method: 10. How were the design strategies measured? The design strategies were measured using five different tasks that were selected because they relate to work done in a typical office setting. Total time for the experiment was 3 hours and 15 minutes. • Word Memory Task (computer based) Participants were visually presented words on a computer screen; these words were organized into 8 lists with 10 words in each list. Lists were viewed one word at a time; 1000ms each word, and a 1000ms break between words. After being presented all words in a list, the participants were asked to recall all 10 words in the list in free order. The task took 6 minutes to perform. • Information Search Task (computer based) Participants were asked to view a table of objects on the computer screen. Each table consisted of 20 rows and 7 columns. Each row had 1 object and each column described 1 characteristic of that object. “Participants were asked to find the object that met a set of criteria, by using 3 of the 7 columns.” (Jahncke, Hongisto, & Virjonen, 2013) There were 12 questions, with 1 minute allotted per question. The task took 12 minutes to perform. • Math Task (computer based) Participants were asked to solve basic addition problems. These problems consisted of adding 2 triple digit numbers. Problems were presented on the computer screen. There were 25 problems total, each problem had an 18 second time limit. After the 18 seconds were up, the computer would move to the next question. This task took 7 minutes to perform. • Word Fluency Task (computer based) This task consisted of two parts, a semantic fluency and a phonemic fluency. For the semantic fluency part participants were presented with a semantic category and then told to list as many words related to that category as possible in 60 seconds. Participants were asked to do this 2 times. The first part of this task took 2 minutes to perform. For the phonemic fluency part, participants were presented with a letter and instructed to list as many words as possible that start with that letter in 60 seconds. Participants were not allowed to use variants of the same word. Participants were asked to do this 2 times. The first part of this task took 2 minutes to perform. 11. How was the benefit measured? Benefits were measured using the results of the 5 tasks mentioned earlier. • Word Memory Task The study found that the most significant decrease in performance happens between STI condition 0.23 and STI condition 0.34. In fact, participants scored 5% better in STI condition 0.00 than in STI condition 0.34. The study found no significant difference between STI conditions 0.00 and 0.23 and STI conditions 0.34 and 0.71. • Information Search Task BIDS™ Study Evaluation Form
48-729 Fall 2014
The study found a linear reduction in performance from STI conditions 0.16 to 0.71. Participants exposed to the STI condition 0.0 scored 6% better than those in STI condition 0.71. • Math Task The study found that participants scored 3% better when exposed to STI condition 0.00 than to STI condition 0.34. • Word Fluency Task No significant difference was found between any of the STI conditions and the two word fluency tasks. 12. Other measurements: None 13. How was the relationship between the design strategy and benefit determined? Simulation Logistic regression/odds ratio ANOVA Other statistical method: ______________________________ x Linear regression Observed only/no analytical method 14. How were confounding factors addressed? Confounding factors were avoided in the following ways: 1) All participants did all 5 tasks in all 6 STI scenarios. 2) Tasks and STI scenarios were presented to participants in a random order. 3) All participants had a practice session for each of the tasks. 4) All participants were subjected to the same environmental conditions (lighting, heating, CO2 etc.) 5) All participants listed to the same speech soundtrack. Study results 15. What is the quantitative benefit of the design action? The study found that STI conditions at or below 0.34 in a simulated open-office plan led to an increase in individual productivity by as much as 5% for word memory tasks (p < 0.0001), 6% for information searching tasks (p < 0.07), and 3% for math tasks (p < 0.05). The study recommends sound masking, sound absorption panels on cubicle screens, and higher cubicle screens to decrease the STI levels in an open-plan office. 16. List other significant relationships found: According to Figure 5 in the paper it appears that there is a productivity increase when participants do tasks in an STI environment that is not perfectly silent. For 3 out of 5 tasks there appears to be a slight rise in test scores in the 0.16 scenario versus the 0.00 scenario (the silent condition). This suggestions that max productivity may not arise from perfect silence but rather with a minimal amount of background noise. 17. List important non-significant factors: None 18. What are the limitations of the study? One of the major limitations for the study was the lack of STI experiments between 0.34 and 0.71. Because the researchers expected the strongest change to occur between STI 0.00 and 0.34, they only had one outlying measure at 0.71. Future studies should have STI points between 0.34 and 1. There was some overlap between STI conditions 0.16 and 0.23, leading to situations where the two STI conditions were not sufficiently different from one another. The study recommends that all STI levels be more than 0.10 points apart.
BIDS™ Study Evaluation Form
48-729 Fall 2014
Headphones may not be fully representative of an actual open-plan office environment. Future studies should take into account the acoustic environment of real offices: reverberation, materials, layout, and direction and movement of speakers. Lastly, the study mentions that the content of the speech noise was very simple (four word sentences). If the sentences were more complex and/or the sentences were longer, the results from the study may be different. 19. Costs provided in the study: Participants were awarded 500 SEK (~$69 U.S. dollars) at the end of the study. No other cost data was provided.
BIDSâ&#x201E;˘ Study Evaluation Form
48-729 Fall 2014
Applied Acoustics 74 (2013) 307–316
Contents lists available at SciVerse ScienceDirect
Applied Acoustics journal homepage: www.elsevier.com/locate/apacoust
Cognitive performance during irrelevant speech: Effects of speech intelligibility and office-task characteristics Helena Jahncke a,⇑, Valtteri Hongisto b, Petra Virjonen b a b
Environmental Psychology, Faculty of Engineering and Sustainable Development, University of Gävle, SE-801 76 Gävle, Sweden Finnish Institute of Occupational Health, Indoor Environment Laboratory, Turku, Finland
a r t i c l e
i n f o
Article history: Received 16 May 2012 Received in revised form 14 August 2012 Accepted 17 August 2012 Available online 24 October 2012 Keywords: Auditory distraction Speech Transmission Index Working memory Long-term memory
a b s t r a c t In open-plan offices, it is common that colleagues talk in the background while non-auditory tasks need to be performed. The aims of the present study were, first, to investigate how much cognitive performance is impaired by irrelevant background speech with varying degree of speech intelligibility, and second, to determine whether some office-related tasks are more susceptible to distraction than others. We tested a model developed by Hongisto [1] which predicts the decrease of cognitive performance as a function of speech intelligibility, depicted by the Speech Transmission Index (STI). Our results showed that the shape of the STI–performance curve, and the magnitude of the performance decrease, depend on task type. A semantic short-term memory task (i.e. word memory) was more sensitive to disruption by speech than a mathematical task. A word fluency task (i.e. retrieval of information from semantic long term memory) was not influenced by varying speech intelligibility. Moreover, performance on an Information search task was impaired by speech with high intelligibility only. The steepest slope in performance decrease appeared at a lower speech intelligibility value than predicted by Hongisto [1]. It also varied between cognitive tasks. The results of this study are useful in determining an appropriate target level for acoustic design in offices: This work demonstrates that attempts to minimize speech intelligibility will yield increases in cognitive performance with a varying degree, depending on the type of focal task. Ó 2012 Elsevier Ltd. All rights reserved.
1. Introduction Performance in memory based and non-memory based cognitive tasks declines during the exposure to normal levels, 51 dB (LAeq), of office noise compared to lower levels, 39 dB (LAeq) of the same noise [2]. The most disturbing noise source in open-plan offices, both with respect to wellbeing and performance, is other people’s speech (for a review, see Kaarlela-Tuomaala et al. [3]). Sound absorption materials, screens and artificial systems designed to mask sound are commonly used to reduce distraction from irrelevant speech and to improve speech privacy [4]. The overall noise level is not always the most important issue when assessing the acoustic environment in open-plan offices. An important factor is also the speech intelligibility of background speech [1]. According to the interference-by-process account of auditory distraction, the distraction is caused by a conflict that emerges between similar, competing processes [5]. Specifically, involuntary processing of the sound (e.g., interpreting the meaning of speech) produces distraction on tasks that necessitates the same, but deliberate, processes (e.g., interpreting the meaning of written prose). ⇑ Corresponding author. Tel.: +46 26 64 81 24. E-mail address: helena.jahncke@hig.se (H. Jahncke). 0003-682X/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.apacoust.2012.08.007
Tasks that do not involve similar processes as that involved in processing the sound, however, should not be susceptible to distraction by sound. For instance, when a task involves processing of order information, it is vulnerable to sounds which change across time, as the unavoidable processing of the order of changes in the sound interferes with the processing of serial order in the task [6]. In contrast, when a task does not involve serial order processes, it is not susceptible to distraction by order information conveyed by the sound [7]. This irrelevant sound effect (ISE) is well documented, especially for the serial recall task, in which a sequence of numbers is to be remembered and recalled in the order of presentation [8]. Further, when a task involves processing of semantic information (i.e. reading a text) the task is more vulnerable to speech sounds, as speech sounds are also semantically processed [5,9]. In the present study we will incorporate cognitive tasks that are more applied and resemble more typical office work, though they still involve seriation (the process of serially-ordering to be remembered material) and semantic processing, such as a search for information task and a math task. Hongisto [1] has sketched a model of the function between Speech Transmission Index, STI, and performance on some cognitive tasks (e.g. serial recall, proofreading and reading comprehension). The Speech Transmission Index is an objective predictor of
308
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
subjective speech intelligibility [10]. The STI values range between 0 and 1. A more detailed description of STI is presented in Section 2.4.2. The model has been formulated by Hongisto et al. [11] and is shown in Fig. 1. However, the model was based on a review of only a few experimental studies where task performance was measured at a couple of STI levels and the experimental data was collectively fitted to a sigmoidal model. According to Fig. 1, the magnitude of successful task performance will start decreasing when STI exceeds 0.2 and the largest decline will be reached above 0.50 (0.00 = not intelligible; 1.00 = perfectly intelligibly speech). The basic idea of the model has been applied in a recently published international room acoustic measurement standard for open-plan offices [16] and in the Finnish acoustic design guideline for open-plan offices [17]. Some recent studies have, to a certain extent, supported the model [14,15]. Hongisto’s model has some distinctive characteristics. First, it makes a quantitative prediction about how much performance will decrease when the STI of the irrelevant speech increases, the maximum being a 7% decrease (i.e. difference in error percentage between silence and speech) for the tasks summarized. Second, the model assumes a steep linear slope in the middle of the STI-values and a flattening around the high and low STI-values. This implies that a test of the model should compare the linear, quadratic and cubic trends of the function between STI and performance. In the current paper, the foregoing predictions of Hongisto’s model—and the applicability of the model in the broader sense— will be tested using an experimental design with a larger body of cognitive tasks than was originally included in the review of Hongisto [1]. There are two issues that we elaborate in the present study. First, performance in the steep middle segment of the STI
curve, and at which higher and lower STI values the curve levels out, will be studied. Exploring this area would be particularly informative. Second, it was stated by Hongisto [1] that this model might not apply to all kinds of cognitive tasks, and should be tested experimentally in the future within the field of cognitive psychology. It can, for example, be argued that tasks with specific characteristics might be more susceptible to variations in intelligibility of background speech than other tasks. Therefore, the extent to which the tasks involve certain characteristics (e.g. search for relevant information, to count, to remember material and to generate words) may be related to the precise point on the STI curve at which the maximum performance reduction will be reached. The aim of the present study is to further specify the function of speech intelligibility and performance in the STI range from 0.10 to 0.40, where the most significant changes in performance are expected to occur, and see whether the STI–performance functions differ between cognitive tasks. The STI of speech was adjusted by changing the signal-to-noise ratio of a single speaker with speech masking sound, which resembled ventilation noise. We have selected four tasks which tap some fundamental characteristics of office work: to remember material (called Word memory task), to search for relevant information amongst distractors (called Information search task), to count (called Math task), and to generate words (called Word fluency task). Specifically, our hypotheses are the following: Hypothesis 1. The task performance depends on the STI of speech in such a way that the performance decrement starts around STI 0.20 and flattens out above STI 0.50. It is expected that the shape of the performance versus STI function will conform to Fig. 1. Hypothesis 2. The maximum performance decrement achieved at high STI values depends on task type. That is, the maximum performance decrement is not 7% for all task types as Fig. 1 suggests.
2
2. Materials and methods
Performance change [%]
0
2.1. Design
-2
The hypotheses were tested by using a within-participants design with six acoustic conditions (STI 0.00, 0.08, 0.16, 0.23, 0.34, 0.71) and four tasks with different office-work characteristics (verbal fluency, word memory, information searching and counting) as dependent variables.
-4 -6 -8
2.2. Participants
-10 -12 -14
0.0
0.2
0.4
0.6
0.8
1.0
Speech Transmission Index Hongisto's model [1] Ellermeier and Hellbrück (1998) Exp. 2A [12] Ellermeier and Hellbrück (1998) Exp. 2B ]12] Venetjoki et al. (2006) [13] Haka et al. (2009) Task1 [14] Haka et al. (2009) Task2 [14] Schlittmeier et al. (2008) Exp. 1 [15] Schlittmeier et al. (2008) Exp. 2 [15] Fig. 1. The schematic prediction model which gives the change of performance as a function of the STI (modified from Hongisto et al. [11]). Normal performance is obtained when no speech is heard (STI = 0.00), as between nearby office rooms. The highest performance decrease is reached when speech is highly intelligible (STI > 0.70), as between nearby workstations in an open-plan office with poor acoustic privacy. (See above-mentioned references for further information.)
The participants were 47 persons with normal hearing and vision recruited from the University of Gävle. Five of the participants were excluded because one of the persons in one of the groups disrupted the others, leaving 42 (18 female) for the statistical analyses (mean age 24 years). Seven to eight participants were tested at the same time except during the session with disruptions. Participants were informed of the nature of the study before participating and they were rewarded with 500 SEK after their participation. 2.3. Research setting The research was carried out in an office laboratory (60 m2) at the University of Gävle. The room was designed to resemble an open-plan office including windows with artificial outdoor lighting. The workstations were separated by 1.43 m high screens. Each cubicle was 1.24 m wide. Ambient conditions of the indoor environment were kept fairly constant throughout the experiments.
309
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
The room temperature was around 21 °C and the CO2 concentration was always well below 1000 ppm. Vertical luminance was set to 480–520 lx on the height of the seating in each workstation and met the current recommendations in Sweden for work with computers [18]. The background noise level in the room was damped to approximately 30 dB (LAeq) not to disturb the acoustic condition STI 0.00 (silence condition) wherein the sound level played by the headphones was only 33 dB (LAeq). 2.4. Acoustic conditions 2.4.1. Speech sounds The original speech material consisted of 120 different fourword sentences spoken by a female actress and 120 different sentences spoken by a male actor, thus yielding 240 sentences. These sentences were spoken in Swedish and taken from the Hagerman test [19], which is part of a package of standardized tests for speech audiometry. The sentences were balanced as they all contain the same number of objects, names, verbs, etc. (e.g. ‘‘Japanese cars are cheap.’’). The duration of each sentence was three seconds. The speech recording was made in anechoic conditions at a distance of approximately 1 m from the mouth using a 1/2-in. freefield microphone (Brüel & Kjaer 4190). The total amount of anechoic speech material was 12 min. To create a realistic impression of speech heard from a nearby workstation in an open-plan office, the anechoic sentences were re-recorded in a small open-plan office in Turku. The room has been described in Haapakangas et al. [20]. The anechoic monophonic sentence files were played back one after the other through a loudspeaker (Genelec 8020) in one desk and recorded binaurally in another desk with a head and torso simulator (B&K 4100). This evaded moderate reverberation changes and binaural effects due to head movements. In all, the acoustic conditions resembled a situation wherein the participant was sitting in an office desk listening to speech from the opposite desk. The distance between the listener and the speaker was 2.2 m and they were separated by a 1.3 m high screen. Thereafter, the binaural sentence files were arranged into 25 sentence lists (five tasks for five STI levels including speech), where the sentences were separated by a 1000 ms pause. The length of the sentence lists was designed to exceed the time it would take to complete the tasks by at least 5 min. The same sentences were used multiple times in a sentence list to obtain sufficiently long lists. Their presentation order was pseudo-randomized (i.e., no more than three female or three male sentences were presented consecutively and the same sentence was separated by at least three other sentences). 2.4.2. Acoustic conditions One way to depict how well speech is perceived in a certain room acoustic condition is to use the Speech Transmission Index, STI [10]. Speech is perfectly intelligible when STI reaches 1.00 and completely non-intelligible when STI is 0.00. STI reduces with increasing reverberation time and increasing background noise level, because they interfere with the original speech signal. The STI method takes the room reverberation and background noise effects automatically into account. The equations behind the determination of STI are given, for instance, in Larm (at present: Virjonen) and Hongisto [21] and will not be repeated here. Subjective speech intelligibility of a specific STI value is dependent on whether speech items are syllables, words or sentences [1]. For example, perfect speech intelligibility of sentences is achieved at lower STI values than for words or syllables because of the greater semantic content that can be used as cues to perception. Hongisto’s model assumes that the intelligibility of full sentences is of main importance in applied settings as normal speech typically consists of full
sentences. The sigmoid model of Fig. 1 uses the relation of sentence intelligibility and STI as published in [10]. We tested the effects of solo speech at six levels of STI. The planned STI values of the acoustic conditions were 0.00, 0.08, 0.16, 0.23, 0.34 and 0.71, respectively. STI 0.00 was a reference condition where speech was absent. It represented silence during which very silent masking sound was played at almost unnoticeable level of 33 dB (LAeq) corresponding with silent ventilation noise in typical office rooms. The other five conditions contained speech and masking sound of varying pressure level, while maintaining a constant total sound level of 51 dB (LAeq). This level is the typical overall activity noise level in open-plan offices [3]. The STI values were chosen to cover the critical area of STI where a drop in performance was expected to take place (STI below 0.30). STI 0.71 was a reference condition of speech with perfect sentence intelligibility. The STI of the acoustic conditions including speech was changed by varying the signal-to-noise ratio of speech and masking sound (Table 1). The STI distribution in the acoustic conditions containing speech is presented in Fig. 2. The distribution was determined by defining the STI of every sentence individually. To elucidate the normal variation of the STI and speech level between individual sentences, an example is given from the acoustic condition STI 0.71, presented in Fig. 3. The time-averaged spectra of speech and masking sound in the acoustic conditions are presented in Fig. 4.
2.4.3. Determination of STI The STI of speech was determined according to the procedure of Houtgast and Steeneken [22] which is also applied in the measurement standard for open-plan offices [16]. The STI was determined for each binaural sentence. The modulation transfer functions (MTFs) were measured for the workstation pair used in the binaural recording (WinMLS 2004). The early decay time (EDT) of the office room used for binaural recordings in Turku was 0.32, 0.37, 0.36, 0.28, 0.31, 0.32, and 0.27 in octave bands 125–8000 Hz. This represents the early decay time measured in highly sound-absorbing open-plan offices [4]. The speech level of reverberant speech sentences was determined in octave bands (Brüel & Kjaer SQ 14.1.1) for the left and the right ear separately and they were averaged. STI was calculated for each sentence using the measured MTF and the average speech and background noise levels. The variation of STI between sentences (Figs. 2 and 3) is a normal feature of speech. The sentences were short and all phonemes and consonants of the Swedish language were not present in each sentence. Therefore, the octave band spectrum of every sentence was different. This results in minor changes between the STI of the sentences even if the A-weighted sound level of the sentences would be constant. However, the acoustic conditions were successfully implemented since the average STI values and related standard deviations did not overlap significantly between successive acoustic conditions.
Table 1 The description of the six acoustic conditions. The names of the acoustic conditions represent also their measured time-averaged STI values. LA,S = A-weighted level of speech, LA,M = A-weighted level of masking sound, and LA,tot = A-weighted level of the sum of speech and masking. All levels are equivalent sound pressure levels. Acoustic condition 0.00 0.08 0.16 0.23 0.34 0.71
LA,S (dB)
LA,M (dB)
LA,tot (dB)
39.6 43.2 45.3 47.9 51.0
33.0 50.7 50.2 49.6 48.1 33.2
33.0 51.0 51.0 51.0 51.0 51.1
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
0.80
60
0.70
50
Sound level [dB]
Measured STI of sentences.
310
0.60 0.50 0.40 0.30
STI 0.71
STI 0.16
STI 0.34
STI 0.08
STI 0.23
STI 0.00
40 30 20 10
0.20
60
0.10 0.00 0.08
0.16
0.23
0.34
0.71
Acoustic condition (Nominal STI value) Fig. 2. The statistical distribution of STI of individual sentences in the acoustic conditions containing speech. Square: average. Whiskers: standard deviation (67%). Bars: minimum and maximum values.
Sound level [dB]
50 40 30 20 10
1.00
60
70 Speech level
STI
40 30 20
Frequency [Hz]
8000
4000
2000
500
1000
250
125
8000
2000
4000
1000
250
500
10 125
STI
Speech level L A,S [dB]
Sound level [dB]
50
Frequency [Hz]
Fig. 4. The linear time-averaged sound pressure levels of speech (black square) and masking sound (hollow triangle) in the six acoustic conditions at octave bands. These values were used to determine the average STI values.
0.00 300
280
260
240
220
200
180
160
140
120
80
100
60
40
0
20
40
Sentence number of the sound track Fig. 3. A typical example of the variation of the STI and speech level of individual sentences in sound tracks. In this example, the mean STI was 0.71 and the mean speech level was 51 dB (LAeq).
2.4.4. Sound tracks and the addition of masking sounds The final sound tracks, which were played in the experiment, were created by mixing the binaural sentence lists with masking sound of adequate level (Adobe Audition 3.0). The masking sound was anechoic and monaural filtered pink noise (Adobe Audition 3.0). The spectrum of the masking sound was adjusted to 5 dB per octave band (Brüel & Kjaer SQ 14.1.1) as suggested by a recent study by Haapakangas et al. [20]. The sound resembled comfortable ventilation noise. Altogether, 25 sound tracks containing speech were prepared. The 26th sound track represented the silence condition (STI 0.00). The sound tracks were presented via headphones (Philips SHP8900) using the VLC-media player software (VLC 0.9.9, VideoLAN Organization, 2009). The computer managed the starting and stopping of sound playback. After the completion of each task,
the playback of the sound track was stopped until the beginning of the next task. The equivalent sound level of each sound track containing speech was adjusted to 51 dB (LAeq). The level at both ears, respectively, was checked using the head and torso simulator (B&K 4100) in each workstation, using a calibration signal stored at the end of each track. 2.5. Cognitive tasks 2.5.1. Word memory task Memory for words was measured with a simple short-term memory task. Short-term memory has relevance for office workers as they need to keep their memory updated daily with the information of importance. The participants were visually-presented with words from eight lists comprising ten words each. The lists were composed of words drawn from the same semantic category (e.g., ‘‘animals’’). After being presented with a given list of ten words, the participants were required to immediately recall the words from the most recent list in free order. This task puts substantial demands on both encoding and retrieval processes in semantic short-term memory. Each experimental block (i.e. six blocks) consisted of diverse categories: ‘‘fish and other water living
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
creatures’’, ‘‘fruit and vegetables’’, ‘‘animals’’, ‘‘food and beverages’’, ‘‘countries’’ and ‘‘cities in Sweden’’. For the practice session we used the category ‘‘cars’’. The words were presented sequentially on the computer screen, one every 1000 ms and were each visible for 1000 ms. The order of the categories was counterbalanced between participants and acoustic conditions. The scores were the sum of correct answers and the task took approximately 6 min to perform. 2.5.2. Information search task This task was originally designed to have relevance for office work as it taps processes required to search through and understand the contents of a table of information, while successively updating and memorizing information that accords with a target criterion [23]. The table contained 20 rows and seven columns. For each row, an object was presented (e.g., a person, a house or a country), and each column described one aspect of the object (price, location, year, etc.). The participants were asked to find the object that met a set of criteria, by using three of the seven columns (e.g. ‘‘Which building of row house type was built after the 70-s and is the most expensive?’’). Each experimental block (i.e. six blocks) consisted of 12 questions (the same table was used for each question) and the time was limited to 1 min per question before a new question was presented. The scores were the sum of correct answers and the task took 12 min to perform. 2.5.3. Math task This task was a simple addition task where participants were instructed to add triple-digit numbers, e.g. 146 + 309 [23]. Each experimental block (i.e. six blocks) consisted of 25 triple-digit expressions presented one by one together with an answer box. Every second expression had two carry over operations and the others had one carry over operation. The carrying component of mental arithmetic can place demands on inhibition and memory processes of interim results in working memory [24], depending on how the expression is solved. Some kind of episodic or working memory is therefore required (i.e. storage of a solution). Each experimental block (i.e. six blocks) consisted of diverse numbers to sum. The time limit for each expression was 18 s, after which the computer program continued to the next expression if an answer was not registered. The scores were the sum of correct answers. The task took approximately 7 min to perform.
311
type of human dwelling; a type of reading material; a tree; a type of fabric; an alcoholic beverage; a toy; and a type of dance. The scores were the sum of correctly generated words and the task took 2 min to perform. In the Phonemic fluency task the participants were presented with a letter and instructed to generate as many words beginning with that letter as possible during 60 s. However, they were not allowed to generate variants of the same word: e.g., sail, sailing, and they were told to avoid producing proper nouns (e.g. names of places, people and products). The letter selection for the phonemic fluency task was based on a study by Borkowski et al. [27] that defines the difficulty level of the letters according to the frequency of the candidate words. In each of the six versions of the task two letters were presented. One letter was taken from the easy range category of difficulty (i.e. H, D, M, A, B, F) and one from the moderate range category (i.e. O, N, E, G, L, R), to keep the difficulty level fairly constant between the versions of the task. The scores were the sum of correctly generated words and the task took 2 min to perform. 2.6. Procedure The participants were first instructed about voluntarily participation and the anonymity of their results. When the participants were seated in the open-plan office they were told to put on their headphones. They were instructed to wear them during the whole session and adhere to specific instructions. They were further told that they were not allowed to speak with each other and that the experimental instructor would be present in the room during the whole experiment (behind a wall screen). The session started with computerized background questions, followed by a practice period of all tasks for 5 min in silence. Every participant conducted the five tasks in the six acoustic conditions, that is, altogether 30 tasks. The tasks and the acoustic conditions were presented in randomized order generated by the computer for each participant (i.e. the participants had different procedures within the experiment); with the constraint that two of the same tasks were not presented consecutively and not more than three successive tasks occurred in the same acoustic condition. Each task started with an instruction in quiet (without masking). After half of the time the participants had a 15 min break in another room. The whole procedure took about 3 h 15 min. 2.7. Statistical analyses
2.5.4. Word fluency tasks (semantic- and phonemic fluency) The word fluency task (method used by Troyer et al. [25]) consists of two tasks: semantic- and phonemic fluency, which make similar demands on retrieval from long term memory but rely on different activation and search processes [9]. As both tasks are based on long term memory, none of them involve rehearsal of to-be-remembered material. These two tasks are also relevant for office work in the respect that they tap processes to generate written texts and ideas. In more detail, the procedure for the Semantic fluency task required participants to generate (write) as many exemplars from a given semantic category as possible within 60 s. Directly after the first category, another category was presented with the same time limit for responses. We used the category norms of van Overschelde et al. [26] to select categories. For each of the six versions of the task developed for the present experiment, one category was taken from within the category mean potency rank number 33–38 (i.e. mean frequency of generated words was 5.7–5.8 words) and one was taken from within the rank number 44–50 (mean frequency of generated words was 5.2–5.3 words) to keep the difficulty level fairly constant between the versions of the task. The selected categories were: a thing that flies; a substance for flavoring food; a liquid; a weapon; a carpenter´s tool; a
The performance in each STI level was directly converted to percentages of maximum number of correct answers, PSTI (%). For the word memory, information search, and math task, the maximum score was the maximum possible score (i.e. word memory: 80, information search: 12, math: 25). For the two word fluency tasks (semantic-, phonemic fluency), the best score generated by a participant during the acoustic condition where the best performance was achieved, was used as the maximum score (i.e. 30 was used for both tasks). This maximum score was different to the other tasks as there are no obvious limits in word generation. The absolute performance change, DP, was calculated as the difference in performance between the acoustic condition containing speech (non-zero STI values) and silence using the equation: DP = PSTI P0.00. The data presented in Fig. 1 was also determined using this definition. A General Linear Model (GLM, SPSS version 18) with repeated measures design was used for testing the first hypothesis. When Mauchly’s test indicated non-sphericity in the variance–covariance matrix for the within-subject analyses, the Greenhouse–Geisser adjusted degrees of freedom and corresponding p-values are reported for the F-tests. When the form of the STI–performance function was tested, unequal spacing between the STI values was
312
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
adopted by specifying the values (00, 08, 16, 23, 34, 71) in the Polynomial contrast. This was done to avoid inflated steepness between the steps 34 and 71. For testing Hypothesis 2 (7% performance decrement) we used one-way paired t-tests as these comparisons were planned in advance and not a priori.
2
Performance change [%]
0
3. Results 3.1. Main findings over all tasks and acoustic conditions A GLM analysis with acoustic condition and task (6 5) as within-participants factors with unequal spacing between the acoustic conditions revealed a significant main effect of acoustic condition, F (4.2, 170) = 5.20, p < 0.001, partial N2 = 0.11. This indicates that performance varied over acoustic conditions, which is in accordance with Hypothesis 1. Further, there was a significant main effect of task, F (2.2, 89) = 41.4, p < 0.0001, partial N2 = 0.50, indicating that the Math task was the easiest task while the Word memory task was the most difficult one (Fig. 5). As shown in Fig. 6, performance did not change in the STI range 0.00–0.23. This is in agreement with the model of Hongisto [1] (Fig. 1). However, the maximum decrease in performance was already achieved around STI 0.34. Fig. 1 suggested that performance decrement would continue until approximately STI 0.50, above which the curve flattens out. We also analyzed whether the steepest slope of the impaired performance flattened out at one end of the STI-spectrum (quadratic components) or at both ends (cubic components). When analyzing the form of the function for STI and performance (all tasks grouped together) we found a significant linear trend, F (1, 41) = 14.2, p < 0.001, partial N2 = 0.26, and a significant cubic trend, F (1, 41) = 4.2, p < 0.05, partial N2 = 0.10. As indicated by Fig. 6, there is a linear decrease in performance between STI 0.16 and 0.34. Further, the significant cubic trend indicates a knee
70
Percentage of correct answers [%]
-4 -6 -8 -10 -12 0.0
0.2
0.4
0.6
0.8
1.0
Speech Transmission Index A. Average of all tasks B. Average of Math, Word memory and Information search tasks Fig. 6. Task-averaged performance change between acoustic conditions. Curve B is comparable with the model of Fig. 1 since it contains only the tasks where there was a significant difference in performance between acoustic conditions.
around STI 0.16 and another knee from STI 0.34 where the curve flattens out. The STI–performance function for each task separately is shown in Fig. 5. Next, we investigated each task independently and tested the form of the function for STI and performance, and whether and to what extent the lowest STI-values differed from the higher values (STI 0.23, 0.34 and 0.71), with planned comparisons. 3.2. Word memory task The analysis with acoustic condition as within-subjects factor revealed a significant effect of acoustic condition on performance (P (%)), F (5, 205) = 6.6, p < 0.0001, partial N2 = 0.14. There was a significant linear trend when taking unequal spacing between the acoustic conditions into account, F (1, 41) = 23.5, p < 0.001, partial N2 = 0.36; and with a tendency towards a significant cubic trend, F (1, 41) = 3.5, p < 0.07, partial N2 = 0.08 (see Fig. 5). The paired-comparisons (planned, one-tailed) showed that performances in the acoustic conditions from STI 0.00 to 0.23 were significantly different from those in STI 0.34 to 0.71 (see Table 2). There was no significant difference between STI 0.00 and 0.23. The major reduction in word memory performance takes place from STI 0.23 to 0.34, which supports the results from the first general analysis with all tasks grouped together (Fig. 5). There was no significant difference
75
65 60 55 50 45 40
Table 2 The t-test comparisons (t-values, one-tailed) of performance between the acoustic conditions in the Word memory task. Asterisks describe the statistical significance level of the difference between the STI values. The t-values for the planned comparisons are bolded.
35 30
-2
0.0
0.2
0.4
0.6
0.8
1.0
Speech Transmission Index Math Phonemic fluency Information search Semantic fluency Word memory Fig. 5. The STI–performance function for the five cognitive tasks.
** ***
Acoustic condition
0.08
0.16
0.23
0.34
0.71
0.00 0.08 0.16 0.23 0.34 0.71
0.1
1.2 1.1
1.5 1.3 0.3
3.9⁄⁄⁄ 4.1⁄⁄⁄ 2.7⁄⁄ 2.5⁄⁄
4.3⁄⁄⁄ 3.6⁄⁄⁄ 3.3⁄⁄⁄ 2.4⁄⁄ 0.5
p < .01. p < .001.
313
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
between STI 0.71 and 0.34 which confirm that the ultimate performance drop is achieved at much lower STI values than Hongisto [1] suggests. Further, the performance decreased with about 5% from baseline (STI 0.00) to the most critical STI-value for performance (i.e. STI 0.34).
Table 4 The t-test comparisons (one-tailed) of performance between the acoustic conditions in the Information search task. Asterisks describe the statistical significance level of the difference between the STI values. The t-values for the planned comparisons are bolded.
3.3. Math task The analysis with acoustic condition as the within-subjects factor revealed a significant effect of acoustic condition on performance (P (%)), F (5, 205) = 2.5, p < 0.05, partial N2 = 0.06. This effect had a significant cubic trend when taking unequal spacing into account, F (1, 41) = 8.8, p < 0.01, partial N2 = 0.18; but no significant linear trend F (1, 41) = 1.5, p > 0.05, partial N2 = 0.04 (see Fig. 5). The paired-comparisons showed that performance in the acoustic conditions 0.08, 0.16, and 0.23 were not significantly different from each other (see Table 3). However, performance in all of them was significantly better than in STI 0.34, where the worst performance was reached. The performance in silence (STI 0.00) also tended to be better than with STI 0.34. The performance difference between STI 0.71 and 0.34 was not statistically significant and one cannot, therefore, state that performance at STI 0.71 would be better than at STI 0.34 as the tendency of the curve might indicate. In the Math task the performance decrease was less than 3% from baseline (STI 0.00) to the most critical STI-value for performance (i.e. STI 0.34).
3.4. Information search task The analysis with acoustic condition as within-subjects factor revealed no significant effect of acoustic condition on performance (P (%)), F (5, 205) = 1.3, p > 0.05, partial N2 = 0.03. However, there was a tendency for a linear trend when taking unequal spacing into account, F (1, 41) = 3.4, p = 0.07, partial N2 = 0.08. As seen in Fig. 5, performance tended to follow a linear reduction from STI 0.16 to STI 0.71. The paired-comparisons showed that the acoustic conditions 0.00, 0.08, 0.16 and 0.23 did not differ significantly from each other (see Table 4). The comparison revealed, however, that the participants’ performance was significantly poorer in STI 0.71 than in STI 0.00 and 0.16. The significant decrease in performance from STI 0.00 to STI 0.71 was about 6%.
3.5. Word fluency tasks There were no significant effects (F < 1 and p > 0.05) of acoustic condition on the two Word fluency tasks (semantic- and phonemic fluency).
Table 3 The t-test comparisons (t-values, one-tailed) of performance between the acoustic conditions in the Math task. Asterisks describe the statistical significance level of the difference between the STI values. The t-values for the planned comparisons are bolded. Acoustic condition 0.00 0.08 0.16 0.23 0.34 0.71 * ** ***
p < .05. p < .01. p < .001.
0.08 0.5
0.16 2.0* 1.4
0.23 0.8 0.3 1.9
0.34 1.5 2.0⁄ 3.3⁄⁄⁄ 2.7⁄⁄
0.71 0.3 0.7 2.1⁄ 0.9 1.2
*
Acoustic condition
0.08
0.00 0.08 0.16 0.23 0.34 0.71
1.2
0.16
0.23
0.34
0.71
0.1 1.4
0.4 0.8 0.6
0.9 0.1 1.0 0.6
2.2⁄ 0.8 1.8⁄ 1.6 0.9
p < .05.
4. Discussion We addressed two issues in Hongisto’s [1] prediction model. First, we tested the general STI (Speech Transmission Index)– performance function for a variety of office tasks. Second, we investigated whether the average maximum performance decrement differed in magnitude between the office-tasks. 4.1. The general function of speech intelligibility and performance We found a main effect of acoustic condition on task performance in accordance with Hongisto’s [1] model, in two out of five tasks. In a further task we also found a difference in performance between the highest and lowest STI condition. The average of these three tasks is presented in Fig. 6, curve B. It can be compared to the curve of Fig. 1, which also represents an average result over tasks selected whereby a significant performance change was reported in the literature. However, the general form of the STI–performance functions deviated somewhat from Hongisto’s model. The steepest slope of overall performance decrease was located between 0.23 and 0.34 and the difference between 0.34 and 0.71 was negligible. These results partly disclaim Hypothesis 1, as Hongisto’s model suggests that the steep slope of performance deterioration flattens out later just above STI 0.50. Also, the type of function was different depending on which task was performed, which was not considered in the Hongisto’s model. For instance, for the Math task performance seems to drop off first from STI 0.16. However, the spread of the scatter points in Hongisto’s model is wide and our results at STI 0.71 do not differ more from the model than the data point from the study by, e.g. Venetjoki et al. [13]. 4.2. The influence of task characteristics on the maximum performance decrement The STI–performance functions looked different for the Word memory task and the Math task. Because of this, we investigated the difference between acoustic conditions in more detail and sought to determine whether the magnitude of maximum performance decrement differed between task characteristics. The higher STI-values (STI 0.23, 0.34 and 0.71) made a difference to task performance only in certain tasks. In the Word memory task, performance decreased about 5% from baseline (STI 0.00) to the most critical STI-value for performance (i.e. STI 0.34). As the Word memory task places high demands on both encoding and retrieval processes in semantic short-term memory, the results are compatible with the interference-by-process view, according to which auditory distraction is a result of a conflict between similar processing of the sound and the focal task [9]. The more intelligible the background speech was, the greater the performance impairment in this semantic short-term memory task.
314
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
In the Math task, there was less than a 3% decrease in performance from baseline to the most critical STI-value for performance (i.e. STI 0.34). This task, consisting of simple addition, was obviously not as vulnerable to varying levels of speech intelligibility, compared to the Word memory task. In the Information search task participants performed significantly worse in STI 0.71 compared to STI 0.00 (silence), between which the decrease in performance was greater than 6%. This is in line with a study by Haka et al. [14] which shows that some tasks (i.e. operation span and serial memory) are not impaired until STI gets higher than 0.50. For the Word fluency tasks there were no significant differences between acoustic conditions. One explanation for the invulnerability of these tasks to distraction is that they are based on long-term memory retrieval and do not need encoding or rehearsal of a to-be remembered material. Studies have shown that division of attention at encoding reduces memory performance markedly, while division of attention at retrieval has almost no effect on memory performance [28]. This is in line with an alternative explanation of auditory distraction, which states that sounds interrupt cognitive processes due to attentional capture [29]. Therefore, tasks with high demands on the encoding processes may be more vulnerable to distractions from background speech. In summary, we have support for Hypothesis 2, as the average performance decrement depended on task type. The Word memory task, based on semantic short-term memory, was most vulnerable to intelligible speech, while there was no effect on the tasks based on long-term memory retrieval. There are, however, other feasible explanations to why the tasks applied in the present study reacted differently to the acoustic conditions. For instance, the complexity of the tasks and the use of different individual strategies for solving the tasks may render the tasks differentially sensitive to disruption via the irrelevant speech. 4.3. Limitations 4.3.1. The variety of STI-values and tasks One of the limitations of the present study is that we did not have measurement points between STI 0.34 and 0.71 since we expected that the strongest performance change takes place between 0.00 and 0.34, and 0.71 was just the reference condition with perfect speech intelligibility. The validity of this assumption should be questioned and put to test in future studies. It is likely that participants use more effort in high STI levels to maintain their performance [15]. However, self-rated effort or subjective workload was not measured in this study and we cannot say whether there was a balance between performance and effort over all STI levels and tasks. However, it is possible that the participants used more energy/resources to maintain their performance level in laboratory experiments because they are under continuous surveillance. There is a need to explore a greater variety of tasks and discover a way to categorize them in terms of the mechanisms or processes they involve. One suggestion is to vary the tasks according to a given dimension (e.g. demands on rapidity or memory load) and investigate their differentially susceptibility to irrelevant speech. It is noted that there is a limitation in the possibility of generalizing the results of the present study to other tasks, since the STI–performance function seems to differ significantly between tasks. 4.3.2. Uncertainties in the acoustic conditions The original aim was that the acoustic conditions would not overlap. As described in Section 2, the STI value of individual sentences within a specific acoustic condition changed significantly because the sentences were not fully balanced in respect with speech spectrum. The outcome was that the STI of some
sentences in the acoustic condition STI 0.16 overlapped with the standard range of the acoustic condition STI 0.23 (Fig. 2). Therefore, the acoustic conditions STI 0.16 and 0.23 were perhaps not sufficiently different from each other. This may have reduced our possibilities to make clear discrimination of task performance in the STI range between 0.08 and 0.34. In future studies, it is suggested that the difference of nearby STI levels should not be less than 0.10. Another question concerning the acoustic conditions is the uncertainty related with headphone listening. In this study, special care was taken that the headphone levels were correctly calibrated. However, the sound field produced to the ear channel depends strongly on the individual fitting of headphones. We tried to address this by giving the participants specific instructions on how to put them on before the experiment started. The withinparticipant design also assured the same circumstances between acoustic conditions. Changing to listening tests with loudspeakers and individual testing was not an alternative, since it would have been more time consuming. One reason why the greatest performance loss was achieved already at STI 0.34 and not above 0.50, as suggested by Fig. 1, might be that the content of sentences was very simple. In addition, the four-word sentences were very well pronounced. The sentence intelligibility of normal speech containing longer sentences (i.e. unpredictable content) and casual pronouncing is less intelligible. Therefore, we suggest that the subjective intelligibility associated with the reported STI values is higher than would be achieved with normal speech having the same speech-to-noise ratio. If this is indeed the case, the effects of low intelligible speech on performance may have been overestimated. Further, the original speech material was limited and the same sentences were used repeatedly, although randomly. The structure of each four-word sentence was also constant. These predictable syntactical structures may have helped the subjects to habituate to the acoustic stimulus [30]. It is also possible that the results would have been different if the speech was not so monotonic. 4.4. Practical implications of the study The present study design is unique since it is one of the first experimental studies that contains several levels of speech intelligibility in the STI range where the largest change of performance was expected to occur according to the model of Hongisto [1]. Evidence from the field [4] and laboratory measurements [31,32] show that it is possible to design the open-plan office in such a way that an STI below 0.50 is possible to achieve in a nearby workstation (2.5 m), or, at the least, in the second nearest workstation (5 m), with proper room acoustic design. Proper room acoustic design requires that sufficient absorption, high screens and appropriate speech masking is used. The speech masking levels used in this study should not be used as guidelines for designing open-plan offices. Quite high masking sound levels, up to 51 dB (LAeq), were used to create the lowest STI values. The focus in this study was to give scientific support for the STI–performance function and specify the model for a variety of tasks. Suggested masking levels for open-plan offices range between 40 and 45 dB (LAeq) [20,33]. The present results also imply that demands on tasks involving short term memory are more affected by irrelevant background speech than tasks that can rely on retrieval of information from long-term memory. This should be taken into account when designing the office. Workers doing mainly simple tasks may be less distracted by the irrelevant speech of an open-plan office than those doing demanding memory tasks. Similarly, better room acoustic designs, quiet open-plan office areas or private rooms are required for the optimal performance of more demanding
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
tasks. In practice, this logic has already been tested in some offices [33] and this study yields support for the idea that better room acoustics in critical task areas might solve some problems in employee satisfaction and productivity.
315
Health was a part of the TOTI project (User oriented office spaces) which was funded by Tekes and 10 companies.
References 4.5. Future research needs The present analysis and results can serve as a platform for further investigations of the Hongisto’s [1] model of the STI– performance function. There is, for example, a need to sort out the details of the shape of the curve and the amount of performance decrement with more STI-values and tasks that, for example, can be classified with regard to semantic content, complexity and memory demands. Studies are also required that can rule out whether semantic or changing-state effects (e.g., attributable to the acoustic variability of the speech signal) explain the STI–performance function by including a control condition with meaningless speech. It has been suggested that the STI of speech can be used to describe the disruptive potency of speech only in such conditions where the masking sound does not have temporal variations [20]. In this study, the masking sound was filtered pink noise and free from temporal variations. If the masking sound varies with time, like babble or other fluctuating sounds, STI may no longer fully explain the phenomenon observed here. However, the most typical masking sounds in open-plan offices are ventilation or road traffic sounds. In the future, studies containing time-varying maskers should be undertaken as there is evidence that fluctuating sounds are much more effective speech maskers than constant sounds [20].
5. Conclusions This study aimed to clarify the model of Hongisto [1] describing the dependence of the performance decrement of cognitive tasks on the STI (Speech Transmission Index) of irrelevant speech. This study examined the effect of five different STI levels on the performance of five cognitively different tasks. The most important outcome was that the maximum performance decrease was already achieved at STI 0.34 with two out of five tasks while Hongisto [1] suggested that it is not achieved until STI exceeds 0.50. A maximum performance decrement above STI 0.50, suggested by Hongisto, was observed only with one task. Two tasks were not at all affected by speech. Attempts to reduce the STI in open-plan offices by way of altering room acoustics are strongly recommended if the cognitive work requires concentration and high short-term memory load. The target value of STI should be less than 0.50 between the speaker and the listener. In addition, the positive effect of acoustic design on work performance depends on the task type (i.e., the cognitive processes that underpin the task). Acknowledgements The international research cooperation between the authors of this article has received funding from the European Union’s Seventh Framework Programme (FP7/2007-2013) under Grant Agreement No. 226442. We would like to thank Staffan Hygge, John Marsh, Anders Kjellberg and Patrik Sörqvist for helpful comments on earlier drafts of this paper. For the first author the research was conducted in partial fulfilment of the requirements for a doctoral degree. This research was partly funded by AFA Insurance; the University of Gävle (Sweden) and Finnish Institute of Occupational Health. The work of the Finnish Institute of Occupational
[1] Hongisto V. A model predicting the effect of speech of varying intelligibility on work performance. Indoor Air 2005;15:458–68. [2] Jahncke H, Hygge S, Halin N, Green A-M, Dimberg K. Open-plan office noise: cognitive performance and restoration. J Environ Psychol 2011;31: 373–82. [3] Kaarlela-Tuomaala A, Helenius R, Keskinen E, Hongisto V. Effects of acoustic environment on work in private office rooms and open-plan offices – longitudinal study during relocation. Ergonomics 2009;52:1423–44. [4] Virjonen P, Keränen J, Hongisto V. Determination of acoustical conditions in open-plan offices – proposal for new measurement method and target values. Acta Acust Unit Acust 2009;95:279–90. [5] Marsh JE, Hughes RW, Jones DM. Auditory distraction in semantic memory: a process-based approach. J Mem Lang 2008;58:682–700. [6] Perham N, Marsh JE, Jones DM. Syntax and serial recall: how language supports short-term memory for order. Quat J Exp Psychol 2009;62:1285–91. [7] Jones DM, Macken WJ. Irrelevant tones produce an irrelevant speech effect: implications for phonological coding in working memory. J Exp Psychol: Learn Mem Cogn 1993;19:369–81. [8] Jones DM, Morris N. Irrelevant speech and serial recall: implications for theories of attention and working memory. Scand J Psychol 1992;33:212–29. [9] Marsh JE, Jones DM. Cross-modal distraction by background speech: what role for meaning? Noise Health 2010;12:210–6. [10] IEC 60268-16:2003. Sound system equipment—Part 16: Objective rating of speech intelligibility by speech transmission index; 2003. [11] Hongisto V, Haapakangas A, Haka M. Task performance and speech intelligibility – a model to promote noise control actions in open offices. In: 9th International congress on noise as a public health problem (ICBEN), Mashantucket, Connecticut, USA, July 21–25 2008. p. 418–25. [12] Ellermeier W, Hellbrück J. Is level irrelevant in irrelevant speech? Effects of loudness, signal-to-noise ratio, and binaural unmasking. J Exp Psychol Hum Percept Perform 1998;24:1406–14. [13] Venetjoki N, Kaarlela-Tuomaala A, Keskinen E, Hongisto V. The effect of speech and speech intelligibility on task performance. Ergonomics 2006;49: 1068–91. [14] Haka M, Haapakangas A, Keränen J, Hakala J, Keskinen E, Hongisto V. Performance effects and subjective disturbance of speech in acoustically different office types – a laboratory experiment. Indoor Air 2009;19:454–67. [15] Schlittmeier S, Hellbrück J, Thaden R, Vorländer M. The impact of background speech varying in intelligibility: effects on cognitive performance and perceived disturbance. Ergonomics 2008;51:719–36. [16] ISO 3382-3:2012. Acoustics—measurement of room acoustic parameters—Part 3: Open-plan offices. Genève (Switzerland): International Organization for Standardization; 2012. [17] FiSIAQ. Classification of indoor climate. Espoo (Finland): Finnish Society of Indoor Air Quality and Climate; 2008 [available in English]. [18] Swedish Standard, designation SS-EN 12464-1. Ljus och belysning – Belysning av arbetsplatser – Del 1: Arbetsplatsens utformning [Light and lighting – Lighting of work places – Part 1: Workplace Design]. ICS code 91.160.10; 2003. [19] Hagerman B. Sentences for testing speech intelligibility in noise. Scand Audiol 1982;11:79–87. [20] Haapakangas A, Kankkunen E, Hongisto V, Virjonen P, Oliva D, Keskinen E. Effects of five speech masking sounds on performance and acoustic satisfaction – implications for open-plan offices. Acta Acust Unit Acust 2011;97:641–55. [21] Larm P, Hongisto V. Experimental comparison between speech transmission index, rapid speech transmission index, and speech intelligibility index. J Acoust Soc Am 2006;119:1106–17. [22] Houtgast T, Steeneken HJM. A review of the MTF concept in room acoustics and its use for estimating speech intelligibility in auditoria. J Acoust Soc Am 1985;77:1069–77. [23] Jahncke H, Halin N. Performance, fatigue and stress in open-plan offices: the effects of noise and restoration on hearing impaired and normal hearing individuals. Noise Health, 2012; Accepted for publication. [24] Fürst AJ, Hitch GJ. Separate roles for executive and phonological components of working memory in mental arithmetic. Mem Cogn 2000;28:774–82. [25] Troyer AK, Moscovitch M, Winocur G, Alexander MP, Stuss D. Clustering and switching on verbal fluency: the effects of focal frontal- and temporal-lobe lesions. Neuropsychologia 1998;36:499–504. [26] van Overschelde JP, Rawson KA, Dunlosky J. Category norms: an updated and expanded version of the Battig and Montague (1969) norms. J Mem Lang 2004;50:289–335. [27] Borkowski JG, Benton AL, Spreen O. Word fluency and brain damage. Neuropsychologia 1967;5:135–40. [28] Naveh-Benjamin M, Craik FIM, Perretta JG, Tonev ST. The effects of divided attention on encoding and retrieval processes: the resiliency of retrieval processes. Quart J Exp Psychol 2000;53:609–25. [29] Cowan N. Attention and memory: an integrated framework. Oxford: Oxford University Press; 1995.
316
H. Jahncke et al. / Applied Acoustics 74 (2013) 307–316
[30] Sörqvist P, Nöstl A, Halin N. Working memory capacity modulates habituation rate: evidence from a cross-modal auditory distraction paradigm. Psychon Bull Rev 2012;19:245–50. [31] Keränen J, Hongisto V, Hakala J, Oliva D. Speech privacy in an open-plan office with different room acoustic conditions, Forum Acusticum, paper 135, 1897– 1900, 27 June–1 July 2011, Aalborg, Denmark.
[32] Virjonen P, Keränen J, Helenius R, Hakala J, Hongisto V. Speech privacy between neighbouring workstations in an open office – a laboratory study. Acta Acust Unit Acust 2007;93:771–82. [33] Hongisto V, Haapakangas A, Helenius R, Keränen J, Oliva D. Acoustic satisfaction in an open-plan office before and after the renovation. In: Proceedings of Euronoise 2012, June 10–13, Prague. p. 654–9.
Sound Masking in the Office R E DUCI NG NOI S E DI STRACTION S TO I NCR EAS E WOR KE R PRODUCTIVITY
Quiet Technologyâ&#x201E;˘ sound masking decreases distractions to improve performance. An ideal sound-masking solution makes speech beyond a
Z
12-to-16-foot radius unintelligible so a person can concentrate on his or her work, collaborate with colleagues, and be more productive.
Solution Essay / 2003
What We Know Office workers who participated in a 2002 study on privacy-related issues cited “overheard conversation” as their biggest complaint (Herman Miller, 2002). While people can get used to many office noises, to the point where their brains “tune out” the distractions, some find it nearly impossible to disregard intelligible human speech (Banbury and Berry, 1997, 1998). Without sound masking, human speech can be intelligible up to 50 feet away. Other researchers have documented the impact of noise on employee stress (Cohen, 1980; Evans and Johnson, 2000) and short-term memory (Jones, 1999). is ongoing skepticism regarding these claims that has limited Early sound-masking systems installed in buildings in the 1960s
adoption of sound masking by North American businesses.
simulated the sound of air moving by electronically filtering random noise produced by gas-discharge vacuum tubes. Loudspeakers
Two factors often make the cost-benefit relationship for in-ceiling
in the ceiling distributed the amplified noise signal throughout
systems prohibitive for organizations, especially small and medium-
the office. However, making human speech unintelligible required
sized ones, and building owners. First, typical in-ceiling systems
a volume level so high that the sound masking itself became a
have hardware-plus-installation costs of $1.00 to $2.00 per square
distracting annoyance.
foot. In addition to these costs, the services of a trained acoustical engineer are needed to oversee installation and fine-tuning of
In the 1970s, electronically generated sound masking using
the system to coordinate with the acoustical properties of various
frequency generators that shaped sound to better mask speech
architectural elements in the space (e.g., reflectivity of ceiling, floor
became more practical and worked well when installed correctly.
coverings, lights), as well as the furniture layout once workers
Ten years later, researchers began studying 1/f noise, the
occupy the space.
phenomenon also known as “flicker” or “pink” noise. Targeting “pink” noise to match the frequencies of human speech raised the threshold of audibility just enough to mask intelligibility without requiring the higher volumes used in earlier systems.
Second, in-ceiling systems are considered by many to be, in effect, permanent fixtures within the architectural space even though a leasee who installs such a system could take the components when vacating a space. So, building leasees must be
In-ceiling sound-masking systems began using this targeted
convinced that the cost of adding these systems to their leasehold
spectrum of sound. This approach involved sound engineers
improvements will yield real benefits whether they eventually take
working with architects to design appropriate sound-masking
the hardware or leave it. Building owners are reluctant to offer
layouts for new and significantly renovated buildings. The design
sound masking as part of the base building package because they
typically incorporated sound-masking units placed on a grid pattern
are unable to increase rents proportionately to cover costs. And,
in the ceiling plenum in order to achieve uniform coverage.
because in-ceiling units are considered part of the building, they
The benefits of sound-masking systems have been studied
problem of recouping the investment.
must be depreciated over a 30-year schedule, which adds to the extensively. Independent research has documented productivity gains of 8 to 38 percent, job satisfaction increases of 125 to 174 percent, and reductions in stress up to 27 percent. Yet, there
Z
Sound Masking in the Office Solution Essay / 2
In-ceiling sound masking results in costly, inefficient coverage because speakers are placed in the ceiling grid before the layout of the space is determined.
Therefore An effective sound-masking system incorporates the advantages
Quiet Technology™ sound masking attaches to furniture, so efficient coverage results from targeting only those areas of the space that require sound masking.
Quiet Technology achieves these results by delivering an Articulation Index (AI) of .2 or less, meaning that few words (20 percent or less) spoken by people 12 to 16 feet away are intelligible.
of a human-speech-targeting sound spectrum (i.e., pink noise), the ability to integrate with furniture, and no need for an acoustical
Because people working in highly collaborative environments stand
designer. The integration feature addresses cost controls through
to gain the most from Quiet Technology sound masking, it made
targeted application (only where it’s needed and not in hallways,
sense to focus initial development efforts on the Herman Miller
copy/mail rooms, conference rooms, etc.), ease of reconfiguration
furniture that is particularly adept at supporting collaboration—the
without the need to go above the ceiling, and a seven-year
Resolve® system. Incorporating the soundmasking advantages of
depreciation schedule equal to the furniture. Additionally, it provides
Quiet Technology would reduce noise distractions to enhance even
the ability to take the system along when removing furniture in
further the furniture’s ability to foster collaboration and productivity.
a leasing situation. And it is less dependent on the acoustical properties associated with a building’s ceiling height and material
Design Solution
type because of the opportunity for precision placement. Most
Working with the developer of Quiet Technology, Cambridge Sound
important, research into the value of sound masking demonstrates
Management, Inc., Herman Miller incorporated a patented sound-
a link between this technology and productivity increases among
masking spectrum into an effective system for delivering masking
workers engaged in cognitive activities.
sounds. Small speakers embedded in the “petals” that attach easily
Design Problem In undertaking the development of economical, effective sound
to the top of Resolve tall poles broadcast the patented spectrum. While straightforward in application, able to be retrofitted, and
masking that attaches to its systems furniture products, Herman
more economical than in-ceiling systems, Herman Miller’s Quiet
Miller sought to capitalize on recent advances in the application of
Technology sound-masking system more importantly addresses
sound masking.
the issue of impact on worker productivity. To determine the level of this impact, Herman Miller contracted an independent research
Quiet Technology sound masking incorporates a patented sound
group, Ergometrica, to test Quiet Technology sound masking
spectrum that has proven more effective than previous systems in
under scientifically controlled conditions on Resolve furniture and
decreasing the radius of potential distraction to 12 to 16 feet from
65-inch-high traditional systems environments.
40 to 50 feet in the typical office without sound masking.
Z
Sound Masking in the Office Solution Essay / 3
Lack of Sound Distractions
Ability to Focus
Short-Term Memory/Accuracy
In conducting the tests, the research team used the ∆Productivity
The research team recruited knowledge workers from the Boston
Index™ (∆PI), which compares information inputs to labor output
metropolitan area. They divided 136 participants— 49 percent
by measuring the rate at which workers perform three cognitive
men, 51 percent women—randomly among the test groups. Other
operations that are typical of knowledge work. This method used
characteristics of participants included:
standard measures of performance and specific variables for four
• Work experience in knowledge industries
different office environments.
• Minimum of eight months experience working in open-plan
The general theory behind ∆PI states that ∆P = f (Vh + Ve), i.e.,
• Routine problem solving at work
change in productivity (∆P) can be measured by comparing
• Experience using Web browsers
offices.
the performance of a set of cognitive operations (Vh) that are characteristic of typical office work, and the influence of a set of
Testing software captured the demographics of the test takers,
workplace variables (Ve). The three standard cognitive operations
their speed of performance and task accuracy, as well as their self-
are defined as “attention,” “data translation,” and “short-term
reported performance. An analysis of the data found statistically
memory.” The four measures of task performance are “speed,”
significant improvement in worker productivity on a number
“accuracy,” “retention,” and “self-reported performance” (attitude).
of measures for those tested in sound-masked environments. Perhaps most important among these findings was that when
Researchers built four environments to simulate working
Quiet Technology sound masking was present, test subjects
conditions in large open-plan offices. These four environments
reported improved performance independent of the type of
included Resolve with sound masking (AI = <0.2), Resolve without
furniture environment in which they took the test.
sound masking (AI = >0.6), and 65-inch-high traditional panel systems with (AI = <0.2) and without sound masking (AI = >0.6)
Aggregate numbers for Resolve and 65-inch-high traditional
in a specialized test facility free from echoes and reverberations.
systems furniture with Quiet Technology sound masking compared
They located speakers at appropriate locations and then broadcast
to these furniture systems without sound masking indicated the
prerecorded, calibrated, male-female conversations similar to those
following: 43.6 percent improvement in sounds not considered
heard in a typical office. Research subjects could take as long as
distracting, 34.3 percent increase in ability to focus on tasks, and
they wanted to complete a series of tasks presented through Web-
16.5 percent improvement in accuracy in performing required tasks.
based interactive software. Based on a seven-point scale, with seven being the best possible score, the mean value for all subjects in the Resolve environment
Z
Sound Masking in the Office Solution Essay / 4
with Quiet Technology reporting on the issue of sounds not considered to be distracting was 4.84. Those in the 65-inchhigh traditional systems environment with Quiet Technology had a mean value of 3.97. Thus, subjects reported greater satisfaction in the very open environment of the Resolve system compared to the more enclosed nature of the 65-inch-high traditional system. Without Quiet Technology, Resolve and the 65-inch-high traditional system had negligible differences in mean value scores for the three key measurements. In addition, test participants adapted to Quiet Technology sound masking immediately. The subjectsâ&#x20AC;&#x2122; improved performance on a half-hour test indicates that people need little or no time to acclimate to this form of acoustic privacy. Measurement
Resolve with Quiet Technology compared to Resolve system without sound masking
Lack of sound distractions
51.4% improvement
Ability to focus on tasks
47.7% improvement
Short-term memory/accuracy
9.6% improvement
The resulting total increase in productivity of 35 percent is calculated by assigning equal weighting to each measure in this equation: Quiet Technology productivity increase = 51.4%(.33) + 47.7%(.33) + 9.6%(.33) = 35%.
Z
Sound Masking in the Office Solution Essay / 5
References Ergometrica (2002)—independent research contractor to Cambridge Sound Management, Inc.—“Empirical evaluation of the impact on worker productivity of Herman Miller’s Resolve furniture system,” Report of a fiveweek research project conducted under scientifically controlled conditions.
Thomas Horrall, B.S., University of Wisconsin. Mr. Horrall is a physicist and acoustical engineer at Acentech, Inc., (formerly BBN Acoustics) in Cambridge, MA. As codeveloper of the Quiet Technology sound-masking system adapted for use with Resolve furniture, he supervised the preparation of the acoustical test conditions used in this project.
Evans, G.W. and Johnson, D. (2000). “Stress and open-office noise,” Journal of Applied Psychology, 85:5, 779-783. Herman Miller, Inc. (2002), “Systems Audit Research: A Comprehensive Study of Herman Miller Systems Furniture From the End User’s Perspective.” Internal report. Jones, D., (1999). “The cognitive psychology of auditory distraction: 1997 Broadbent Lecture.” British Journal of Psychology, 90, 167-187. Banbury, S. and Berry, D. (1998). “Disruption of office-related tasks by speech and office noise.” British Journal of Psychology, 89, 499-517.
For more information about our products and services or to see a list of dealers, please visit us at www.HermanMiller.com or call (800) 851 1196. © 2003 Herman Miller, Inc., Zeeland, Michigan ® l and Resolve are among the registered trademarks of Herman Miller, Inc. ™ Quiet Technology is among the trademarks of Cambridge Sound Management, Inc. ™ ∆Productivity Index is among the trademarks of The Remington Group.
Banbury, S. and Berry, D. (1997). “Habituation and dishabituation to office noise.” Journal of Experimental Psychology: Applied, 3, 181-195. Armstrong, Inc. (1997). “A case study of office speech noise distraction and work productivity.” Summary of a six-month longitudinal field study at a customer site. Cohen, S. (1980). “After effects of stress on human performance and social behaviour: a review of research and theory,” Psychological Bulletin, 88, 82-108.
Credits Elaine Lewis, A.B., Wellesley College; M.P.S., Cornell University; Ph.D., Rennselaer Polytechnic Institute. Dr. Lewis, who codeveloped the ∆Productivity Index™ methodology is a tenured associate professor at Boston University, where she was codirector of the Project for Interdisciplinary Research in Information. Dr. Lewis has directed numerous studies on usability and office system efficiency for large corporations, including United Airlines and Apple Computer. She is currently president of Ergometrica in Cambridge, MA. Peter H. Lemieux, B.A., Harvard University; Ph.D., Massachusetts Institute of Technology. Dr. Lemieux, president of Cyways, Inc., was a senior lecturer at the Massachusetts Institute of Technology and has taught research methods and statistical methodology at the University of Rochester and the Massachusetts Institute of Technology. Dr. Lemieux codeveloped with Dr. Lewis the Web-enabled software for the ∆Productivity Index methodology used in this project. David Sykes, A.B., University of California-Berkeley; M.A. and Ph.D., Cornell University. Dr. Sykes was a professor at Boston University, where he and Dr. Lewis codirected the Project for Interdisciplinary Research in Information. He is managing director of The Remington Group, Cambridge, MA, and vice president/strategy of Cambridge Sound Management.
Z
Sound Masking in the Office Solution Essay / 6
Case Study 4 THE EFFECTS OF THE NINTENDO WII EXERCISE PROGRAM ON CHRONIC WORK-RELATED LOW BACK PAIN IN INDUSTRIAL WORKERS
M. Yvonne Hidle CARNEGIE MELLON UNIVERSITY | VIVIAN LOFTNESS
Exercise at Work = Health + Reduced Absenteeism Results Before and after 8-Week Nintendo Wii Exercise Program
Health â&#x20AC;&#x201C; Park et al 2013 (industrial)1
In a 2011 LBP clinical study that analyzed claims data from 40 self-insured employers and 211,551 patients in the United States. Ivanova et al identifies a 234.4% decrease in employer costs for employees without LBP compared to employees with LBP (p < 0.0001), at roughly $1,856 per employee. First cost increase:
$230 / employee
(Renovations, equipment)
Annual time cost:
Before: 63.12 After: 73.12
Self-Reported Survey
Emotional Wellbeing (p < 0.05)
Physical Tests
In an 8-week intervention study with 24 workers at a K Tire Factory, Park et al identifies a 13% decrease in Lower Back Pain (LBP) (p < 0.05), a 6.11% increase in back strength (p < 0.05) and a 15.8% increase in self-reported emotional wellbeing (p < 0.05) among factory workers who participated in 50 minutes of physical therapy and a 30 minute Nintendo Wii Fitness program 3 times a week compared to their baseline.
Before: 58 After: 68.12
Energy (p < 0.05)
Before: 41.25 After: 44.37
General Health (p < 0.05)
Balance Ability (statistically insignificant)
Before: 8.75 After: 10.62 Before: 54.48 After: 57.81
Back Strength (p < 0.05)
Before: 6.75 After: 5.87
Pain (p < 0.05)
0
Baseline Condition
20 40 60 80 Physical Test and Survey Scores
After 8-Week Nintendo Wii Exercise Program Intervention
$3,100 / employee
(80 minutes x 3 times a week)
Annual indirect cost savings:2
$1856 / employee
ROI:
55.7%
1 Park, J., Lee, S., & Ko, D. (2013). The Effects of the Nintendo Wii Exercise Program on Chronic Work-related Low Back Pain in Industrial Workers. Journal of Physical Therapy Science, 985-988. 2 Ivanova, J., Birnbaum, H., Schiller, M., Kantor, E., Johnstone, B., & Swindle, R. (2011). Real-world practice patterns, health-care utilization, and costs in patients with low back pain: The long road to guideline-concordant care. The Spine Journal, 622-632.
F14-Productivity, Health and the Quality of Buildings M. Yvonne Hidle Overview: 1. Article citation: Park, J., Lee, S., & Ko, D. (2013). The Effects of the Nintendo Wii Exercise Program on Chronic Work-related Low Back Pain in Industrial Workers. Journal of Physical Therapy Science, 985-988. 1.1 Supporting article citation: Ivanova, J., Birnbaum, H., Schiller, M., Kantor, E., Johnstone, B., & Swindle, R. (2011). Real-world practice patterns, health-care utilization, and costs in patients with low back pain: The long road to guideline-concordant care. The Spine Journal, 622-632. 2. Study objective: This study seeks to understand the effects of a Nintendo Wii exercise program on industrial workers with lower back pain. 3. What type of study is it? Simulation study Building case study Cross sectional (multiple building) study Before and after study
x
Lab experiment Field experiment Meta-analysis (multiple studies) Other: _____________________________
4. When did the study occur? (eg., month(s)/year) 8 week intervention study, the time of year was not specified. 5. Where did the study occur? (eg., “Xerox headquarters, Boston”, “office buildings in Paris”) The intervention study took place in a K Tire Factory. 6. How many people? 24 People characteristics: All employees in the study were diagnosed with chronic Lower Back Pain (LBP) by a physician and had lived with LBP for over 3 months. None of these employees had undergone any surgical treatment for disc herniation, spina bifida, or spinal stenosis. The average age of the participants was 44 years old, and the average length of service at the K Tire Factory was 17 years.
BIDS™ Study Evaluation Form
48-729 Fall 2014
7. How many buildings? 1 Building characteristics: K Tire Factory; an industrial factory that manufactures tires. 8. What design strategies were compared in the study? Three design strategies were compared in three participant groups. Each group consisted of 8 individuals. All groups were given physical therapy that consisted of hot packs, interferential current therapy, and deep heat with ultrasound. This physical therapy lasted for a total of 50 minutes. 1) Control group: No additional intervention. 2) LSE (Lumber Stabilization Exercise Group): Exercises for lumber stabilization. Participants were asked to hold 7 different positions for 15 seconds in 3 sets. Total time was 30 minutes. 3) NWE (Nintendo Wii Exercise Group): Nintendo Wii sports program games. Participants played exercise games on the Wii and took a 2 minute break every 10 minutes. Total time was 30 minutes. Participants were treated 3 times every week, each session for the control group was 50 minutes. Each session for the LSE and NWE groups was 80 minutes (50 minutes of physical therapy and 30 minutes of exercise). 9. What design strategy does the study support? The study supports the use of the Nintendo Wii fitness program and lumber stabilization exercises for industrial workers with LBP. Study method: 10. How were the design strategies measured? The design strategies were measured after the 8-week intervention. Participants of the study “were evaluated for pain, back strength, balance ability, and health-related Quality of Life (QOL) with the following tools; visual analog scale (VAS) for pain, isometric lifting strength for back strength, one-legged stand test for balance ability, and the RAND-36 Health Status Inventory for health-related QOL.” (Park, Lee, & Ko, 2013) The physical tests provided quantitative results for pain, back strength, and balance ability. The RAND-36 provided self-reported physical and mental health. To get a baseline, all participants underwent both the physical tests and the self-reported survey before the 8-week intervention started. 11. How was the benefit measured? The study does not give a specific cost savings or other quantitative measurement of the benefit. It does, however, mention that the Wii Fitness Program could be used to reduce LBP symptoms and thereby improve the health and functionality of industrial workers. A supporting study was found to get monetary benefits. 12. Other measurements: None
BIDS™ Study Evaluation Form
48-729 Fall 2014
13. How was the relationship between the design strategy and benefit determined? Simulation Logistic regression/odds ratio x ANOVA Other statistical method: ______________________________ Linear regression Observed only/no analytical method 14. How were confounding factors addressed? 1) The study divided participants up into 3 groups, one of which was a control group. All groups received physical therapy through agent modalities for 50 minutes, 3 times a week. 2) All participants had LBP for at least 3 months and were diagnosed by a physician. Study results 15. What is the quantitative benefit of the design action? The study found a significant decrease in pain among the LSE and NWE groups, but not the control group (p < 0.05). In addition, the study also found that there was an increase in back strength among all three groups, with the LSE and NWE groups having the best results (p < 0.05). Lastly, there was an improvement in balance ability for all three groups, but the results for the NWE group in regards to balance ability were considered insignificant. In regards to self-reported emotional problems, energy/fatigue, and emotional wellbeing, the Control Group and the LSE group noted no significant changes, but the NWE group showed significant improvement (p < 0.05). Participants in the NWE group had a 13% decrease in pain (p < 0.05), a 6.11% increase in back strength (p < 0.05), a 17.45% increase in self-reported energy (p < 0.05), and a 15.8% increase in emotional wellbeing (p < 0.05). 16. List other significant relationships found: None 17. List important non-significant factors: None 18. What are the limitations of the study? The benefits of the Nintendo Wii Fitness program could have been quantified better if none of the participants had undergone any physical therapy. 19. Costs provided in the study: None
BIDSâ&#x201E;˘ Study Evaluation Form
48-729 Fall 2014
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
Health – Park et al 2013 Exercise at Work = Health + Reduced Absenteeism In an 8-week intervention study with 24 workers at a K Tire Factory, Park et al identifies a 13% decrease in Lower Back Pain (LBP) (p < 0.05), a 6.11% increase in back strength (p < 0.05) and a 15.8% increase in self-reported emotional wellbeing (p < 0.05) among factory workers who participated in a 30 minute Nintendo Wii Fitness program 3 times a week compared to their baseline. In a 2011 LBP clinical study that analyzed claims data from 40 self-insured employers and 211,551 patients in the United States. Ivanova et al identifies a 234.4% decrease in employer costs for employees without LBP compared to employees with LBP (p < 0.0001), at roughly $1,856 per employee. Overview In an 8-week intervention study with 24 industrial workers at a K Tire Factory, Park et al observed the effects of a Nintendo Wii Fitness program on individuals with diagnosed Lower Back Pain (LBP). All participants had been previous diagnosed with LBP and had been living with the condition for at least 3 months. None of the participants had undergone any surgical treatments for disc herniation, spina bifida, and spinal stenosis. The average age of the participants was 44 and the average length of service at the K Tire Factory was 17 years. Participants were divided into 3 equal groups: Control Group, LSE (Lumber Stabilization Exercise Group), and NWE (Nintendo Wii Exercise Group). Each group consisted of 8 individuals and all participants underwent physical therapy with agent modalities consisting of hot packs (30 minutes), interferential current therapy (15 minutes), and deep heat with ultrasound (5 minutes). The Control Group had no additional intervention other than the physical therapy. The LSE group underwent exercises for lumber stabilization. These exercises consisted of holding 7 different positions for 15 seconds in 3 sets. Such positions included “standard hand and knees” and “standard back bridge.” In the NWE group, participants selected and played exercise games using the Wii Sports Program. These games included Wakeboard, Frisbee Dog, Jet Ski, and Canoe games. 2 minute breaks were taken every 10 minutes. Over the course of 8 weeks, participants were treated 3 times every week. Participants in the Control Group had 50 minute sessions, whereas the participants in the LSE and NWE groups had sessions lasting 80 minutes (50 minutes of physical therapy and 30 minutes of exercise). Study Method At the end of the 8 week study, all participants were required to undergo several tests. A series of physical tests provided quantitative results for pain, back strength, and balance ability; whereas a survey provided self-reported physical and mental health. For the physical measurements, participants in the study underwent a visual analog scale (VAS) for pain, isometric lifting strength test to assess back strength, and a one-legged standing test to assess balance ability. In addition, participants were asked to fill out the RAND-36 Heath Status Inventory for health-related QOL. Study Results The study found a significant decrease in pain among the LSE and NWE groups, but not the control group (p < 0.05). In addition, the study also found that there was an increase in back strength among all three groups, with the LSE and NWE groups having the best results (p < 0.05). Lastly, there was an improvement in balance ability for all three groups, but the results for the NWE group in regards to balance ability were considered insignificant. In regards to self-reported emotional problems, energy/fatigue, and emotional wellbeing, the Control Group and the LSE group noted no significant changes, but the NWE group showed significant improvement (p < 0.05). Participants in the NWE group had a 13% decrease in pain (p < 0.05), a 6.11% increase in back strength (p < 0.05), a 17.45% increase in self-reported energy (p < 0.05), and a 15.8% increase in emotional wellbeing (p < 0.05).
BIDS™ Revised Case Study Template M. Yvonne Hidle
November 2014
BIDS Life Cycle Assumptions The 2014 BIDS factory baseline assumes a 100,000 square foot building, 100 employees with an average salary of $31,000. The first cost increase is $9.54/SF for renovations and equipment and $9,300/employee for time lost due to exercise during work time for the first year; this totals to $18,842 per employee. Annual health and absenteeism savings per employee was found in a supporting study by Ivanova et al. The study found that individuals with LBP had a total of $2,606 worth of indirect costs compared to $750 for the control group (p < 0.0001). “Medically related absenteeism costs accounted for 67.8% of indirect costs.” (Ivanova, Birnbaum, Schiller, Kantor, Johnstone, & Swindle, 2011) Indirect costs were considered costs companies had to pay on the behalf of the employee, such as payments for disability days and estimated medically related absenteeism. All costs were calculated over the course of 12 months. In a baseline building, a 234.4% decrease in employer costs due to daily physical therapy among employees with LBP using the Nintendo Wii Fitness program, yields annual savings of $1,856/employee, resulting in an ROI of 9.85%. First cost increase per square foot: Baseline first cost increase: First cost increase per employee: Annual time cost per employee: Annual health and absenteeism savings per employee: 2011) Baseline annual savings: NPV of the savings: ® EVA : ROI:
$0.23 $22,951.54 $229.52 $3,100
(See appendix) (See appendix) (See appendix) (See appendix)
$1856.00
(Ivanova, Birnbaum, Schiller, Kantor, Johnstone, & Swindle,
$928,000 $7,058,368 $7,032,316.46 55.7%
($1856.00 x 500 employees) ($1856.00 x 500 x 7.606) (NPV of the savings – FCI + Annual time cost) (Savings per employee / FCI per employee + Annual time cost)
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
November 2014
Limitations No limitations were mentioned in the study. However, more participants and a longer intervention period could provide better and more accurate results in future studies. In addition, the physical therapy through agent modalities provided for all groups could have potentially masked the actual results from the Nintendo Wii intervention and should be excluded from future studies.
References Ivanova, J. I., Birnbaum, H. G., Schiller, M., Kantor, E., Johnstone, B. M., & Swindle, R. W. (2011). Real-world practice patterns, health-care utilization, and costs in patients with low back pain: The long road to guideline-concordant care. The Spine Journal , 622-632. Park, J.-H., Lee, S.-H., & Ko, D.-S. (2013). The Effects of the Nintendo Wii Exercise Program on Chronic Work-related Low Back Pain in Industrial Workers. Journal of Physical Therapy Science , 985-988.
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle
November 2014
Appendix Finalized Calculations Total Cost Total cost for equipment and renovations
$22,951.54
Building Area BIDS Baseline
100,000
SQFT
Cost per SQFT Total Cost / Total Building Area
$0.23
Total Employees BIDS Baseline
100
Total
Cost per Employee
$229.52
Average Salary BIDS Baseline
$31,000
Year
(Total Cost / Total Employees)
Equipment Nintendo Wii Fit Bundle
$449.99
Per Console
x 5 consoles
Pandora's Cube
$549.99
Total Cost Per Television
$2,249.95 x 5 televisions
Best Buy
Total cost
$2,749.95
20
Feet
Ceiling Height
12
18 $17,952
Feet
Total Cost of Room Walls Partitions
$4.70
Per SQFT
C10101265060
5/8" F.R. drywall, 3 5/8" Metal Stud Framing @16" O.C., 5/8" F.R. drywall
Total Cost
$4,286.40
Bundle includes console, games, controls, and accessories
Television 50 inch screen
Interior Construction Room Width Room Depth
Wall area Interior Finishes Paint
$0.87
Per SQFT
C30102300060
Wall area Carpet tile
$4.24
Total cost Per SQFT
$793.44 C30204100060
Floor area Acoustic tile
$3.82
Total Cost Per SQFT
$1,526.40 C30204100060
Total Cost
$1,375.20
Per SQFT
D30201041520
Total Cost
$2,761.20
Ceiling area HVAC Large Hydronic Heating System
Using this to get the assembly cost for the ductwork system. Ceiling area
$7.67
Feet
Painting, interior on plaster and drywall, brushwork, primer & 1 coat Nylon, fusion bonded, 18"x18", or 24"x24" 24oz 5/8" fiberglass board 24"x48" tee grid suspended support
149,000 S.F., 2,400 K.W. 8 floors
BIDSâ&#x201E;˘ Revised Case Study Template M. Yvonne Hidle Fire Protection Wet Pipe Sprinkler System
November 2014
$2.85
Ceiling area
Per SQFT
D40104100640
Total Cost
$1,026.00
Light Hazard, steel black, sch. 40, 50,000 s.f.
Electrical Florescent Lighting
$7.52
Per SQFT
D50202080640
Ceiling area Wall Switches
$2.51
Total Cost Per SQFT
$2,707.20 D50201300400
Wall area Receptacles
$3.90
Total Cost Per SQFT
$903.60 D50201100680
Wall area Data and Voice System
$3.25
Total Cost Per SQFT
$1,404.00 D50309200102
Total Cost
$1,168.20
2 data/voice outlets per 1000 s.f. x5 for 10 data/voice outlets per 1k sf
Per Console
x 52 weeks in a year
BIDS Baseline
Total Cost
$3,100.00
Wall area
Employee Time 240 minutes lost work (80 minutes x 3 times a week)
$596.15 $119.23 $14.90 $0.25
a week a day a hour a minute
$59.62
Flourescent Fixtures 23 fixtures per 1600 S.F. 10 per 1000 S.F.
20 per 1000 S.F., 2.4 watts per S.F.
The Effects of the Nintendo Wii Exercise Program on Chronic Work-related Low Back Pain in Industrial Workers
J. Phys. Ther. Sci. 25: 985â&#x20AC;&#x201C;988, 2013
Ji-Hyuk Park, OT, PhD1), Sang-Heon Lee, OT, PhD2)*, Dae-Sik Ko, PT, MPH3) 1) Department
of Occupational Therapy, College of Health Science, Yonsei University of Occupational Therapy, College of Medical Science, Soon Chun Hyang University: 646 Eupnae-ri, Shinchang-myeon, Asan-si, Chungnam 336-745, South Korea. TEL: +82 41-530-3089, FAX: +82 41-530-3089 3) Department of Health Science, Graduate School, Chosun University 2) Department
Abstract. [Purpose] The purpose of this study was to investigate the effects of a Nintendo Wii exercise program on chronic work-related LBP compared with stability exercise. [Methods] Twenty-four workers participated in this study. All of the participants were diagnosed with chronic LBP by a physician. Participants were randomly assigned to three groups: a control group (CG), lumbar stabilization exercise group (LSE), and Nintendo Wii exercise group (NWE). Participants were treated 3 times a week for 8 weeks. Each session lasted 30 minutes. [Results] The results demonstrated that exercise programs improved significantly physical functions related to LBP. In health-related QOL, the Nintendo Wii exercise program significantly improved both the mental and physical health composites, but other groups had significant improvement only in the physical health composite. [Conclusion] The Nintendo Wii exercise program could be a biopsychosocial intervention for work-related LBP in factory workers. Key words: Low back pain, Exercise, Nintendo Wii (This article was submitted Mar. 4, 2013, and was accepted Mar. 29, 2013)
INTRODUCTION Work-related low back pain (LBP) is a common workrelated musculoskeletal disorder. In Korea, work-related low back pain makes up 36.7% of the total reimbursement from industrial accident compensation insurance1). In Western countries, LBP is one of the biggest single reasons for absence from work 2). The annual prevalence of LBP has been reported at 38% across the globe3). LBP commonly attenuates within weeks, but 24â&#x20AC;&#x201C;50% of cases may recur within 1 year4). Chronic LBP decreases productivity in the workplace because the recurrence rate is over 60%5). Thus, treatment of chronic LBP is important in work populations. Chronic LBP induces various problems, including physical, psychosocial, and quality of life features5, 6). Chronic LBP reduces muscle strength, endurance, flexibility, and balance ability. Among psychological aspects, depression is one of the most common symptoms in chronic LBP7). Psychological symptoms include depression, anxiety, and low self-efficacy, which negatively affect quality of life (QOL)8). Exercise is the main component in treating chronic LBP, but repetitive exercise for strengthening has limitations in increasing motivation for exercise programs and for treating psychosocial symptoms. Recently, the Nintendo Wii exercise program has been used for therapeutic effect in various populations9, 10). The *To
whom correspondence should be addressed. E-mail: sangheon@sch.ac.kr
Nintendo Wii console is an instrument of biofeedbackbased exercise. A user controls a game character in a virtual environment using a remote controller with a motion sensor. A user performs similar movements to real sports activities such as swinging a tennis racket or rowing a boat. The Nintendo Wii is commercially available and inexpensive. Game components motivate participants be involved in an exercise program through a sense of fun. A previous study reported that the Nintendo Wii exercise program improved upper extremity function in stroke patients10). The Nintendo Wii exercise program has been shown to improve cognitive function, balance ability, and visual perception in various populations11, 12). However, there have been few studies on the effect of the Nintendo Wii exercise program in chronic LBP. The purpose of this study was to investigate the effect of the Nintendo Wii exercise program on chronic work-related LBP compared to stability exercise. The hypothesis of this study was that the Nintendo Wii exercise program would significantly improve health-related quality of life (QOL) as well as physical function, including core strength, flexibility, endurance, and balance. Health-related QOL includes mental health and emotional factors. SUBJECTS AND METHODS Twenty-four workers from the K tire factory participated in this study. The inclusion criteria were as follows: 1) diagnosed with chronic LBP by a physician; 2) over 3
986 J. Phys. Ther. Sci. Vol. 25, No. 8, 2013 Table 1. General characteristics of subjects
Table 2. Changes in pain and physical function
CG (n=8)
LSE (n=8)
NWE (n=8)
Age Height Weight Years of service
45.50±5.34 169.43±2.07 68.90±5.78 18.75±7.38
43.37±5.42 170.57±3.23 72.50±7.09 16.00±6.88
44.12±5.48 170.48±2.87 71.92±7.1 17.00±7.59
Mean ± SD. * p<0.05. CG, control group; LSE, lumber stabilization exercise; NWE, Nintendo Wii exercise
months with LBP; and 3) no experience with surgical treatments for disc herniation, spina bifida, or spinal stenosis. The average age was 44 years old. The average length of service was 17 years (Table 1). Participants were randomly assigned to three groups: a control group (CG), lumbar stabilization exercise group (LSE), and Nintendo Wii exercise group (NWE). There were no significant differences among groups in height, weight, age, or years of service (p>0.05). In each group, physical therapy was provided using physical agent modalities such as a hot pack (30 min), interferential current therapy (15 min), and deep heat with ultrasound (5 min). Subsequently, an additional exercise program was provided in the LSE and NWE groups, but not in the control group. Participants were treated 3 times a week for 8 weeks. Each session lasted 30 minutes. Stabilization exercise included 7 positions based on the back bridge, hands and knees, and side bridge as follows13): 1) standard back bridge, 2) back bridge with alternate leg straight, 3) back bridge with one leg, 4) standard hands and knees, 5) hands and knees with alternate arm straight forward, 6) hands and knees with alternate leg straight backward, and 7) side bridge. Participants in the LSE group were asked to maintain each position for 15 seconds for 3 sets. The Nintendo Wii sports program was used for the Nintendo Wii exercise program, including the wakeboard, Frisbee dog, jet ski, and canoe games. Participants in the NWE group controlled a virtual character on the screen by swinging, rowing, and tilting remote controllers with motion sensors. Participants chose which Nintendo Wii sports program they performed and took a 2 minute break every 10 minutes. Before and after 8 weeks of intervention, the participants of each group were evaluated for pain, back strength, balance ability, and health-related QOL with the following tools; visual analog scale (VAS) for pain14), isometric lifting strength for back strength15, 16), One-legged Stand Test for balance ability16, 17), and the RAND-36 Health Status Inventory (RAND-36) for health-related QOL18). Back strength was defined as the maximum isometric lifting weight15). Balance was evaluated by measuring how long the subject could stand on one leg17). The RAND-36 includes a physical and mental health summary score derived from the 36-item Short-Form Health Survey (SF-36) items. The physical health category includes physical functioning, role limitations resulting from physical health, pain, and general health perceptions. The mental health category includes emotional well-being, role limitations resulting from emotional problems, social functioning, and energy/fatigue. One-way ANOVA was used to test group differences in pre-
Category Pain
Back strength Balance ability
Group CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8)
Pre-test
Post-test
6.50±1.30 6.62±0.74 6.75±1.38 53.73±22.07 54.43±12.72 54.48±10.31 8.62±3.20 8.50±3.07 8.75±1.83
5.75±0.88 4.87±0.83* 5.87±1.12* 55.58±23.20* 60.16±13.58* 57.81±9.07* 10.75±3.61* 11.37±3.62* 10.62±3.73
Mean ± SD. * p<0.05. CG, control group; LSE, lumber stabilization exercise; NWE, Nintendo Wii exercise
test values. Paired t-test was used to compare pre- and posttest values within each group. RESULTS Pain decreased significantly in the LSE and NWE groups, but not in the control group (p<0.05) (Table 2). All groups demonstrated a significant increase in back strength (p<0.05) (Table 2). The control and LSE groups showed a significant improvement in balance, but the NWE group did not. In the physical health composite of QOL, all three groups improved significantly in pain and general health (p<0.05) (Table 3). Only the LSE group improved significantly in physical functioning and role limitations due to physical health (p<0.05) (Table 3). In the mental health composite of QOL, the control and LSE groups showed no significant changes, but the NWE group improved significantly in terms of role limitations due to emotional problems, energy/fatigue, and emotional well-being (p<0.05) (Table 4). There were no significant differences in the pretest values among groups (p>0.05). DISCUSSION Work-related musculoskeletal disorders are a significant issue in industry. LBP is one of the most common workrelated disorders in the world3). The most common therapy for LBP includes physical agent modalities and stabilization exercise13). However, these common approaches for LBP do not consider mental and cognitive aspects that are important factors in LBP in addition to physical factors. Recently, the Nintendo Wii exercise program has been employed in various populations, and positive effects have been reported with regard to physical and mental health11, 19). However, there have been few studies on the effects of the Nintendo Wii exercise program on work-related LBP. Thus, the purpose of this study was to investigate the effects of the Nintendo Wii exercise program on work-related low back pain in factory workers. Participants were randomly assigned to three groups: control, stabilization exercise program, and Nintendo Wii exercise program. The results demonstrated that exercise programs significantly improved physical function-related LBP. In health-related QOL, the Nintendo
987 Table 3. Changes of QOL in the physical health composite of the RAND-36 Category
Group CG (n=8) LSE (n=8) NWE (n=8)
Physical functioning
Role limitations due to physical health
Pain
General health
Pre-test
Post-test
57.50±26.32 48.75±19.59 53.12±13.34
58.12±26.98 57.50±17.92* 54.37±12.37
CG (n=8)
67.25±20.87
68.12±21.66
LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8)
71.12±9.89 69.62±14.75 30.00±14.14 42.50±19.82 32.50±19.08 45.00±18.12 50.00±12.24 41.25±18.85
81.50±13.56* 73.50±16.65 36.25±10.60* 57.50±12.81* 40.00±14.14* 50.00±17.32* 64.37±11.78* 44.37±18.60*
Mean ± SD. * p<0.05. CG, control group; LSE, lumber stabilization exercise; NWE, Nintendo Wii exercise
Table 4. Changes of QOL in the mental health composite of the RAND-36 Category Role limitations due to emotional problem
Energy/Fatigue
Emotional wellbeing
Social functioning
Group CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8) CG (n=8) LSE (n=8) NWE (n=8)
Pre-test 71.00±19.84 78.00±13.23 68.75±15.90 62.75±22.77 68.87±15.51 58.00±23.68 70.62±7.28 65.00±9.25 63.12±6.51 59.50±19.69 69.00±14.82 61.00±18.22
Post-test 73.00±15.36 80.25±10.83 81.12±10.84* 62.00±16.85 70.50±10.92 68.12±17.26* 72.50±5.34 66.87±7.98 73.12±7.52* 59.62±18.74 72.00±3.93 69.12±9.35
Mean ± SD. * p<0.05. CG, control group; LSE, lumber stabilization exercise; NWE, Nintendo Wii exercise
Wii exercise program showed significant improvements in both the mental and physical health composites, but other groups showed significant improvement only in the physical health composite. Exercise programs are essential to reducing pain and improving QOL in work-related LBP. In the control group, no exercise program was provided outside of pain control approaches such as hot pack, ICT, and ultrasound. These physical agent modalities reduce pain physiologically, but there was no significant change demonstrated in this study because daily work causes cumulated LBP. The short-term effect of physical agent modalities did not significantly reduce work-related LBP. Although both stabilization exercise and Nintendo Wii exercise significantly reduced work-related LBP, stabilization exercise increased muscle strength more than Nintendo Wii exercise because stabilization exercise focused more on the low back muscles compared with the Nintendo Wii exercise. Additionally, only stabilization exercise significantly increased balance. The stabilization exercise program demonstrated more effective
increases in physical function than the Wii exercise program. Previous studies have reported that stabilization exercises for the low back significantly improved the strength of the low back muscles, thus reducing LBP13). However, a physical exercise program only did not improve return to work 20). In addition to physical function, LBP decreases the mental and physical aspects of health-related QOL. The Nintendo Wii exercise program significantly improved both the physical and mental composites of healthrelated QOL. In previous studies, LBP has been shown to affect not only physical function, but also aspects of mental health5, 8). For example, chronic LBP is related to depression 21). In this study, stabilization exercise did not significantly improve the mental health composite of health-related QOL. The Nintendo Wii exercise program influenced mental health as well as physical health. Other studies have demonstrated that rehabilitation programs using a Nintendo Wii console improved mental and cognitive functions in various populations, which may be related to the mental health aspects of health-related QOL11, 19). The gaming
988 J. Phys. Ther. Sci. Vol. 25, No. 8, 2013 components of Nintendo Wii sports encourage participants to enjoy exercise and require mental and cognitive functions, which may be possible reasons for improvement in the mental health composite. However, the Nintendo Wii exercise program was less effective than stabilization exercise in improving the physical function and physical health composites of QOL. In terms of clinical implications, the Nintendo Wii exercise program could be a biopsychosocial intervention for work-related LBP in factory workers. Recent research supports biopsychosocial approaches for LBP rather than stabilization exercise based on a biomedical model22). Low-back stabilization exercise has been the most commonly prescribed exercise for LBP, but only 8–15% of LBP was related to a patho-anatomical diagnosis23). Although low-back stabilization exercise was previously included in therapeutic interventions for work-related LBP24), stabilization exercise did not improve the mental health composite of QOL. The stabilization exercise was based on the biomechanical aspect, not the biopsychosocial aspect. The Nintendo Wii exercise program significantly improved the physical and mental health aspects of QOL. Even though the Nintendo Wii exercise program demonstrated a lower effect on physical function and physical health QOL than stabilization exercise, the Nintendo Wii exercise program improved the mental health composite of QOL, which stabilization exercise barely improved. Future studies are needed on various effects of Nintendo Wii exercise programs, including social and psychological factors, for work-related LBP in factory workers. ACKNOWLEDGEMENT This work was supported by the Soonchunhyang University Research Fund (No. 20110695). REFERENCES 1) Korea Occupational Safety and Health Agency: Annual Report. Incheon: KOSHA, 2010. 2) Andersson GB: Epidemiological features of chronic low-back pain. Lancet, 1999, 354: 581–585. [Medline] [CrossRef] 3) Brage S, Ihlebaek C, Natvig B, et al.: Musculoskeletal disorders as causes of sick leave and disability benefits. Tidsskr Nor Laegeforen, 2010, 130: 2369–2370. [Medline] [CrossRef] 4) Hoy D, Brooks P, Blyth F, et al.: The Epidemiology of low back pain. Best Pract Res Clin Rheumatol, 2010, 24: 769–781. [Medline] [CrossRef] 5) Sterud T, Tynes T: Work-related psychosocial and mechanical risk factors for low back pain: a 3-year follow-up study of the general working popula-
tion in Norway. Occup Environ Med, 2013, 70: 296–302. [Medline] 6) Pincus T, Santos R, Breen A, et al.: A review and proposal for a core set of factors for prospective cohorts in low back pain: a consensus statement. Arthritis Rheum, 2008, 59: 14–24. [Medline] [CrossRef] 7) Licciardone JC, Gatchel RJ, Kearns CM, et al.: Depression, somatization, and somatic dysfunction in patients with nonspecific chronic low back pain: results from the Osteopathic Trial. J Am Osteopath Assoc, 2012, 112: 783–791. [Medline] 8) Mok LC, Lee IF: Anxiety, depression and pain intensity in patients with low back pain who are admitted to acute care hospitals. J Clin Nurs, 2008, 17: 1471–1480. [Medline] [CrossRef] 9) Wuang YP, Chiang CS, Su CY, et al.: Effectiveness of virtual reality using Wii gaming technology in children with Down syndrome. Res Dev Disabil, 2011, 32: 312–321. [Medline] [CrossRef] 10) Saposnik G, Teasell R, Mamdani M, et al.: Effectiveness of virtual reality using Wii gaming technology in stroke rehabilitation: a pilot randomized clinical trial and proof of principle. Stroke, 2010, 41: 1477–1484. [Medline] [CrossRef] 11) Pompeu JE, Mendes FA, Silva KG, et al.: Effect of Nintendo Wii-based motor and cognitive training on activities of daily living in patients with Parkinson’s disease: a randomised clinical trial. Physiotherapy, 2012, 98: 196–204. [Medline] [CrossRef] 12) Clark R, Kraemer T: Clinical use of Nintendo Wii bowling simulation to decrease fall risk in an elderly resident of a nursing home: a case report. J Geriatr Phys Ther, 2009, 32: 174–180. [Medline] [CrossRef] 13) Imai A, Kaneoka K, Okubo Y, et al.: Trunk muscle activity during lumbar stabilization exercises on both a stable and unstable surface. J Orthop Sports Phys Ther, 2010, 40: 369–375. [Medline] 14) Hawksley H: Pain assessment using a visual analogue scale. Prof Nurse, 2000, 15: 593–597. [Medline] 15) Hattori Y, Ono Y, Shimaoka M, et al.: Test-retest reliability of isometric and isoinertial testing in symmetric and asymmetric lifting. Ergonomics, 1998, 41: 1050–1059. [Medline] [CrossRef] 16) Ko DS, Lee CG, Kim GY, et al.: The effect of lumbar stabilization exercise on motor capacity and pain in chronic low back pain workers. J Sports Lesiure Stud, 2009, 35: 1021–1028. 17) Franchignoni F, Tesio L, Martino MT, et al.: Reliability of four simple, quantitative tests of balance and mobility in healthy elderly females. Aging (Milano), 1998, 10: 26–31. [Medline] 18) Hays RD, Sherbourne CD, Mazel RM: The RAND 36-Item Health Survey 1.0. Health Econ, 1993, 2: 217–227. [Medline] [CrossRef] 19) Rosenberg D, Depp CA, Vahia IV, et al.: Exergames for subsyndromal depression in older adults: a pilot study of a novel intervention. Am J Geriatr Psychiatry, 2010, 18: 221–226. [Medline] [CrossRef] 20) Hagen EM, Odelien KH, Lie SA, et al.: Adding a physical exercise programme to brief intervention for low back pain patients did not increase return to work. Scand J Public Health, 2010, 38: 731–738. [Medline] [CrossRef] 21) Ellegaard H, Pedersen BD: Stress is dominant in patients with depression and chronic low back pain. A qualitative study of psychotherapeutic interventions for patients with non-specific low back pain of 3-12 months’duration. BMC Musculoskelet Disord, 2012, 13: 166. [Medline] [CrossRef] 22) O’Sullivan P: It’s time for change with the management of non-specific chronic low back pain. Br J Sports Med, 2012, 46: 224–227. [Medline] [CrossRef] 23) Waddell G: The back pain revolution. New York: Churchill Livingstone, 2004. 24) Davarian S, Maroufi N, Ebrahimi I, et al.: Trunk muscles strength and endurance in chronic low back pain patients with and without clinical instability. J Back Musculoskelet Rehabil, 2012, 25: 123–129. [Medline]
The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
Clinical Study
Real-world practice patterns, health-care utilization, and costs in patients with low back pain: the long road to guideline-concordant care Jasmina I. Ivanova, MAa,*, Howard G. Birnbaum, PhDb, Matt Schiller, BAb, Evan Kantor, BSb, Bryan M. Johnstone, PhDc, Ralph W. Swindle, PhDc a
Analysis Group, Inc., 10 Rockefeller Plaza, 15th Floor, New York, NY 10020, USA b Analysis Group, Inc., 111 Huntington Ave, 10th Floor, Boston, MA 02199, USA c Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN 46285, USA Received 28 September 2010; revised 2 February 2011; accepted 23 March 2011
Abstract
BACKGROUND CONTEXT: Treatment guidelines suggest that most acute low back pain (LBP) episodes substantially improve within a few weeks and that immediate use of imaging and aggressive therapies should be avoided. PURPOSE: Assess the actual practice patterns of imaging, noninvasive therapy, medication use, and surgery in patients with LBP, and compare their costs to those of matched controls without LBP. STUDY DESIGN: A retrospective analysis of claims data from 40 self-insured employers in the United States. PATIENT SAMPLE: The study sample included 211,551 patients, aged 18 to 64 years, with one LBP diagnosis or more (per Healthcare Effectiveness Data and Information Set specification) during 2004 to 2006, identified from a claims database. Patients had continuous eligibility for 12 months or more after their index LBP diagnosis (study period), for 6 months or more before their index diagnosis (baseline period), and no other LBP diagnosis during the baseline period. Patients with LBP were matched to a random cohort of patients without LBP by age, gender, employment status, and index year. OUTCOMES MEASURES: Physiological measures (eg, imaging and diagnostic tests), functional measures (eg, pharmacologic and nonpharmacologic treatment for LBP, health-care resource use), and direct (medical and prescription drug) and indirect (disability and medically related absenteeism) costs were assessed within the year after the LBP diagnosis. METHODS: Univariate analyses described treatment patterns and compared baseline characteristics and study period costs. RESULTS: Patients with LBP had significantly higher rates of baseline comorbidities and resource use compared with controls. Of patients with LBP, 41.6% had imaging mean (median) [standard deviation] 34.3 (0) [78.6] days after the LBP diagnosis. Most patients with LBP (69.4%) used medications starting 51.9 (8) [86.2] days after the diagnosis. Opioids were commonly prescribed early (41.6% of patients; after 82.8 (25) [105.9] days). Of patients with LBP, 2.05% had surgery during the study period. Patients with LBP were likely to have chiropractic treatment first, followed by pharmacotherapy with muscle relaxants and nonsteroidal anti-inflammatory drugs. Except for less
FDA device/drug status: Not approved for this indication (drug classes used in low back pain treatment). Author disclosures: JII: Consulting: Eli Lilly and Company (Salary); Relationships Outside the One Year Requirement: Eli Lilly and Company (Salary). HGB: Consulting: Eli Lilly and Company (Salary); Relationships Outside the One Year Requirement: Eli Lilly and Company (Salary). MS: Consulting: Eli Lilly and Company (Salary); Relationships Outside the One Year Requirement: Eli Lilly and Company (Salary). EK: Consulting: Eli Lilly and Company (Salary); Relationships Outside the One Year Requirement: Eli Lilly and Company (Salary). BMJ: Stock Ownership (including options, warrants): Eli Lilly and Company; Trips/Travel: Eli Lilly and Company. RWS: Stock Ownership (including options, warrants): Eli Lilly and Company; Trips/Travel: Eli 1529-9430/$ - see front matter Ă&#x201C; 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.spinee.2011.03.017
Lilly and Company (None); Other Relationships: Eli Lilly and Company (Salary). The disclosure key can be found on the Table of Contents and at www. TheSpineJournalOnline.com. Research funding was provided from Eli Lilly and Company to Analysis Group, Inc. All the authors are salaried employees and have received no compensation specific for this study. JII, HGB, MS, EK are employees of Analysis Group, Inc.; BJ was an employee of Eli Lilly and Company at the time of the study; and RS is an Eli and Company employee. * Corresponding author. Analysis Group, Inc., 10 Rockefeller Plaza, 15th Floor, New York, NY 10020, USA. Tel.: (212) 492-8100; fax: (212) 492-8188. E-mail address: jivanova@analysisgroup.com (J.I. Ivanova)
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622–632
623
surgery, these findings also held for patients with only nonspecific LBP. Patients with LBP had higher mean direct costs compared with controls ($7,211 vs. $2,382, respectively; p!.0001), with surgery patients having mean direct costs of $33,931. CONCLUSIONS: Contrary to clinical guidelines, many patients with LBP start incurring significant resource use and associated expenses soon after the index diagnosis. Achieving guidelineconcordant care will require substantial changes in LBP practice patterns. Ó 2011 Elsevier Inc. All rights reserved. Keywords:
Low back pain; Treatment pattern; Costs
Introduction Low back pain (LBP) is a common reason for seeking medical care and is associated with substantial direct (health care) and indirect (employer-paid work-loss) costs [1]. Estimates of the direct cost burden in the United States range considerably, with the most recent estimate at $86 billion in incremental health-care costs [2–7]. Large variations in diagnostic tests and treatments as well as the resulting costs have been reported previously, but patients seem to experience similar outcomes [8,9]. Low back pain could be classified into the following broad categories: nonspecific LBP, back pain potentially associated with radiculopathy or spinal stenosis, or back pain potentially associated with another specific spinal cause [10]. Guidelines by the American College of Physicians and the American Pain Society (ACP/APS) recommend that physicians should not routinely use imaging or diagnostic tests in patients with nonspecific LBP [10]. Routine imaging is not recommended because most imaging findings are nonspecific, and routine imaging does not improve clinical outcomes [11]. However, diagnostic imaging and testing is recommended for patients with LBP with severe or progressive neurologic deficits or when serious underlying conditions are suspected [10]. Because pain often improves on its own within the first month among patients who seek medical care [12], physicians are to advise patients to stay active and discuss self-care options initially [10]. Acetaminophen and nonsteroidal antiinflammatory drugs (NSAIDs) are recommended as firstline medications [10]. For patients who do not improve with self-care, physicians are advised to consider the addition of nonpharmacologic therapies, such as spinal manipulation, intensive interdisciplinary rehabilitation, exercise therapy, acupuncture, massage therapy, yoga, cognitive-behavioral therapy, or progressive relaxation [10]. This study uses retrospective claims data analyses to characterize patients with LBP, describe the real-world practice patterns related to their treatment, and compare their costs with those of matched controls without LBP. The specific objectives included comparing the characteristics of patients with a new episode of diagnosed LBP with those of demographically matched control patients without LBP diagnosis; describing the time to diagnostic/imaging procedures and nonsurgical treatment before surgery and the time to surgery among patients with LBP; comparing health-care costs of patients with LBP with those of
demographically matched controls as well indirect costs and describing health-care costs among subgroups of patients with LBP. Our hypotheses were that patients with LBP would likely have more comorbidities, higher resource use, and costs compared with demographically matched controls without LBP and that LBP surgery would be associated with substantial costs. Methods Data source The study sample was selected from a nationwide, privately insured claims database covering over 8 million beneficiaries (including employees, spouses, and dependents) from 40 US-based companies for services provided from 1999 to 2006. The database contains deidentified information on patients’ demographics (eg, age, gender), enrollment history, and medical and pharmacy claims. Utilization of medical services was recorded with dates of service, associated diagnoses (up to two codes, using the codes for International Classification of Diseases, Ninth Edition), performed procedures (Current Procedural Terminology), place of service codes, and actual amounts paid to providers. The database also includes pharmacy claims with prescribed medications identified by National Drug Code, date of prescription fill, days of supply, quantity, and actual payment amounts. Disability claims that report work-loss dates and payments to employees are available for a subset of these companies. Sample selection Patients with LBP The study sample was drawn from patients aged 18 to 64 years who had at least one LBP diagnosis, as defined by Healthcare Effectiveness Data and Information Set (HEDIS) [13], during 2004 to 2006 to evaluate more recent treatment patterns and costs. Patients were required to have at least 6 months of continuous benefit eligibility before the initial (index) LBP diagnosis (baseline period that could begin as early as July 1, 2003) without another LBP diagnosis during the baseline period and 12 months of continuous benefit eligibility after the index LBP diagnosis (study period). A subset of LBP employees in companies providing disability data was also identified.
624
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
Context Guidelines founded upon the currently available evidence base have been established for the treatment of acute low back pain. Using a large database, the authors aim to assess whether such guidelines are generally followed and costs relative to non-LBP patients. Contribution Among the questionable practices identified by the authors, a third of acute low back pain patients underwent X-rays; 70% received prescription medications, most often opioids. Costs for these patients were three times greater than for those of matched controls. Implication Guidelines for non-chronic low back pain generally suggest reassurance of the favorable natural history of the acute episode, over the counter analgesia/NSAIDs as needed, and progressive return to activities. The findings of this article suggest that more aggressive and costly management strategies are commonly employed despite such recommendations. While the methodology used clearly has limitations (surely some patients had real red flags, already chronic LBP by history, etc.); it is likely the authorsâ&#x20AC;&#x2122; findings are not purely the result of methodological flaws, and apparent large-scale deviation from best-practice guidelines has been observed. These findings are consistent with the studies of patients with acute lower back pain in Canada, Spain, and Australia. â&#x20AC;&#x201D;The Editors
Controls without LBP Controls were identified among patients aged 18 to 64 years who had no LBP diagnosis during 1999 to 2006. Controls were required to have at least 18 months of continuous benefit eligibility after July 1, 2003 (6-month baseline and 12-month study period). The index date was defined as the date 6 months into the most recent 18-month eligibility period. Matching patients with LBP to controls Patients with LBP were matched 1:1 to beneficiaries without an LBP diagnosis by age, gender, employment status, and index date within 1 year. Patient characteristics Patient characteristics compared between the study and control groups included variables on which the study and control patients were matched (eg, age, gender, employment
status), baseline comorbidities, and severity indicators assessed over the 6-month baseline period. Chronic low back pain (CLBP) was defined as LBP present for more than 3 months [10]. Patients who developed CLBP were identified as having another LBP diagnosis that was between 3 and 6 months of the first diagnosis. Several back pain diagnostic severity categories used elsewhere were constructed for the index LBP diagnosis [14,15]. Baseline comorbidity severity indicators included the Charlson Comorbidity Index (CCI) [16,17], the Chronic Disease Score [18], and other comorbidities associated with LBP (eg, osteoarthritis, fibromyalgia, hypertension, depression, and anxiety disorder) [19]. Other proxies for severity included baseline period use of opioids, reported in previous studies to be associated with higher comorbidity burden and higher use of health-care services, and baseline period use of corticosteroids, a risk factor for fractures [10,19]. Imaging, diagnostic tests, and LBP treatments In the primary analysis, treatment patterns were described among all patients with LBP. As a sensitivity analysis, treatment patterns were also described among the subgroup of patients with LBP who had an index diagnosis of nonspecific backache based on Cherkin et al. [14] and who did not have a diagnosis of spinal stenosis, herniated disc, cancer, trauma, or intravenous (IV) drug use [13]. Among patients with LBP and spinal stenosis, herniated disc, cancer, trauma, or IV drug use, the use of imaging procedures or invasive treatment may be appropriate. Table 1 summarizes the codes used to identify imaging, diagnostic tests, noninvasive treatments, and surgery relevant to LBP treatment based on practice guidelines. Timing to imaging, diagnostic testing, noninvasive treatment, and LBP-related medication use was described during the study period. For patients who underwent LBP surgery during the study period, only imaging, testing, and treatments occurring before surgery were reported to describe nonsurgical treatment of LBP and avoid reporting treatment use directly as a result of the surgery. Imaging procedures such as magnetic resonance imaging, computed tomography scan, X-ray, and radiography were identified based on CPT codes specified by HEDIS [8]. Other diagnostic procedures (eg, electromyography, nerve conductor velocity) were based on published guidelines for LBP diagnosis [5,10]. Noninvasive treatments for LBP described in practice guidelines and reported here included chiropractic treatment (defined as acupuncture, massage, spinal manipulation, and other chiropractic procedures), cognitive behavioral therapy, exercise therapy, and physical modalities (ie, interferential therapy, low-level laser therapy, lumbar support, shortwave diathermy, traction, transcutaneous electric nerve stimulation, and ultrasound) [20]. Low back pain treatment medications included prescribed acetaminophen, NSAIDs, skeletal muscle relaxants,
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
625
Table 1 Codes to identify LBP, imaging, diagnostic tests, noninvasive treatment, and surgery Diagnosis or procedure codes
CPT or ICD-9-CM code
LBP (per HEDIS)
ICD-9-CM diagnosis codes: 721.3, 722.10, 722.32, 722.52, 722.93, 724.02, 724.2, 724.3, 724.5, 724.6, 724.7, 724.71, 724.79, 738.5, 739.3, 739.4, 846.0, 846.1, 846.2, 846.3, 846.8, 846.9, 847.2
Imaging MRI X-ray CT scan Lumbosacral plain radiology
CPT CPT CPT CPT
codes: codes: codes: codes:
72141, 72010, 72131, 72100,
72142, 72146, 72147, 72148, 72149, 72156, 72158 72020, 72052, 72114, 72120, 72200, 72202, 72220 72132, 72133 72110
Diagnostic tests C-reactive protein Electromyography Erythrocyte sedimentation rate Nerve conductor velocity
CPT CPT CPT CPT
codes: codes: codes: codes:
86140, 96002, 85032, 95900,
86141 96003, 96004, 95872 85041 95903, 95924
Nonpharmacologic therapy Chiropractic treatment (spinal manipulation, massage, acupuncture) Cognitive behavioral therapy or progressive relaxation Exercise therapy Physical therapy Surgery Fusion
Discectomy Laminectomy Other surgery
CPT codes: 22505, 98940, 98941, 98942, 97780, 97781, 97810, 97811, 97813, 97814, 97124 CPT codes: 22505, 97124, 97780, 97781, 97810, 97811, 97813, 97814, 98940, 98941, 98942 CPT codes: 97110, 97112, 97113, 97116, 97139 CPT codes: 97014, 97032, 64999, 97026, 97024, 97140, 97014, 97032, 97035 CPT codes: 22532, 22533, 22534, 22558, 22585, 22612, 22614, 22630, 22632, 22800, 22802, 22804, 22808, 22810, 22812, 22830 ICD-9-CM procedure codes: 81.00, 81.06, 81.07, 81.08, 81.09 CPT codes: 63030, 63035, 63042, 63044 ICD-9-CM procedure codes: 80.5 CPT codes: 63005, 63011, 63012, 63017, 63047, 63048, 63185, 63190 ICD-9-CM procedure codes: 03.30, 03.09 CPT codes: 22818, 22819, 22842, 22851, 63295, 20930, 20931, 20936, 20937, 22102, 63085, 63086, 63087, 63088, 63090, 63091, 63710 ICD-9-CM procedure codes: 03.02, 03.6, 78.69
CPT, Current Procedural Terminology; LBP, low back pain; ICD-9-CM, International Classification of Diseases, Ninth Edition, Clinical Modification; MRI, magnetic resonance imaging; CT, computed tomography.
antidepressants, antiepileptic drugs, benzodiazepines, tramadol, opioids, and systemic corticosteroids [21]. Drug classes were identified by National Drug Codes obtained from a pharmacy database (Master Drug Data Base [Medi-Span, Indianapolis, IN, USA]) and using Healthcare Common Procedure Coding System codes in the case of corticosteroid injections. Surgical procedures for LBP described here included fusion, laminectomy, discectomy, and other surgical procedures.
absenteeism costs were calculated based on the number of work-loss days associated with the use of medical services multiplied by employeesâ&#x20AC;&#x2122; daily wage obtained from the eligibility file. All costs were calculated over the 12-month study period and inflated to 2006 US dollars using the Consumer Price Index for Medical Care (310.1 in 2004, 323.2 in 2005, and 336.2 in 2006) [22].
Direct and indirect cost calculation
Univariate analyses were used to compare patients with LBP and the employed subset to their matched controls without LBP. Categorical variables were compared using McNemar tests. Continuous variables such as age and CCI were compared using t tests. Resource use counts, such as work-loss days and costs, were compared using nonparametric Wilcoxon signed-rank tests. The study reported the overall costs of patients with LBP rather than the incremental costs associated with LBP. Therefore, no risk-adjusted direct cost estimates (adjusting for differences in baseline comorbidities) were provided for patients with LBP and controls.
Direct costs were based on reimbursements (ie, paid amounts) from third-party payers to health-care providers for inpatient, emergency department, outpatient (eg, outpatient surgery), other ancillary services (eg, physical therapy, laboratory services), and prescription drugs. Patient outof-pocket costs were not included in direct costs. Indirect costs were calculated for the subset of employees from companies providing disability data and included actual employer payments for disability days plus estimated costs for medically related absenteeism. Medically related
Statistical analyses
626
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
Table 2 Baseline characteristics of patients with LBP and controls Patients
Employees
Patients with LBP (N5211,551)
Controls (N5211,551)
p Value
Employees with LBP (N591,194)
Controls (N591,194)
p Value
Baseline characteristics
[1]
[2]
[1] vs. [2]
[3]
[4]
[3] vs. [4]
Age, mean (SD)
45.7 (11.6)
45.7 (11.6)
1.0000
46.1 (10.8)
46.1 (10.8)
1.0000
Female (%)
54.4
54.4
1.0000
39.5
39.5
1.0000
Employed (%)
61.3
61.3
1.0000
100.0
100.0
1.0000
Chronic LBP (%)
16.7
0.0
!.0001
17.6
0.0
!.0001
Specific LBP diagnostic group (Cherkin et al.)* (%) Herniated disc Possible degenerative changes Spinal stenosis Possible instability Nonspecific backache Miscellaneous
5.2 8.9 1.3 1.5 75.5 16.9
0.0 0.0 0.0 0.0 0.0 0.0
!.0001 !.0001 !.0001 !.0001 !.0001 !.0001
5.4 8.4 1.2 1.5 75.1 17.6
0.0 0.0 0.0 0.0 0.0 0.0
!.0001 !.0001 !.0001 !.0001 !.0001 !.0001
LBP severity (Webster et al.) (%) More severe Less severe
14.0 86.0
0.0 0.0
!.0001 !.0001
14.3 85.7
0.0 0.0
!.0001 !.0001
LBP-related comorbidities (%) Osteoarthritis Fibromyalgia Anxiety Depression Hypertension
4.5 3.0 3.1 5.5 15.3
1.4 0.48 1.0 1.9 8.0
!.0001 !.0001 !.0001 !.0001 !.0001
3.9 2.7 2.8 5.0 15.1
1.1 0.37 0.86 1.5 7.1
!.0001 !.0001 !.0001 !.0001 !.0001
Charlson comorbidity index, mean (SD)
0.26 (0.82)
0.12 (0.56)
!.0001
0.25 (0.81)
0.11 (0.55)
!.0001
Chronic disease score, mean (SD)
2.0 (2.5)
1.7 (2.2)
!.0001
1.9 (2.4)
1.7 (2.2)
!.0001
Prior systemic corticosteroid use (%)
9.9
2.6
!.0001
9.7
2.1
!.0001
Prior use of tramadol or opioids (%)
26.7
6.5
!.0001
25.7
5.5
!.0001
LBP, low back pain; SD, standard deviation. * The sum of percentages of patients in different LBP diagnostic groups exceeds 100% because patients with multiple claims with different LBP diagnoses on index date could be categorized into multiple LBP diagnostic groups.
Resource use rates before surgery and time to first resource use were also stratified by CLBP and surgery status. All analyses were conducted using SAS version 9.1 (SAS Institute Inc., Cary, NC, USA). p Values less than .05 were considered to indicate statistically significant differences.
LBP-related comorbidities and higher values for severity indicators, such as CCI and Chronic Disease Score. Patients with LBP also had about four times higher rates of baseline-period corticosteroid and prescription opioid use. Treatment patterns
Results Patient characteristics The study sample included 211,551 patients with LBP meeting the selection criteria; 91,194 were employed in companies providing disability data. Table 2 presents baseline characteristics for the study sample. Patients with LBP were on average 45.7 years old and 54.4% were women. Most patients had index LBP diagnosis classified as nonspecific backache [14] (75.5%) or less severe LBP [15] (86.0%). Compared with matched controls without LBP, patients with LBP had significantly higher rates of baseline-period
Although guidelines suggest that in most patients who seek medical care back pain symptoms improve within a month, only 15.8% of patients had only the initial LBP diagnosis, and no subsequent back pain-relevant care past 1 month after the initial diagnosis suggesting that symptoms may persist even if patients experience improvement. Among patients with LBP, 16.7% had a subsequent LBP diagnosis from 3 to 6 months after the index LBP diagnosis and were classified as having CLBP. Only 2.05% of patients underwent back surgery within the subsequent year. Treatment pattern characteristics within a year after the index LBP diagnosis and before surgery are reported in Table 3. Resource use stratified by CLBP and surgery status is reported in Table 4.
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
627
Table 3 Treatment pattern among beneficiaries with LBP during the year after index LBP diagnosis and before surgery Patients with LBP (N5211,551)
Nonspecific backache patients* (N5106,737)
Treatment and diagnostic/imaging procedures
%
Days from LBP diagnosis to first claim, mean (median) [SD]
%
Days from LBP diagnosis to first claim, mean (median) [SD]
Any imaging procedure MRI X-ray imaging Discography CT
41.6 18.0 32.2 0.2 1.4
34.3 63.6 34.7 149.4 81.3
(0) [78.6] (14) [95.9] (0) [80.0] (138) [104.1] (28) [104.5]
32.4 8.5 27.4 0.0 0.4
33.5 74.6 29.9 137.2 63.1
(0) [79.0] (19) [102.8] (0) [76.2] (129) [105.4] (13) [99.6]
Any diagnostic procedure Electromyography C-reactive protein Erythrocyte sedimentation rate Nerve conductor velocity
14.1 3.8 4.6 8.4 4.5
127.5 120.2 146.7 132.5 120.2
(100) [113.6] (83) [110.1] (134) [115.1] (110) [115.5] (83) [111.0]
11.6 2.3 4.0 7.5 2.9
127.8 129.6 143.3 127.3 130.1
(102) [116.2] (98) [114.3] (130) [117.5] (102) [117.2] (101) [114.2]
Any LBP-related medication use Acetaminophen NSAIDs Antidepressants Benzodiazepines Antiepileptic drugs Skeletal muscle relaxants Opioids Nonnarcotic opioids (tramadol) Narcotic opioids Systemic corticosteroids
69.4 8.0 35.0 22.2 11.8 8.4 27.0 41.6 6.6 39.2 16.9
51.9 111.5 73.5 79.2 107.6 94.2 52.1 82.8 101.1 86.7 117.6
(8) [86.2] (71) [112.8] (21) [100.4] (40) [91.7] (65) [107.3] (49) [101.4] (1) [92.7] (25) [105.9] (51) [111.2] (29) [107.4] (83) [114.6]
67.5 6.7 34.3 19.5 9.9 6.2 28.5 37.2 5.4 34.9 14.1
49.4 107.7 66.4 81.4 108.2 95.7 41.4 80.6 95.0 84.7 123.1
(3) [86.5) (64) [114.3] (9) [99.4] (39) [94.4] (66) [108.4] (49) [103.1] (0) [85.7] (17) [107.1] (38) [111.9] (22) [108.7] (94) [115.7]
Any noninvasive treatment for LBP Chiropractic therapy Cognitive therapy Exercise therapy Physical therapy
49.5 39.2 8.4 22.7 34.0
39.1 33.2 108.3 66.2 50.1
(0) [81.3] (0) [76.8] (64) [109.2] (14) [98.2] (3) [90.2]
40.2 30.5 7.4 17.4 26.7
43.0 35.5 112.3 61.9 49.2
(1) [85.8] (0) [80.1] (70) [111.1] (11) [97.4] (3) [90.9]
122.2 142.9 107.4 136.7 146.0
(90) [105.7] (125) [109.7] (71) [99.7] (112) [105.9] (129) [111.9]
153.2 144.9 80.4 147.6 159.0
(138) [115.5] (130) [117.2] (38) [106.7] (146) [114.1] (146) [115.5]
Any surgery for LBP Fusion Discectomy Laminectomy Other surgery for LBP
2.05 0.67 0.98 0.57 0.84
0.31 0.14 0.01 0.03 0.27
LBP, low back pain; SD, standard deviation; MRI, magnetic resonance imaging; CT, computed tomography; NSAID, nonsteroidal anti-inflammatory drug. * Patients with index diagnosis of nonspecific backache based on Cherkin et al., without study period diagnoses of herniated disc or spinal stenosis, without baseline or study period diagnosis of cancer, trauma, or IV drug use.
A high proportion of patients with LBP (41.6%) had imaging within a year after the LBP diagnosis and before any LBP surgical procedure. The distribution of days to first imaging procedure among patients with imaging was highly skewed. Although the first imaging procedure observed in the claims data occurred mean 34.3 days after the index LBP diagnosis, approximately 60% of patients with imaging had an imaging procedure on the same day as the index diagnosis and 80% had an imaging procedure within a month from the diagnosis. X-rays were the most common imaging procedure performed in 32.2% of all patients (Table 3). Patients who subsequently underwent surgery were highly likely to have undergone imaging (Table 4). Most patients with LBP (69.4%) were treated with LBPrelated prescription medication during the study period before any surgical procedure. Among patients who were treated with LBP-related medications, the first prescription was filled a mean of 51.9 days after the index LBP
diagnosis. At least half of the patients who filled an LBPrelated prescription, however, filled their first prescription within 8 days after the LBP diagnosis. Patients first tended to be treated with skeletal muscle relaxants (mean [median]: 52.1 [1] days), followed by prescription NSAIDs (73.5 [21] days), antidepressants (79.2 [40] days), and opioids (82.8 [25] days). Opioids were the most frequently prescribed LBP-related medication (41.6%). As was the case with imaging, patients who would ultimately have surgery were more likely to receive prescription medications and receive them earlier in the course of treatment than nonsurgery patients (Table 4). Among patients with LBP, 49.5% received noninvasive treatment for LBP, mean (median), 39.1 (0) days after the index LBP diagnosis. Chiropractic treatment was the most common noninvasive treatment (39.2%), with the first chiropractic treatment occurring the same day as the index diagnosis for most patients receiving chiropractic care.
628
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
Table 4 Treatment pattern among beneficiaries with LBP during the year after index LBP diagnosis and before surgery by CLBP and surgery status CLBP (N535,295) Surgery patients, N52,152 Treatment and diagnostic/ imaging procedures
%
Non-CLBP (N5176,256) Patients without surgery, N533,143
Days to first claim, mean (median)
%
Days to first claim, mean (median)
Surgery patients, N52,179 %
Patients without surgery, N5174,077
Days to first claim, mean (median)
%
Days to first claim, mean (median)
(2) (8) (15) (188.5) (28)
38.0 14.2 29.4 0.0 0.9
29.9 58.0 28.5 134.1 56.9
(0) (8) (0) (68) (6)
126.9 117.8 144.7 131.5
(97) (69) (131) (108)
Any imaging procedure MRI X-ray imaging Discography CT
94.1 81.6 82.8 5.3 19.7
31.4 49.1 63.4 133.0 106.8
(4) (18) (28) (126) (99)
54.2 30.2 41.5 0.4 2.5
50.5 82.0 51.7 166.0 113.6
(4) (50) (1) (153) (91)
85.9 69.8 69.4 1.4 8.3
34.6 51.6 57.4 163.3 86.7
Any diagnostic procedure Electromyography C-reactive protein Erythrocyte sedimentation rate Nerve conductor velocity
27.9 19.3 3.0 7.9
99.5 109.3 75.3 89.6
(80) (94) (59) (70)
18.2 6.2 5.5 9.8
136.0 132.4 159.2 142.1
(117) (111.5) (153) (126)
21.2 14.6 1.9 6.4
81.1 95.1 76.3 65.5
(35) (42) (30) (28)
13.1 3.0 4.5 8.2
15.9
99.1 (75)
7.4
130.5 (110)
11.2
85.7 (38)
3.7
Any LBP-related medication Acetaminophen NSAIDs Antidepressants Benzodiazepines Antiepileptic drugs Skeletal muscle relaxants Opioids Nonnarcotic opioids (tramadol) Narcotic opioids Systemic corticosteroids
84.8 13.5 43.7 24.0 15.8 17.3 41.4 73.7 15.2
22.6 64.3 37.7 43.7 69.4 67.6 41.2 43.1 67.9
(3) (32) (16) (27) (32) (42) (10) (14) (46)
70.6 9.4 37.7 26.2 14.9 12.1 28.4 45.5 9.5
58.3 119.1 86.5 78.6 108.3 97.6 72.8 86.6 108.1
(16) (95) (48) (40) (71) (56) (27) (40) (79)
74.5 10.6 31.0 17.9 13.4 10.7 34.6 62.0 9.3
22.5 65.6 41.9 40.9 63.0 60.2 43.3 41.5 78.6
70.6 33.7
47.2 (15) 49.8 (20)
42.8 19.9
90.6 (46) 119.6 (95)
59.4 28.5
Any noninvasive treatment Chiropractic treatment Cognitive therapy Exercise therapy Physical therapy
46.7 34.4 6.1 29.4 37.8
31.3 33.5 53.8 44.7 36.4
77.8 67.9 10.6 40.1 57.3
30.0 27.7 105.3 68.5 50.0
32.1 22.9 4.2 16.2 24.2
Any surgery for LBP
100.0
(8) (8) (21) (25) (14)
140.9 (129)
(0) (0) (59) (25) (5)
100.0
(3) (22) (11) (23) (21) (24) (5) (7) (30.5)
118.8 (71)
69.0 7.6 34.4 21.5 11.1 7.6 26.5 40.3 5.9
51.4 111.5 71.7 80.2 108.7 94.5 48.2 83.7 100.4
42.4 (8) 60.8 (14)
37.9 16.0
87.7 (28) 120.2 (85)
32.5 35.1 54.6 41.6 39.5
44.4 34.0 8.1 19.4 29.6
42.3 35.3 109.9 66.0 50.5
(4) (2) (21) (9.5) (6)
(7) (69) (16) (40) (65) (48) (0) (23) (42)
(0) (0) (67) (12) (2)
103.8 (54)
LBP, low back pain; CLBP, chronic low back pain; MRI, magnetic resonance imaging; CT, computed tomography; NSAIDs, nonsteroidal antiinflammatory drugs.
Unlike imaging and medication, future surgery patients tended not to receive early chiropractic treatments, and they were less likely to receive chiropractic care than patients with LBP who did not undergo surgery (Table 4). The 2.05% of patients undergoing LBP surgery within a year after the index diagnosis had surgery a mean (median) of 122.2 (90) days after the index LBP diagnosis. Contrary to guidelines recommending that physicians should discuss the risks and benefits of surgery with patients with persistent and disabling symptoms [23], 50.3% LBP surgery patients had surgery before they could be classified as having CLBP (ie, having LBP persisting for more than 3 months); among patients having surgery and not classified as having CLBP, the surgery occurred within 54 days (median) of the initial LBP diagnosis (Table 4). Among patients with LBP with surgery, 70.8% had at least one herniated disc diagnosis in claims data before surgery, 55.7% had a degenerative changes diagnosis, 36.2% had a spinal stenosis diagnosis, and 11.1% had only a nonspecific backache diagnosis (data not shown).
The sensitivity analysis evaluating treatment patterns among the subgroup of patients who had a nonspecific backache index diagnosis and who did not have spinal stenosis, herniated disc, cancer, trauma, or IV drug use shows that 9.7% of patients with an index diagnosis of nonspecific backache were later diagnosed with herniated disc or spinal stenosis. Of the patients initially diagnosed with nonspecific backache, 26.3% also had at least one diagnosis of cancer, trauma, or IV drug use during the baseline or study periods and were excluded. Among the subgroup of patients who had a nonspecific backache index diagnosis and who did not have spinal stenosis, herniated disc, cancer, trauma, or IV drug use (n5106,737; 50.5% of patients with LBP), the imaging and noninvasive treatment patterns were similar to that of the overall LBP population (Table 3). Among the subgroup patients, 32.4% had at least one imaging procedure a mean (median) of 33.5 (0) days after the index LBP diagnosis; X-rays were the most common imaging procedure performed in 27.4% of patients. Low back painâ&#x20AC;&#x201C;related medication use was also
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622â&#x20AC;&#x201C;632
629
Table 5 Direct and indirect costs of patients and employees with LBP and controls All patients
Costs and work-loss days
Employees
Patients with LBP, N5211,551
LBP controls, N5211,551
[1]
[2]
p Value
Patients with LBP, N591,194
LBP controls, N591,194 [4]
p Value
[1] vs. [2]
[3]
$519 ($1,771)
!.0001
$1,178 ($3,143)
$450 ($1,795)
!.0001
$293 ($1,125)
$75 ($404)
!.0001
$242 ($965)
$56 ($320)
!.0001
$991 ($2,685)
$443 ($1,641)
!.0001
$936 ($2,818)
$394 ($1,710)
!.0001
$5,927 ($17,861)
$1,863 ($10,307)
!.0001
$5,714 ($16,774)
$1,642 ($9,225)
!.0001
$1,878 ($12,501) $259 ($1,041)
$601 ($6,895) $70 ($477)
!.0001 !.0001
$1,696 ($11,286) $245 ($1,017)
$524 ($6,424) $68 ($496)
!.0001 !.0001
$3,790 ($9,335)
$1,192 ($6,020)
!.0001
$3,773 ($9,321)
$1,050 ($5,206)
!.0001
$7,211 ($18,695)
$2,382 ($10,785)
!.0001
$6,892 ($17,550)
$2,091 ($9,689)
!.0001
Drug costs, mean (SD) LBP-related medications Nonâ&#x20AC;&#x201C;LBP-related medications
$1,284 ($3,124)
Medical costs, mean (SD) Inpatient costs Emergency department costs Outpatient/other costs Total direct costs
[3] vs. [4]
Indirect costs, mean (SD) Medically related absenteeism costs Disability costs
$2,606 ($4,876)
$750 ($2,525)
!.0001
$1,768 ($2,539)
$471 ($1,322)
!.0001
$838 ($4,214)
$279 ($2,113)
!.0001
Total costs
$9,498 ($19,892)
$2,841 ($10,752)
!.0001
15.5 (35.3)
4.9 (17.5)
!.0001
8.4 (10.4)
2.2 (5.7)
!.0001
7.1 (34.5)
2.7 (16.5)
!.0001
Work-loss days, mean (SD) Medically related absenteeism days Disability days LBP, low back pain; SD, standard deviation.
common (67.5%), 49.4 (3) days after the LBP diagnosis. Opioids were the most frequently prescribed LBP-related medication (37.2%), received 80.6 (17) days after the LBP diagnosis. Among patients with LBP with nonspecific backache, 40.2% received noninvasive treatment for LBP 43.0 (1) days after the index LBP diagnosis. Chiropractic treatment was the most common noninvasive treatment (30.5%). Among the subgroup of patients with nonspecific backache, the surgery rate was 0.31%, with 0.14% of patients undergoing spinal fusion. Surgery was performed a mean (median) of 153.2 (138) days after the LBP diagnosis.
Among the 4,331 patients with LBP who underwent surgery, the mean annual direct costs were $33,931 (standard deviation [SD]: $42,018); 94.4% of all costs were attributable to medical services costs and 65.3% of those were incurred in an inpatient setting. Mean annual direct costs were $38,868 (SD: $47,755) in surgery patients with CLBP and $29,054 (SD: $34,783) in surgery patients without CLBP (ie, no subsequent LBP diagnosis 3 to 6 months after the initial LBP diagnosis). Among patients with LBP who did not undergo surgery, mean direct costs were $8,513 (SD: $18,921) in patients with CLBP and $6,298 (SD: $17,139) in those without CLBP (data not shown).
Direct costs Patients with LBP had about 3 times higher average direct costs compared with controls ($7,211 vs. $2,382, respectively; p!.0001). Outpatient and other medical services costs accounted for 52.6% of direct costs, inpatient for 26.0%, and medications for 17.8%. All direct cost components were significantly higher among patients with LBP compared with controls. Employees with LBP also had significantly higher direct costs compared with used controls ($6,892 vs. $2,091, respectively; p!.0001) (Table 5).
Work-loss and indirect costs Compared with controls without LBP, employees with LBP had significantly more mean disability days (7.1 vs. 2.7 days, respectively; p!.0001) and medically related absenteeism days (8.4 vs. 2.2 days, respectively, p!.0001). Overall, employees with LBP had over three times the number of work-loss days as their controls (15.5 vs. 4.9 days) (Table 5). On average, total indirect costs for employees with LBP were $2,606 compared with $750 for controls (p!.0001).
630
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622–632
Medically related absenteeism costs accounted for 67.8% of indirect costs. The total direct and indirect costs were over three times higher among employees with LBP compared with controls ($9,498 vs. $2,841, respectively; p!.0001) (Table 5). Among the 1,937 employees with LBP who underwent surgery, the average annual indirect costs were $6,670 and the total direct and indirect costs were $39,057 (data not shown).
Discussion These findings suggest that contrary to clinical guidelines, many patients with LBP immediately incur significant resource use and, as a result, accumulate fairly substantial expenses soon after an initial LBP diagnosis. The ACP/APS clinical guidelines recommend that most patients with nonspecific LBP should receive self-care options and be reevaluated in a month [10]. Routine use imaging or diagnostic tests in patients with nonspecific LBP is not recommended. Acetaminophen and NSAIDs are recommended as first-line medications [10]. For patients who do not improve with self-care, physicians are advised to consider the addition of nonpharmacologic therapies, such as spinal manipulation, intensive interdisciplinary rehabilitation, exercise therapy, acupuncture, massage therapy, yoga, cognitive-behavioral therapy, or progressive relaxation [10]. Furthermore, surgery is not recommended for nonspecific LBP unless symptoms have been present for at least 1 year, and there are no serious psychiatric or medical comorbidities or other risk factors associated with poor surgical outcomes [23]. Previous research has also found low concordance to clinical guidelines in Canada, Australia, and Israel [24–26]. In one prior study, different methods of knowledge transfer of clinical guidelines to physicians and patients were all found to be ineffective at improving guideline concordance [24]. Prior LBP research has also found that evidence-based care resulted in lower cost of treatment than usual care [27]. This study found that diagnostic and treatment interventions are typical in the first month. For example, 32.2% of all patients with LBP received X-rays, with at least 50% receiving them the same day as the initial LBP diagnosis. Low back pain–related medication is received by 69.4% of patients, with at least 50% of all those patients filling a relevant prescription within 8 days of the initial diagnosis. At least half of the 41.6% of patients receiving opioids fill the prescription within 25 days of the initial visit. Noninvasive treatments are received by 49.5% of patients, with the median start days for chiropractic, physical, and exercise therapy being 0, 3, and 14 days after initial diagnosis, respectively. Finally, the median number of days to surgery for the 2.05% of patients with LBP receiving surgery is 90 days after the initial LBP visit. Among patients with nonspecific backache, the imaging and noninvasive therapy
pattern is similar to the overall LBP population, but the surgery rate is lower. These findings confirm previous study findings that LBP is associated with substantial economic burden [2] and extend previous findings pointing to how ‘‘frontloaded’’ treatment costs tend to be, even among patients with nonspecific backache. Patients with LBP had about three times higher direct costs on average compared with age and gender matched controls ($7,211 vs. $2,382, respectively; p!.0001). Among employees with LBP, indirect costs were also over three times higher compared with employed controls ($2,606 vs. $750, respectively; p!.0001). Not only is LBP costly but also substantial costs are incurred in the initial 3 months after diagnosis. Nonsurgery patients whose treatment episode took less than 90 days (ie, patients without CLBP) averaged $6,298 in total direct costs annually and those nonsurgery patients with CLBP averaged $8,513 annually. Patients without CLBP receiving surgery averaged $29,054 in direct annual costs and those with CLBP and surgery averaged $38,868. Most patients within the study sample had an index LBP diagnosis classified as nonspecific backache or back pain with no neurologic findings (75.5%) or less severe symptoms (86.0%), which is consistent with the range of less severe LBP or nonspecific backache of 74% to 85% in published studies [10,15,28–30]. Of the patients with LBP in this study, 50.5% had an index diagnosis of nonspecific backache without any subsequent diagnosis of herniated disc or spinal stenosis and without diagnoses of cancer, trauma, or IV drug use. Back pain with neurologic findings was diagnosed in 6.6% of patients. Thus, overall, few subjects in this LBP population appeared to meet criteria for the early aggressive diagnostic and treatment timetables often observed even among patients with nonspecific backache. A limitation to generalizing from the patterns in this study is that administrative claims data do not capture over-the-counter medication use and self-care. Thus, medication and home exercise regimens are likely to be underestimated, particularly in the early treatment period when they are apt to be most recommended. This study refers to all imaging, diagnostic tests, noninvasive treatments, and surgery relevant to LBP treatment based on practice guidelines as LBP related. However, patients could have had any of these treatments for other underlying conditions. Moreover, the treatment patterns here may only be representative among the privately insured. Another limitation is that the 6-month washout period before the index LBP diagnosis may be insufficient to identify newly diagnosed patients with LBP. Thus, the study sample may include patients newly diagnosed with LBP and patients with a recurrent episode of LBP. Patients could have also experienced LBP for a while before seeking medical care and receiving an LBP diagnosis. The study was also limited in its ability to distinguish a true new LBP episode from exacerbations of CLBP. It would be informative to understand
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622–632
if patients with exacerbations of long-standing CLBP accrue costs and resource utilization similar to those with new LBP. Finally, like all studies based on administrative databases, this study was limited in its ability to ascertain clinical confounders at the level of precision achievable from well-designed studies that directly obtain data from subjects and clinical records. Claims data rely on the accuracy of diagnosis and procedure coding and have no information about the clinical presentation of the condition (eg, symptom severity). In addition to the LBP diagnosis, symptom persistence and severity are important in determining the appropriateness of specific treatment. Clinically, these findings suggest that implementation of the ACP/APS guidelines may require a substantial behavioral change from clinicians’ usual practice patterns for LBP, if effective care and cost savings are to be realized. Education in the guidelines may need to be joined with systematic monitoring and feedback on the type, timing, and duration of diagnostic and treatment use. Implementation of LBP guidelines may be amenable to using prior authorization processes to limit the level of early use of certain diagnostic and treatment practices. These findings suggest that imaging is far too routine, given that most patients have the LBP diagnoses classified as nonspecific backache or less severe LBP. Opioids are used early and frequently although they are a second-line medication associated with serious harms related to their abuse potential. Finally, although the rates of surgery are a seemingly low 2.05% and 88.9% of the time in the presence of appropriately severe diagnosis, patients with LBP who underwent surgery incurred over four times the direct costs compared with those who did not undergo surgery. Because many patients appeared to undergo surgery before less invasive and less risky interventions had been given an adequate trial, this could be an important target of guideline implementation improvement efforts.
Conclusions In conclusion, this study suggests that contrary to clinical guidelines indicating that typically LBP will resolve within a few weeks, patients with LBP and nonspecific LBP, in particular, often have imaging procedures, chiropractic treatment, medication use, and surgery soon after the index LBP diagnosis. Opioids are the most frequently prescribed LBP-related medication. In at least half of the patients who undergo surgery, the procedure is performed within 3 months from the index diagnosis. The study also confirms that LBP is associated with a significant early cost burden. These data suggest success in implementing guidelineconcordant care may entail substantial practice change from current norms and that more research is needed to understand what factors drive early use of second-line medications and surgery.
631
Acknowledgments The authors thank Roger Chou, MD, Oregon Evidencebased Practice Center, Portland, OR, USA, for his comments and suggestions on the article.
References [1] Deyo RA, Mirza SK, Martin NI. Back pain prevalence and visit rates: estimates from U.S. national surveys, 2002. Spine 2006;31: 2724–7. [2] Dagenais S, Caro J, Haldeman S. A systematic review of low back pain cost of illness studies in the United States and internationally. Spine J 2008;8:8–20. [3] Druss BG, Marcus SC, Olfson M, Pincus HA. The most expensive medical conditions in America. Health Aff (Millwood) 2002;21:105–11. [4] Luo X, Pietrobon R, Sun SX, et al. Estimates and patterns of direct health care expenditures among individuals with back pain in the United States. Spine 2004;29:79–86. [5] Ricci JA, Stewart WF, Chee E, et al. Back pain exacerbations and lost productive time costs in United States workers. Spine 2006;31:3052–60. [6] Rizzo JA, Abbott TA III, Berger ML. The labor productivity effects of chronic backache in the United States. Med Care 1998;36:1471–88. [7] Martin BI, Deyo RA, Mirza SK, et al. Expenditures and health status among adults with back and neck problems. JAMA 2008;299:656–64. [8] Cherkin DC, Deyo RA, Loeser JD, et al. An international comparison of back surgery rates. Spine 1994;19:1201–6. [9] Carey TS, Garrett J, Jackman A, et al. The outcomes and costs of care for acute low back pain among patients seen by primary care practitioners, chiropractors, and orthopedic surgeons. The North Carolina Back Pain Project. N Engl J Med 1995;333:913–7. [10] Chou R, Qaseem A, Snow V, et al. Diagnosis and treatment of low back pain: a joint clinical practice guideline from the American College of Physicians and the American Pain Society. Ann Intern Med 2007;147:478–91. [11] Chou R, Fu R, Carrino JA, Deyo RA. Imaging strategies for low-back pain: systematic review and meta-analysis. Lancet 2009;373:463–72. [12] Pengel LH, Herbert RD, Maher CG, Refshauge KM. Acute low back pain: systematic review of its prognosis. BMJ 2003;327:323. [13] National Committee for Quality Assurance (NCQA). Use of imaging studies for low back pain. In: HEDIS Technical Update, vol 2. Washington (DC): National Committee for Quality Assurance, 2007:148–50. [14] Cherkin DC, Deyo RA, Volinn E, Loeser JD. Use of the International Classification of Diseases (ICD-9-CM) to identify hospitalizations for mechanical low back problems in administrative databases. Spine 1992;17:817–25. [15] Webster BS, Verma SK, Gatchel RJ. Relationship between early opiod prescribing for acute occupational low back pain and disability duration, medical costs, subsequent surgery, and late opiod use. Spine 2007;32:2127–32. [16] Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–83. [17] Romano PS, Roos LL, Jollis J. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data. J Clin Epidemiol 1993;46:1075–9. [18] Von Korff M, Wagner EH, Saunders K. A chronic disease score from automated pharmacy data. J Clin Epidemiol 1992;45:197–203. [19] Rhee Y, Taitel MS, Walker DR, Lau DT. Narcotic drug use among patients with lower back pain in employer health plans: a retrospective analysis of risk factors and health care services. Clin Ther 2007;29(Suppl):2603–12. [20] Chou R, Huffman LH. Nonpharmacologic therapies for acute and chronic low back pain: a review of the evidence for an American Pain
632
[21]
[22]
[23]
[24]
J.I. Ivanova et al. / The Spine Journal 11 (2011) 622–632 Society/American College of Physicians clinical practice guideline. Ann Intern Med 2007;147:492–504. Chou R, Huffman LH. Medications for acute and chronic low back pain: a review of the evidence for an American Pain Society/ American College of Physicians clinical practice guideline. Ann Intern Med 2007;147:505–14. Bureau of Labor Statistics. Consumer price index for medical care. Available at: http://www.bls.gov/cpi/#data. Accessed January 19, 2011. Chou R, Loeser JD, Owens DK, et al, American Pain Society Low Back Pain Guideline Panel. Interventional therapies, surgery, and interdisciplinary rehabilitation for low back pain: an evidence-based clinical practice guideline from the American Pain Society. Spine 2009;34:1066–77. Bishop PB, Wing PC. Knowledge transfer in family physicians managing patients with acute low back pain: a prospective randomized control trial. Spine J 2006;6:282–8.
[25] Williams CM, Maher CG, Hancock MJ, et al. Low back pain and best practice care: a survey of general practice physicians. Arch Intern Med 2010;170:271–7. [26] Hermoni D, Borkan JM, Pasternak S, et al. Doctor-patient concordance and patient initiative during episodes of low back pain. Br J Gen Pract 2000;50:809–10. [27] McGuirk B, King W, Govind J, et al. Safety, efficacy, and cost effectiveness of evidence-based guidelines for the management of acute low back pain in primary care. Spine 2001;26:2615–22. [28] Vogt MT, Kwoh CK, Cope DK, et al. Analgesic usage for low back pain: impact on health care costs and service use. Spine 2005;30:1075–81. [29] Ritzwoller DP, Crounse L, Shetterly S, Rublee D. The association of comorbidities, utilization and costs for patients identified with low back pain. BMC Musculoskelet Disord 2006;7:72. [30] van Tulder MW, Assendelft WJ, Koes BW, Bouter LM. Spinal radiographic findings and nonspecific low back pain. A systematic review of observational studies. Spine 1997;22:427–34.