Water, sanitation and hygiene (WASH) are important for health and wellbeing. In Timor-Leste, the rates of access to WASH are some of the lowest in the world, particularly in remote areas. The national government’s response to improving sanitation in remote areas is a community-based approach called ‘PAKSI’ (Planu Aksaun Komunidade, Saneamentu no Ijiene) which is also implemented by non-governmental organisations. The research documented in this thesis sought to understand the implementation and outcomes of the PAKSI component of the WaterAid program in Manufahi Municipality, Timor-Leste from 2013 to 2014 as part of the ‘WASH for Worms’ cluster-randomised control trial. PAKSI is based on Community-Led Total Sanitation (CLTS): an approach which aims to inspire rural communities to lead themselves to become open defecation free (ODF) through triggering collective feelings of disgust and shame about open defecation. CLTS is popular; however, there are few rigorous studies about the approach, especially in the context of Timor-Leste. Three studies were conducted as part of this research to better understand PAKSI: the Timor-Leste adaptation of CLTS. The first study compared the official descriptions of PAKSI (and CLTS) to the way it was implemented and described by 19 representatives of the government, WaterAid and intervention communities as well as Dr Kamal Kar and Dr Robert Chambers. The findings suggested that, officially, PAKSI and CLTS are quite similar; however, the implementation of PAKSI in the research sites diverged from these official descriptions in that the approach was more facilitator-led than community-led. The second study used quantitative methods with 65 households to measure the changes in sanitation-related knowledge, attitudes and practices after the WaterAid program. Whilst toilet coverage increased over the course of the study, and open defecation decreased, the proportion of participants who reported feeling disgusted at the sight of faeces in the open did not change. This suggests that a) the WaterAid PAKSI intervention did not trigger feelings of disgust (which is what PAKSI and CLTS are officially purported to do) and b) whilst disgust might be a necessary motivator for becoming ODF, on its own it is insufficient. The third study explored the possible factors influencing whether households become or continue to be ODF, through interviews with 21 participants including community leaders, household representatives and implementing staff. The experiences of participants who had a disability, an age-related impairment or were from female-headed households were sought to be included. Several factors were identified, including competing priorities for finances and other resources, and participants feeling like they were doing what they had been ‘told to do’. The 17 behavioural factors in the RANAS model of behaviour change provided a framework for analysing what motivates people in remote parts of Timor-Leste to change their sanitation behaviour. The findings suggest that motivators other than disgust and embarrassment, such as affordability of toilets and ‘being told what to do’, are relevant. Further research should investigate whether other approaches such as government regulation and strengthened sanitation markets are effective for bringing about ODF communities in Manufahi Municipality, Timor-Leste.

Disability disaggregation of education management information systems (EMIS) is vital to inform policies and resourcing for disability-inclusive education and evaluate progress towards targets. The approach to disaggregation must use a valid and reliable method for identifying children with disabilities. The UNICEF/Washington Group Child Functioning Module (CFM) is recommended by the United Nations for identifying children with disabilities and has been recommended for disaggregating education program data by disability. In the context of an education sector support program in Fiji, this research aimed to validate a method for disaggregating Fiji’s EMIS by disability. A cross-sectional diagnostic accuracy study was undertaken in which teacher and parent CFM responses for 472 primary-aged students were compared to reference standard clinical assessments in five domains: vision, hearing, musculoskeletal, speech and cognition. Receiver operating characteristic curves (depicting the trade-off between sensitivity and specificity) were constructed and optimal cut-off points and inter-rater reliability were assessed. Nested survey data on learning and support needs were analysed to explore whether combining CFM data on activity and participation data with data on environmental factors related to LSN (educational adjustments, assistive technology and personal assistance requirements) more accurately identifies children with disabilities. The study produced a range of novel findings. Diagnostic accuracy of parent observations related to seeing, walking and speaking was stronger than that of teachers, however teacher accuracy was very acceptable. Conversely, for cognitive domains teacher accuracy was far stronger than parents. The CFM domains seeing, hearing, walking and speaking showed “good” to “excellent” accuracy, however remembering and focusing attention showed only “fair” to “poor” accuracy. The domain learning was “good” with teachers as respondents, but only “fair” with parent respondents. As a whole, the CFM had “fair” accuracy (area under the Receiver Operating Characteristic curve: 0.763 parent responses, 0.786 teacher responses). Severe impairments were reported relatively evenly across CFM response categories “some difficulty”, “a lot of difficulty” and “cannot do at all”. If the cut-off level for identifying children with disabilities were “a lot of difficulty”, nearly 40% of children with moderate clinical impairments and 28% of children with severe impairments would miss out on services as they were reported as having “some difficulty”. On the other hand, the rates of false positives would be very high if the cut-off “some difficulty” were used. Combining data from the CFM with LSN data shows potential to increase the accuracy of domain-specific disability identification and, crucially, identification of children with disabilities amongst those reported as having “some difficulty” on the CFM. The CFM alone is not accurate enough for the purpose of disaggregating Fiji’s EMIS by disability. The choice of cut-off level and the mixture of severity of impairments reported across response categories are particular challenges for the CFM. Combining CFM data with data on educational adjustments, assistive technology and personal assistance requirements could improve disability identification accuracy. Follow-up verification visits are required to confirm funding eligibility due to inherent risks of tools based on self-report.

For Indonesian women and men from all social strata, the milestones of heterosexual marriage and parenthood are not only symbolic of the transition to adulthood, but also represent the path to an idealised, normative future, otherwise known as the ‘happily ever after’. This thesis explores the happily ever after trajectories of 20 predominantly middle class, Muslim women in urban Yogyakarta, Indonesia, with a focus on their reproductive experiences and decisions. Specifically, it explores women’s negotiations of family planning, infant feeding and work. The ethnographic research that informs this thesis involved in-depth interviews with 20 women; focus group discussions with seven primary health workers and seven breastfeeding counsellors; semi-structured interviews with five key community stakeholders; and participant observation. Fieldwork was undertaken over 18 months and employed a grounded theory approach to generate key themes for analysis. In exploring the themes that emerged as most important to my women informants, I deployed the concepts of reproductive agency, reproductive modernity and reproductive politics. Key findings highlight how women’s reproductive agency in relation to family planning, infant feeding and juggling reproductive work with paid work or study may be both supported and constrained in their everyday negotiations within their intimate circles and with health workers, employers and educational institutions. The thesis also theorises how reproductive modernity is intrinsic to women’s reproductive experiences and decisions, especially in relation to their preferences for non-hormonal and non-biomedical contraceptive methods. It also critiques the formal political agendas that intersect with women’s attainment of happily ever after, by informing their decisions to breastfeed, their balancing of reproductive and productive work or study, and their family planning negotiations. This thesis presents a new analytical framework for understanding women’s reproductive experiences and decisions at multiple levels, including their interpersonal relationships, societal norms and formal reproductive politics (the WRED framework). The application of the WRED framework demonstrates its utility in the development of multi-level, translational recommendations that take into account the complexity of factors influencing women’s reproductive health and reproductive agency. The thesis addresses a notable gap in research into the reproductive experiences and decisions of middle class women in Indonesia and extends prior research on family planning in the Indonesian context. It adds to recent scholarship challenging the biomedical categorisation of certain contraceptives as ‘modern’ and others as ‘traditional’. The thesis expands feminist critiques of breastfeeding promotion and unpacks the effects of discourses of morality in relation to women’s reproductive rights in Indonesia. It also expands the concept of reproductive agency by considering a form of collective agency for middle class women seeking to juggle their reproductive and productive roles. This thesis concludes that the path to happily ever after for the middle class Indonesian women in this study was only partially attained through marriage and parenthood. There were many other essential ingredients contributing to women’s happily ever afters, including post-secondary education and a career; more companionate, equal marriages; contraceptive choice and a preference for reproductive health over fertility control; and pragmatic religious piety.

Type 2 Diabetes Mellitus (T2DM) is recognised as one of the most challenging public health problems of the 21st century with the number of people with T2DM rapidly increasing worldwide. While the efficacy of interventions in preventing and improving T2DM outcomes has been well established, there has generally been quite poor uptake, implementation and translation in populations, particularly in Low and Middle-Income Countries (LMICs). This thesis aims to evaluate the implementation of behavioural interventions for the prevention and self-management of T2DM and to identify the key factors required to improve their uptake in real-world settings. This Ph.D. thesis includes a systematic review of real-world diabetes prevention programs implemented in the last 15 years as well as a comprehensive implementation evaluation of two cluster randomised controlled trials: 1) the Kerala Diabetes Prevention Program (K-DPP), India and 2) the Australasian Peers for Progress – Diabetes Project (AusPfP-DP), Australia. This thesis utilises two existing evaluation frameworks (RE-AIM and PIPE) and has developed a novel evaluation logic model to identify the provider-, participant- and community-level factors that are most relevant for the successful implementation of real-world diabetes programs. The provider-level factors are further investigated. The systematic review identified that program intensity plays a major role in weight loss outcomes. However, low-intensity interventions that have high uptake—both in terms of good coverage of invitees and their willingness to accept the invitation—can still have considerable impact in lowering diabetes risk in a population, even when effectiveness in weight loss is low or moderate. The findings from the implementation evaluation of K-DPP demonstrated that the intervention was feasible and acceptable in changing lifestyle behaviours in high-risk individuals. The intervention was delivered as per protocol and there was high adherence to the intended design and delivery of the intervention by peer leaders. The communities’ trust in the local partnering institute and the commitment of the local political leaders were important in facilitating the high uptake of the program. The findings of the implementation evaluation of AusPfP-DP show that peer support programs are feasible, acceptable and can be used to supplement treatment for patients motivated to improve behaviours related to diabetes. However, program planners need to focus on the participation component in designing future programs. Finally, the use of the two evaluation frameworks allowed a comprehensive evaluation of a diabetes prevention and self-management program, from the provider-, participant-, community- and public health- perspectives. The research contained in this Ph.D. thesis highlights several important findings required to improve the uptake and wider implementation of diabetes prevention and self-management lifestyle interventions in real-world settings. Future implementation research needs to identify effective recruitment and program implementation strategies for targeting both reach and program participation. Rigorous evaluation should be a central feature of the implementation of prevention and self-management programs. More rigorous reporting of program components is needed to evaluate programs to assess the practical value of the diabetes interventions.

Background In recent decades, the prevalence of cardiovascular diseases (CVDs) around the world has been increasing, and researchers have paid more attention to understand the epidemiology of CVDs to provide reference for healthcare professionals and policy makers. China is one of the countries where rapid development of economy occurs accompanied by changes in people’s socioeconomic status and lifestyle, and is also facing the challenge of increasing burden of CVDs. This challenge exists not only in the rising prevalence of CVDs and increasing deaths from CVDs, but also in the low awareness and treatment rates of some CVDs. This Master research will focus on one important risk condition of CVDs, abnormal blood lipid levels or called dyslipidaemia, as a window to help understand more about the cardiovascular epidemics in China in recent years, based on secondary data analysis of a representative cohort study called China Health and Retirement Longitudinal Study. (CHARLS). Objectives This Master research aims to provide a reliable estimation of the prevalence, awareness, treatment and control rates of dyslipidaemia in middle-aged and elderly population in China, and analyse their related influential factors. Methods This research includes two phases of analysis: Phase 1: A cross-sectional study of the CHARLS baseline survey to investigate the prevalence, awareness, treatment and control rates of dyslipidaemia, using the blood test data from CHARLS. Phase 2: A cohort study of the CHARLS baseline and follow-up surveys to estimate the changes in prevalence, awareness and treatment rates of dyslipidaemia and to investigate related influential factors for the development, awareness and treatment uptake of dyslipidaemia. Results In the cross-sectional study of the baseline data, the weighted prevalence, awareness, treatment and control rates of dyslipidaemia were 43.8% (95%CI: 42.8, 44.8%), 22.7% (95%CI: 20.9, 24.6%), 13.5% (95%CI: 11.7, 15.2%) and 5.1% (95%CI: 4.3, 5.8%), respectively. Urban residents had significantly higher prevalence, awareness and treatment rates of dyslipidaemia compared with rural residents, while the control rates did not differ significantly. In the cohort study of the baseline and two follow-up surveys, the estimated prevalence of dyslipidaemia did not increase much from 2011 to 2015, while the awareness and control rates in both urban and rural areas increased significantly. Waist circumference (cm, OR: 1.04, 95%CI: 1.02, 1.07) was identified as the most important predictors for the development of dyslipidaemia during the observation. Urban residence (OR: 1.50, 95%CI: 1.10, 2.05) and higher education (OR: 1.69/1.97/1.64 for primary/secondary/high school or higher compared with illiterate education level) were found to be significantly associated with awareness of dyslipidaemia, while awareness of dyslipidaemia (OR: 5.75, 95%CI: 3.95, 8.36) was directly associated with the treatment uptake of dyslipidaemia. Conclusions Dyslipidaemia was found to be widespread in China with high prevalence, but its awareness, treatment and control rates remained relatively low, indicating its poor detection and inadequate control. However, the awareness and treatment rates were gradually increasing over years, possibly showing the outcomes of China’s healthcare reform and public health promotion programs. Obesity was a crucial contributor of dyslipidemia, and therefore management of dyslipidemia could refer to obesity management for policy making. Urban residence and higher education were important for awareness of dyslipidemia, while awareness was important for treatment uptake, so public education programs of health promotion were needed to raise awareness of dyslipidemia and better manage dyslipidemia epidemic in China.

Ageing of the Australian population affects the residential aged care system, yet the structure and dynamics of the system remain uncertain. A comprehensive model of residential care based on the individual perspective of residential aged care events is missing. Thus, older Australians, government and care providers have only a limited model of aged care actions. This study uses big administrative unit record data on aged care assessments (ACAP), aged care appraisals (ACFI) and unit record survey data (SDAC) to identify factors associated with aged care events in older persons’ trajectories towards and through residential care. To achieve this goal and broaden understanding of Australian residential care, this study uses the following steps: (1) modelling of the probability of entry to and exit from residential care, with a multi-state model of transitions between levels of care needs; (2) modelling the applications for aged care and approvals for entry to residential care; (3) derivation of transition and mortality assumptions for individual care needs that can be used in a projection model; (4) estimation of life expectancy in residential care based on needs for assistance; and (5) assessing the quality of Australian data on aged care for statistical modelling and projections of residential care demand. The results show that health factors, above all needs for assistance, determine the speed and direction of a person’s progression towards institutional care. Probabilities of aged care events, transition rates and life expectancy estimates, derived in this study, provide a comprehensive picture of Australian residential aged care. These findings are expected to have important implications for residential aged care policies in Australia in terms of having better understanding and more accurate predictive power for the future of aged care.

This thesis explores youth sexuality, sexual risk and sexual resilience among young iTaukei women attending university in Fiji. An adapted ecological model is developed to enable the theorisation of sexual risk and sexual resilience in these women’s lives. The thesis illustrates how hierarchies of sexual risk are produced through the interaction of local cultures, sex and gender ideologies, global processes, complex sexual subjectivities and sexual desires. It demonstrates the need for a multi-level analysis to inform effective STI/HIV prevention strategies.

Introduction: Despite an intense national discussion in India during 2010 – 2012, progress towards universal health coverage (UHC) has stalled. Coverage of the entire population is still a challenge, especially effective coverage of the poor. Through the mechanism of microfinance-based self-help groups (SHGs), poor women and their families are provided not only with access to finance in a way that is understood to improve livelihoods, but also in many cases with a range of basic health services. With 93 million people organised nationally, SHGs provide an established organised network that can potentially be used to extend health coverage. Through a combination of quantitative and qualitative research approaches, this thesis aims to explore the potential for existing microfinance networks, using SHGs with attached health programs, to contribute to improved health coverage for the poor. Methods: A mixed-methods approach was used to address the study aim. A review of published evidence on the role of microfinance programs in improving health outcomes was conducted. This was followed by analysis of a national survey dataset to assess the impact of the presence of an SHG at village level on key health indicators at the individual level. Finally, a mixed methods study to assess the effect of combining a health program with microfinance-based SHGs was undertaken. This mixed methods study comprised two rounds of surveys to collect pre-test and post-test data with matched comparison groups and subsequent qualitative investigation to better understand the interconnections between SHGs, health programs and health. Results: The presence of SHGs was associated with significantly higher odds of women delivering their babies in an institution, feeding colostrum to their newborns, having knowledge of modern family planning methods and using family planning products and services. Additionally, the inclusion of a health program within microfinance-based SHGs was associated with further improvements in health behaviours, including facility-based deliveries, feeding newborns colostrum and having a toilet at home. However, the SHG health program led to no significant reduction in diarrhoea among children and no effect in reducing household money spent on health care. Conclusion: Capitalising on SHGs with health programs to improve the health of poor women and their families is an avenue worth investigating further. These established organised networks of SHGs provide an administrative apparatus to more effectively reach poor women and their families with essential health programs. Public health planners could leverage SHGs to increase the proportion of the population enjoying health coverage and make progress in relation to financial coverage and utilisation of existing publically-financed health protection schemes, although a lot more work is needed to optimise these possibilities.

Hospitals are resource intensive with detrimental environmental effects. This thesis explored hospital environmental resource use and resultant waste production in the operating room and intensive care unit. It examined safely reducing the use of equipment; compared reusable versus single use items and their resultant life cycles of carbon, water etc; and explored the effects of recycling. Steam sterilisation's large carbon and water footprint and how staff use sterilisers were quantified. Several examples of environmental and financial savings were found.

Introduction: Contracting, regarded as one of innovative approaches, has been used to address these issues. As part of larger national reform strategies to address poor public services, Cambodian government has piloted internal contracting as an approach to improve the management of district health system and services. The approach means parties from the same legal entity enter into a contract that is relational and based on trust and existing relationship. In Cambodia, a third of health districts, nominated as “Special Operating Agency - SOA”, are granted a semi-autonomous status with a greater degree of flexibility for decision-making. They receive additional funding to provide staff with financial incentives, recruit additional staff and engage communities and users. Using the Cambodian reform as a case study, this thesis examines the effectiveness of this approach in improving the management and provision of district health services. The effectiveness is judged by its being appropriate to and implementable in the context and able to overcome constraints and improve the conditions of staff and organisations, and the performance of service provision. It describes the design of the internal contracting arrangement, examines the implementation, changes in staff income and motivation, and assesses changes in the provision of primary health services. Methods: I reviewed the literature and available documents related to the reform, analysed data from interviews with officials and from a cross-sectional survey of 266 primary health care staff. Routine service data on child immunisation, antenatal care at second visit, newborn delivery by trained staff and outpatient consultation from four SOA districts were used to assess changes in primary care services before and after the reform. Findings: The implementation of internal contracting enabled facilities to maintain service delivery outputs elevated through other forms of contracting before the introduction of the current reform. The additional funding, carefully managed and utilised, made it possible to pay staff incentive; it along with performance contract made it possible to improve accountability addressing their underperformances and the functioning of health facilities. The incentive, albeit low, contributed to staff income and motivation. Coupled with job monitoring, incentive dissuaded many from private practice. Several challenges remained. System, tools and resources for management of performance and incentive and for contract monitoring were inadequate. Resource provision has been improved, but transparency of funding flow and information sharing was insufficient to tackle delays and cuts. Improvement in working conditions and organisational conditions was not optimal and their deficiencies continued to constrain staff performance. Analysis of provision primary care services (outpatient consultation, child immunisation, antenatal care and newborn delivery) showed mixed results, with most of the services increasing and stabilising after the introduction of the internal contracting. Conclusions: Internal contracting, when carefully packaged and designed, is effective in enhancing and sustaining service delivery for the long term. Adequate systems and tools for implementing performance contracts and managing resources are pre-requisites for improving staff performance and service delivery. Fine-tuning of the governance structure and the arrangements to ensure integrity of resource management and improve monitoring function would likely result in increased effectiveness of the local health system and service provision.

Between 1996 and 2008 global funding for the treatment of human immuno-deficiency virus (HIV) and acquired immune deficiency syndrome (AIDS) increased from US$300 million to an estimated US$15.6 billion. Much of this money was directed to a small number of countries such as Zambia in sub-Saharan Africa, where HIV/AIDS constitutes a major health, social and economic threat. Although the necessity and urgency of responding to the HIV epidemics in these countries was not in question, the exceptional levels of HIV funding and the rapidity of the scale-up of HIV- services did reignite a debate regarding the impact of disease-specific programs on recipient countries’ health systems. Notwithstanding the high profile nature of this debate, little empirical research exists to inform policy makers or programmers in their efforts to meet the dual aims of improving disease-specific health outcomes and simultaneously strengthening health systems. Meeting a gap in the literature, this study examines the impact of introducing donor-funded HIV services into the Zambian health system, focusing specifically on the impact on primary health ‘micro-systems’. The conceptual framework for this study draws from theory developed in the application of complexity science and systems thinking to health systems analysis, which suggests that health systems are characterised by the interconnectedness of their component parts. The multi-disciplinary framework theorises that interactions between system ‘hardware’ and system ‘software’ influence mechanisms of accountability and trust, and through these, the quality and responsiveness of service delivery within health micro-systems. This approach challenges the implicit assumptions of more reductionist frameworks, which suggest that health systems – and particularly micro-level systems – are a simple composite of individual ‘building blocks’. This study adopted a multi-case study design, with four Zambian health centres purposefully selected based on the presence of an established HIV department (more than 3 years old), and urban, peri-urban and rural characteristics. Case data collected in each facility included facility audits, direct observation of facility operations and interviews with patients, staff, and District and non-government officials. Data were triangulated and analysed for each case first, and cross-case analysis subsequently carried out to improve the analytical generalisability of the findings. The findings from this study demonstrate that the rapid scale-up of HIV services in Zambia, which focused predominantly on investing in health system hardware, acted unevenly on mechanisms of accountability and trust and had mixed outcomes on the four health centres’ overall functionality. It was revealed, for example, that the short-term gains in health worker performance achieved through investment in system hardware for HIV services were difficult to sustain, as the lack of investment in underlying mechanisms of accountability such as improved answerability and enforceability or stronger patient-provider trust, enabled perverse work norms to flourish in ways that undermined quality and responsiveness of care. The study points to the critical importance of accounting for the ideas, values and norms of actors in the health system (system software) in order to plan and deliver disease-specific interventions that achieve both their programmatic aims as well as producing long-term, system-strengthening effects. The study constitutes an important contribution to the field of health policy and systems research providing empirical evidence of the complex, social and adaptive nature of health micro-systems and demonstrating the critical value of the hardware-software construct for analysing mechanisms of effect in this domain.

Background: The extraordinarily high thresholds of adherence (≥95% of doses taken) currently deemed essential for good control of HIV has proved daunting for patients and challenging to clinicians in antiretroviral therapy (ART) programmes. Typical average adherence in other chronic diseases is between 50%-70%. Non-adherence to ART leads to the possibility of increased viral load, threatening the survival of individual patients and potentially raising the incidence of HIV in the general population. Adherence to ART is further complicated by lack of a gold standard tool to measure it. This study was conducted in Zimbabwe, one of the eight Southern African countries with generalised HIV epidemics (estimated prevalence of 15.3% in 2007). ART roll out began in 2004 as part of responses to address the epidemic. The ART programme is being implemented within an unprecedented humanitarian crisis. Aim: The study aimed to contribute to public health policy and program development by documenting adherence levels and assessing patient, community and health delivery factors influencing adherence to anti-retroviral therapy among patients in rural Zimbabwe. Method: The study design was mixed methods in two phases. The qualitative phase comprised six focus group discussions (FGDs) with 48 participants and 39 semi-structured interviews with the community and patients on ART, and one FGD with health staff; this phase was conducted to gain insight into influences on adherence in a specific context. The quantitative phase was a cross sectional survey with 180 patients on first line ART using an interviewer administered questionnaire that included three measures to estimate levels of adherence (visual analogue scale [VAS], seven day recall and unannounced pill counts [UPC] and questions exploring influences on adherence. Qualitative data were thematically analysed. Quantitative data were analysed using descriptive statistics, Chi-square and T tests to ascertain associations between variables Predictors of adherence were identified through binary logistic modelling. Results: Enablers of adherence identified qualitatively were: availability and accessibility of food; improved quality of life; hope of living longer; psychosocial support; and reliable drug supplies. Obstacles to adherence were: poverty, food unavailability and inaccessibility, alcohol use, denial and non-disclosure of HIV diagnosis, stigma and discrimination, lack of psychosocial support, cultural beliefs and practices, and lifelong duration of ART. Quantitative results showed levels of ≥95% adherence varied by method of measurement: 80.7%, 64.5% and 44.4% by VAS, seven day recall and UPC, respectively. Predictors of suboptimal adherence were: reported feelings of depression [OR 4.3, 95% CI 1.85-9.75]; need for a break from taking ARVs [OR 3.9, 95% CI 1.22- 12.15]; being female [OR 2.6, 95% CI 1.08-6.11]; feeling of not being listened to by health workers [OR 3.2, 95% CI 1.20-8.34]; having completed at least secondary education [OR 5.0, 95% CI 2.22-11.13]; perceived experience of stigma and discrimination as a disadvantage of disclosure [OR 2.5, 95% CI 1.15-5.20]; and difficulties travelling to the health centre for collection of ARVs and clinical review [OR 2.4, 95% CI 1.14-5.10]. Conclusions: Variations in adherence by measure indicate a need for further development of feasible tools. Patients alone do not hold the key to adherence. The complex problem of non-adherence should be urgently addressed from a multilayered, collaborative and complementary perspective that considers implications of the current humanitarian crisis. A range of context appropriate recommendations for patients, community and health provider levels are identified.