Transfers between health facilities of people living with diabetes attending primary health care services in the Western Cape Province of South Africa: A retrospective cohort study.

OBJECTIVES: Transfers between health facilities of people living with HIV attending primary health care (PHC) including hospital to PHC facility, PHC facility to hospital and PHC facility to PHC facility transfers occur frequently, affect health service planning, and are associated with disengagement from care and viraemia. Data on transfers among people living with diabetes attending PHC, particularly transfers between PHC facilities, are few. We assessed the transfer incidence rate of people living with diabetes attending PHC, and the association between transfers between PHC facilities and subsequent HbA1c values. METHODS: We analysed data on HbA1c tests at public sector facilities in the Western Cape Province (2016-March 2020). Individuals with an HbA1c in 2016-2017 were followed-up for 27 months and included in the analysis if ≥18 years at first included HbA1c, ≥2 HbA1cs during follow-up and ≥1 HbA1c at a PHC facility. A visit interval was the duration between two consecutive HbA1cs. Successive HbA1cs at different facilities of any type indicated any transfer, and HbA1cs at different PHC facilities indicated a transfer between PHC facilities. Mixed effects logistic regression adjusted for sex, age, rural/urban facility attended at the start of the visit interval, disengagement (visit interval >14 months) and a hospital visit during follow-up assessed the association between transfers between PHC facilities and HbA1c >8%. RESULTS: Among 102,813 participants, 22.6% had ≥1 transfer of any type. Including repeat transfers, there were 29,994 transfers (14.4 transfers per 100 person-years, 95% confidence interval [CI] 14.3-14.6). A total of 6996 (30.1%) of those who transferred had a transfer between PHC facilities. Visit intervals with a transfer between PHC facilities were longer (349 days, interquartile range [IQR] 211-503) than those without any transfer (330 days, IQR 182-422). The adjusted relative odds of an HbA1c ≥8% after a transfer between PHC facilities versus no transfer were 1.20 (95% CI 1.05-1.37). CONCLUSION: The volume of transfers involving PHC facilities requires consideration when planning services. Individuals who transfer between PHC facilities require additional monitoring and support.

Diagnostic test accuracy of artificial intelligence in screening for referable diabetic retinopathy in real-world settings: A systematic review and meta-analysis.

Retrospective studies on artificial intelligence (AI) in screening for diabetic retinopathy (DR) have shown promising results in addressing the mismatch between the capacity to implement DR screening and increasing DR incidence. This review sought to evaluate the diagnostic test accuracy (DTA) of AI in screening for referable diabetic retinopathy (RDR) in real-world settings. We searched CENTRAL, PubMed, CINAHL, Scopus, and Web of Science on 9 February 2023. We included prospective DTA studies assessing AI against trained human graders (HGs) in screening for RDR in patients with diabetes. Two reviewers independently extracted data and assessed methodological quality against QUADAS-2 criteria. We used the hierarchical summary receiver operating characteristics (HSROC) model to pool estimates of sensitivity and specificity and, forest plots and SROC plots to visually examine heterogeneity in accuracy estimates. From our initial search results of 3899 studies, we included 15 studies comprising 17 datasets. Meta-analyses revealed a sensitivity of 95.33% (95%CI: 90.60-100%) and specificity of 92.01% (95%CI: 87.61-96.42%) for patient-level analysis (10 datasets, N = 45,785) while, for the eye-level analysis, sensitivity was 91.24% (95%CI: 79.15-100%) and specificity, 93.90% (95%CI: 90.63-97.16%) (7 datasets, N = 15,390). Subgroup analyses did not provide variations in the diagnostic accuracy of country classification and DR classification criteria. However, a moderate increase was observed in diagnostic accuracy in the primary-level healthcare settings: sensitivity of 99.35% (95%CI: 96.85-100%), specificity of 93.72% (95%CI: 88.83-98.61%) and, a minimal decrease in the tertiary-level healthcare settings: sensitivity of 94.71% (95%CI: 89.00-100%), specificity of 90.88% (95%CI: 83.22-98.53%). Sensitivity analyses did not show any variations in studies that included diabetic macular edema in the RDR definition, nor studies with ≥3 HGs. This review provides evidence, for the first time from prospective studies, for the effectiveness of AI in screening for RDR in real-world settings. The results may serve to strengthen existing guidelines to improve current practices.

Positive attitudes toward adoption of a multi-component intervention strategy aimed at improving HIV outcomes among adolescents and young people in Nampula, Mozambique: perspectives of HIV care providers.

BACKGROUND: Service providers' attitudes toward interventions are essential for adopting and implementing novel interventions into healthcare settings, but evidence of evaluations in the HIV context is still limited. This study is part of the CombinADO cluster randomized trial (ClinicalTrials.gov NCT04930367), which is investigating the effectiveness of a multi-component intervention package (CombinADO strategy) aimed at improving HIV outcomes among adolescents and young adults living with HIV (AYAHIV) in Mozambique. In this paper we present findings on key stakeholder attitudes toward adopting study interventions into local health services. METHODS: Between September and December 2021, we conducted a cross-sectional survey with a purposive sample of 59 key stakeholders providing and overseeing HIV care among AYAHIV in 12 health facilities participating in the CombinADO trial, who completed a 9-item scale on attitudes towards adopting the trial intervention packages in health facilities. Data were collected in the pre-implementation phase of the study and included individual stakeholder and facility-level characteristics. We used generalized linear regression to examine the associations of stakeholder attitude scores with stakeholder and facility-level characteristics. RESULTS: Overall, service-providing stakeholders within this setting reported positive attitudes regarding adopting intervention packages across study clinic sites; the overall mean total attitude score was 35.0 ([SD] = 2.59, Range = [30-41]). The study package assessed (control or intervention condition) and the number of healthcare workers delivering ART care in participating clinics were the only significant explanatory variables to predict higher attitude scores among stakeholders (ß = 1.57, 95% CI = 0.34-2.80, p = 0.01 and ß = 1.57, 95% CI = 0.06-3.08, p = 0.04 respectively). CONCLUSIONS: This study found positive attitudes toward adopting the multi-component CombinADO study interventions among HIV care providers for AYAHIV in Nampula, Mozambique. Our findings suggest that adequate training and human resource availability may be important in promoting the adoption of novel multi-component interventions in healthcare services by influencing healthcare provider attitudes.

Mobility during the post-partum period and viraemia in women living with HIV in South Africa.

BACKGROUND: We investigated the association between travel and viraemia in post-partum women with human immunodeficiency virus on antiretroviral therapy (ART). METHODS: Data are from a trial of post-partum ART delivery strategies. Women who initiated ART during pregnancy, were clinically stable with a viral load (VL) <400 copies/ml and were <10 weeks post-partum were enrolled at a primary care antenatal clinic in Cape Town, South Africa. Study visits at 3, 6, 12, 18 and 24 months post-partum included questions about travel, defined as ≥1 night spent outside of the city, and VL testing. Generalised mixed effects models assessed the association between travel and subsequent VL ≥400 copies/ml. RESULTS: Among 402 women (mean age 29 y, 35% born in the Western Cape), 69% reported one or more travel events over 24 months. Being born beyond the Western Cape (adjusted odds ratio [aOR] 2.03 [95% confidence interval {CI} 1.49 to 2.77]), duration post-partum in months (aOR 1.03 [95% CI 1.02 to 1.05]) and living with the child (aOR 0.60 [95% CI 0.38 to 0.93]) were associated with travel. In multivariable analyses, a travel event was associated with a 92% increase in the odds of a VL ≥400 copies/ml (aOR 1.92 [95% CI 1.19 to 3.10]). CONCLUSIONS: Interventions to support women on ART who travel are urgently required.

Transfer of Patients on Antiretroviral Therapy Attending Primary Health Care Services in South Africa.

BACKGROUND: Patients stable on antiretroviral therapy (ART) may require transfer between health care facilities to maintain continuous care, yet data on the frequency, predictors, and virologic outcomes of transfers are limited. METHODS: Data for all viral load (VL) testing at public sector health facilities in the Western Cape Province (2011-2018) were obtained. Participant inclusion criteria were a first VL between 2011 and 2013, age >15 years at ART initiation, and >1 VL within 5 years of ART initiation, of which ≥1 was at a primary health care facility. Two successive VLs taken at different facilities indicated a transfer. We assessed predictors of transfer using generalized estimating equations with Poisson regression and the association between transfer and subsequent VL> 1000 copies/mL using generalized mixed effects. RESULTS: Overall 84,814 participants (median age at ART initiation 34 years and 68% female) were followed up for up to 4.5 years after their first VL: 34% (n = 29,056) transferred at least once, and among these, 26% transferred twice and 11% transferred thrice or more. Female sex, age <30 years, and first VL > 1000 copies/mL were independently associated with an increased rate of transfer [adjusted rate ratio 1.24, 95% confidence interval (CI): 1.21 to 1.26; 1.34, 95% CI: 1.31 to 1.36; and 1.42, 95% CI: 1.38 to 1.45, respectively]. Adjusting for age, sex, and disengagement, transfer was associated with an increased relative odds of VL > 1000 copies/mL (odds ratio 1.35, 95% CI: 1.29 to 1.42). CONCLUSIONS: Approximately one-third of participants transferred and virologic outcomes were poor post-transfer. Stable patients who transfer may require additional support to maintain adherence.

Pregnancy outcomes in women living with HIV and HIV-negative women in South Africa: Cohort analysis based on bias-corrected gestational age.

BACKGROUND: Antiretroviral therapy (ART) use during pregnancy may be associated with adverse outcomes, but findings have been inconsistent, at least in part due to unreliably estimated gestational age. OBJECTIVE: To quantify the association between HIV status, ART initiation timing and adverse birth outcomes, with reliably assessed gestational age at booking, in a public sector primary care facility in Cape Town, South Africa. METHODS: Pregnant women, HIV-negative or living with HIV (WLHIV), were enrolled at first antenatal care visit and followed through delivery. Ultrasound-assessed gestational age was deemed the gold standard. Based on quantitative bias analysis for outcome misclassification, gestational age by non-ultrasound assessment was corrected using multiple overimputation, which deals with missing data and measurement error simultaneously. Using bias-corrected gestational age, birth outcomes were compared between WLHIV and HIV-negative women, and among WLHIV who initiated ART before versus during pregnancy, further divided into trimesters. RESULTS: Of 3952 women enrolled, 37% were WLHIV (mostly using tenofovir + emtricitabine + efavirenz). Last menstrual period (LMP)-based gestational age was identified to be biased, and LMP measures were thus corrected using multiple overimputation. Comparing WLHIV and HIV-negative women, adjusted risk ratio (aRR) of overall pregnancy loss was 1.26 (95% confidence interval [CI] 0.98, 1.61); aRR of preterm delivery was 1.02 (95% CI 0.88, 1.20); aRR of small for gestational age infants was 1.43 (95% CI 1.14, 1.80). Among WLHIV, outcomes were similar by ART initiation timing. CONCLUSIONS: In this routine care cohort, risk of SGA, and possibly of pregnancy loss, was increased in WLHIV compared with HIV-negative women, with no evidence of increased risk of preterm delivery. Further research is needed to improve mechanistic understanding of the contribution of ART to adverse birth outcomes to optimize treatment for pregnant WLHIV and ensure optimal maternal and infant outcomes.

A review of simulation models for the long-term management of type 2 diabetes in low-and-middle income countries.

INTRODUCTION: The burden of type 2 diabetes is steadily increasing in low-and-middle-income countries, thereby posing a major threat from both a treatment, and funding standpoint. Although simulation modelling is generally relied upon for evaluating long-term costs and consequences associated with diabetes interventions, no recent article has reviewed the characteristics and capabilities of available models used in low-and-middle-income countries. We review the use of computer simulation modelling for the management of type 2 diabetes in low-and-middle-income countries. METHODS: A search for studies reporting computer simulation models of the natural history of individuals with type 2 diabetes and/or decision models to evaluate the impact of treatment strategies on these populations was conducted in PubMed. Data were extracted following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and assessed using modelling checklists. Publications before the year 2000, from high-income countries, studies involving animals and analyses that did not use mathematical simulations were excluded. The full text of eligible articles was sourced and information about the intervention and population being modelled, type of modelling approach and the model structure was extracted. RESULTS: Of the 79 articles suitable for full text review, 44 studies met the inclusion criteria. All were cost-effectiveness/utility studies with the majority being from the East Asia and Pacific region (n = 29). Of the included studies, 34 (77.3%) evaluated the cost-effectiveness of pharmacological interventions and approximately 75% of all included studies used HbA1c as one of the treatment effects of the intervention. 32 (73%) of the publications were microsimulation models, and 29 (66%) were state-transition models. Most of the studies utilised annual cycles (n = 29, 71%), and accounted for costs and outcomes over 20 years or more (n = 38, 86.4%). CONCLUSIONS: While the use of simulation modelling in the management of type 2 diabetes has been steadily increasing in low-and-middle-income countries, there is an urgent need to invest in evaluating therapeutic and policy interventions related to type 2 diabetes in low-and-middle-income countries through simulation modelling, especially with local research data. Moreover, it is important to improve transparency and credibility in the reporting of input data underlying model-based economic analyses, and studies.

A comparative analysis and review of how national guidelines for chronic disease monitoring are made in low- and middle-income compared to high-income countries.

BACKGROUND: Understanding how clinical practice guidelines and recommendations are adopted in high-income and low-income settings will help contextualise the value and validity of recommendations in different settings. We investigate how major guidelines and recommendations are developed for management and monitoring of post-diagnosis treatment for three important chronic diseases: HIV, hypertension and type 2 diabetes mellitus (T2DM). METHODS: Eligible guidelines were searched for using PubMed, Google, and health ministry websites for all three conditions. Only guidelines published from 2010 to 2020 were included. The source of the guidelines, year of most recent guideline, and basis of the guidelines were assessed. Additionally, recommendations, the strength of the recommendation and the quality of the evidence for treatment goals of non-pregnant adults and the frequency of monitoring were also extracted and assessed. RESULTS: Of the 42 countries searched 90%, 71% and 60% had T2DM, hypertension and HIV guidelines outlining targets for long-term management, respectively. Most T2DM guidelines recommend an HbA1c target of ≤7.0% (68%) or ≤6.5% (24%) as the ideal glycaemic target for most non-pregnant adults, while hypertension guidelines recommend blood pressure (systolic blood pressure/diastolic blood pressure) targets of <140/90 mm Hg (94%) and <130/80 mm Hg (6%). Of the identified HIV guidelines, 67% define virological failure as a viral load >1000 copies/mL, with 26%, mostly HICs, defining virological failure as a viral load >200 copies/mL. Recommendations for the frequency of monitoring for any diagnosed patients were available in 18 (55%) of the hypertension guidelines, 25 (93%) of HIV guidelines, and 27 (73%) of the T2DM guidelines. Only a few of the guidelines provide the strength of the recommendation and the quality of the evidence. CONCLUSIONS: Most guidelines from LMICs are adopted or adapted from existing HIC guidelines or international and regional organisation guidelines with little consideration for resource availability, contextual factors, logistical issues and general feasibility.

Mother-child separation among women living with HIV and their children in the first four years postpartum in South Africa.

OBJECTIVES: Mother-child pairs may separate during early life, yet the health impacts thereof are unclear. We explored the patterns and impact of separation among women living with HIV (WLHIV) and their children in South Africa. METHODS: WLHIV who had initiated antiretroviral therapy (ART) during pregnancy received HIV viral load (VL) testing and completed a timeline questionnaire of mother-child separation since delivery at 3-5 years post-partum. Health care usage was abstracted from routine medical records. We examined associations between separation and (a) maternal health outcomes (engagement in HIV care and HIV viral suppression, [VS]) and (b) child health outcomes (post-breastfeeding HIV testing and immunisation completion), using logistic regression. RESULTS: Of 346 mother-child pairs (median maternal age at antenatal ART initiation, 28 years), 24% were ever separated (median time to first separation 20 months, interquartile range [IQR] 9, 31). Most separated children were living with their grandmothers (65/83, 78%). Mothers who ever separated were younger, and more likely to be employed, and to reside in informal housing than those who never separated. Any separation reduced the odds of VS ≤ 50 copies/mL at four years post-partum (odds ratio 0.57; 95% CI 0.34-0.93); associations were similar for VL ≤ 1000 copies/mL and maternal engagement in care. No association was found between separation and child confirmatory HIV testing or immunisation completion. CONCLUSIONS: In this setting, mother-child separation is common in the first four years of life and appears associated with suboptimal maternal outcomes. Further research is required to understand the drivers and implications of mother-child separation.

OBJECTIFS: Les couples mère-enfant peuvent être séparés au début de la vie, mais les effets sur la santé ne sont pas clairs. Nous avons exploré les modèles et l'impact de la séparation chez les femmes vivant avec le VIH (PVVIH) et leurs enfants en Afrique du Sud. MÉTHODES: Les PVVIH qui avaient initié un traitement antirétroviral (ART) pendant la grossesse ont subi un test de charge virale du VIH (CV) et ont rempli un questionnaire chronologique sur la séparation mère-enfant depuis l'accouchement à 3-5 ans post-partum. L'utilisation des soins de santé a été extraite des dossiers médicaux de routine. Nous avons examiné les associations entre la séparation et (a) les résultats pour la santé maternelle (engagement dans les soins du VIH et la suppression virale du VIH [SV]) et (b) les résultats pour la santé de l'enfant (dépistage du VIH après l'allaitement et achèvement de la vaccination), en utilisant la régression logistique. RÉSULTATS: Sur 346 couples mère-enfant (âge maternel médian au début de l'ART prénatal, 28 ans), 24% ont été séparés (délai médian jusqu'à la première séparation 20 mois, intervalle interquartile [IQR] 9 à 31). La plupart des enfants séparés vivaient avec leurs grands-mères (65/83, 78%). Les mères qui ont vécu une telle séparation étaient plus jeunes et plus susceptibles d'avoir un emploi et de vivre dans un logement informel que celles qui ne l'ont jamais vécue. Toute séparation réduisait les chances de SV ≤50 copies/mL à quatre ans après l'accouchement (odds ratio 0,57; IC95%: 0,34-0,93); les associations étaient similaires pour la CV ≤1000 copies/mL et l'engagement de la mère dans les soins. Aucune association n'a été trouvée entre la séparation et le test de confirmation du VIH chez l'enfant ou l'achèvement de la vaccination. CONCLUSIONS: Dans ce contexte, la séparation mère-enfant est courante au cours des quatre premières années de la vie et semble associée à des résultats maternels sous-optimaux. Des recherches supplémentaires sont nécessaires pour comprendre les facteurs et les implications de la séparation mère-enfant.

ART Adherence, Resistance, and Long-term HIV Viral Suppression in Postpartum Women.

HIV+ South African women who achieved viral suppression during routine antenatal care, but later experienced a viremic episode (viral load >1000 copies/mL), were examined for presence of antiretrovirals and classified as "nonadherers" or "suboptimal adherers." Women were tested for drug resistance mutations (DRMs) at several time points and underwent viral load testing 36-60 months postpartum. Suboptimal adherers were more likely to have DRM detected during their viremic episode (P = .03) and at a subsequent viremic time point (P = .05). There was no difference in levels of viral suppression 36-60 months later in women with DRM detected vs women who had no evidence of DRM (P = .5).

Mixed-method estimation of population-level HIV viral suppression rate in the Western Cape, South Africa.

INTRODUCTION: There are few population-wide data on viral suppression (VS) that can be used to monitor programmatic targets in sub-Saharan Africa. We describe how routinely collected viral load (VL) data from antiretroviral therapy (ART) programmes can be extrapolated to estimate population VS and validate this using a combination of empiric and model-based estimates. METHODS: VL test results from were matched using a record linkage algorithm to obtain linked results for individuals. Test-level and individual-level VS rates were based on test VL values <1000 cps/mL, and individual VL <1000 cps/mL in a calendar year, respectively. We calculated population VS among people living with HIV (PLWH) in the province by combining census-derived midyear population estimates, HIV prevalence estimates and individual level VS estimates from routine VL data. RESULTS: Approximately 1.9 million VL test results between 2008 and 2018 were analysed. Among individuals in care, VS increased from 85.5% in 2008 to 90% in 2018. Population VS among all PLWH in the province increased from 12.2% in 2008 to 51.0% in 2017. The estimates derived from this method are comparable to those from other published studies. Sensitivity analyses showed that the results are robust to variations in linkage method, but sensitive to the extreme combinations of assumed VL testing coverage and population HIV prevalence. CONCLUSION: While validation of this method in other settings is required, this approach provides a simple, robust method for estimating population VS using routine data from ART services that can be employed by national programmes in high-burden settings.

Antiretroviral Adherence, Elevated Viral Load, and Drug Resistance Mutations in Human Immunodeficiency Virus-infected Women Initiating Treatment in Pregnancy: A Nested Case-control Study.

BACKGROUND: Elevated viral load (VL) early after antiretroviral therapy (ART) initiation appears frequently in pregnant and postpartum women living with human immunodeficiency virus; however the relative contributions of pre-ART drug resistance mutations (DRMs) vs nonadherence in the etiology of elevated VL are unknown. METHODS: Within a cohort of women initiating ART during pregnancy in Cape Town, South Africa, we compared women with elevated VL after initial suppression (cases, n = 80) incidence-density matched to women who maintained suppression over time (controls, n = 87). Groups were compared on pre-ART DRMs and detection of antiretrovirals in stored plasma. RESULTS: The prevalence of pre-ART DRMs was 10% in cases and 5% in controls (adjusted odds ratio [aOR], 1.53 [95% confidence interval {CI}, .4-5.9]); all mutations were to nonnucleoside reverse transcriptase inhibitors. At the time of elevated VL, 19% of cases had antiretrovirals detected in plasma, compared with 87% of controls who were suppressed at a matched time point (aOR, 131.43 [95% CI, 32.8-527.4]). Based on these findings, we estimate that <10% of all elevated VL in the cohort may be attributable to pre-ART DRMs vs >90% attributable to ART nonadherence. CONCLUSIONS: DRMs account for a small proportion of all elevated VL among women occurring in the 12 months after ART initiation during pregnancy in this setting, with nonadherence appearing to drive most episodes of elevated VL. Alongside the drive for access to more robust antiretroviral agents in resource-limited settings, there is an ongoing need for effective strategies to support ART adherence in this patient population.

Optimal timing of viral load monitoring during pregnancy to predict viraemia at delivery in HIV-infected women initiating ART in South Africa: a simulation study.

INTRODUCTION: HIV viral load (VL) monitoring is a central tool to evaluate ART effectiveness and transmission risk. There is a global movement to expand VL monitoring following recent recommendations from the World Health Organization (WHO), but there has been little research into VL monitoring in pregnant women. We investigated one important question in this area: when and how frequently VL should be monitored in women initiating ART during pregnancy to predict VL at the time of delivery in a simulated South African population. METHODS: We developed a mathematical model simulating VL from conception through delivery using VL data from the Maternal and Child Health - Antiretroviral Therapy (MCH-ART) cohort. VL was modelled based on three major compartments: pre-ART VL, viral decay immediately after ART initiation and viral maintenance (including viral suppression and viraemic episodes). Using this simulation, we examined the performance of various VL monitoring schema in predicting elevated VL at delivery. RESULTS AND DISCUSSION: If WHO guidelines for non-pregnant adults were used, the majority of HIV-infected pregnant women (69%) would not receive a VL test during pregnancy. Most models that based VL monitoring in pregnancy on the time elapsed since ART initiation (regardless of gestation) performed poorly (sensitivity <50%); models that based VL measures in pregnancy on the woman's gestation (regardless of time on ART) appeared to perform better overall (sensitivity >60%). Across all permutations, inclusion of pre-ART VL values had a negligible impact on predictive performance (improving test sensitivity and specificity <6%). Performance of VL monitoring in predicting VL at delivery generally improved at later gestations, with the best performing option a single VL measure at 36 weeks' gestation. CONCLUSIONS: Development and evaluation of a novel simulation model suggests that strategies to measure VL relative to gestational age may be more useful than strategies relative to duration on ART, in women initiating ART during pregnancy, supporting better integration of maternal and HIV health services. Testing turnaround times require careful consideration, and point-of-care VL testing may be the best approach for measuring VL at delivery. Broadening the scope of this simulation model in the light of current scale up of VL monitoring in high burden countries is important.

Differentiated models of care for postpartum women on antiretroviral therapy in Cape Town, South Africa: a cohort study.

BACKGROUND: The numbers of women initiating lifelong antiretroviral therapy (ART) during pregnancy and postpartum is increasing rapidly, presenting a burden on health systems and an urgent need for scalable models of care for this population. In a pilot project, we referred postpartum women who initiated ART during pregnancy to a community-based model of differentiated ART services. METHODS: Eligible women (on ART for at least 3 months with viral load (VL)<1000 copies/mL) were offered a choice of two ART models of care: (i) referral to an existing system of community-based 'adherence clubs', operated by lay counsellors with medication collection every 2-4 months; or (ii) referral to local primary healthcare clinics (PHC) with services provided by clinicians and medication collection every 1-2 months (local standard of care for postpartum ART). For evaluation, women were followed through 6-months postpartum with VL testing separate from either ART service. RESULTS: Through September 2015, n = 129 women were enrolled (median age, 28 years; median time postpartum, 10 days). Overall, 65% (n = 84) chose adherence clubs and 35% (n = 45) chose PHCs; there were no demographic or clinical predictors of this choice. Location of service delivery was commonly cited as a reason for choice by women selecting either model of care; shorter waiting times, ability to receive ART from lay counsellors and less frequent appointments were motivations for choosing adherence clubs. Among women choosing adherence clubs, 15% never attended the service and another 11% attended the service but were not retained through six months postpartum. Overall, 86% of women (n = 111) remained in the evaluation through 6 months postpartum; in this group, there were no differences in VL<1000 copies/mL at six months postpartum between women choosing PHCs (88%) vs. adherence clubs (92%; p = 0.483), but women who were not retained in adherence clubs were more likely to have VL≥1000 copies/mL compared to those who remained (p = 0.002). DISCUSSION: Adherence clubs may be a valuable model for postpartum women initiating ART in pregnancy, with good short-term outcomes observed during this critical period. To support optimal implementation, further research is needed into patient preferences for models of care, with consideration of integration of maternal and child health services, while ART adherence and retention require ongoing consideration in this population.