Impact of health interventions on educational outcomes

Page 1

Scottish Collaboration for Public Health Research and Policy Final report Form

SCPHRP reference number:

Project title: Impact of health interventions on educational outcomes: an exemplar study of the management of breech infants. Start date: 1 January 2012

Finish date: 31 December 2012

Investigators: Jill Pell

Lucy Reynolds

Rachael Wood

Carole Morris

Albert King

Anthea Springbett

Danny Mackay 1. Summary We linked three databases from the education sector, held by ScotXed, with one (maternity) database from the health sector (held by ISD) to provide Scotland-wide, individual-level data on pregnancy, school performance, examination results and destination after school. Two other databases (the CHI database and birth registrations) also needed to be used to facilitate this linkage. The linkage provided information on children who had attended Scottish schools from 2006 to 2011. Linkage to health records was achieved for over 90%/ Of the 803,275 children on whom we had data from both school and pregnancy, 674,705 had been singleton, live-born infants. Of these 2,130 (0.3%) had been breech infants delivered vaginally, 13,054 (1.9%) had been breech infants delivered by elective caesarean section and 461,571 (68.4%) had been cephalic infants delivered vaginally. The percentage of breech infants who were delivered vaginally fell from 23% among children who started school in 2006 to 7% among children who started in school in 2011. Of the children born by breech vaginal delivery, 3.0% had had a low APGAR scores (≤3) compared with only 0.4% of those born by either cephalic vaginal delivery or elective caesarean section for breech presentation (p<0.001). The corresponding figures for special educational need were 6.5% compared with 2.8% and 2.6% respectively (p<0.001), and for examination passes above standard grade were 41.7% compared with 48.6% and 43.0% respectively (p<0.001). The associations remained significant following adjustment for potential confounders. Overall, 0.4% of SEN could be attributed to breech vaginal delivery but this percentage fell from 0.9% among children who started school in 2006 to 0.3% among those who started in 2011. Our study confirms that it is feasible to link Scotland-wide, individuallevel data across sectors to undertake public health research.


2. Original aim To conduct the first Scotland-wide linkage of childhood health and education data and demonstrate its utility by studying the impact of guidelines changing the mode of delivery of breech infants on their educational outcomes. Specific objectives: 1. To test the governance processes, including access and approvals, pertaining to cross-sector linkages 2. To test the technical feasibility, completeness and accuracy of data linkage across two sectors 3. To demonstrate the utility of cross-sector linkage in measuring the impact of a health intervention on educational outcomes.

3. Methodology Governance The data flows required for this study are shown in Appendix 1. NHS West of Scotland Research Ethics Service confirmed that NHS ethical approval was not required for this study as no identifiable data were transferred to the University of Glasgow. The following permissions/approvals were secured before any data were shared for linkage/analysis: -

-

-

A data sharing agreement between the Scottish Government, Education Analytical Services division, ScotXed unit (ScotXed) and the University of Glasgow (the University) covering the release of pupil identifiers from ScotXed for linkage purposes and the release of data on pupil support, attainment, etc for analytical purposes A data sharing agreement between the Scottish Qualifications Agency (SQA) and the University of Glasgow covering the release of pupil identifiers from the SQA for linkage purposes A data processor agreement between the University of Glasgow and NHS National Services Scotland Information Services Division (ISD) to enable ISD to receive the pupil identifier data and conduct the linkage on behalf of the University Approval from the Community Health Index Advisory Group (CHIAG) to allow ISD to transfer pupil names derived from the CHI database back to ScotXed to check the quality of the linkage Approval from the Privacy Advisory Committee to allow ISD to undertake the linkage and release the relevant health data to the University for analysis

Securing these permissions took from December 2011 to June 2012 inclusive.


Throughout the duration of the study, the core research team met monthly to monitor progress against original objectives and timelines. Datasets The datasets used in the study are described in Table 1. Table 1: Datasets used in the study Dataset name

Description

Information obtained

Datasets used in linkage ScotXed Pupil census 20062011 inclusive SQA Qualifications database ISD CHI database NRS Birth registrations

Annual census of all pupils in LA maintained and grant aided primary and secondary schools All pupils entered for an SQA examination All patients registered with a Scottish GP Statutory birth registration records

Pupil identifiers for linkage: Scottish Candidate Number (SCN), date of birth, postcode, gender Pupil identifiers for linkage: SCN, forename, surname, date of birth, postcode, gender Used to identify pupils’ CHI numbers Used to link pupils’ CHI numbers to their mothers’ CHI numbers

Datasets used to identify breech babies and their educational outcomes ScotXed Pupil census 20062011 inclusive

Annual census of all pupils in LA maintained and grant aided primary and secondary schools

Educational outcomes: record of additional educational need, need type Confounders for inclusion in analysis: level of English, looked after status, deprivation status, etc

ScotXed Pupil attainment data 2006-2011 inclusive ScotXed School leavers destination survey 2007-2011 inclusive

Pupil attainment in all Scottish qualifications supplied to ScotXed by the SQA

Educational outcomes: attainment at each SCQF level

Annual survey of all children leaving school

Educational outcomes: destination after leaving school

NHS Scotland record of obstetric care

Identifying breech babies and their mode of delivery: presentation, mode of delivery Short term health outcomes: Apgar score Confounders for inclusion in analysis: maternal age, height, smoking, medical conditions, deprivation status, parity, infant gestation, birth weight, sex, admission to neonatal unit, etc

ISD SMR02


Data linkage This study essentially involved linkage of educational outcome data on all children in school from 2006 onwards to their mothers’ delivery records. The delivery records allow identification of children’s presentation at delivery (breech vs cephalic) and their mode of delivery (vaginal vs Caesarean section). Linkage of these two data sources therefore allows the risk of poor educational outcomes (need for additional educational support or poor qualification attainment) to be compared for different groups of babies (breech babies born vaginally; breech babies born by elective Caesarean section; cephalic babies born vaginally). The delivery and education records also provide information on a range of other factors that may be associated with poor educational outcomes, such as prematurity, low birth weight, deprivation, etc, hence these factors can be taken into account in the analysis. An annual census of all children in Scottish local authority maintained or grant aided primary or secondary schools is conducted at the start of every school year. Data are returned by schools to local authorities then to ScotXed. To enable linkage between education and health data, ScotXed transferred a restricted set of identifiers on all children included in the pupil censuses conducted in 2006 (for children in school in academic year 2006/07) to 2011 inclusive to ISD. The available identifiers were: - Scottish Candidate Number (SCN – the unique pupil identifier used by the education sector) - Pupil date of birth - Pupil gender - Pupil home postcode ScotXed separately hold pupil names but they are not permitted to release these for statistical/research purposes. First, ISD examined the quality of the pupil identifier data. ISD then matched the available identifiers against the Community Health Index (CHI) database to identify children’s CHI numbers. The CHI database includes the up to date details of all patients registered with a Scottish GP. The CHI number is the unique patient identifier used by the health sector: as well as being recorded on the CHI database it is included on all the national health datasets held by ISD hence can be used to identify individuals’ health records. ISD’s usual procedures for identifying CHI numbers require individuals’ names as well as date of birth, gender, and postcode. Modification of existing algorithms, and attention to the quality of the links made (i.e. was the correct CHI linked to a child’s SCN), was therefore required. Full technical details of the linkage process are available on request. Matching against the CHI database is a probabilistic rather than an all or nothing process. The linkage algorithms essentially identify the ‘best match’ CHI for each child and indicate how likely that is to be the correct CHI by returning an overall linkage score for the match. In this case, the linkage score depended on how closely matching each of the


three available identifiers (DOB, gender, postcode) were. The linkage algorithms also return the difference between this best match score and that of the next best matching CHI (parameter termed ‘delta score’) to give an indication of how closely rivalled any putative match is. Initially, the identifiers available in each year’s pupil census were matched as separate, stand alone, files (‘method A’). The returned CHIs were categorised into discrete ‘match categories’ depending on linkage and delta scores, and the categories that were deemed acceptable matches were determined. The pupil census data were then merged to create one record per pupil that included all identifier data from the 2006 to 2011 censuses combined. In this file, each pupil/SCN could have between one and six sets of identifiers. This combined file was then used to match pupils/SCNs included in each of the pupil censuses to the CHI database (‘method B’). This approach overcame the problem of occasional error in recording of pupil identifiers. It also ensured that a pupil’s most recently recorded postcode was available for matching against the current and previous postcodes held in the CHI database. As it is unusual to conduct linkage studies without names being available, two exercises were undertaken to assess the quality of the linkages made and the impact of not having names. Firstly, ISD extracted the names associated with the best match CHIs from the CHI database and returned these along with the children’s SCNs to ScotXed. ScotXed then compared these names against the names held for those SCNs in their records using previously developed standardisation and matching algorithms. The proportion of names that matched (and hence the presumed proportion of correct CHIs that had been identified for the pupils/SCNs) was noted. This checking was done for pupils included in the 2006 census when matching was done using method A and method B. Secondly, we used full pupil identifier data from the SQA to assess the difference that having pupil names available would make to the linkage. The SQA holds a database of all children who have been entered for a Scottish qualification such as Standard Grade or Higher exam (hence younger children will not be included). The SQA were able to transfer full pupil identifiers on the children with an active record in their database (i.e. SCN, forename, surname, DOB, gender, home postcode) to ISD for linkage. The subset of pupils in the SQA dataset that were also in the 2006 and 2011 pupil censuses (i.e. same SCN and DOB) were matched to the CHI database using ISD’s usual matching algorithms. The completeness and accuracy of the matches achieved using the full set of identifiers was then compared to that achieved using the restricted identifiers available from ScotXed/pupil census. Once the SCN-CHI links were finalised for all pupils in school 2006/07-2011/12 inclusive, ISD generated a SCN-CHI-unique study identifier (ID) key. The SCN and ID were returned to ScotXed. ScotXed then extracted the relevant education variables and sent these along with just the ID (and no other identifiers) to the University.


The SMR02 is the mother’s delivery record hence it contains the mother’s rather than the child’s CHI. ISD receives a record of all statutory birth registrations from National Records of Scotland and uses the identifiers available to append both the child’s and the mother’s CHI numbers onto these records. Using this ‘look up’ facility, the children’s CHI numbers were linked to their mothers’ CHI numbers. Maternal CHI and date of birth/delivery were then used to identify the relevant SMR02 records. ISD extracted the required health variables from the SMR02 records and sent these along with the ID to the University. The University then merged the health and education variables using the IDs provided to form the final analysis files. Statistical analyses The analyses were restricted to singleton children. Multiple pregnancies (twin, triplet etc) were excluded because it is impossible to determine birth order in the linked dataset and therefore attribute the correct birthweight, APGAR score etc to the correct sibling. The analyses also excluded children who had had emergency deliveries. Breech infants delivered vaginally were compared with breech infants delivered by elective caesarean section and cephalic infants delivered vaginally. Continuous variables were summarised by the median and inter-quartile range. Univariate comparisons were performed using the Kruskal-Wallis test, Chi square test and Cuzick’s test for trend for continuous, categorical and ordinal data respectively. Gestational age at birth was defined was completed weeks of gestation on the basis of the estimated date of delivery recorded in each woman’s clinical record. Gestational age has been confirmed by ultrasound in the first half of pregnancy in more than 95% of women in the United Kingdom since the early 1990’s. Previous miscarriage was defined as previous delivery of a conceptus, showing no signs of life before 24 weeks gestation, excluding therapeutic abortions. Previous therapeutic abortion was defined as previous therapeutic termination of pregnancy, by any means, prior to 24 weeks gestation. The outcomes examined were: - five minute APGAR score - school record of special educational need - level of examination attainment The associations were analysed using univariate and multivariable binary and ordinal logistic regression analyses. The covariates included in the multivariable analyses were infant sex, maternal age and height, marital status, area deprivation index, parity, birth weight centile, previous spontaneous and therapeutic deliveries. The p values for all hypothesis tests were two-sided with statistical significance assumed at p<0.05. The odds ratios and prevalence were used to derive population attributable fractions associated with breech vaginal delivery overall and by year. All statistical analyses were undertaken using STATA v10.1 (Stata Corporation,TX, USA).


4. Results Governance We were able to use existing governance processes to undertake cross-sectoral linkage of health and education data for research purposes, although securing all the necessary permissions was extremely time consuming and delayed extraction of data by six months Data linkage The number of records included in each pupil census, and the quality of the identifier data, is shown in Table 2. The number of children included in the pupil census has fallen slightly year on year. The quality of the pupil identifiers included in the censuses appears good, with only around 0.02% of records containing a clearly invalid date of birth, gender, or postcode. The results of matching the pupil census identifiers to the CHI database using method A and method B are shown Table 3, Table 4 and Figure 1. Table 5. provides a description of each of the match categories. Method B (using the identifiers available from all censuses to link pupils included in any one year’s census to the CHI) increased the proportion of children for whom an acceptable match category CHI was found, particularly in the earlier years’ censuses (i.e. close to 2006). Using method B, an acceptable CHI was found for 93% of all children in school between 2006/07 and 2011/12. Due to the higher linkage completeness obtained, method B was used when preparing the final analysis files. The results of checking the names from the best match CHIs against the names held by ScotXed are shown in Table 6. Method A correctly identified 93.9% of subjects whilst method B correctly identified 97.5%. Both methods correctly identified more than 99% of subjects in the acceptable match categories. The results of matching the subset of children included in the SQA database that were also included in the 2006 and 2011 pupil censuses are shown in Table 7. It can be seen that when full identifiers including children’s names are available, an acceptable CHI is found for almost all (≥99.9%) children. The lack of names therefore does result in a noticeable drop in the completeness of linkage achieved. The drop is most marked for the earlier census years and when linkage method A is used.


Table 2. Number of records in each pupil census and quality of pupil identifier data.

Total number of records Invalid DOB, gender, and/or postcode Duplicate DOB, gender, and postcode

2006

2007

2008

2009

2010

2011

703,500

693,181

682,603

677,813

673,975

671,264

118

124

128

130

106

81

(0.02%)

(0.02%)

(0.02%)

(0.02%)

(0.02%)

(0.01%)

6,436

6,429

6,394

6,343

6,408

6,376

(0.91%)

(0.93%)

(0.94%)

(0.94%)

(0.95%)

(0.95%)


Table 3. Children included in the pupil censuses 2006 to 2011: number and percentage linked to a CHI number, by match category Method A: Pupil identifiers contained in each year’s census treated as a separate files and linked independently to the CHI database Match category

2006

2007

2008

2009

2010

2011

2006-2011*

N

%

N

%

N

%

N

%

N

%

N

%

N

%

A

70,406

10.0

73,079

10.5

75,553

11.1

78,588

11.6

81,348

12.1

83,248

12.4

462,222

11.3

B

445,526

63.3

448,455

64.7

451,260

66.1

457,823

67.5

462,720

68.7

464,817

69.2

2,730,601

66.6

C

61,613

8.8

60,204

8.7

58,609

8.6

57,744

8.5

56,901

8.4

56,032

8.3

351,103

8.6

D

8,139

1.2

7,873

1.1

7,539

1.1

7,455

1.1

7,294

1.1

7,062

1.1

45,362

1.1

E

12,231

1.7

12,305

1.8

12,566

1.8

12,762

1.9

12,960

1.9

12,956

1.9

75,780

1.8

F

2,605

0.4

2,448

0.4

2,270

0.3

1,993

0.3

1,759

0.3

1,575

0.2

12,650

0.3

G

18,826

2.7

16,981

2.4

15,193

2.2

13,089

1.9

11,459

1.7

10,662

1.6

86,210

2.1

H

589

0.1

546

0.1

487

0.1

414

0.1

365

0.1

323

0.0

2,724

0.1

I

339

0.0

323

0.0

313

0.0

312

0.0

305

0.0

323

0.0

1,915

0.0

J

1,723

0.2

1,678

0.2

1,568

0.2

1,495

0.2

1,476

0.2

1,472

0.2

9,412

0.2

K

33

0.0

30

0.0

39

0.0

34

0.0

30

0.0

29

0.0

195

0.0

L

4,552

0.6

3,915

0.6

3,328

0.5

2,792

0.4

2,239

0.3

1,909

0.3

18,735

0.5

M

76,918

10.9

65,344

9.4

53,878

7.9

43,312

6.4

35,119

5.2

30,856

4.6

305,427

7.4

Total in acceptable match categories

607,115

86.3

609,040

87.9

610,424

89.4

616,692

91.0

621,481

92.2

623,396

92.9

3,688,148

89.9

Total

703,500

100.0

693,181

100.0

682,603 100.0 677,813 100.0 673,975 100.0 671,264 100.0 4,102,336 100.0 * For method A, figures for 2006-2011 combined are the sum of figures for each individual census Acceptable match categories are A, B, C, D, F, and G


Table 4.Children included in the pupil censuses 2006 to 2011: number and percentage linked to a CHI number, by match category Method B: Pupil identifiers from all censuses (2006 to 2011 inclusive) combined before linkage to the CHI database Match category

2006

2007

2008

2009

2010

2011

2006-2011*

N

%

N

%

N

%

N

%

N

%

N

%

N

%

A

79,275

11.3

81,677

11.8

83,358

12.2

85,240

12.6

86,635

12.9

87,283

13.0

119,104

11.8

B

493,689

70.2

493,452

71.2

490,600

71.9

489,463

72.2

486,644

72.2

482,465

71.9

708,300

70.0

C

62,825

8.9

60,328

8.7

57,718

8.5

55,826

8.2

54,337

8.1

53,037

7.9

85,897

8.5

D

7,735

1.1

7,288

1.1

6,829

1.0

6,602

1.0

6,342

0.9

6,155

0.9

10,460

1.0

E

12,693

1.8

12,742

1.8

12,903

1.9

12,932

1.9

13,010

1.9

12,971

1.9

18,298

1.8

F

1,404

0.2

1,295

0.2

1,221

0.2

1,157

0.2

1,134

0.2

1,168

0.2

2,067

0.2

G

10,501

1.5

9,250

1.3

8,405

1.2

7,797

1.2

7,628

1.1

7,903

1.2

15,257

1.5

H

309

0.0

273

0.0

247

0.0

240

0.0

231

0.0

239

0.0

450

0.0

I

194

0.0

204

0.0

216

0.0

237

0.0

275

0.0

332

0.0

485

0.0

J

946

0.1

958

0.1

975

0.1

1,059

0.2

1,196

0.2

1,428

0.2

2,191

0.2

K

15

0.0

17

0.0

20

0.0

20

0.0

24

0.0

26

0.0

39

0.0

L

1,958

0.3

1,592

0.2

1,304

0.2

1,183

0.2

1,158

0.2

1,242

0.2

2,788

0.3

M

31,956

4.5

24,105

3.5

18,807

2.8

16,057

2.4

15,361

2.3

17,015

2.5

46,249

4.6

Total in acceptable match categories

655,429

93.2

653,290

94.2

648,131

94.9

646,085

95.3

642,720

95.4

638,011

95.0

941,085

93.0

Total

703,500

100.0

693,181

100.0

682,603 100.0 677,813 100.0 673,975 100.0 671,264 100.0 1,011,585 100.0 * For method B, figures for 2006-2011 combined are based on the total number of individual children (i.e. unique study IDs) in school at any point over the six years Acceptable match categories are A, B, C, D, F, and G


Table 5. Description of match categories Best match CHI: match between pupil census identifiers and CHI database DOB

Gender

Postcode

Best match CHI compared to next † best match CHI

A

Exact

Exact

Exact

Unrivalled

B

Exact

Exact

Exact

Distant rival

C

Exact

Exact

Exact

Intermediate rival

D

Exact

Exact

Exact

Close rival

E

Exact

Exact

Exact

Tied

F

Exact

Exact

Close

Unrivalled

G

Exact

Exact

Close

Rivalled

H

Exact

Exact

Close

Tied

I

Close

Exact

Exact

Unrivalled

J

Close

Exact

Exact

Rivalled

K

Close

Exact

Exact

Tied

Unrivalled

Match category

L M

Other combination of close matches/high overall linkage score

Acceptable match category

Other – considered a non-match  † Unrivalled means the next best CHI had a much lower linkage score than the best match CHI. Tied means that the next best CHI had the same linkage score as the best match CHI Close match on postcode indicates 6 out of 7 characters agreed Close match on DOB indicates 2 out of 3 (of DD, MM, YY) agreed


Figure 1: Percentage of children in each pupil census 2006-2011 linked to an acceptable match category CHI, by method

% of children in pupil census linked to an acceptable match category CHI

100.0%

90.0%

80.0% Method B Method A 70.0%

60.0%

50.0% 2006

2007

2008

2009

2010

2011

Pupil census year

Method A: Pupil identifiers contained in each year’s census treated as a separate file and linked independently to the CHI database Method B: Pupil identifiers from all censuses (2006 to 2011 inclusive) combined before linkage to the CHI database


Table 6. Number and percentage of all children included in the 2006 pupil censuses that had the correct CHI identified, by match category and linkage method

Match category

A B C D E F G H I J K L M Total in acceptable match categories Total

Linkage method A

Linkage method B

2006

2006

N

%

63,762 397,079 53,937 68,90 10,661 2,188 15,015 383 184 914 12 3,190 34,703

99.4 99.5 99.5 96.7 95.3 93.6 90.4 73.8 61.5 59.4 42.9 79.8 52.3

588,918 538,871

93.9 99.2

N 72,595 444,917 55,167 6,546 11,155 1,064 7,527 175 99 439 5 1,077 11,165

% 99.5 99.5 99.5 97.4 95.8 93.6 90.6 72.6 63.9 57.5 50.0 76.1 51.9

611,931 97.5 587,816 99.4 The linked census data contained 703,500 records. Of these 76,095 SCNs were not available for validation for operational reasons. The total number of records available for validation was 627,405. The best match CHI is assumed to be correct if the name extracted from the CHI database matched that held by ScotXed for that pupil/SCN

Method A: Pupil identifiers contained in each year’s census treated as a separate file and linked independently to the CHI database Method B: Pupil identifiers from all censuses (2006 to 2011 inclusive) combined before linkage to the CHI database


Table 7. Subset of children included in the 2006 and 2011 pupil censuses that were also included in the SQA database. Results of linkage to CHI using full SQA identifiers and restricted (no names) pupil census identifiers Children in 2006 pupil census and SQA database

Children in 2011 pupil census and SQA database

(N=528,469)

(N=242,406)

Linkage method

Full pupil identifiers from SQA and standard ISD linkage algorithms Restricted pupil identifiers from pupil census: linkage method A Restricted pupil identifiers from pupil census: linkage method B

N with acceptable link to CHI

% with acceptable link to CHI

N with acceptable link to CHI

% with acceptable link to CHI

527,933

99.90

242,287

99.95

460,065

87.1

227,276

93.8

490,239

92.8

233,017

96.1

Acceptable link to CHI when using full identifiers defined as linkage score above ISD’s standard thresholds Acceptable link to CHI when using restricted identifiers defined as match categories A, B, C, D, F, and G Method A: Pupil identifiers contained in each year’s census treated as a separate file and linked independently to the CHI database Method B: Pupil identifiers from all censuses (2006 to 2011 inclusive) combined before linkage to the CHI database


Figure 2: Children included in the analysis 2006 - 703,500 2007 - 693,181 2008 - 682,603 2009 - 677,813 2010 - 673,975 2011 - 671,264 ↓ 1,011,585 ↓ 941,085 (93.0%) ↓ 838,498 (82.9%) ↓ 833,883 (82.4%) ↓ 811,860 (80.3%) ↓ 803,275 ↓ 674,705

Total number of records in the pupil censuses (includes >1 record for some pupils)

Total number of individual children (ie unique study ID) in any pupil census 2006-2011 Number for whom an acceptable CHI number could be found (linkage method B) Number with birth registration record available (NB some children born outwith Scotland or , rarely, child CHI not seeded onto birth registration record) Number with maternal CHI available (NB maternal CHI not seeded onto some birth registration records) Number with SMR02 record identified (NB some children not born in hospital, incomplete SMR02 returns from some units, date of delivery incorrectly recorded on SMR02) Number with key variables available (on SMR2, pupil census and SQA) Singleton, live-born infants (ie eligible for statistical analyses)

Statistical analyses Of the 803,275 children on whom we had data from both school and pregnancy, 674,705 were singleton, live-born infants. Of these 2,130 (0.3%) had been breech infants delivered vaginally, 13,054 (1.9%) had been breech infants delivered by elective caesarean section and 461,571 (68.4%) had been cephalic infants delivered vaginally. There were significant differences in maternal and obstetric characteristics according to presentation and mode of delivery (Table 1). Infants who presented breech were delivered at an earlier gestation and, therefore, had lighter bodyweights. With regard to crude outcomes, breech infants delivered vaginally had significantly lower APGAR scores than both cephalic infants delivered vaginally and breech infants delivered by caesarean section, and were significantly more likely to have a record of special educational need (Table 1). They were also less likely to acquire examination passes at higher or advanced higher level than breech infants delivered by caesarean section (Table 8).

15


Table 8. Comparison of case mix and crude outcomes by mode of delivery

Cephalic vaginal N=461,571 N (%)

Breech vaginal N=2,130 N (%)

Breech caesarean N=13,054 N (%)

P value

Infant sex

Female Male

230,175 (49.9) 231,396 (50.1)

1,163 (54.6) 967 (45.4)

7,230 (55.4) 5,824 (44.6)

<0.001

Married

No Yes Missing

263,776 (59.0) 183,372 (41.0) 14,423

1,294 (62.2) 788 (37.8) 48

8,082 (63.8) 4,581 (36.2) 391

<0.001

SIMD5

1 (deprived) 2 3 4 5 (affluent) Missing

131,009 (28.5) 97,204 (21.1) 84,551 (18.4) 77,433 (16.8) 69,807 (15.2) 1,567

666 (31.4) 434 (20.5) 362 (17.1) 330 (15.6) 329 (15.5) 9

3,264 (25.1) 2,660 (20.5) 2,427 (18.7) 2,486 (19.1) 2,170 (16.7) 47

<0.001

Parity

Nulliparous Multiparous Missing

288,778 (62.6) 171,613 (37.2) 1,180

1,429 (67.1) 690 (32.4) 11

5,932 (45.4) 7,075 (54.2) 47

<0.001

Gestation (weeks)

24-27 28-32 33-36 37-39 40 41 42 43

293 (0.06) 1,813 (0.39) 16,662 (3.6) 165,079 (35.8) 157,061 (34.0) 102,182 (22.1) 18,020 (3.9) 461 (0.10)

93 (4.4) 135 (6.3) 328 (15.4) 885 (41.6) 464 (21.8) 193 (9.1) 30 (1.4) 2 (0.09)

8 (0.06) 83 (0.64) 474 (3.6) 10,914 (83.6) 1149 (8.8) 366 (2.8) 57 (0.44) 3 (0.02)

<0.001

Previous 0 spontaneous 1 abortions 2+ missing

370,177 (80.2) 71,255 (15.4) 20,086 (4.4) 53

1,706 (80.1) 324 (15.2) 100 (4.7) 0

10,234 (78.4) 2,091 (16.0) 726 (5.6) 3

<0.001

Previous therapeutic abortions

0 1 2+ missing

410,922 (89.0) 43,374 (9.4) 7,225 (1.6) 50

1,872 (87.9) 224 (10.5) 34 (1.6) 0

11,629 (89.1) 1,215 (9.3) 1,215 (9.3) 3

0.508

APGAR Score

0-3 4-7 8-10 missing

2,005 (0.4) 7,525 (1.6) 448,235 (97.9) 3,806

63 (3.0) 181 (8.6) 1,851 (88.4) 35

54 (0.4) 135 (1.0) 12,734 (98.5) 131

<0.001

SEN

Yes

10,495 (2.8)

109 (6.5)

280 (2.6)

16


Highest exam level

Maternal age Maternal height Birthweight

No missing

368,216 (97.2) 82,860

1,573 (93.5) 448

10,518 (97.4) 2,256

Access 1-3 Standard Higher Adv higher missing

8,808 (3.9) 121,081 (53.0) 73,556 (32.2) 25,062 (11.0) 233,064

61 (4.6) 716 (53.7) 410 (30.7) 147 (11.0) 796

196 (3.2) 2,927 (48.2) 2,214 (36.4) 741 (12.2) 6,976

(years)

Median (IQR) 28 (24-32)

Median (IQR) 28 (24-32)

Median (IQR) 29 (26-33)

<0.001

(cm)

163 (158-167)

163 (158-167)

162 (157-167)

<0.001

(g)

3,420 (3,090-3,740)

2,980 (2,500-3,340)

3,240 (2,940-3,565)

<0.001

<0.001 <0.001

N number; SIMD Scottish Index of Multiple Deprivation; SEN special educational need; Adv advanced; IQR interquartile range; cm centimetres; g grammes On univariable logistic regression analysis, children who had presented breech and were delivered by elective caesarean section had significantly higher 5-minute APGAR scores than breech children who had been delivered by vaginal delivery; as did children who had presented cephalic (Table 2). Some of the association was explained by confounding but the differences remained statistically significant after adjustment (Table 9). Table 9. Univariable and multivariable ordinal logistic regression of the association between mode of delivery and 5 minute APGAR score

(4-10) referent to (0-3) OR (95% CI)

(8-10) referent to (0-7)

P value

OR (95% CI)

P value

<0.001 <0.001

<0.001 <0.001

Univariable Breech vaginal Breech CS Cephalic vaginal

1.0 7.39 (5.12,10.66) 7.05 (5.46, 9.09)

<0.001 <0.001

1.0 8.88 (7.30, 10.81) 6.20 (5.42, 7.10)

Multivariable* Breech vaginal Breech CS Cephalic vaginal

1.0 3.43 (2.32, 5.07) 2.72 (2.04, 3.61)

<0.001 <0.001

1.0 5.46 (4.41, 6.75) 3.50 (3.00, 4.09)

*adjusted for infant sex, maternal age and height, marital status, area deprivation index, parity, birth weight centile, previous spontaneous and therapeutic deliveries. OR odds ratio; CI confidence interval; CS caesarean section 17


Similarly, children who had presented breech were significantly less likely to have a record of special educational need if they had been delivered by elective caesarean section rather than vaginal delivery (Table 10). This association remained statistically significant after adjusting for potential confounders (Table 10). Table 10. Univariable and multivariable binary logistic regression of the association between mode of delivery and a record of special educational need Univariable

Breech vaginal Breech CS Cephalic vaginal

Multivariable*

OR (95% CI)

P value

OR (95% CI)

P value

1.0 0.38 (0.31,0.48) 0.41 (0.34, 0.50)

<0.001 <0.001

1.0 0.65 (0.51, 0.83) 0.61 (0.49, 0.75)

0.001 <0.001

*adjusted for infant sex, maternal age and height, marital status, area deprivation index, parity, birth weight centile, previous spontaneous and therapeutic deliveries. On univariate analysis, children who had presented breech and were delivered vaginally achieved significantly lower levels of examination passes up to Higher grade that those who presented breech but were delivered by elective caesarean section (Table 11). The association was no longer statistically significant after adjusting for potential confounders (Table 11). Table 11. Univariable and multivariable ordinal logistic regression of the association between mode of delivery and highest examination level attained Highest examination level (Access1,2,3) V (Gen/Cred/Stan Grade) & (Higher) & (Adv Higher) OR (95% CI) P value

(Access 1,2,3) & (Gen/Cred/Stan Grade) V (Higher) & (Adv Higher) OR (95% CI) P value

(Access1,2,3) & (Gen/Cred/StanGrade) & (Higher) V (Adv Higher) OR (95% CI) P value

Univariable Breech vaginal Breech CS Cephalic vaginal

1.0 1.43 (1.07, 1.93) 1.20 (0.92, 1.55)

0.015 0.175

1.0 1.32 (1.17, 1.49) 1.06 (0.95, 1.18)

<0.001 0.302

1.0 1.12 (0.93, 1.32) 0.99 (0.84, 1.18)

0.233 0.952

Multivariable* Breech vaginal Breech CS Cephalic vaginal

1.0 1.04 (0.77, 1.41) 1.05 (0.81, 1.37)

0.789 0.705

1.0 1.06 (0.93, 1.20) 1.03 (0.91, 1.15)

0.412 0.654

1.0 0.87 (0.72, 1.06) 0.97 (0.81, 1.16)

0.162 0.751

*adjusted for infant sex, maternal age and height, marital status, area deprivation index, parity, birth weight centile, previous spontaneous and therapeutic deliveries. Overall, 0.4% of records of special education need could be attributed to vaginal delivery of breech infants (Table 12). The percentage of breech infants who were delivered vaginally fell from 23% among children who started school in 2006 to 7% among children who started in school in 2011. Therefore, the percentage of records of special educational need that could be attributed to vaginal delivery of breech infants fell from 0.9% among children who started school in 2006 to 0.3% among those who started school in 2011.

18


Table 12. Percentage of cases of special educational need attributable to vaginal breech delivery

Univariate

Multivariate

%

95% CI

%

95% CI

2006 2007 2008 2009 2010 2011

1.5 0.9 0.8 0.7 0.8 0.5

1.0-2.0 0.6-1.2 0.6-1.1 0.5-0.9 0.5-0.1 0.3-0.6

0.9 0.5 0.5 0.4 0.4 0.3

0.5-1.4 0.3-0.8 0.2-0.7 0.2-0.6 0.2-0.6 0.1-0.4

Overall

0.6

0.4-0.8

0.4

0.2-0.5

5. Discussion We successfully linked Scotland-wide, individual-level data across sectors and demonstrated the utility of such linked data in undertaking public health research. The determinants of health are wide; they extend beyond health sector factors to wider societal influences such as education, wealth, employment, housing, environment and culture. Similarly, the impact of health extends beyond morbidity and mortality to other outcomes such as education, wealth, employment, housing and environment. Therefore, public health research, whether epidemiological or intervention studies, requires access to cross-sectoral data. The Scottish Collaboration for Public Health Research and Policy emphasised the importance of multisector linkage studies as a powerful and efficient way of conducting high impact, policy relevant research in its March 2011 workshop. It also specifically noted the lack of multi-sector linkage studies focusing on child health in its Early Life Working Group’s environmental scan [1]. Randomised controlled trials have shown that infants who present breech at term have significantly reduced risk of prenatal mortality, neonatal mortality and serious neonatal morbidity if delivered by elective Caesarean section rather than vaginal breech delivery [2]. Guidelines were issued to obstetricians resulting in a dramatic change in practice in favour of delivery of breech infants by elective Caesarean. More recently, observational evidence suggests that the benefits of elective Caesarean delivery may be even greater for breech infants requiring delivery preterm [3]. The studies conducted to date have focused on the short term health outcomes of the intervention in terms of reduced perinatal, neonatal and maternal mortality and morbidity. However, it is known that pregnancy complications can result in longterm neurodevelopmental problems, such as learning and language difficulties, impaired cognitive function, and behavioural problems [4-7]. In our study, as well as corroborating the adverse effect of breech vaginal delivery and short terms outcomes (5 minute APGAR score), we were able to demonstrate longer term adverse effects (special educational need). We were also able to show that the contribution of breech delivery to special educational need has fallen over time with a reduction in the percentage of breech infants delivered vaginally.

19


Historically, it has been impossible to link health and education data at a Scotland-wide level due to unclear governance structures. In a previous study linking health and education data, we were required to approach local authorities individually and obtained permission from 19 out of 32 [4]. The Scottish Government has recently committed to facilitating cross-sectoral linkage of data and clearer governance of access to national level education data for research purposes is now available. 6. Conclusions Our pilot project has demonstrated the feasibility and usefulness of cross-sectoral record linkage and confirms the importance of efforts in this area. There is scope to improve the efficiency of the processes governing research projects based on the linkage of administrative data. Within the health sector, seeking approval from both CHIAG and PAC is time consuming and duplicates effort: consideration should be given to aligning these processes more closely. This study has demonstrated that it is technically feasible to link administrative data from the health and education sectors to levels of completeness and accuracy that are acceptable for at least some research purposes without having children’s names available. This has potentially important implications for the future of cross-sector linkage based research in Scotland. Having names available increases the completeness and accuracy of linkage that can be achieved but increases the risks to data subjects’ privacy. Whether or not full identifiers should be used for future research projects will depend on the precise details of the datasets involved and the research questions being addressed. Consideration should be given to how the findings of this project could inform efficient linkage of data from the health and education sectors for future projects. ISD and ScotXed should work with the newly established National Data Linkage Service to consider relevant options, for example maintenance of complex keys between SCN and CHI that facilitate (re)linkage whilst posing minimal privacy risks.

7. References [1] Geddes R, Haw S, Frank J (2010). Interventions for promoting early child development for health: an environmental scan with special reference to Scotland. Scottish Collaboration for Public Health Research and Policy; Edinburgh. [2] Hoffmeyr GJ, Hannah ME. Planned caesarean section for term breech delivery. Cochrane Database Syst Rev 2003;(3):CD000166. [3] Robiloio PA, Boe NM, Danielsen B, Gilbert WM. Vaginal versus caesarean delivery for preterm breech presentation of singleton infants in California: a population-based study. J Reprod Med 2007;52(6):473-9. [4]. Mackay DF, Smith GCS, Dobbie R, Pell JP (2010). Gestational Age at Delivery and Special Educational Need: Retrospective Cohort Study of 407,503 schoolchildren. PLoS Med 7(6): e1000289. doi:10.1371/journal.pmed.1000289. [5] Kerr-Wilson CO, Mackay DF, Smith GCS, Pell JP. Meta-analysis of the association between preterm delivery and intelligence. J Public Health 2011;Mar 9 [Epub ahead of print] [6] Foulder-Hughes LA, Cooke RW. Motor, cognitive and behavioural disorders in children born preterm. Dev Med Child Neurol 2003;45:97-103. 20


[7] Rickards AL, Kitchen WH, Doyle LW, Ford GW, Kelly EA et al. Cognition, school performance, and behaviour in very low birth weight and normal birth weight children at 8 years of age: a longitudinal study. J Dev Behav Pediatr 1993;14:363-68.

21


Appendix 1. Data flows required for linkage of health and education data to explore the educational outcomes of breech babies

SQA: Scottish Qualifications Authority ScotXed: Scottish Government, Education Analytical Services division, ScotXed unit ISD: NHS National Services Scotland Information Services Division SCN: Scottish candidate number CHI: Community health index number ID: unique pupil identifier derived for this study

22


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.