SHS Vol 3, No 1 (2018)

MAY 2018 | VOLUME 2 | ISSUE 3

Communicable diseases in South Africa: The battle continues Listeriosis in the City of Johannesburg, South Africa HIV and tuberculosis prevention and control in South Africa: An overview

EDITOR IN CHIEF Debashis Basu

MAY 2018 | VOLUME 2 | ISSUE 3 Official journal of the South African Public Health Medicine Association, an affiliate of the South African Medical Association

EDITORIAL BOARD Leegail Adonis Michael Asuzu Chauntelle I Bagwandeen Lilian Dudley Francis Hyera Willem Kruger Shan Naidoo Petrus G Rautenbach PUBLISHED BY THE HEALTH AND MEDICAL PUBLISHING GROUP CEO AND PUBLISHER Hannah Kikaya

EDITORIALS 48

Diseases of public health importance in South Africa D Basu

49

No more business as usual: A case for qualitative approaches in the fight against TB S Grande

ARTICLES 52

HIV and tuberculosis prevention and control in South Africa: An overview W D F Venter

55

Listeriosis in the City of Johannesburg, South Africa Insur­ance pilot districts P Manganye, B Desai, R Bismilla

59

Organising HIV ageing-patient care in South Africa: An implementation science approach D Croce, D Mueller, G Rizzardini, U Restelli

63

PUBLIC HEALTH NOTEBOOK Hypothesis tests for the difference between two population proportions using Stata B V Girdler-Brown, L N Dzikiti

CHIEF OPERATING OFFICER Diane Smith EXECUTIVE EDITOR Bridget Farham PRODUCTION MANAGER Emma Jane Couzens MANAGING EDITORS Claudia Naidu Naadia van der Bergh ONLINE MANAGER Gertrude Fani TECHNICAL EDITOR Kirsten Morreira SENIOR DESIGNER Clinton Griffin HEAD OFFICE Health and Medical Publishing Group (Pty) Ltd Block F, Castle Walk Corporate Park, Nossob Street, Erasmuskloof Ext. 3, Pretoria, 0181 Tel. 012 481 2069 Email: dianes@hmpg.co.za CONTACTS Website and online submissions: www.shsjournal.org.za Publisher website: www.hmpg.co.za Editorial queries: shseditor@hmpg.co.za Tel: +27 (0)21 532 1281 ISSN 2312-9360

Published by the Health and Medical Publishing Group (Pty) Ltd Company registration 2004/022032/07, a subsidiary of the South African Medical Association Suite 11 Lonsdale Building, Gardener Way, Pinelands, 7405 Tel: +27 (0)21 532 1281 | Email: publishing@hmpg.co.za Creative Commons Attribution – Non Commercial Works Licence (CC BY-NC 4.0)

EDITORIAL

Diseases of public health importance in South Africa South Africa (SA) is facing a quadruple burden of disease: the HIV/AIDS epidemic, along with a high burden of tuberculosis (TB); high maternal and child mortality; high levels of violence and injuries; and a growing burden of non-communicable diseases.[1] Our focus in this issue is on communicable diseases, which still remain of major public health concern in developing countries.

Many factors are contributing to the emergence of communicable diseases, including climate change, globalisation and urbanisation, and most of these are to some extent caused by humans.[2] These factors have a significant impact on the epidemiology of these diseases, and on the capacity to effectively prevent, control and treat them with the scarce resources available in developing countries. Three articles in this issue deal with TB and HIV, while a fourth, from Johannesburg, SA, describes the experience of dealing with the recent outbreak of listeriosis.

Statistics SA’s 2016 mortality report has found that TB remains the top cause (6.5%) of natural deaths among SA’s population. HIV came in as the fifth, at 4.8%.[3] Several decades of steady improvement in life expectancy, on the back of improved sanitation and housing, safer food and access to vaccines and improved medical care, were rapidly reversed during the 1980s and 1990s, as HIV and associated TB prevalence increased rapidly. Although both epidemics are complex, and evolving, the huge successes around treatment, and more limited ones in the identification of effective prevention techniques in the HIV field, have fuelled new resolve in the TB research and public health world, long constrained by a lack of research and programme funding, and arguably, by a lack of ambition.[4] In the near future, people living with HIV will become more complicated patients. The progressive ageing of the HIVinfected population means that an increasing number of patients will have one or more comorbidities not directly related to HIV infection, and/or correlated with the side-effects of treatment. This situation will require integrated, specialised ambulatories for the main comorbidities.[5] The clinical evaluation of these patients will

require a new strategy, as risk reduction through a health-system process where detection, treatment and cure are the work of clinical staff and hospitals may not be enough. More detailed analysis of community behaviours based on geography, language and culture might be necessary.[6] Since October 2018, a significantly high number of listeriosis cases have occurred in Gauteng and some other provinces in SA. The case fatality rate in the City of Johannesburg (CoJ), Gauteng Province, was particularly high. Officials from the CoJ share their experiences in the management of this outbreak.[7] The last part of this issue includes the abstracts presented both orally and as posters at the Gauteng Health Research and Innovation Summit held in February 2018. The summit deliberated on a number of public health challenges, and explored possible solutions to these challenges that could be implemented.

Deb Basu Editor South Afr J Pub Health 2018;2(3):48. DOI:10.7196/SHS.2018.v2.i3.72

1. Bradshaw D, Groenewald P, Laubscher R, et al. Initial burden of disease estimates for South Africa, 2000. S Afr Med J 2003;93(9):682-688.

2. Lindahl JF, Grace D. The consequences of human actions on risks for infectious diseases: A review. Infect Ecol Epidemiol 2015;5(1):10. https://doi.org/10.3402%2Fiee.v5.30048 3. Statistics South Africa. Mortality and causes of death in South Africa, 2016: Findings from death notification. http://www.statssa.gov.za/publications/P03093/P030932016.pdf (accessed 30 March 2018).

4. Venter WDF. HIV-TB prevention and control in South Africa: An introduction. S Afr J Pub Health 2018;2(3):52-54. https://doi.org/10.7196/SHS.2018.v2.i3.61 5. Croce, D Mueller, G Rizzardini, U Restelli. Organising HIV ageing-patients care in South Africa: An implementation science approach. S Afr J Pub Health 2018;2(3):59-62. https:// doi.org/10.7196/SHS.2018.v2.i3.67 6. Grande S. No more business as usual: A case for qualitative approaches in the fight against TB. S Afr J Pub Health 2018;2(3):49-51. https://doi.org/10.7196/SHS.2018.v2.i3.70 7. Manganye P, Desai B, Daka M, Bismilla R. Listeriosis in the City of Johannesburg. S Afr J Pub Health 2018;2(3):55-58. https://doi.org/10.7196/SHS.2018.v2.i3.73

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

48

May 2018

This open-access article is distributed under Creative Commons licence CC-BY-NC 4.0.

EDITORIAL

No more business as usual: A case for qualitative approaches in the fight against TB Setting the stage Decades of efforts to mitigate the global burden of tuberculosis (TB) have led to it rematerialising in even more burdensome and troubling multidrug-resistant (MDR) strains. Traditional approaches to TB control across the globe, and in South Africa (SA) specifically, reflect decades of biomedical research focused primarily on clinical surveillance and the reduction of new infections. [1-3] Reduction of disease through a health-system process, where detection, treatment and cure become the work of clinical staff and hospitals, are only optimal under social conditions where screening, seeking care, follow-up and adherence to treatment are consistent with norms in communities. What if community behaviours vary based on geography, language and culture? How would standardised clinical practices address the various differences across groups?

Given the challenges associated with the current surveillance and control mechanisms to arrest TB, not to mention MDR-TB, across SA, the emphasis on clinical-based approaches is neither sufficient nor appropriate for capturing the sociocultural aspects of TB-related illness.[4-6] Health services researchers who have learned from medical anthropologists and sociologists have adopted qualitative community-based approaches to address these very real gaps in sociocultural understanding. The following editorial calls attention (Table 1) to four characteristics of community-based qualitative approaches that may effectively challenge the ‘business as usual’ approaches to TB control, and guide more appropriate and relevant TB intervention design.

Today’s consensus is that despite decades of research, intervention and spending, the global TB epidemic is still a social disease. TB continues to attack vulnerable communities across the globe.

Bacteria thrive in areas where people live in close proximity. They hide in the spaces where people live and spend most of their time – eating and sleeping, visiting with family and chatting with friends. While the health costs of the disease are easy to quantify, the social costs often go unnoticed. There are reams of data about the unremitting impact of TB, MDR-TB, and extensively drug-resistant (XDR) TB in the developing world, as well as in rural, more isolated parts of the globe. Money from USAID as part of the Stop TB Partnership has been spent to further methods of delivering better services (mainly medicines) to individuals in these communities, and yet reports suggest that these efforts remain grossly underfunded. Leaders at high political and ministerial levels have raised their collective voices for interventions to end the growing TB burden.

This public commitment to change is readily available to global health professionals, and can be found in the report published by the Stop TB Partnership called ‘Paradigm Shift’. This report outlines in detail how current research and funding practices must be redefined and repositioned to support three main pillars to seeing the end of TB: (i) integrated patient-centred care and prevention; (ii) bold policies and supportive systems; and (iii) intensified research and innovation.[7] There are myriad examples of how researchers have contributed to the reduction of global TB, not to mention

Table 1. Four characteristics of community-based qualitative approaches to improving TB treatment and services Characteristic Person-centred

Participatory

Preference-based

Persistent

Approach method Purpose Elicitation interviews; Contextualises collection and cognitive interviews validation of data within lived experiences of individuals in communities.

Community-based Levels power hierarchies by participatory emphasising shared goals and research; stakeholder responsibilities for research analysis and implementation of new services; engages diverse stakeholders Implementation Provides communities with science; shared most up-to-date evidence to decision-making foster informed conversations with the intention of finding optimal service and treatment solutions Applied Formal commitment between ethnography researchers and communities that emphasises co-leadership

Relevance Key takeaway Consistent with 1st pillar of End TB Strategy;[7] Transfers setting of ensures external validity wof findings evidence from lab to community Meets demand of 3rd pillar of End TB Strategy[7] to promote innovative and rapid intervention designs

Engages groups in the codesign of innovative contextdriven strategies

Feasibility and acceptability of interventions are necessary for sustainability; sustainability is only possible if strategies match community preferences (needs and wants)

Any intervention designed in the absence of evidence and community choice risks failure

Addresses 2nd pillar of End TB Strategy[7] to ensure bold strategies and commitment to communities; research designed with purposes of application and improving outcomes has sustainability as main feature

Bold action requires ambition and commitment to communities

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

49

May 2018

EDITORIAL the fact that much of this work has led to innovative approaches and potentially novel pharmaceuticals to reinforce prevention measures, but despite these relatively impressive effects, those who perform health services research in infectious diseases know that we are still losing the battle with TB in the form of multi-drug resistance.

The global plan to end TB led by the Stop TB Partnership[7] follows the ambitious 90-90-90 programme, where in theory, 90% of people with TB are reached, 90% of key populations where need is greatest are screened, and there is a 90% success rate in treatment.[4] This partnership readily acknowledges a need to change the ‘businessas-usual’ practice, calling instead for patient-centred prevention, bold supporting policies and intensified research and innovation.

In SA, while ambitious, it is possible that this programme reinforces the same old approach that has included screening, treating and follow-up, as described in traditional approaches that collect incidence and prevalence data on TB. The reality is that while many across SA recognise the need for novel approaches, the questions of method and approach are limited by strict patterns consistent with clinical outcomes-based approaches, and are in dire need of modification.

Qualitative approaches offer new insights into relevant challenges posed by current gaps in TB control. Arguably, the scope of qualitative methods and their relevance towards optimising approaches to fighting TB are best characterised by four features: they are person-centred, participatory and preferencebased, and require perseverance. While other approaches eschew bias for parsimony, objectivism and clear comparative controls, qualitative methods look at the social world as messy and fully of diverse, subjective and naturally unfamiliar phenomena. The lived experiences of individuals, in how they form groups, come together in networks to support or defend each other, relate to one another and are impacted and impact their environment are just a few of the related areas of inquiry appropriate for qualitative methods. By design, the methods permit the use and analysis of diverse forms of data, and embrace multiple modes of data collection. Qualitative, community-based methods are flexible, as well as specifically useful to the right context. Social scientists from disciplines such as anthropology and sociology have adopted and expertly applied these methods in ways that have led to deep insights into human behaviour, not to mention the worlds of communication and medicine. Ironically, the strengths inherent to qualitative methods are also the characteristics criticised by the most strident positivists.

Person-centred Qualitative community-based approaches are by design embedded in the communities in which they are applied. In order for TB research to be impactful, before description, it must first be dedicated to understanding the cultural and behavioural aspects of how people with TB, and those who are at highest risk of acquiring TB, actually live. In medical sociology, the phrase ‘lived experience’ is used to define those attributes of human behavior that encapsulate the social world of people in their day-to-day lives. Author Ursula K Le Guin once said that ‘all knowledge is local,’ which is critical to this argument about the necessity of finding methods that are

50

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

person-centred. Individual knowledge is often constructed in space and environment. These concepts are naturally a part of one’s community, and where one lives, and was born and raised.

Determining which aspects of TB are most locally relevant, and how the bacteria spread in communities requires intimate knowledge of local habits, customs and structures.

Collecting this type of data, which requires local knowledge, is the purpose of qualitative methods, and in particular the focus of elicitation interviews and cognitive interviews. These types of interviews require the investigator to be versed in multiple skills in community engagement and interviewing.

Whether collecting data independently or in teams, the skills necessary for doing person-centred work require an ability to train others, approach research more locally, and analyse data with community partners. This kind of approach runs counter to many objectivist approaches to data collection, yet underscores one of the fundamental strengths of qualitative inquiry – the understanding that all knowledge is local.

Participatory To achieve person-centred approaches, the researcher must put aside professional constraints associated with hierarchies, and abandon so-called rigorous objectivism, to embrace the natural phenomenon of partnership. Participatory methods, historically associated with anthropology, see their modern origins in Paulo Freire’s codesigned educational methods.[8] These approaches have also been highlighted in more contemporary epistemologies, where researchers and communities collaborate to develop more effective interventions for education, health promotion and public health.[9,10] The principles of community-based research, or community-based participatory research, have also been successfully applied in SA in HIV prevention and treatment.[11] What participatory methods offer, beyond the idea of partnership with communities, is the multistakeholder approach to developing research questions, and to data collection. By partnering with stakeholders in the community and beyond, researchers can incorporate multi-level, multi-perspective data that can uniquely inform service delivery.[12,13]

Preference-based The importance of offering choices in treatment and modes of receiving treatment is often overlooked and misunderstood as reflecting an uneducated point of view. As mentioned already, if we work from the premises that all knowledge is local, and that a person-centred approach is ethically appropriate, then we must acknowledge that individual choice matters. People living in diverse communities see the world differently, and most likely have independent experiences that uniquely characterise their social worlds. If we further recognise that our job as health professionals is to educate as well as treat, then we must agree on some level that providing communities with evidence permits the examination of choice.[14,15] One might only have to look at the availability of treatment for TB. Whether an individual receives TB medicine through a directly observed treatment, short-course (DOTS) or community health-worker model, the data support both approaches. How can we as health professionals determine the optimal approach for

EDITORIAL receiving treatment if we don’t provide individuals with the choice to decide which option they prefer, given the evidence?

Persistent In a world of medicine and clinical science where hypothesis testing, counts and percentages, objective principles of measurement and minimising error and limiting bias hold sway, qualitative methods face a tough path towards relevancy. Yet I argue (Table 1) that it is indeed this unique role in the science of medicine where a qualitative method, ethnography in this case, offers deeper insight into some of the most challenging of questions that perplex the most accurate process and outcome measures. Ethnography is widely accepted as one of the most appropriate qualitative methods for capturing the unique cultural and situational characteristics of illness.[16,17] Illness, the common marker of human physical distress, has been defined by and subject to specific cultural and social orientations. Ethnography in general, and applied ethnography specifically, is well-positioned to explore sociocultural issues related to TB treatment and care delivery.[18,19] TB is by nature a social disease, and as others have pointed out, requires a specific methodological skill set to properly understand it. TB’s impact has been widely experienced by the poor, rural and unseen. In more ways than one, applied ethnography, framed by stakeholder engagement, participant observation and collaborative approaches to data collection, holds the most promise for supporting the development of long-term treatment approaches to mitigate the TB burden.[20]

Conclusion New approaches to TB research that are patient-centred, participatory, preference-based and persistent offer the most robust set of methodologies to change current ‘business-as-usual’ paradigms in health services research on TB control. The history of TB is a timeless story characterised by frustration, forced labor and migration, fear and filth. For decades, large investments have been made in mitigating the progression of the illness, and while some methods have seen improvements in incidence, the rise of multi-drug resistant strains has caused many to question these traditional approaches. The time is now for business as usual to stop. Those who are most likely to acquire TB, and be unable to fight against its effects, are people who are already unhealthy and malnourished, and who live in close proximity to one another. This is a perfect recipe for disaster in rural areas of the world, where resources are limited and traditional methods of care delivery have a history of ineffective implementation. The methods described here in brief offer some insight into qualitative approaches that might capture more relevant data to shape more applicable and useful TB interventions in the future.

S Grande Dartmouth Institute for Health Policy and Clinical Practice, Geisel School of Medicine at Dartmough, NH, USA South Afr J Pub Health 2018;2(3):49-51. DOI:10.7196/SHS.2018.v2.i3.70 1. Ho MJ. Sociocultural aspects of tuberculosis: A literature review and a case study of immigrant tuberculosis. Soc Sci Med 2004;59(4):753-762. https://doi.org/10.1016/j. socscimed.2003.11.033 2. Naidoo P, Peltzer K, Louw J, et al. Predictors of tuberculosis (TB) and antiretroviral (ARV) medication non-adherence in public primary care patients in South Africa: A cross sectional study. BMC Public Health 2013;13(1):396-406 https://doi.org/10.1186/14712458-13-396 3. Benson VS, Patnick J, Davies AK, Nadel MR, Smith RA, Atkin WS. Colorectal cancer screening: A comparison of 35 initiatives in 17 countries. Int J Cancer 2008;122(6):1357-1367. https:// doi.org/10.1002/ijc.23273 4. World Health Organization. Global Tuberculosis Report. Geneva: WHO, 2016 5. Dhavan P, Dias HM, Creswell J, et al. An overview of tuberculosis and migration. Int J Tuberc Lung Dis 2017;21(6):610-623. https://doi.org/10.5588/ijtld.16.0917 6. Ortblad KF, Salomon JA, Bärnighausen T, Atun R. Stopping tuberculosis: A biosocial model for sustainable development. Lancet 2015;386(10010):2354-2362. https://doi. org/10.1016/S0140-6736(15)00324-4 7. The Global Plan to End TB 2016-2020: The Paradigm Shift. Stop TB Partnership. http://www.stoptb.org/assets/documents/global/plan/GlobalPlanToEndTB_ TheParadigmShift_2016-2020_StopTBPartnership.pdf (accessed 20 April 2018).

8. Freire P. Pedagogy of the oppressed. New York: Continuum International, 1970.

9. Israel BA, Schulz AJ, Parker EA, Becker AB. Review of community-based research: Assessing partnership approaches to improve public health. Ann Rev Public Health 1998;19:173-202. https://doi.org/10.1146/annurev.publhealth.19.1.173 10. Wallerstein NB, Duran B. Using community-based participatory research to address health disparities. Health Promot Pract 2006;7(3):312-323. https://doi. org/10.1177/1524839906289376 11. Mosavel M, Simon C, van Stade D, Buchbinder M. Community-based participatory research (CBPR) in South Africa: Engaging multiple constituents to shape the research question. Soc Sci Med 2005;61(12):2577-2587. https://doi.org/10.4135/9781412971942. n77 12. Glanz K, Bishop D. The role of behavioral science theory in development and implementation of public health interventions. Ann Rev Public Health 2010;31(1):399418. https://doi.org/10.1146/annurev.publhealth.012809.103604 13. Lambert H, Mckevitt C. Anthropology in health research: From qualitative methods to multidisciplinarity. BMJ 2002;325(7357):210-213. https://doi.org/10.1136/bmj.325.7357.210 14. Llewellyn-Thomas HA, Crump RT. Decision support for patients: Values clarification and preference elicitation. Med Care Res Rev 2013;70(Suppl 1):S50-S79. https://doi. org/10.1177/1077558712461182 15. Katz JN. Preferences, disparities, and the authenticity of patient choices. J Rheumatol 2003;68(Suppl 1):S12-S14.

16. Leslie M, Paradis E, Gropper MA, Reeves S, Kitto S. Applying ethnography to the study of context in healthcare quality and safety. BMJ Qual Saf 2014;23(2):99-105. https://doi. org/10.1136/bmjqs-2013-002335 17. Reeves S, Kuper A, Hodges BD. Qualitative research methodologies: Ethnography. BMJ 2008;337(07_3):a1020. https://doi.org/10.1136/bmj.a1020 18. Adams LV, Basu D, Grande SW, et al. Barriers to tuberculosis care delivery among miners and their families in South Africa: An ethnographic study. Int J Tuberc Lung Dis 2017;21(5):571-578. https://doi.org/10.5588/ijtld.16.0669 19. Savage J. Ethnographic evidence: The value of applied ethnography in healthcare. J Res Nurs 2006;11(5):383-393. https://doi.org/10.1177/1744987106068297 20. Lassiter LE. Collaborative ethnography and public anthropology. Curr Anthropol 2005;46(1):83-106. https://doi.org/10.1086/425658

Accepted 6 March 2018.

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

51

May 2018

ARTICLE

HIV and tuberculosis prevention and control in South Africa: An overview W D F Venter, FCP, MMed, DTM&H Wits Reproductive Health and HIV Institute, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa Corresponding author: W D F Venter (fventer@wrhi.ac.za)

HIV and tuberculosis (TB) remain major challenges in South Africa, despite free, and widely available, effective prevention and treatment.

While modest successes have been enjoyed in terms of prevention, these started from an already very high infection level, and new infections of both diseases are still among the highest in the world. This overview explores the complex and evolving epidemiology of HIV and associated TB, and the possible mechanisms for achieving control of the epidemics.

South Afr J Pub Health 2018;2(3):52-54. DOI:10.7196/SHS.2018.v2.i3.61

It is difficult to overstate the impact of HIV and tuberculosis (TB) on the health of populations in sub-Saharan Africa. In this series, I provide an overview of the status of the prevention and control of both diseases. Several decades of steady improvement in life expectancy, on the back of improved sanitation and housing, safer food and access to vaccines and improved medical care, were rapidly reversed during the 1980s and 1990s as HIV and associated TB prevalence increased rapidly. South Africa (SA), paradoxically protected by its politically and socially restricted borders during apartheid, then saw a rapid rise in HIV prevalence during the 1990s, to the point where it is now similar to that in surrounding epidemic countries.[1-3] The broad availability of potent antiretroviral (ARV) therapy for the treatment of HIV, with a subsequent improvement in the tolerability and effectiveness of the drug cocktail, has dramatically reduced the morbidity and mortality associated with HIV, and for the first time in decades, we have seen an associated slight decline in new diagnoses of TB.[2,3] However, control of these two closely interlinked diseases has remained a frustrating challenge at several levels. Common, serious, transmittable, ostensibly preventable and usually treatable, the public health importance of dealing with the dual epidemics is very clear, but the prevention of HIV and TB has proven to be remarkably difficult, with very limited success in sub-Saharan Africa.

However, better understanding of epidemiology, better interventions for infection control and chemoprophylaxis and better treatment and care packages for both diseases have given new hope to prevention efforts. In particular, efforts to reduce the toxicity of the medication used to treat both diseases, especially for HIV and multidrug-resistant TB, have begun to bear fruit. The fact that both diseases are effectively and rapidly rendered nontransmissible by effective therapy has led to focused attention

on improving treatment packages. Earlier and better diagnosis of TB, long neglected in favour of poorly sensitive microscopy, has inspired the entire field, with more ambitious attention now paid to drug development and system delivery.[4,5]

Understanding the rise and fall of HIV HIV in SA in the 1980s looked very similar to that in the USA and some European countries: largely confined to white haemophiliacs (infected with blood procured from the USA, before the discovery of HIV in their blood supply), men-who-have-sex-with-men (MSM) and a small number of intravenous drug users. However, the epidemic exploded in the early 1990s in the general community, owing to heterosexual sexual, pregnancy and breastfeeding transmission, which have contributed to the vast majority of new infections since then.[1] Initial efforts to contain the HIV epidemic were constrained by delayed epidemiological recognition of this rapid rise of new infections, especially in KwaZulu-Natal (KZN) Province, where political instability appeared to fuel a startling rise in incidence, from less than 1% of pregnant women attending antenatal clinics being affected, in sentinel surveys, to some areas showing over 60%.[1] Historically, HIV prevention relied on general education campaigns and steadily improved condom distribution. As data emerged, prevention of mother-to-child transmission using ARVs since 2001 has been a huge public health victory, decreasing transmission levels from 70 000 per year 10 years ago, to less than 3 000 in 2016.[6] Medical male circumcision, shown to be highly effective in 2005, has been steadily scaled up across the country.

The control of viral load in HIV-positive patients with ARVs was shown to completely stop sexual and breastfeeding transmission, allowing the decision to move to ‘test and treat’ models, where ARVs are offered irrespective of level of immune dysfunction.[7,8]

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

52

May 2018

ARTICLE Pre-exposure prophylaxis, where ARVs are given to HIV-negative people at risk for HIV, has recently become an option in SA.[9] Finally, SA’s blood supply has, crucially, remained very safe throughout the HIV-era, in contrast to many other countries in Africa. New emerging issues, such as increased intravenous drug use, and increased ease in finding sexual partners using social media apps, may challenge current prevention programmes with new and evolving microepidemics.

There are inexplicable significant differences in the evolution of prevalence by province. The increase in KZN was followed rapidly by similar increases in Mpumalanga and Gauteng, while the Northern Cape and Western Cape have remained far less affected. Currently, almost all the other SA provinces have similar high prevalence rates, with the exception of these two.[4,5] HIV new incident infections now appear to be on a downward trend. SA has a rich source of independently collected data that chronicles the HIV incident evolution over the last two decades.

The country has several sources of this data, including from the Human Sciences Research Council surveys, the pregnant women survey conducted by the Department of Health, and audits of death certificates nationally, by Statistics SA.[10-12] In addition, more geographically constrained research projects, such as the Africa Centre, the Centre for the AIDS Programme of Research in Africa (CAPRISA) and the HIV Incidence Provincial Surveillance System (HIPSS), all in KZN, have documented interesting changes in demography and access to care in the last decade.[13-15] Most studies appear to demonstrate a decrease in new infections, especially among younger people, although the absolute number remains very high.

More recently, the widespread availability of effective ART has confounded some of these studies, as people live for decades on treatment, achieving near-normal life expectancies, with a resulting steady increase in prevalence. Simple and cheap antibody-based assays were previously used to document changes in epidemiology, but accurately documenting new incident infections has become necessary, to understand where new microepidemics are occurring.

These tests, including techniques using ‘detuned’ ELISA assays, antibody avidity assays and pooled RNA remain controversial in terms of interpretation, and are often expensive, posing new challenges to surveillance efforts.

Understanding the rise of TB in SA TB remains the number one killer in SA, where its incidence is the second highest in the world, largely owing to the scale of the HIV epidemic, which renders HIV-positive people at increased risk of TB, even in the presence of ARV therapy. A newly described epidemic among healthcare workers points to inadequate ventilation within health facilities, which adds another at-risk population to the traditionally recognised groups, such as prisoners and miners. Poor access to effective isonicotinylhydrazide (INH) prevention therapy for people with HIV has meant that this population continues to contribute to new infection rates.[4,5] TB has been effectively brought under control within most developed countries, with large numbers of developing countries outside of sub-Saharan Africa now making progress in epidemic

53

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

control. Reports that demonstrate global improvements in TB control gloss over the fact that sub-Saharan Africa is actually still experiencing an overall rise in numbers. As with HIV, the epidemiology is not straightforward, with the Western Cape Province having the highest prevalence despite having a far lower HIV burden than the rest of the country. The places with lower HIV rates predictably report lower HIV-TB coinfection rates. There is even high variation between districts, and while some of these findings may be due to reporting issues, this does not fully explain the differences, which differ fourfold between the highest- and lowestincident districts.[3,5] The reporting of TB is complex, as it has relied traditionally on poorly sensitive sputum tests, which lose sensitivity even further in the presence of HIV, so that accurate numbers are hard to derive. Autopsy studies suggest that TB is underdiagnosed, even in academic institutions.

In parallel, the area has seen the rise of multidrug-resistant TB, made more apparent by the rise of sensitive new polymerase chain reaction-based technologies that allow rapid diagnosis of resistance. While traditionally blamed on poor patient adherence, it has increasingly been recognised that other factors are probably responsible for this rise, including inadequately potent regimens with inadequate serum concentrations of active drug, inadequate ventilation and infection control in health facilities, and more widespread community and health facility transmission than previously thought.

A national survey is planned in 2018 that may allow more accurate tracking of the epidemiological evolution of TB in SA. While traditional TB control programmes often trumpet the fact that ‘TB can be cured’, long-term consequences even in cured TB often include impact on affected organs, including lung, bone and brain; the term ‘respiratory cripple’ needs little explanation, and is a consequence in a significant number of TB survivors.[5] In terms of research, TB is an orphan when compared with HIV when it comes to research money and effort. However, this is changing, with the first new diagnostics and chemotherapy agents in decades being tested and rolled out. The pace is still slow, but the development of shorter, more effective treatment regimens for both drug-sensitive and drug-resistant TB looks promising, and may contribute substantially to better epidemic control. However, far more so than HIV, TB is a disease of poverty, and social and economic improvement is probably a key component of control, going forward.[3]

Conclusions Both the HIV and TB epidemics are complex and still evolving. The huge victories around treatment, and more limited ones in the identification of effective prevention techniques in the HIV field, have fuelled new resolve within the TB research and public health world, long constrained by a lack of research and programme funding and, arguably, by a lack of ambition.

Both epidemics require attention from a public health perspective; the HIV programme alone consumes a massive proportion of the health budget, and TB has a huge impact on the nation’s health, even in people successfully treated. Understanding

ARTICLE the nuances of the epidemiology, with proper scientific surveys, will allow more nuanced and effective prevention targeting.

In the future, this series will look at these aspects of prevention, reviewing the current status of the field in terms of medical efficacy, the current and future impact on each epidemic in the SA context, and the implementation challenges of each for public health programmes.

Acknowledgements. None.

Author contributions. Sole author.

Funding. WDFV is supported by the United States Agency for International Development (USAID), the SA Medical Council and Unitaid.

Conflicts of interest. WDFV is part of the OPTIMIZE consortium, which evaluates new antiretroviral approaches to improve access to treatment, which has included drug donations for studies. He has accepted honoraria from multiple pharmaceutical manufacturers for talks and participation on advisory boards.

1. Simelela N, Venter WDF, A brief history of South Africa’s response to AIDS. S Afr Med J 2014;104(3):249-251. https://doi.org/10.7196/SAMJ.7700 2. Joint United Nations Programme on HIV/AIDS (UNAIDS). 90-90-90: An ambitious treatment target to help end the AIDS epidemic. Geneva: UNAIDS, 2014. http://www.unaids.org/ sites/default/files/media_asset/90-90-90_en.pdf (accessed 5 May 2018).

3. World Health Organization. Global tuberculosis report 2016. Geneva: WHO, 2016. http:// www.who.int/tb/publications/global_report/en/ (accessed 5 May 2018).

4. South African National AIDS Council. Draft South African National Strategic Plan 2017 2022. Pretoria: SANAC, 2016. http://nsp.sanac.org.za/2017/02/01/the-draft-of-the-newnsp-2017-2022-is-now-ready-for-review (accessed 5 May 2018).

5. Health Systems Trust. District Health Barometer 2015/2016. http://www.hst.org.za/ publications/Pages/-District-Health-Barometer-201516.aspx (accessed 5 May 2018).

6. AfricaCheck. Are fewer than 6 000 babies born HIV+ every year in SA, as Zuma said?

AfricaCheck, 2017. https://africacheck.org/reports/fewer-6000-babies-born-hiv-every-year-sazuma-said/ (accessed 5 May 2018).

7. Abdool Karim SS, Abdool Karim Q. Antiretroviral prophylaxis: A defining moment in HIV control. Lancet 2011;378(9809):e23-e25. https://doi.org/10.1016/s0140-6736(11)61136-7 8. Cohen MS, Chen YQ, McCauley M, et al. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med 2011;365(20):493-450. https://doi.org/10.1056/nejmc1110588 9. Eakle R, Venter WDF, Rees H, Pre-exposure prophylaxis for HIV prevention: Ready for prime time in South Africa? S Afr Med J 2013;103(8):515-516. https://doi.org/10.7196/samj.6937 10. Shisana O, Rehle T, Simbayi LC, et al. South African National HIV Prevalence, Incidence and Behaviour Survey, 2012. Cape Town: HSRC Press, 2014.

11. Statistics South Africa. Mid-Year Population Estimates, 2015. Cape Town: StatsSA, 2015. https://www.statssa.gov.za/publications/P0302/P03022015.pdf (accessed 5 May 2018).

12. National Department of Health, South Africa. The 2013 National Antenatal Sentinel HIV Prevalence Survey South Africa. Pretoria: NDoH, 2013. https://www.health-e.org.za/wpcontent/uploads/2016/03/Dept-Health-HIV-High-Res-7102015.pdf (accessed 5 May 2018).

13. Abdool Karim Q, Baxter C, Birx D. Prevention of HIV in adolescent girls and young women: Key to an AIDS-free generation. J Acquir Immune Defic Syndr 2017;75(Suppl 1):S17-S26. https://doi.org/10.1097/qai.0000000000001316 14. Grobler A, Cawood C, Khanyile D, Puren A, Kharsany ABM. Progress of UNAIDS 9090-90 targets in a district in KwaZulu-Natal, South Africa, with high HIV burden, in the HIPSS study: A household-based complex multilevel community survey. Lancet HIV 2017;4(11):e505-e513. https://doi.org/10.1016/s2352-3018(17)30122-4 15. Tanser F, Bärnighausen T, Grapsa E, Zaidi J, Newell M-L. High coverage of ART associated with decline in risk of HIV acquisition in rural KwaZulu-Natal, South Africa. Science 2013;339(6122):966-971. https://doi.org/10.1126/science.1228160

Accepted 2 March 2018.

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

54

May 2018

RESEARCH

Listeriosis in the City of Johannesburg, South Africa P Manganye, Nat Dip Public Health (SA); B Desai, MB BCh, MMed; M Daka, PhD; R Bismilla, LLM, RCP Irel, LLM, RCS Irel City of Johannesburg Health Department, South Africa Corresponding author: P Manganye (PeterMan@joburg.org.za)

Listeriosis is a food-borne disease caused by food contaminated with the Listeria monocytogenes (L. monocytogenes) bacterium. L. monocytogenes is found in soil, vegetation and water. There are six species of Listeria, but only L. Monocytogenes causes disease in humans.

It is a relatively rare disease, with 0.1 - 10 cases per million people per year, depending on the country or region of the world. The World Health Organization believes that South Africa’s (SA’s) current listeria outbreak is the largest ever in the world. The National Institute of Communicable Diseases reported that as of 28 February 2018, there had been 943 laboratory-confirmed cases of listeriosis in SA, and 176 deaths from the disease. As of March 2018, the City of Johannesburg (CoJ) has had a total of 251 cases (26% of total cases), with an incidence of 51 cases per 1 million, and a case fatality rate of 15%. The age group 15 - 49 is the most badly affected, followed by neonates >28 days old.

A detailed outbreak preparedness and response plan to prevent listeriosis and promote good hygiene was developed which emphasised the fact that the main preventive measure is to always ensure that good basic hygiene is followed. The CoJ is committed to continuing the management and control of listeriosis according to the National Department of Health communicable disease guidelines and surveillance policy, which includes the provision and management of primary healthcare to all patients presenting with suspected listeriosis at facilities, and conducting regular preventive and promotive activities/measures to create community awareness.

South Afr J Pub Health 2018;2(3):55-58. DOI: 10.7196/SHS.2018.v2.i3.73

Listeriosis is a food-borne disease caused by food contaminated with the Listeria monocytogenes (L. monocytogenes) bacterium.

L. monocytogenes is found in soil, vegetation and water. Vegetables can become contaminated by way of the soil or from manure used as fertiliser. Some animals carry the bacteria and might contaminate their meat and dairy products. Processed foods, such as soft cheeses and cold cuts, can become contaminated during processing. Unpasteurised milk could be unsafe to consume.[1] L. monocytogenes has been recognised as an animal pathogen since the early part of the 20th century. It is widespread in nature, in soil, decaying vegetation and the bowels of many mammals. The first human outbreak was reported in Canada in 1983, proving that indirect transmission from animals to humans was possible. In that outbreak, cabbages, stored in the cold over the winter, were contaminated with L. Monocytogenes through exposure to infected sheep manure. There are six species of Listeria, but only L. monocytogenes causes disease in humans.[2] It is a relatively rare disease, with 0.1 - 10 cases per million people per year, depending on the country or region of the world.

Although the number of cases of listeriosis is small, the high death rate associated with infection makes it a significant public health concern.[1] The World Health Organization believes that South Africa’s (SA’s) current listeria outbreak is the largest ever in the world. The second largest outbreak occurred in 2011, with a total of 147 cases reported in the USA. Italy also had a large outbreak in 1997.[1]

SA situation The first documented outbreak in SA occurred between August 1977 and April 1978 (14 cases reported in Johannesburg). Since then, there have been sporadic cases. Listeriosis was not then recognised as a notifiable disease, and therefore it could not be picked up by the routine surveillance system. After the recent outbreak, the National Department of Health has made it a notifiable medical condition.[3] The National Institute of Communicable Diseases (NICD) reported that as of 28 February 2018, there had been 943 laboratoryconfirmed cases of listeriosis in SA, and 176 deaths from the disease. The distribution of cases per province was: Eastern Cape, 48; Free State, 33; Gauteng, 555; KwaZulu-Natal, 65; Limpopo, 47; Mpumalanga, 46; Northern Cape, 5; North West, 27; and Western Cape, 116.

On 4 December 2017, the NICD reported that whole genome sequencing had been performed on 189 clinical L. monocytogenes isolates, and 15 sequence types (STs) identified; 71% (134/189) of the isolates were of a single ST (ST6). It was reported that the isolates in the ST6 cluster were very closely related, which suggests that most cases in this outbreak have had exposure to a widely available, common food type/source.[4] A media statement on 4 March 2018 by the Minister of Health confirmed that the source of the recent outbreak hasdbeen confirmed to be the Enterprise food-production facility in Polokwane.[5]

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

55

May 2018

RESEARCH City of Johannesburg situation The City of Johannesburg (CoJ) is the capital of Gauteng Province, the economic hub of SA. The city has a population of 4.9 million people spread across seven geographical regions (Fig. 1).

Fig. 2 shows the number of laboratory-confirmed listeriosis cases per year from January 2013 to December 2017. As can be seen, there were very few cases of listeriosis in SA before 2017.

To date, the CoJ has had a total of 251 cases (26% of total cases), an incidence of 51 cases per 1 million, and a case fatality rate of 15%. The age breakdown of listeriosis cases in the CoJ is described in Fig. 3. The figure shows that taken together, the age group 15 - 49 is the most badly affected, followed by neonates >28 days old.

These cases were spread across the seven regions as shown in Table 1, and hospitals as shown in Table 2.

An outbreak response team was activated in the CoJ. A detailed outbreak preparedness and response plan to prevent listeriosis and promote good hygiene was developed (Table 3), which included the following: • prevention and health promotion activities, focusing on the following target groups: community; food premises/food handlers; and health workers • development of pamphlets and posters, and distribution of frequently asked question (FAQ) documents • briefing sessions for environmental health practitioners and professional nurses, conducted on 12 December 2017 • training of environmental health practitioners on 1 February 2018 • food samples taken from different food stores and outlets in the city by environmental health practitioners.

The prevention and health promotion activities emphasised the fact that the main preventive measure is to always ensure that good basic hygiene is followed. This includes:

Fig. 1. The seven regions of the City of Johannesburg.

743

800 Cases, n

Table 1. Listeriosis cases and deaths per CoJ region, 1 January 2017 to 28 March 2018

600 400 200

9

25

40

42

2013

2014

2015

2016

0 2017

Year

Fig 2. Laboratory-confirmed listeriosis cases in South Africa, 1 January 2013 - 31 December 2017.

Region A B C D E F G Unknown Total

Cases, n 14 17 10 60 8 29 16 97 251

Deaths, n 1 4 1 12 1 9 4 8 40

CoJ = City of Johannesburg

70 60

Cases and deaths, n

60

Cases

Table 2. Listeriosis deaths per hospital

Deaths

50 40 30

28

25

20

21

17 11

10

5

1

31

7 0

1

2

1

2

18 11

9 2

4

10 1

5

2

16

7 0

0-

28

da ys 15 610 11 -1 5 16 -2 0 21 -2 5 26 -3 0 31 -3 5 36 -4 0 41 -4 5 46 -5 0 51 -5 5 56 -6 0 > Un 61 kn ow n

0

6

15

Age groups, years

Fig. 3. Age breakdown of laboratory-confirmed listeriosis cases and deaths in the City of Johannesburg, 1 January 2017 - 28 February 2018.

56

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

Hospital Life Carstenhof Clinic Charlotte Maxeke Johannesburg Academic Hospital Chris Hani Baragwanath Hospital Helen Joseph Hospital/Coronation Hospital/Rahima Moosa Mother And Child Hospital Life Healthcare Brenthurst Clinic Life Healthcare Flora Netcare Milpark Hospital Netcare Olivedale Hospital Zola Jabulani District Hospital

Deaths, n 1 5 17 9

Total

40

1 2 2 1 2

L ine listing of any suspected cases; follow-up on old and new cases reported; health talks on listeriosis; distribution of health information material Conduct radio talks; draft newspaper statements; Jozinet stories; CoJ messages

Activity Give briefing on listeriosis; provide FAQ document; distribute listeriosis guidelines and information material; give presentations on listeriosis; email information to GPs Social mobilisation; health education to informal and formal food handlers; door-to-door household visits giving information to communities; distribution of information material; communication (media, radio and newspapers) Education to farm owners and workers on listeriosis

Responsible party EHPs; surveillance officers; CoJ outbreak response team; health promoters; epidemic preparedness operations managers oJ and Gauteng C Province; Departments of Environmental Health and Health Promotion; WBOTs

EHPs

linic nurses; CoJ outbreak C team; health promoters

Communications officer

Target group All CoJ healthcare facility staff members; NGOs/CBOs/FBOs; traditional/faith healers

ommunity in informal C settlements; community at large

Farm owners and workers

Community at large

Various media platforms

I nformed farm owners and workers; reduced food contamination at farm level Well-investigated cases; well-informed community; reduced number of listeriosis cases

Information material

I nformation, facts and figures on listeriosis

Well-informed audience

reating awareness; wellC informed community; reduced number of listeriosis cases; healthy community

Pamphlets; posters; loudhailers; health promotion vehicle; press releases

ase investigation form; C information material

Outcomes Stakeholder alert; informed stakeholders; well-managed suspected listeriosis cases; referrals done in time

Required resources Updated guidelines; FAQ document; pamphlets

FAQ = frequently asked question; GP = general practitioner; NGO = non-governmental organisation; CBO = community-based organisation; FBO = faith-based organisation; EHP = environmental health practitioner; CoJ = City of Johannesburg; WBOT = wardbased outreach team; Jozinet = CoJ website.

Communication

ealth facilities, regional and H central offices

Farms

F ormal and informal settlements (including suburbs) taxi ranks; bus stations; Park Station

Area/ward All regions

Table 3. Listeria outbreak response plan for prevention and hygiene promotion, December 2017 - ongoing

RESEARCH

• using only pasteurised dairy products • separating raw and cooked food, and thoroughly cooking raw foods from animal sources, such as beef, pork or poultry • keeping food at safe temperatures • using safe water and raw materials • washing hands before preparing food, before eating and after going to the toilet • washing and decontaminating kitchen surfaces and utensils regularly, particularly after preparing raw meat, poultry and eggs, including in industrial kitchens • washing raw vegetables and fruit thoroughly before eating.

Those at high risk of listeriosis were advised to avoid the following foods: • raw or unpasteurised milk, or dairy products that contain unpasteurised milk • soft cheeses (e.g. feta, goat’s milk, brie) • foods from delicatessen counters (e.g. prepared salads, cold meats) that have not been heated/reheated adequately • refrigerated pâtés.

Conclusion

The CoJ is committed to continue the management and control of listeriosis according to the National Department of Health communicable disease guidelines and surveillance policy, which includes the provision and management of primary healthcare to all patients presenting with suspected listeriosis at facilities, and conducting regular preventive and promotive activities/measures to create community awareness.

Acknowledgements. We acknowledge the staff working in the CoJ, the management and the member of the mayoral committee (MMC) for Health, Councillor M Phalatse, for their active contribution during the outbreak.

Author contributions. All the authors contributed actively in the preparation of this article and worked on this project.

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

57

May 2018

RESEARCH Funding. The work done for this project was funded by the CoJ. Conflicts of interest. None.

1. World Health Organization. Listeriosis: Fact Sheet. Geneva: WHO, 2018. http://www.who. int/mediacentre/factsheets/listeriosis/en/ (accessed 20 February 2018).

2. Orsi RH, Wiedmann M. Characteristics and distribution of Listeria spp., including Listeria species newly described since 2009. Appl Microbiol Biotechnol 2016;100:5273-5287. https://doi.org/10.1007/s00253-016-7552-2 3. National Department of Health. National Health Act No. 61 of 2003. Regulations relating to the surveillance and the control of notifiable medical conditions. Government Gazette No. 41330:630(15). 2017.

58

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

4. National Institute of Communicable Diseases. Situation Report on Listeriosis Outbreak, South Africa, 2017. Pretoria: NICD 2017. http://www.nicd.ac.za/index.php/situationreport-on-listeriosis-outbreak-south-africa-2017/ (accessed 20 February 2017).

5. National Department of Health. Media statement by the Minister of Health Dr Aaron Motsoaledi regarding the update on the Listeriosis outbreak in South Africa. Pretoria: NDoH, 2018. http:// www.health.gov.za/index.php/component/phocadownload/category/439 (accessed 5 March 2018).

Accepted 6 April 2018.

ARTICLE

Organising HIV ageing-patient care in South Africa: An implementation science approach D Croce,1 BEng, MBA; D Mueller,2 MEng; G Rizzardini,3 MD; U Restelli,1 MEcon, PhD Centro di Ricerca in Economia e Management in Sanità e nel Sociale, Faculty of Management Engineering, Università Carlo Cattaneo LIUC, Castellanza, Italy 2 Charlotte Maxeke Medical Research Cluster (CMeRC), University of the Witwatersrand, Johannesburg, South Africa 3 Divisione di Malattie Infettive and Cattedra di Immunologia, Università di Milano, Milan, Italy 1

Corresponding author: D Mueller (dbmueller7@yahoo.de)

The rollout of efficient antiretroviral therapy in many countries, including South Africa (SA), has transformed HIV into a manageable chronic condition, and led to rising life expectancies among people living with HIV/AIDS. As a result, in Africa and elsewhere, there have been a number of reports on multiple comorbidities from non-communicable diseases in those living and ageing with HIV. We conducted a desktop review of studies conducted in SA and other countries on medical service administration and organisation, and planning of interventions aimed at people ageing with HIV, and acquiring non-communicable diseases due to ageing. Furthermore, older adults with HIV have, as a group, unique issues relating to medication compliance, and are more likely to have issues such as polypharmacy and cognitive impairment.

One approach to tackling these issues could be an integrated care, instead of a specialised clinical approach. However, an integrated approach requires a strong commitment from all parties, investment in patient and clinician education and relationship management among providers, services and funders. Although there is no doubt that great progress has been made in extending services for HIV prevention, care and treatment in the last decade, substantial gaps remain in terms of what we know is working, and what we are really achieving with the various programmes. To address this issue, we suggest the use of an implementation science framework, to improve the efficiency and effectiveness of these programmes. Policy- and decision-makers in SA and other parts of Africa will need to put further concerted effort and greater emphasis on targeted care, in particular for older adults.

South Afr J Pub Health 2018;2(3):59-62. DOI:10.7196/SHS.2018.v2.i3.67

The number of HIV-infected individuals in South Africa (SA) was estimated by the Joint United Nations Programme on HIV and AIDS (UNAIDS) to be between 6.4 million and 7.8 million in 2016,[1] with an infection-rate prevalence of 16.6% - 21.0% among the population aged 15 - 49 years.

In terms of the management of the infection, in relation to the 90-90-90 treatment cascade, in 2016, UNAIDS estimated that the mean percentage of infected people aware of their status was 86% (between 78% and 95%); of these people, 65% (between 59% and 72%) were receiving antiretroviral treatment (ART); and the mean percentage of patients receiving ART with viral suppression was 81% (between 74% and 89%).[1] In 2016, in SA, the most common cause of premature death was HIV/AIDS (and also the most common cause of death overall), with a 50.6% decrease from 2005, while diabetes was the sixth most common cause of premature death (fourth of the overall causes of death), with an increment of +5.3% since 2005, while ischaemic heart disease was the seventh most common cause of premature death (second of the overall causes of death) with a 10.9% decrease since 2005.[2] In terms of disability-adjusted life years, in 2016, HIV/AIDS was

the most common cause of death and disability combined, diabetes the sixth-most, and ischaemic heart disease the seventh.[2] This data shows how the combination of these pathologies and risk factors might lead to an increased complexity of patient conditions.

Chronic diseases and HIV burden According to Young et al.,[3] ART for HIV is associated with an increased risk of developing metabolic syndrome (a cluster of obesity, hyperglycaemia, dyslipidaemia and hypertension), and therefore also a predisposition to type 2 diabetes and cardiovascular disease.

Aboud et al.[4] found that insulin resistance is common in HIVinfected people, particularly those treated with protease inhibitor therapy. The prevalence of hyperglycaemia and diabetes mellitus is significantly higher in people with HIV infection treated with antiretrovirals (ARVs) than in the general population.

Carr et al.[5] mention that patients who were HIV proteaseinhibitor naive had similar body composition to healthy men, whereas for HIV-infected people treated with protease inhibitor, this was associated with substantially lower total body fat (13.2 v. 18.7 kg

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

59

May 2018

ARTICLE in protease-inhibitor-naive patients; p=0.005), and with significantly higher total cholesterol and triglyceride levels.

Lipodystrophy in HIV-infected patients is associated with insulin resistance and its metabolic complications, such as impaired glucose tolerance, diabetes, hypertriglyceridaemia and low serum high-density lipoprotein cholesterol levels.[6] The combination of central lipohypertrophy, dyslipidaemia and insulin resistance is associated with accelerated rates of atherosclerosis and other potentially significant long-term effects resulting from highly active antiretroviral therapy (HAART).[7] Grinspoon et al.[8] found that metabolic and body-fat abnormalities are common among HIV-infected adults receiving nucleoside-analogue and protease-inhibitor therapy. There are suggestions that HAART increases risk of cardiovascular disease.

Diet, lifestyle modification and use of lipid-lowering and insulinsensitising regimens may be useful in specific situations.

Pao et al.[9] state that people with HIV infection have metabolic abnormalities that resemble metabolic syndrome (hypertriglyceridaemia, low high-density lipoprotein cholesterol and insulin resistance), which are known to predict increased risk of cardiovascular disease (CVD). CVD has become more prominent among HIV-infected individuals. Patients with HIV had significantly higher CVD mortality than the general population in all age groups up until age 65 (based on demographic characteristics such as sex, race/ethnicity and borough of residence).[10] Hanna et al.[10] further found that after age 65, CVD mortality rate was similar or greater in the general population compared to that of the HIV population. The CVD mortality rate was highest among viraemic patients, but still elevated among the virally suppressed (<400 copies/mL) compared with the general population.

Age-associated comorbidities (especially cardiovascular and renal disease) were more prevalent among HIV-infected patients than HIV-uninfected patients. Comorbidity was associated with cardiovascular risk factors, but also with HIV infection, immunodeficiency and, to a lesser extent, systemic inflammation and prior high-dose ritonavir use.[11] In comparison with HIVuninfected controls, all age-associated non-communicable comorbidities (AANCCs), especially peripheral arterial cardiovascular disease and impaired renal function, were more prevalent among HIV-infected participants. In addition to recognised cardiovascular risk factors, HIV infection and longer time spent with severe immunodeficiency increased the risk of a higher composite AANCC burden. However, there was a less pronounced contribution from residual inflammation, immune activation and prior high-dose ritonavir use.

and overweight were less frequent, while lipid metabolism abnormalities, such as hypercholesterolaemia, total cholesterol, low-density lipoprotein cholesterol and triglycerides, were more frequent than in the general population.

The overall clinical condition of HIV-infected patients may therefore lead to further decline owing to concomitant pathologies related to HIV infection, with an increased clinical, humanistic and economic burden for both the patient and the health services.

The presence of multiple chronic pathologies would have clinical consequences in terms of decreased quality of life for the patients, and both higher direct medical costs and indirect costs to the health service due to the provision of healthcare services, and lower working capacity, respectively, leading to absenteeism and further productivity loss in caregivers.

Patients affected by multiple chronic diseases are complex, and it is necessary to implement specific care pathways, to avoid drug interactions and to identify possible synergies in the diagnostic and curative pathways of the affected pathologies. The management of these patients, in fact, might require multiple services for the monitoring and care of each individual pathology, with increasing costs involved in managing possible events related to the clinical evolution of each one.

The costs correlated with the management of diabetes in SA were investigated by Atun et al.[13] in 2017, identifying an annual cost per person of USD140 in the public sector, and USD1 400 in the private sector, resulting in high levels of inequity.

The outpatient management of hypertension was estimated in 2001 to lead to a yearly cost per person of USD169.28 (all values at 2012 rates) in SA.[14,15] The costs in SA of further cardiovascular events are reported in the literature as: ischaemic heart disease in terms of coronary artery bypass grafting procedure USD22 500.46 per patient; catheter-based revascularisation inpatient treatment USD9 324.18 per patient; average inpatient treatment direct medical costs, USD11 093.74 per patient; annual per-patient cost for postcoronary heart disease with coronary artery bypass grafting USD2 558.88 the first year, and USD1 181.02 in subsequent years; and the annual per-patient cost for postcoronary heart disease without coronary artery bypass grafting USD2 952.56 the first year, and USD1 653.43 in subsequent years.[14,15] Regarding stroke, the cost of inpatient visits for stroke care per patient are estimated at USD16 992.95.[14,15] To manage the increasing level of complexity of patients affected by multiple chronic pathologies, the integration of the care of diabetes and hypertension in HIV care models would improve the management of HIV-infected patients.[13]

The SA context

The hospital organisation problem: Integrated v. specialised clinical approach

In 2011, Julius et al.[12] published a monocentric analysis of the prevalence of metabolic diseases among 304 HIV-infected patients of the HIV clinic at the Charlotte Maxeke Johannesburg Academic Hospital, focusing on hypertension, diabetes, obesity and dyslipidaemia. Regarding hypertension, the authors identified a slightly higher prevalence in HIV-infected women than women in the general population, and a twofold higher prevalence in HIV-infected males than males in the general population. Obesity

60

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

We performed a comparative analysis of both approaches, the integrated and the specialised clinical, to address the problem of organisational (and clinical) appropriateness. The most suitable dimensions of the comparison are patient perception, hospital management, clinical approach and specialised output (Table 1).

Another dimension to consider is the human factor. Doctors have different attitudes, professional capacities and amount of

ARTICLE Table 1. Hospital organisation: Comparative analysis of integrated and specialised clinical approaches Dimension Patient perception Hospital management Clinical approach Specialised output

Specialised (multiple ambulatories) Multiple waiting queues, multiple doctors/nurses in relation to time for overall performance, increased stigma risk Number of doctors (contract, control, performance check), hospital infrastructure (space) Follows doctorsâ&#x20AC;&#x2122; attitude; integration of specialised treatment Multiple

time to dedicate; hence, the choice of hospital organisation is complicated.

Implementation science provides a solution to this challenge.

It is the study of methods to promote the integration of research findings and evidence into healthcare policy and practice. Therefore, it supports innovative approaches to identifying, understanding and overcoming barriers to the adoption, adaptation, integration, scale-up and sustainability of evidence-based interventions, tools, policies and guidelines.[16]

Support for implementation science In the last few decades, the application of clinical research findings to the implementation and dissemination of routine practice, for the benefit of both patients and the public, has come under the spotlight. This reflects the collective realisation that findings from clinical studies have not uniformly resulted in changes in the practices of healthcare providers or patients, nor have they always yielded improvements in health outcomes.[17,18] Implementation science is the study of what happens after adoption occurs, especially in organisational settings. Typical questions addressed are: where does the current emphasis on dissemination and implementation science come from? How is new media altering the diffusion of new practices, programmes, and beliefs? Collective knowledge of the diffusion-of-innovations paradigm has given way to a focus on those paradigmatic concepts, in purposive tests of how to best disseminate and implement evidence-based health practices, programmes and policies. This has long been an objective in trying to spread effective innovations for improved global health, as well as for domestic healthcare and public health.[19] The Consolidated Framework for Implementation Research (CFIR) offers an overarching typology to promote implementation theory development, and verification about what works where and why across multiple contexts.[20] The CFIR is composed of five major domains: intervention characteristics, outer setting, inner setting, characteristics of the individuals involved and the process of implementation. Eight constructs compose the intervention area (e.g. evidence strength and quality), four constructs are related to outer setting (e.g. patient needs and resources), twelve constructs are related to inner setting (e.g. culture, leadership engagement), five constructs are related to individual characteristics (e.g. knowledge of and belief about the intervention, self-efficacy, individual stage of change, individual identification with organisation, and other personal traits such as ambiguity or motivation) and eight constructs are related to process (e.g. plan, evaluate and reflect).

Curran et al.[21] propose three methods for blending design components of clinical effectiveness and implementation research.

Integrated (single ambulatory) Risk of one single contact, doctor time constraint per single (complicated) visit Single doctor relation, long queue Doctor with broad specialisation Single

For them, an effectiveness-implementation hybrid design is one that uses an a priori dual focus in assessing clinical effectiveness and implementation. The proposed hybrid types are: (i) testing the effects of a clinical intervention on relevant outcomes, while observing and gathering information on implementation; (ii) dual testing of clinical and implementation interventions/strategies; and (iii) testing an implementation strategy while observing and gathering information on the clinical interventionâ&#x20AC;&#x2122;s impact on relevant outcomes.

However, implementation science studies have their challenges.

Both Link4Health and Engage4Health studies,[22,23] conducted in public HIV programmes in Swaziland and Mozambique, faced difficulties in assessing outcomes owing to incomplete electronic records, missing health records and issues of missing data around parameters such as linkage of care, patients no longer in care and mortality data. Nevertheless, implementation science research results in generalisability, and gives relevance to scalable delivery modes.[25] Powell et al.[17} propose further tools for implementation science strategies, and this comprehensive approach can be adapted to the SA context.

Conclusions and next steps In the near future, PLHIV will begin to experience numerous complications. The progressive ageing of the HIV-infected population means that an increasing number of patients will have one or more comorbidities not directly related to HIV infection, and/or correlated with the side-effects of drugs.

The clinical complexity of the HIV-infected patient therefore requires a programme of care that can deal with the medical, psychosocial and functional aspects of the disease, as well as all other complications characteristic of elderly persons.[26] The intensity of treatment (shorter intervals between visits and referrals to multispecialist centres) required for these patients needs to be personalised, through shared pathways between the HIV specialist and other specialists (e.g. cardiovascular and renal).

To manage this trend we suggest two technical solutions: integrated ambulatory care for HIV patients in which doctors and other healthcare professionals manage HIV and the main comorbidities, or different specialised ambulatory care for the main comorbidities.

Additionally, robust policy settings, appropriate infrastructure and improved service capabilities need further attention.[25] Therefore, the evaluation of the patientâ&#x20AC;&#x2122;s clinical picture through the global risk assessment of a particular HIV-associated comorbidity through successive levels of assessment is fundamental, and actions may require specialist intervention.[26] In hospital organisation, the final decision will depend upon the availability of human resources. Doctors, specialised staff and their

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

61

May 2018

ARTICLE attitudes towards the management of these treatments will make all the difference. Hence, managers will play a key role.

Since 2010, in SA, new HIV infections have decreased by 49%, and AIDS-related deaths by 29%.[1] However, to achieve better outcomes, all partners need to optimise the implementation of existing prevention and treatment interventions, and use tools such as the implementation science framework to measure the efficient utilisation of various resources.

Acknowledgements. The authors thank the other members of the research team.

Author contributions. All authors contributed equally.

Funding. None.

Conflicts of interest. None.

1. Joint United Nations Programme on HIV/AIDS. AIDSinfo indicators. Geneva: UNAIDS, 2016. http://aidsinfo.unaids.org (accessed 2 November 2017).

2. Institute for Health Metrics and Evaluation. Global Health Data Exchange. http://ghdx. healthdata.org/ (accessed 3 November 2017).

3. Young F, Critchley JA, Johnstone LK, Unwin NC. A review of comorbidity between infectious and chronic disease in sub-Saharan Africa: TB and diabetes mellitus, HIV and metabolic syndrome, and the impact of globalization. Global Health 2009;5(1):9. https:// doi.org/10.1186/1744-8603-5-9 4. Aboud M, Elgalib A, Kulasegaram R, Peters B. Insulin resistance and HIV infection: A review.

Int J Clin Prac 2007;61(3):463-472. https://doi.org/10.1111/j.1742-1241.2006.01267.x 5. Carr A, Samaras K, Burton S, et al. A syndrome of peripheral lipodystrophy, hyperlipidaemia and insulin resistance in patients receiving HIV protease inhibitors. AIDS 1998;12(7):F51-F58. https://doi.org/10.1097/00002030-199807000-00003 6. Chen D, Misra A. Lipodystrophy in human immunodeficiency virus-infected patients. J Clin Endocrinol Metab 2002;87(11):4845-4856. https://doi.org/10.1210/jc.2002-020794 7. Falutz J. Therapy insight: Body-shape changes and metabolic complications associated with HIV and highly active antiretroviral therapy. Nat Clin Prac Endocrinol Metab 2007;3(9):651-661. https://doi.org/10.1038/ncpendmet0587 8. Grinspoon S, Carr A. Cardiovascular risk and body fat abnormalities in HIV-infected adults.

N Engl J Med 2005;352(1):48-62. https://doi.org/10.1056/nejmra041811 9. Pao V, Lee GA, Grunfeld C. HIV therapy, metabolic syndrome, and cardiovascular risk. Curr Atheroscler Rep 2008;10(1):61-70. https://doi.org/10.1007/s11883-008-0010-6 10. Hanna DB, Ramaswamy C, Kaplan RC, et al. Trends in cardiovascular disease mortality among persons with HIV in New York City, 2001 - 2012. Clin Infect Diseases 2016;63(8):11221129. https://doi.org/10.1093/cid/ciw470 11. Schouten J, Wit FW, Stolte IG, et al. for the AGEhIV Cohort Study Group. Cross-sectional comparison of the prevalence of age-associated comorbidities and their risk factors between HIV-infected and uninfected individuals: The AGEhIV cohort study. Clin Infect Diseases 2014;59(12):1787-1797. https://doi.org/10.1093/cid/ciu701 12. Julius H, Basu D, Ricci E, et al. The burden of metabolic diseases amongst HIV positive patients on HAART attending the Johannesburg Hospital. Curr HIV Res 2011;9(4):247-252. https://doi.org/10.2174/157016211796320360

62

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

13. Atun R, Davies JI, Gale EAM, et al. Diabetes in sub-Saharan Africa: From clinical care to health policy. Lancet Diabetes Endocrinol 2017;5(8):622-667. https://doi.org/10.1016/ S2213-8587(17)30181-X 14. Brouwer ED, Watkins D, Olson Z, Goett J, Nugent R, Levin C. Provider costs for prevention and treatment of cardiovascular and related conditions in low- and middle-income countries: A systematic review. BMC Public Health 2015;15(1):1183. https://doi. org/10.1186/s12889-015-2538-z 15. Gaziano TA, Steyn K, Cohen DJ, Weinstein MC, Opie LH. Cost-effectiveness analysis of hypertension guidelines in South Africa: Absolute risk versus blood pressure level.

Circulation 2005;112(23):3569-3576. https://doi.org/10.1161/circulationaha.105.535922 16. Sturke R, Harmston C, Simonds RJ, et al. A multi-disciplinary approach to implementation science: The NIH-PEPFAR PMTCT implementation science alliance. J Acquir Immune Defic Syndr 2014;67(Suppl 2):S1-S3. https://doi.org/10.1097/qai.0000000000000323 17. Powell BJ, Waltz TJ, Matthew JC, et al. A refined compilation of implementation strategies: Results from the Expert Recommendations for Implementing Change (ERIC) project.

Implement Sci 2015;10:21. https://doi.org/10.1186/s13012-015-0209-1 18. Bonham AC, Solomon MZ, Mittman B, et al. Implementation Science and Comparative Effectiveness Research, Comparative Effectiveness Research in Health Services. Part of the series Health Services Research. New York: Springer US, 2016:181-203. http://doi. org/10.1007/978-1-4899-7600-0_11 19. Dearing JW, Kee KF. Historical roots of Dissemination and Implementation Science. In: Brownson RC, Colditz GA, Proctor EK, eds. Dissemination and Implementation Research in Health: Translating Science to Practice. Oxford: Oxford University Press, 2012.

20. Dissemination and Implementation Research in Health: Translating Science to Practice. Oxford Scholarship Online, 2012. https://doi.org/10.1093/acprof:oso/9780199751877.003.0003 21. Damschroder LJ, Aron DC, Keith RE et al. Fostering implementation of health services research findings into practice: A consolidated framework for advancing implementation science. Implement Sci 2009;4:50 https://doi.org/10.1186/1748-5908-4-50 22. Curran GM, Bauer M, Mittman B, et al. Effectiveness-implementation hybrid designs: combining elements of clinical effectiveness and implementation research to enhance public health. Med Care 2012;50(3):217-226. https://doi.org/10.1097/ MLR.0b013e3182408812 23. McNairy ML, Lamb M, Gachuhi AB, et al. Evaluation of the effectiveness of a combination strategy on linkage and retention among HIV positive individuals in Swaziland: The Link4Health Study. PLOS Med 2017;14(11):e1002420. https://doi.org/10.1371/journal. pmed.1002426 24. Elul B, Lamb MR, Lahuerta M, et al. A combination intervention strategy to improve linkage to and retention in HIV care following diagnosis in Mozambique: A cluster-randomized study. PLOS Med 2017;14(11):e1002433. https://doi.org/10.1371/journal.pmed.1002433 25. Barnabas RV, Celum C. Closing the gaps in the HIV care continuum. PLoS Med 2017;14(11):e1002443. https://doi.org/10.1371/journal.pmed.1002443 26. Australian Healthcare & Hospitals Association. Integrated Healthcare: Policy Pathways and Pitfalls. AHHA, 2014. https://ahha.asn.au/sites/default/files/docs/policy-issue/ahha_ integrating_care_-_policy_pathways_and_pitfalls_1.pdf (accessed 28 November 2017).

27. Andreoni M. Comorbidities in HIV-infected patients in reference to the new Italian guidelines.

Cure 2016;165-175. https://www.cure-journal.com/wp-content/uploads/2016/04/CURE_ n5_Andreoni.pdf (accessed 1 March 2018).

Accepted 6 March 2018.

ARTICLE

Hypothesis tests for the difference between two population proportions using Stata B V Girdler-Brown, FCPHM, FFPH, Hons BComm (Econ); L N Dzikiti, MSc School of Health Systems and Public Health, Faculty of Health Sciences, University of Pretoria, South Africa Corresponding author: B V Girdler-Brown (brendangirdlerbrown@gmail.com)

This educational article outlines the main methods available, using Stata statistical software, for testing hypotheses about the equality of population proportions, using sample-derived data. The article focuses on how to select the most appropriate test to use, the relevant Stata statistical software commands and the interpretation of the Stata output obtained following these commands. Both single-sample and two-sample hypothesis tests are covered.

South Afr J Pub Health 2018;2(3):63-68. DOI:10.7196/SHS.2018.v2.i3.71

In the previous edition of the journal we presented an overview of hypothesis testing for the difference between two population means, using Stata (StataCorp, USA) statistical software. In that article, we dealt with numerical data.[1] For those wishing to read further at this introductory level we recommend the text by Pagano and Gauvreau.[2] In this article, we will give a similar overview, but for testing the difference between two population proportions. We will be dealing with binary variables where, at an individual level, a characteristic is either present (coded as 1) or absent (coded as 0). For each individual there is a characteristic of interest/outcome variable, such as lung cancer, with only two possible states, namely lung cancer present or lung cancer absent.

There is also, for hypothesis testing, a second classifying/ exposure variable, which is also binary and coded 1 or 0. This second variable identifies the two groups that must be compared.

An example might be smoker/non-smoker, etc.

The data would be laid out as in Table 1, for a sample of 100 study participants (lung cancer coded as 1 if present, 0 if absent; smokers coded as 1 if a smoker, 0 if a non-smoker).

Table 1. Long-format data entry for proportions Participant_id 1 2 3 4

Lung_cancer 1 1 0 0

Smoker 1 1 0 1

…

…

…

79 80

0 1

1 0

We might summarise the information available from this table into counts of participants in a 2 × 2 contingency table (Table 2).

From Table 2 one can see that, in this sample of 100 people, the proportion of smokers with lung cancer is 15/45 (0.3), while the proportion of non-smokers with lung cancer is 5/55 (0.09). The proportion with lung cancer appears to be higher among smokers than non-smokers.

However, this apparent difference may be due to a sampling error. If I were to draw a different sample of 100 people at random, then the same difference might not be observed in the second sample.

The hypothesis test would involve the relationship between the outcome state and the exposure state. For example, the hypothesis may be articulated as an answer to the question: ‘Is there a difference in the proportions of smokers and non-smokers who develop lung cancer?’ An appropriate null hypothesis would be that the proportions that develop lung cancer (πsmokers and πnon-smokers) are equal, and may be written in three different ways: H0: πsmokers = πnon-smokers; or; H0: πsmokers - πnon-smokers = 0; or H0: πsmokers / πnon-smokers = 1 (Note the use of the Greek π in these null hypothesis statements.

This reminds us that the hypothesis test is testing a hypothesis about the study population parameters from which the samples have been drawn.)

Table 2. Contingency table (2 × 2) for lung cancer cases/noncases by smoking status

Smoking history?

Total

Yes No

Lung cancer present?

Yes No 15 30 5 50 20 80

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

100

63

May 2018

ARTICLE Scope of the article The foci of this article are on the selection of the most appropriate test; the Stata statistical software commands to use in order to perform the test; and the interpretation of Stata output for the test.

We have assumed that the reader possesses an understanding of the principles of statistical hypothesis testing.

In this example, only the logit transformation method yields a 90% CI that does not include the null value of 0.78; the normal approximation and binomial exact methods produced 90% CIs that overlap with the null value of 0.78. Hence we would often fail to reject the null hypothesis using these methods, when we should, in fact, have rejected the null hypothesis.

A single-sample test, large sample size

Single-sample test, small sample size

A single-sample hypothesis test involving a proportion would involve, for example, the comparison of a sample-based population proportion estimate with a given gold standard or target.

For example, in 2018, there may be a target of 78% for voluntary HIV testing among pulmonary tuberculosis (TB) patients who make use of public sector health facilities, and who do not have a record of a previously positive HIV test result.

Official surveillance data in a rural district might show that this target has been met (or exceeded). However, a researcher might want to perform a study to determine the coverage that is based on carefully collected and verified information. The researcher could draw a simple random sample of 200 patients listed in the district’s TB register, and then look for laboratory confirmation of the testing that has taken place. (S)he finds that 73% of the patients in her sample have in fact had an HIV test performed during the course of their anti-TB treatment. This result, 73%, is clearly below the target of 78%. However, could this difference be due to sampling error?

The null hypothesis (the null value would be the gold standard, 0.78, since this is a single-sample test) is: H0: πtested = 0.78 (if one is interested in any difference from 0.78, either <0.78 or >0.78, a so-called ‘two-tail’ test); or H0: πtested ≥0.78 (if one is only concerned about the possibility that the target has not been met, a so-called ‘single-tail’ test).

When the conditions that nπ and n(1-π) are both ≥5 have not both been met, then one may peform the binomial test (or ‘bitest’) in Stata. This test is based on the expected and observed number of successes for a given number of trials. The lower the number of trials, the more poorly the result of this test will compare with the prtest. Stata provides the calculated p-values for the observed number of successes under the null hypothesis. As the number of trials increases, the bitest and prtest will produce similar results for the p-values, even though the bitest results are estimated from whole numbers of successes, while the prtest results are obtained from the proportion treated as a continuous variable. There is no CI obtained from the bitest.

If the sample size is large enough such that np and n(1-p) are both ≥5, then one may use a single sample z-test to test these null hypotheses. In Stata, this z-test is called a ‘prtest’, and the same command is used for both a single-sample test and a two-sample test. There is also an immediate command, ‘prtesti’, that may be used when the data are not already entered in the usual 1/0 format.

The Stata output provides a confidence interval (CI) that is wholly derived from the sample information, and a z-score p-value that is derived on the assumption that the null hypothesis is true.[3] This p-value should be used to decide on statistical significance. There may be discrepancies between this p-value and the CI. For example, assume that one has a sample proportion of 0.73 (sample size = 200) for a null hypothesis that π = 0.78. Assuming a single-tail hypothesis test, the p-value is the probability that a random sample size of 200 will have a proportion of 0.73 or less if the null hypothesis is true. The value of p is found to be 0.044. Hence one rejects the null hypothesis (if α = 0.05).

However, the 90% CI (90% since we are dealing with a single-tail test and want to know the upper and lower significance levels for rejection) comes to 0.678 - 0.782 (normal approximation method) or 0.674 - 0.781 (exact binomial method). The 90% CI comes to 0.674 - 0.779 if one uses the logit transformation method to estimate the 90% CI.

64

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

Two-sample tests Suppose we have collected data from non-pregnant female patients with listeriosis, and from healthy non-pregnant female controls.

We find that 45/100 of the patients indicated that they had eaten uncooked polony during the 2 weeks prior to the onset of symptoms. Sixty of 200 controls also indicated that they had eaten uncooked polony during the 2 weeks before being interviewed.

Consider the set of results in Table 3 (fictitious data). It can be seen that the proportion of those with listeriosis who ate polony is 0.45 (45/100, 45%) The proportion who ate polony among the controls is 0.30 (60/200, 30%).

Is there a statistically significant difference between these two proportions? Might the difference that we observe be due to sampling error?

There are two variables in this situation, and both are binary. We are interested in the proportion of those classed as listeriosis patients who have a history of polony consumption, v. the proportion of those who do not have listeriosis and who have a history of polony consumption.

There are three main ways in which these questions may be addressed in Stata. The first is a two-sample prtest (which, as pointed out, is Stata’s name for a z-test of two proportions), the second is a χ2 test and the third is Fisher’s exact test.

The prtest The two-sample prtest is based on the assumption that both samples are ‘large’ (i.e. that np and n(1-p) are both ≥5 for each Table 3. Fictitious data for listeriosis v. polony eaten (casecontrol study)

Polony eaten Totals

Yes No

Yes 45 55 100

Listeriosis No 60 140 200

Totals 105 195 300

ARTICLE of the samples, where n is the sample size and p is the sample proportion with the outcome of interest). One should only perform a prtest if these large sample conditions are met. The reason for this is that the prtest is a normal approximation test that treats the mean of the 1 and 0 values as if it were a continuous variable with a normal probability sampling distribution.

This is only approximately acceptable if the sample sizes are sufficiently large. The null hypothesis is either: H0: πlisteriosis = πcontrols ; or H0: πlisteriosis - πcontrols = 0 The two-sample prtest will yield a p-value as well as a CI for the difference between the two proportions.

The χ2 test for independence, two binary variables One of the other two alternative tests that are available in Stata is the χ2 test. This test is performed on the count data in the 2 × 2 contingency table illustrated in Table 3. Table 3 contains the actual count data (whole numbers) in each cell. It also shows the row and column totals. In the χ2 test, the null hypothesis is that the exposure variable (eating polony) observed counts are independent of the outcome variable (listeriosis v. control) observed counts.

To do this, Stata first estimates what the expected cell values would be if the null hypothesis is true. These expected cell values are estimated using the observed row and column totals as a given, and then expected cell values are assigned using simple probability theory. For the χ2 test result to be valid for a 2 × 2 table, all these calculated expected cell values should be ≥5. In Stata, one is able to request that Stata show the expected cell values, so that one can then check and ensure that this important condition has been met.

If one or more of the expected cell values in a 2 × 2 table is/are <5, then one should not rely on the p-value that Stata has presented.

Instead, one should perform Fisher’s exact test on the data. The Stata commands for both these tests are presented below.

For large sample situations the prtest, χ2 test and Fisher’s exact test will all give very similar p-values. For smaller samples, the χ2 and Fisher’s exact tests will usually agree fairly well; for very small samples with an expected cell value <5, the Fisher’s exact result will be quite different from that for the χ2 test, and the Fisher’s exact test p-value should be used. Where the conditions for a prtest are not met, it is recommended that the χ2 (or Fisher’s exact) test be used rather than the prtest.[4] One of the drawbacks of using either the χ2 test or the Fisher’s exact test is that, while one obtains a p-value, one does not obtain a CI for the difference between the two proportions.

Performing the analyses using Stata statistical software Data layout Irrespective of whether one is performing a prtest, a χ2 test or a Fisher’s exact test, there will be an outcome variable and an exposure variable.

Both should be coded as 1 (factor present) or 0 (factor absent).

Stata commands (given between < and >; when typing the command omit < and >) 1. For single-sample tests: The following commands are presented for the single-sample prtest, where GS = the gold standard or target against which you are comparing actual performance. The output will give p-values for both single-tail and two-tail tests. Remember that np and n(1 – p) must both be ≥5. <prtest variable = GS> Where variable is the name of the 1/0 variable and GS is the gold standard or target proportion.

Should you require a 90% CI instead of the default 95% CI then use: <prtest variable = GS, citype(90)> Should you wish to use the immediate command: <prtesti n p GS> Where n is the sample size, p is the sample proportion (between 0 and 1) and GS is the gold standard proportion. Again, you may add in ‘… citype(90) after the GS if you want to obtain a 90% CI. If nπ and/ or n(1 – π) is <5, then one may no longer use the prtest. Instead, one makes use of the binomial test. Only whole numbers are allowed. <bitest variable==GS> Where variable is the name of the binary variable coded as 1/0 and GS is the gold standard proportion (between 0 and 1). Please note the use of the double equal sign for this command.

For the immediate command: <bitesti trials successes GS> Where ‘trials’ is the sample size (a whole number), successes is the number of those who have the outcome of interest (also a whole number) and GS is the gold standard proportion (between 0 and 1).

2. The commands are presented for the two-sample prtest, with data stored in the long format (i.e. one binary variable indicating the presence or absence of the outcome of interest, and another indicating, for each participant, which comparison group that person belongs to).

The output will show p-values as well as 95% CIs for the difference between the population proportions of the two comparison groups.

Again, remember that n1p1; n1(1 – p1); n2 p2 and n2 (1 – p2) must all be ≥5 (n1 and p1 refer to the numbers and proportions in the first comparison group; n2 and p2 do the same for those in the second comparison group). <prtest variable1, by(variable2)> Where variable1 is the outcome variable and variable2 is the group identifier variable for the groups being compared. The immediate command is: <prtesti na pa nb pb> Where na and pa refer to the number of people in group A and the proportion with the outcome; and nb and pb refer to the number and proportion in group B. 3. Next, the commands are presented for the two-sample prtest with data stored in the wide format. <prtest variable_a = (variable_b)> Where variable_a is the binary (1/0) outcome measure for group

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

65

May 2018

ARTICLE A, and variable_ is the binary (0/1) outcome measure for those in group B. 4. The commands for the χ2 test in Stata are as follows (data must be in the long format for this command): <tab variable1 variable2, chi2> Where variable1 is the outcome variable and variable2 is the group identifier variable for the groups being compared.

Should you wish to also see the expected cell count values in order to decide whether to rather perform Fisher’s exact test (if any one or more of the expected cell values is/are <5), then use the following command. This command will give you the expected value cell counts as well as the χ2 test result: <tab variable1 variable2, expected chi2> The corresponding immediate commands are: <tabi a b \ c d, chi2> and <tabi a b \ c d, expected chi2> Where a, b, c, and d represent the cell counts for the 2 × 2 table (Table 4).

5. Should you decide, after finding that an expected cell value is <5 that you would prefer to perform Fisher’s exact test, then simply substitute ‘exact’ for ‘chi2’ in any of the above commands. In fact, Stata allows one to ask for all the results in a single step, and then one can just decide which test to rely on and which p-value to use, without having to repeat the commands. The following command, for example, would yield a great deal of information in a single step: <tab variable1 variable2, expected chi2 cchi2 exact> This would result in the following information being made available: the expected cell values, the χ2 p-value, the individual cell χ2 values, and the exact test p-values. The immediate command equivalent would be: <tabi a b \ c d, expected chi2 cchi2 exact>

Stata version 14 outputs

hypothesis that the proportion tested is >0.78. The Ha: p!=0.78 shows the p-value for a two-tail test of the alternative hypothesis that the proportion tested is not equal to 0.78.

Since we are only interested in/concerned about the possibility that the proportion tested fails to meet the target of 0.78 we would concern ourselves with the single-tail test result, p=0.044. We would then reject the null hypothesis and conclude that we have probably failed to reach the target of 78%.

Example 2: Single-sample bitest (small samples) Let us assume that, instead of a sample of 200 as we had in example 1, we only had a sample of 18 TB patients. The records show that 14 of these patients had undergone HIV testing. Have we met the target of 78% tested?

Recall that, for the prtest to be valid, we must meet the condition that np and n(1 – p) must both equal or exceed 5 (n is the sample size and p is the proportion that were tested). In this case, np = 18 × (14/18) = 14; n(1 – p) = 18 × (4/18) = 4. Clearly the conditions required for the prtest have not been met. We therefore resort to the binomial test.

The output in Fig. 2 has been obtained from Stata for the binomial test carried out using the immediate command option (numbers are small so this will be the most common situation).

If you had these data entered in Stata at the individual level as 1s and 0s, and if you called this variable ‘tested’, then the following Stata command may be used to obtain the same results: <bitest tested=0.78> Notice that there are no CIs presented.

In Fig. 2, ‘k’ is the number of successes. Stata has used the formula for calculating binomial probabilities for different numbers of successes from 18 trials if the null hypothesis of p=0.78 is true.

We see that if H0 is true, then the probability of obtaining 13 or fewer success is 0.361. Clearly we have no grounds in this case to reject the null hypothesis. The deviation from 78% success could easily be due to a sampling error.

(NB Stata version 15 outputs will be almost identical).

Example 1: Single-sample prtest (large samples) Using the immediate command for the TB HIV testing example with 200 TB patients, 146 (73%) of whom had undergone HIV testing, the following output was obtained given a target of 78% (0.78) (Fig. 1).

If the individual level data had been entered into Stata as 1s (tested) and 0s (not tested), and if the variable has the name ‘tested’ then the Stata command would be: <prtest tested=0.78> The Ha: p<0.78 shows the p-value for a single-tail test of the alternative hypothesis that the proportion tested is <0.78. The Ha: p>0.78 shows the p-value for a single-tail test of the alternative Table 4. A generic 2 x 2 contingency table

Exposed?

66

May 2018

Yes No

Outcome?

Yes No a b c d

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

. prtesti 200 .73 .78 One-sample test of proportion Variable

Mean

x

.73

x: Number of obs =

[95% Conf. Interval]

.0313927

.6684715

p = proportion(x) Ho: p = 0.78

.7915285 z =

Ha: p < 0.78 Pr(Z < z) = 0.0439

200

Std. Err.

Ha: p != 0.78 Pr(|Z| > |z|) = 0.0878

-1.7070

Ha: p > 0.78 Pr(Z > z) = 0.9561

Fig. 1. Stata output for a single-sample test of proportion. . bitesti 18 13 .78 N 18

Observed k 13

Expected k 14.04

Pr(k >= 13) = 0.813374 Pr(k <= 13) = 0.361298 Pr(k <= 13 or k >= 16) = 0.569714

Assumed p

Observed p

0.78000

0.72222

(one-sided test) (one-sided test) (two-sided test)

Fig. 2. Stata output for a binomial test (immediate command).

ARTICLE Notice as well that for the binomial probabilities these are worked out for whole numbers of successes only. It is not possible to have, say, 13.3 people vaccinated.

As the expected value vaccinated under the null hypothesis is 14.04 (0.78 × 18) this should be rounded down to ≤13 (14 – 1) for those results falling below the expected value and rounded up to 16 or more (15 + 1) for those results exceeding the expected value.

Example 3: Two-sample prtest (large samples) The Stata command and output in Fig. 3 was obtained using the case-control study data summarised in Table 3.

The 95% CI for the difference between the two proportions (–0.26 to –0.03) was calculated by Stata using the sample difference (–0.15), the standard error for the sample difference and the normal approximation.

The p-value (0.0102 for the two-tail test option) was calculated on the assumption that the null hypothesis is true (i.e. that the true difference in the proportions = 0).

Should you not have the data entered into Stata as an individual level 1/0 variable, then the immediate command in Fig. 4 would obtain the same results as those presented above (you would first need to work out the proportions with listeriosis among the polony-eating group and the non polony-eating . prtest listeriosis, by(polony) Two-sample test of proportions

0: Number of obs = 1: Number of obs =

Variable

Mean

0 1

.2820513 .4285714

Std. Err. .032225 .0482945

diff

-.1465201 under Ho:

.0580587 .0570614

z

-2.57

P>|z|

0.010

[95% Conf. Interval] .2188914 .3339159

.3452112 .523227

-.2603131

-.0327272

diff = prop(0) - prop(1) Ho: diff = 0 Ha: diff < 0 Pr(Z < z) = 0.0051

z =

Ha: diff != 0 Pr(|Z| < |z|) = 0.0102

. tab listeriosis polony, expected chi2 Key frequency expected frequency

Consumed polony in previous two weeks 0 1

Total

0

140 130.0

60 70.0

200 200.0

1

55 65.0

45 35.0

100 100.0

Total

195 195.0

105 105.0

300 300.0

Pearson chi2(1) =

6.5934

-2.5678

Ha: diff > 0 Pr(Z > z) = 0.9949

Fig. 3. Stata output for a two-sample test of proportions.

Listeriosi s case

195 105

Pr = 0.010

Fig. 4. Stata output for a χ2 test of independence (large sample).

. tab tb hiv, expected chi2 exact Key frequency expected frequency

Tuberculos is

HIV sero-status 0 1

Total

0

7 4.2

6 8.8

13 13.0

1

5 7.8

19 16.2

24 24.0

Total

12 12.0

25 25.0

37 37.0

Pearson chi2(1) = Fisher's exact = 1-sided Fisher's exact =

4.1937

Pr = 0.041 0.067 0.048

Fig. 5. Stata output for a small-sample Fisher’s exact test.

group): <prtesti 195 0.2821 105 0.4286>

Example 4: Chi square (χ2) test The Fig. 4 output was obtained from Stata for the listeriosis and polony data from Table 3. The output was obtained using the names of the data variables with the data entered into a Stata data set at the individual level.

Firstly, notice that the expected cell values (130, 70, 65 and 35) are all >5. We are therefore comfortable using a χ2 square test.

Secondly, there is no 95% CI presented for the difference between the two group proportions. This is because the null hypothesis is that polony consumption and listeriosis are independent of each other.

With this large sample situation, the p-value is almost identical to that obtained from the prtest.

Example 5: Fisher’s exact test Stata output is now presented (Fig. 5) for a sample where one of the expected cell values is <5. This means that we should be cautious about the χ2 results, as they may be misleading. We would rather make use of the Fisher’s exact test results in this case.

As an aside, the general approach is that if >10% of expected cell values are <5, then the χ2 results may be misleading. However, not all people accept this guideline. In the case of a 2 × 2 table, such as that illustrated in the output displayed in Fig. 5, one cell makes up 25% of all the cells. In such a case, a low expected cell value affecting only one cell would be reason to prefer the Fisher’s exact test result.

The χ2 p-value is 0.041, suggesting statistical significance. That for the Fisher’s exact test is 0.067, suggesting statistical nonsignificance.

In the past, many statisticians have made use of the Yates correction factor for discontinuity, especially in cases where numbers are small. This correction factor is not available with Stata. Nowadays, the trend is to use the Fisher’s exact test, rather than invoking the Yates correction.[5] Prior

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

67

May 2018

ARTICLE to the advent of desktop statistical software programmes, when many statistical analyses were done by hand, the Fisherâ&#x20AC;&#x2122;s exact test proved burdensome to use, and alternative approximate methods were popular, but they are rarely used any more. Note here that the expected value for the upper left-hand cell is 4.2 (<5). Therefore it is preferable to rely on the Fisherâ&#x20AC;&#x2122;s exact p-value.

1. Dzikiti LN, Girdler-Brown BV. Parametric hypothesis tests for the difference between two population means. Strengthen Health Syst 2017;2(2):40-46. https://doi.org/10.7196/ SHS.2017.v2.21.60

68

May 2018

SOUTHERN AFRICAN JOURNAL OF PUBLIC HEALTH

2. Pagano M, Gauvreau K. Principles of Biostatistics, 2nd ed. Pacific Grove: Duxbury, 2000.

3. Gauvreaux K. Hypothesis testing proportions. Circulation 2006;114(14):1545-1548. https:// doi.org/10.1161/circulationaha.105.586487 4. Rosner B. Fundamentals of Biostatistics, 6th ed. Belmont: Thomson Brooks/Cole, 2006.

5. Lydersen S. Statistical review: Frequently given comments. Ann Rheum Dis 2015;74(2): 323-325. https://doi.org/10.1136/annrheumdis-2014-206186

Accepted 12 March 2018.