Understanding factors, outcomes and reasons for loss to follow-up among women in Option B+ PMTCT programme in Lilongwe, Malawi




To assess factors, outcomes and reasons for loss to follow-up (LTFU) among pregnant and breastfeeding women initiated on a lifelong antiretroviral therapy (ART) for PMTCT in a large antenatal clinic in Malawi.


We identified all pregnant and breastfeeding women who were initiated on ART between September 2011 and September 2013 and had missed their clinic appointment by at least 3 weeks at Bwaila Hospital, the largest antenatal clinic in Malawi. These women were traced by phone or home visits. Their true status and reasons for ART discontinuation were documented during tracing.


A total of 2930 women started ART for PMTCT; 2458 (84%) pregnant and 472 (16%) breastfeeding, of which, 577 (20%) missed a scheduled clinic appointment. LTFU was associated with younger age, being pregnant, and earlier year of ART initiation. We successfully traced 229 (40%), of whom, 10 (4%) had died. Of the 219 women found alive, 118 (54%) had stopped taking ARV drugs, 67 (30%) had self-transferred to another ART clinic, 13 (6%) had collected drugs from other sources, 9 (4%) had treatment interruptions and 12 (5%) had other outcomes. Reasons cited for stopping ART were travel (38%), lack of transport money (16%), not understanding the initial ARV education session (10%), being too weak/sick (10%), ARV side effects (10%) and other reasons.


Approximately half of the women who were traced were taking ARVs. The study emphasises the need for enhanced post-test counselling strategies, ongoing psychosocial support, provision of incentives and further decentralisation efforts of PMTCT services.



Evaluer les facteurs, les résultats et les raisons des pertes au suivi chez les femmes enceintes et allaitantes engagées dans une thérapie antirétrovirale (ART) continue pour la PTME dans une grande clinique prénatale au Malawi.


Nous avons identifié toutes les femmes enceintes et allaitantes qui ont été engagées dans l’ART entre septembre 2011 et septembre 2013 et ont manqué leur rendez-vous clinique d'au moins trois semaines à l'hôpital Bwaila, la plus grande clinique prénatale au Malawi. Ces femmes ont été recherchées par téléphone ou par des visites à domicile. Leur véritable état et les raisons de l'arrêt de l’ART ont été documentés au cours de leur recherche.


Au total 2930 femmes ont commencé l’ART pour la PTME; 2458 (84%) femmes enceintes et 472 (16%) allaitantes, dont 577 (20%) ont manqué un rendez-vous clinique prévue. La perte au suivi a été associée à un plus jeune âge, être enceinte et l'engagement à l’ART à une année plus antérieure. Nous avons retrouvé la trace de 229 (40%) femmes, dont 10 (4%) étaient décédées. Parmi les 219 femmes retrouvées en vie, 118 (54%) avaient cessé de prendre les médicaments ARV, 67 (30%) s’étaient auto-transférées dans une autre clinique ART, 13 (6%) avaient collecté des médicaments provenant d'autres sources, 9 (4%) avaient des interruptions de traitement et 12 (5%) avaient d'autres résultats. Les raisons invoquées pour l'arrêt de l’ART étaient le voyage (38%), le manque d'argent pour le transport (15%), ne pas avoir compris la session initiale de la formation sur les ARV (10%), être trop faible/malade (10%), les effets secondaires des ARV (10%) et d'autres raisons (16%).


Environ la moitié des femmes qui ont été retrouvées continuaient à prendre des ARV. L’étude met l'accent sur la nécessité de renforcer les stratégies de conseil post-test, le soutien psychosocial continu, plus d'incitations et d'efforts de décentralisation des services de PTME.



Evaluar los factores, resultados y razones para la pérdida durante el seguimiento (PDS) de mujeres embarazadas y lactando que habían iniciado terapia antirretroviral (TAR) de por vida para prevenir la transmisión madre-hijo en una gran clínica de atención prenatal en Malawi.


Hemos identificado a todas las mujeres embarazadas y lactando que habían iniciado TAR entre Septiembre del 2011 y Septiembre del 2013 y que no se presentaron en la cita que tenían programada, con un retraso de al menos 3 semanas, en el Hospital de Bwaila, la clínica prenatal más grande de Malawi. A estas mujeres se les realizó un seguimiento telefónico o mediante visitas a su hogar. Durante el seguimiento se documentó su estatus real y las razones para no continuar con el TAR.


Un total de 2930 mujeres comenzaron TAR para prevenir la transmisión vertical del virus; 2458 (84%) embarazadas y 472 (16%) lactantes, de las cuales 577 (20%) no se presentaron a la cita médica programada. La PDS estaba asociada con ser más joven, estar embarazada y haber comenzado antes el TAR. Seguimos con éxito a 229 (40%), de las cuales 10 (4%) habían muerto. De las 219 mujeres encontradas vivas, 118 (54%) habían dejado de tomar medicación ARV, 67 (30%) se habían auto-transferido a otro centro de TAR, 13 (6%) habían recogido medicamentos de otras fuentes, 9 (4%) habían interrumpido en algún momento el tratamiento y 12 (5%) tenía otros resultados. Las razones mencionadas para haber suspendido el TAR eran viajes (38%), falta de dinero para el transporte (15%), no haber entendido la sesión inicial de educación sobre el TAR (10%), estar muy débil / enferma (10%), efectos secundarios del TAR (10%) y otros motivos (16%).


Aproximadamente la mitad de las mujeres que fueron rastreadas estaban tomando ARVs. El estudio hace énfasis en la necesidad de mejorar las estrategias de aconsejamiento post-prueba, apoyo psicosocial continuo y un mayor esfuerzo de descentralización de los servicios de prevención de la transmisión madre-hijo.


Use of antiretroviral therapy (ART) in women infected with HIV significantly reduces mother to child transmission (MTCT) of HIV. Previous studies demonstrate a reduction in the risk of MTCT of HIV from 25% to <2% in HIV-infected pregnant women who start ART early and breasting feeding women who receive ART (Connor et al. 1994; Cooper et al. 2002; Lallemant et al. 2004; Shapiro et al. 2010).

In 2011, the Malawi Ministry of Health (MoH) embarked on a novel PMTCT programme known as ‘Option B+’. This programme takes a public health approach to promote maternal health and reduce vertical transmission of HIV infections through a test-and-treat approach. Option B+ recommends lifelong ART for all pregnant and breastfeeding women regardless of their CD4 count level or World Health Organisation (WHO) clinical stage and 6 weeks of daily nevirapine for the infant. In 2013, WHO released revised guidelines to include Option B+ for PMTCT (World Health Organisation 2013). Implementation of Option B+ in Malawi resulted in a sevenfold increase in the number of pregnant and breastfeeding women starting ART between the second quarter of 2011 and third quarter 2012 (Centers for Disease Control and Prevention 2013). The national data shows good programme retention: the majority of the women starting ART based on Option B+ are in care at 6 (82.3%) and 12 months (72%) post-ART initiation (Malawi Ministry Of Health 2013).

Although the Malawian Option B+ programme shows overall increased coverage of ART for the purpose of PMTCT, a loss to follow-up (LTFU) up to 24% in the first 6 months was noted in high patient volume facilities and for patients who were prescribed ART on the day they were diagnosed with HIV (Tenthani et al. 2014). Additional studies on LTFU in women starting lifelong ART for PMTCT purposes remain limited. We therefore assessed factors, outcomes and reasons for LTFU of pregnant and breastfeeding women initiated on a lifelong ART at a large ART clinic that uses a real-time, point-of-care Electronic Medical Record (EMR) system (Douglas et al. 2010). As part of routine care, we traced ART defaulters within the clinic catchment area. It is hoped that the findings of this study will assist in developing strategies to achieve higher retention rates in the implementation of PMTCT Option B+ in Malawi and other countries adopting a similar strategy.


Study design, site and population

This retrospective cohort study uses data from a real-time, touch screen-based EMR system for ART patient management and from a patient tracing programme called Back-To-Care (B2C) at Lighthouse Martin Preuss Centre (MPC) and the Family Health Unit (FHU). FHU and MPC are located at Bwaila Hospital, the busiest maternity unit in Malawi's capital, and the two clinics provide integrated ANC, TB, and HIV services. More than 14 000 pregnant women are registered annually in the ANC facility. As of January 2014, MPC provided ART services to 24 377 patients, 13 710 of whom were alive and on ART. In 2013, an average of 109 women was initiated on ART monthly based on Option B+ criteria. Provision of PMTCT services is a collaborative effort, led by the Lilongwe District Health Office and involving other partners, including University of North Carolina Project, Mothers2Mothers, Baobab health Trust and Lighthouse Trust. All pregnant and breastfeeding women who initiated ART at MPC between September 2011 and September 2013 were included in our analysis.

PMTCT services and data collection

During ANC registration visits, all pregnant women with unknown HIV status undergo a group HIV counselling session, followed by ‘opt-out’ provider-initiated HIV antibody rapid testing, and individual post-test counselling. Psychosocial and adherence support for women diagnosed with HIV are also provided on initial and follow-up visits by ‘expert mothers’. All pregnant and breastfeeding women diagnosed with HIV are routinely registered in the EMR system and asked for consent to be traced in case they fail to return for subsequent appointments.

HIV-infected pregnant and lactating women are started on a lifelong ART (fixed dose combination of Tenofovir/Lamivudine/Efavierenz) regardless of their clinical or immunological status in accordance with the Option B+ strategy stipulated in Malawi's Integrated Guidelines for Clinical Management of HIV (Malawi Ministry Of Health 2011). Women are initiated on ART on the same day of HIV diagnosis. After ART initiation, monthly ARV dispensing visits are managed by a clinic nurse within the ANC wing. At each ART dispensing visit, adherence is assessed using self-reported pill counts which are entered in the EMR system. Scheduled clinic appointments are electronically calculated by the EMR system based on the regimen number of tablets newly dispensed and the tablets remaining at present visit. One year after the delivery, the ongoing care of women and infants is routinely transferred to ART clinic of their choice, with the majority remaining at MPC.

Data collection for the B2C programme was previously described (Tweya et al. 2010). In brief, patients who miss their next scheduled appointment by 3 weeks or more are presumed to have run-out of ARVs and defined as LTFU, for the purpose of the B2C programme. The list of LTFU patients is generated by the EMR system and verified by B2C tracers to rule out data errors. Confirmed LTFU patients who consented during ART registration are traced by phone and/or home visits. Tracing outcomes for the B2C programme are recorded as: (i) dead; (ii) ‘self-transfer’ if the patient had arranged the transfer independently; (iii) alive on ART (with ‘treatment interruptions’ if a patient took none or fewer than the prescribed drugs before the interview date, or with ‘uninterrupted therapy’ if the patient was still taking ARVs despite missing appointment); (iv) stopped ART by themselves; (v) refused to be interviewed and (vi) never started ART although they collected drugs. Among women who were found to have stopped ART without clinician direction, the B2C tracers collected further information about reasons for ART discontinuation using a standard questionnaire.

Statistical analysis

STATA 12.0 was used for statistical analysis. Descriptive statistics were used to explore baseline characteristics of the study population. In time to event analysis, observation time began from the date of ART initiation and ended at either death, LTFU (defined as missing a clinic appointment by at least 3 weeks), transfer to other ART facility or censor date of September 30, 2013, whichever occurred first. The Kaplan–Meier methods were used to estimate the proportions of LTFU at 6 and 12 months post-ART initiation. Poisson regression models were used to investigate the incidence of LTFU. Multivariable Poisson regression was used investigate factors associated with LTFU by modelling the association between LTFU and covariates of interest. Unadjusted rate ratios (RR) and adjusted RR are reported, including the 95% confidence intervals (CI). A level of significance of  0.05 was used. Tracing outcomes were described using the follow-up status that was reported by the traced women. Women's reasons were described if they stopped ART on their own.

Ethical considerations

Only LTFU patients who had given consent to tracing at clinic registration were followed up. Data for analysis did not include personal identifiers and was analysed anonymously. The study was approved by the Malawi National Health Science Research Committee in Lilongwe, Malawi, the Ethics Advisory Group of the International Union Against Tuberculosis and Lung Disease in Paris, France and University of North Carolina, Chapel Hill, USA.


Baseline characteristics and loss to follow-up

Between September 2011 and September 2013, 2930 HIV-infected women started ART for PMTCT Option B+ at Bwaila Hospital. Of these, 2458 (84%) were pregnant and 472 (16%) were breastfeeding at ART initiation. The median age at ART initiation was 26 years (interquartile range, IQR 22–30). Information on employment status was available for a total of 2911 (99%) women, of whom 1050 (35%) were employed.

Of all the 2930 women, 577 (20%) missed a scheduled clinic appointment by at least 3 weeks. Of the 577 women, 272 (47%) women only collected ARVs at the time of initiation and did not return, 69 (12%) collected ARVs at two visits, 54 (9%) collected ARVs at three visits and 182 (32%) at four or more visits. Overall incidence of LTFU was 23.5 per 100 person-years of observation time with median follow-up of 8.2 months (IQR 3.1–16.7). Retention probabilities were 85%, 82% and 79% at 3, 6 and 12 months, respectively. Women who were LTFU were more likely to have initiated ART during pregnancy, younger than 25 years, unemployed or started ART in 2011 soon after Option B+ was introduced (Table 1).

Table 1. Baseline characteristics of women who started antiretroviral therapy (ART) for Option B+ at Bwaila Hospital in Lilongwe Malawi between September 2011 and September 2013 (n = 2930)
CharacteristicWomen who were LTFUWomen who not were LTFUP-value
n % n %
  1. LTFU, lost to follow-up.

  2. a

    Employment status includes those working and doing small scale businesses.

  3. b

    Data from September 2011.

  4. c

    Data up to September 2013.

All women57780235320 
Age at ART initiation (years)
Reason for ART initiation
Employment statusa
Year of implementation

Factors associated with loss to follow-up

In unadjusted analysis, the incidence of LTFU varied according to baseline characteristics (Table 2). The incidence of LTF was significantly associated with the age at ART initiation, reason for starting ART, employment status at ART initiation and year of Option B+ implementation. After adjusting for age, reasons for starting ART, employment status and year of Option B+ implementation, the incidence of LTFU remained significantly associated with the age at ART initiation, reason for starting ART and year of Option B+ implementation. However, employment status was no longer statistically significantly associated with LTFU. The probability of LTFU was higher in women who were aged 13–24 years at ART initiation than in those who were 25 or older (adjusted RR = 1.29 95% CI 1.09–1.52). Breastfeeding women were less likely to be LTFU than women who were pregnant at ART initiation (adjusted RR = 0.63 95% 0.49–0.82). Women who started ART in 2011 (the year when Option B+ was introduced) were more likely to be LFTU from care than those who started in 2012 (adjusted RR 1.25; 95% CI 1.06–1.49). The probability of LTFU was lower among women who started ART in 2013 than in women who started ART in 2012 (adjusted RR = 0.41; 95% CI 0.29–0.58). Likelihood of LTFU decreased with increasing year of Option B+ implementation (test for trend < 0.001).

Table 2. Factors associated with incidence of loss to follow-up (LTFU) among women who started antiretroviral therapy (ART) for Option B+ at Bwaila Hospital in Lilongwe Malawi between September 2011 and September 2013 (n = 2930)
CharacteristicsLTFU (n/person-years)UnadjustedP-valueaAdjustedcP-valuea
Rate ratio (95% CI)Rate ratio (95% CI)
  1. a

    P-value for likelihood ratio test.

  2. b

    Data from September 2011.

  3. c

    Adjusted for age, reason for starting ART, year of ART start.

  4. d

    Employment status includes those working and doing small scale businesses.

  5. e

    Data up to September 2013.

Age at ART initiation (years)
13–24242/8871.27 (1.08–1.50)0.0041.29 (1.09–1.52)<0.001
Reason for ART initiation
Breastfeeding67/4170.64 (0.50–0.83)<0.0010.63 (0.49–0.82)<0.001
Employment statusd
Yes131/7390.68 (0.56–0.83)0.0010.87 (0.71–1.07)0.177
Year of implementation
2011b233/7851.25 (1.05–1.48)<0.0011.25 (1.06–1.49)<0.001
2013e38/3840.42 (0.21–0.27)0.41 (0.29–0.58)

Tracing outcomes

Of the 577 LTFU patients, 349 (60%) could not be traced due to incorrect/incomplete addresses (n = 291, 83%), change of residence to outside the clinics' catchment areas (n = 55, 16%) or human resource shortages at clinic (n = 3, 1%). However, there were no significant differences between women who were traced and those who were not traced in terms of age at ART initiation (P = 0.779), breastfeeding/pregnancy (P = 0.889) and year of Option B+ implementation (P = 0.383). Of the 291 women who could not be traced due to incorrect/incomplete addresses, 229 (79%) started in the first year of the Option B+ implementation. Of the 228 (40%) who were successfully traced and 9 (4%) had died. Of the 219 women found alive, 118 (54%) stopped taking drugs, 67 (30%) self-transferred to another ART clinic, 13 (6%) were on uninterrupted therapy (collected drugs from alternative sources e.g. friends, relatives and etc), 9 (4%) had treatment interruptions, 7 (3%) had not started taking drugs despite collecting them and 5 (2%) refused to be interviewed.

Reasons for ART discontinuation

Among the women who were successfully traced, all who stopped ART were also asked about their reasons for ART discontinuation and 111 (94%) provided an explanation. Some women gave more than one reason for discontinuing ART. Frequently cited reasons for stopping ART were travel away from Lilongwe either for work or to visit relatives (n = 42; 38%), lack of money for transport to the clinic (n = 17; 15%), not understanding the initial ARV education session (n = 11; 10%), being too weak/sick to take ART (n = 11; 10%), due to developing adverse side effects (n = 11; 10%), non-disclosure of HIV status to the spouse (n = 9; 8%) and other (n = 49; 44%) (Table 3). Patients who stopped ART were also asked regarding their intention to continue taking ARVs. Information about intent to continue ART was not documented for 11 (9%) of patients. A total of 107 women were counselled and advised to return to the ART clinic with 95 (89%) women saying they would return and restart ART, however, only 27 of these (23%) returned to care.

Table 3. Reasons for discontinuing antiretroviral therapy (ART) women starting ART for PMTCT at Bwaila Hospital in Lilongwe, Malawi
Reasons for ART discontinuationTotal
n %
  1. a

    Percentages are out of those who responded to each question. Some women gave more than one reason for Discontinuing ART.

Total women118 
Forgotten to take ARVs55
Suspected side effects of ARVs1110
Very weak/sick1110
Religious belief55
Travelled away4238
Non-disclosure of HIV status to the spouse98
Transport costs1716
Limited information about ARVs1110
Other reasons4944


In one of the first studies critically evaluating LTFU among women in PMTCT Option B+ programme, we found LTFU was 23.5% per year and was associated with younger age at ART initiation, being pregnant at ART initiation and earlier year of Option B+ implementation. Among women who were LTFU but alive, almost half had stopped ART, leaving their infants at a high risk of HIV infection, and a third self-transferred to another clinic, suggesting an underestimation of national retention in PMTCT programme. Reasons cited for ART discontinuation were travel away from home, lack of transport money, developing adverse side effects, being too sick to travel to clinic and limited understanding of the initial ART education session. These findings highlight barriers to retention and provide insight into strategies that may improve health systems delivery and reduce future LTFU in the Option B+ PMTCT programme.

We found significant LTFU of women from ART within the first year, which may negatively affect the effectiveness of PMTCT. Among women who were LTFU, 47% received ARVs only once and never returned for their appointment, which suggests that a proportion of these women never started ART. LTFU observed in this study was higher than that reported in the general HIV-infected individuals accessing ART for personal health (9.3% per year) (Tweya et al. 2010). Our results also showed an independent association between age and LTFU. Women who were 25 years or older at ART initiation were less likely to be LTFU than younger women. Older women may have more settled lifestyles which allow them to better manage ARVs for PMTCT. In agreement with other previous studies (Kaplan et al. 2008; Tenthani et al. 2014), we found that being pregnant at ART initiation was associated with high LTFU. As it is the case with many other clinics in Malawi, women in our cohort are started on ART on the same day of HIV diagnosis. These pregnant women may have been daunted by the complexity of decisions that have to be made such as after being newly diagnosed as HIV-positive and the need to start lifelong treatment (Stinson & Myer 2012), indicating the need for more effective counselling and support both before and after ART initiation. Currently, post-test counselling is usually done by ART nurses who may have limited time due to high workload. Engaging HIV testing counsellors and ‘expert mothers’ in the provision of HIV test results and post-test counselling may facilitate acceptance of these major life changes over time and improve adherence to PMTCT services.

We also observed a decreasing trend in the likelihood of LTFU with increasing year of programme implementation between 2011 and 2013, likely due to the stabilization of the programme at the hospital, improved knowledge of PMTCT in the community and incentives. In 2012, a community mobilisation and sensitisation campaign was launched to introduce PMTCT Option B+ whose key messages were adherence in PMTCT Option B+, male involvement in reproductive services and early infant diagnosis. Community mobilisation involved community leaders and volunteers, while sensitisations were done mostly in market places. Also, the improved retention might have been due to the vitameal (fortified soya flour) provided by the hospital to breastfeeding women who were in the PMTCT Option B+ programme for their children aged 6–24 months. Such food supplements may have acted as incentives for breastfeeding mothers to remain in care.

Other tracing outcomes provide insight into efforts to further strengthen the programme. Among women successfully traced, around 30% who would have been categorised as defaulters had actually transferred silently to another facility and were still on ART for PMTCT. The need to improve communication of ART patients who transfer between clinics through establishment of data linkages has been highlighted before (Tweya et al. 2013) and remains a high priority intervention. Also, although routine active patient tracing system has proved successful in returning LTFU patients to HIV care (Tweya et al. 2010; Thomson et al. 2011), the proportion of women who returned in our study was very low. Lastly, despite tracing efforts, a sizeable proportion of women (60) could not be traced due to incorrect addresses documented in the patient clinic files. The proportion of women not traced is higher than that observed in the general population of ART patients within the same catchment area (30%). As indicated previously by a study done in Malawi, it is possible that a considerable number of women in this study may have given a false physical addresses because of fear of stigma and discrimination through inadvertent disclosure of their HIV status (Bwirire et al. 2008). Improvements are required in routine PMTCT tracings programme to make them effective such as women newly diagnosed with HIV should be informed of providing correct location details when being given HIV results.

ART discontinuation may be explained by diverse reasons across the continuum of PMTCT care. Almost half of the women LTFU had stopped ART. Although all PMTCT women undergo pre-treatment counselling sessions about the benefits of PMTCT, some women reported having limited knowledge of PMTCT. Given this, women need to be prepared for HIV testing, its implications(Bwirire et al. 2008). Given the combination of these factors, pregnant women, especially those of younger age, need to be provided with adequate post-test counselling about the implications of their results and linked with supportive services, such as peer groups and being mentored by expert mothers. Furthermore, while non-disclosure to partners was reported in literature as key reason for ART discontinuation, very few women cited this in our study. Partner involvement is reported to enhance utilisation of PMTCT services. In particular, previous studies demonstrated increased adherence to PMTCT services and better infant outcomes through partner involvement (Msuya et al. 2008; Aluisio et al. 2011; Panditrao et al. 2011). However, another study (Barigye et al. 2010) did not support this finding. Therefore, further rigorous evaluations are required on the effectiveness of male involvement on PMTCT service.

We also found that lack of transport money and travel away from home as other contributing reasons to ART discontinuation. Decentralisation of HIV services was reported to be associated with improved retention by reducing transportation costs, travel time, and clinic wait times among other benefits. A study in Malawi reported higher LTFU rate at hospitals (9.9%) than at rural health centres (1.5%), such that the use of rural centres resulted in a 77% reduction in risk of attrition (Massaquoi et al. 2009). High burden ANC/ART clinics should actively ask pregnant women to register at health centres that are closer to their residence, and there is need to sufficiently staff these health centres.

Our study had limitations. First, we used routine data from the patient tracing programme and some information was missing. Information on outcomes and reasons for LTFU were only available for the 40% of women that were traced, and these findings may not necessarily be extrapolated to the whole LFTU population. Second, use of ART was partly self-reported and may result in underestimating or overestimating the number of patients that are actually taking ART. Despite these limitations, we believe that the study findings are useful to inform implementation of PMTCT Option B+ in large ART clinics in Malawi and other comparable settings.

In conclusion, we identified barriers to retention in Option B+ programme which comprise travel, lack of transport money, side effects and inadequate knowledge of PMTCT knowledge. ANC/ART clinics can further enhance post-test counselling by engaging HIV testing counsellors and expert mothers for ongoing counselling and psychosocial support. Further decentralisation of PMTCT services with good ANC and maternity services to health centres and the provision of attendance incentives might have a positive effect on retention. We acknowledge the generous financial support provided by the US Agency for International Development.


The authors would like to thank all the staff who collected data at Bwaila Health Unit and Martin Preuss Center clinics. We received funding from the Bill and Melinda Gates Foundation and the USAID-NIH initiative Partnerships for Enhanced Engagement in Research (PEER) Health. We thank Dr Caryl Feldacker from the International Training and Education Center for Health, University of Washington, Seattle, for helpful comments. We acknowledge the generous financial support provided by the US Agency for International Development.