Globally, the number of new HIV-infected children decreased by 70% from 2000 to 2015,1 primarily because of increasing implementation of the Option B+ strategy, which entails lifelong antiretroviral therapy (ART) for all HIV+ pregnant and breastfeeding women.2 However, poor retention of HIV+ mothers and HIV-exposed infants (HEIs) in the service delivery cascade for elimination of HIV mother-to-child transmission of HIV (eMTCT) remains a significant challenge in Sub-Saharan Africa.3–5 This threatens global efforts to achieve eMTCT by 2020. Various factors operating at personal, family, community, and health facility–levels have been associated with poor retention.6–10
The use of short-text messaging service (SMS) to remind mothers about scheduled or missed clinic appointments and the provision of integrated HIV and maternal, neonatal, and child health (MNCH) are potential interventions that can improve retention of mother–infant pairs (MIPs). Previous randomized controlled trials found that SMS reminders increased maternal retention at 8 or 10 weeks postpartum11,12 but did not assess long-term retention. A South African cohort study found no association between the use of SMS reminders and 12-month postpartum maternal retention.13 These previous studies enrolled only mothers with mobile phones, thereby limiting the application of the findings to settings with poor mobile phone ownership.
Studies assessing the effect of integration of HIV and MNCH service delivery on MIP retention have yielded mixed results. A Malawian observational study found higher maternal retention in health facilities providing ART and MNCH services in separate clinics than in those providing integrated HIV and MNCH services,14 but another Malawian study found the opposite results.15 A Kenyan cluster randomized trial found no effect of HIV/MNCH service integration on 6-month maternal retention.16 The conflicting results are likely because of differences in study designs, health system context, duration of maternal follow-up, and definitions of “integration” and “maternal retention.” Nevertheless, provision of HIV and MNCH services at multiple times and access points could potentially result in frequent health facility attendances and out-of-pocket expenditures by women, which may negatively affect long-term retention.
We assessed whether retention of HIV+ mothers and their infants in eMTCT care can be improved by integrating the delivery of HIV and MNCH services and using SMS to trace mothers who miss scheduled health facility visits.
Study Design and Setting
Details of the study's methodology have been described elsewhere.17 Briefly, from May 2013 to August 2016, we implemented the “Promoting Retention among Infants and Mothers Effectively (PRIME)” cluster randomized trial in Salima and Mangochi districts, located in central and southern Malawi, respectively. In 2010, HIV prevalence was estimated at 16.2% among the Yao (predominant tribe in Mangochi) and 9% among the Chewa (predominant tribe in Salima).18 All health facilities in Malawi officially started implementing Option B+ regimen in July 2011. Mobile phone ownership among Malawians was estimated at 36.0% in 2014 and was markedly lower in rural (30.2%) than urban areas (71.8%).19
Table 1 provides details of 3 study arms namely MIP, MIP + SMS, and standard of care (SOC). Briefly, the MIP clinics aimed at combining the delivery of HIV and MNCH services at a single time and point of access. MIP clinics in the study facilities were offered from twice per week to once per month, depending on the volume of registered HIV+ mothers and HEIs and availability of health workers. The SMS reminders, which were sent through community-based volunteers, aimed at expediting communication between facility-based health workers and mothers, most of who did not possess mobile phones. Before study initiation, mobile phones were distributed to each study health facility and community-based volunteers.
Facilities in the SOC (control) arm provided HIV and MNCH services at multiple times and access points and used government-employed health surveillance assistants to track mothers who missed scheduled health facility appointments using client-tracing forms.
According to the Ministry of Health (MoH) HIV management guidelines, HIV+ mothers and HEIs are required to attend HIV clinics at least monthly up to 6 months postpartum and quarterly, thereafter.20 During these visits mothers and infants access various HIV and MNCH services, as described in Table 1. To assess maternal and infant retention and compliance to this visit schedule, our trial had 2 primary outcomes. First, maternal retention was defined as “the proportion of women enrolled antenatally who were retained in HIV care through 12 months postpartum and received all preceding antiretroviral drug refills no later than 14 days of a scheduled refill date.” Second, infant retention was defined as “the proportion of HEIs attending the 12-month scheduled visit plus at least 5 of the 6 preceding scheduled visits, including a visit by 2 months when a dried blood spot (DBS) sample is collected for early infant diagnosis.”
As part of routine monitoring and evaluation, the MoH defines 12-month retention as “attendance of the mothers or infants at the scheduled 12 months postpartum visit.” To compare our trial results with the MoH routine data, we considered the MoH 12-month maternal and infant retention definitions as secondary study outcomes.
Study Population and Study Sample
Our study population was all HIV+ pregnant women on Option B+ ART regimen. We excluded women meeting any of the following criteria:
- not permanent residents of villages surrounding the health facilities or planning to move out of the villages in the following 6–9 months
- unable to give informed consent because of immediate life‐threatening illness or mental disorders
- participating in other studies
- not receiving ART services at the facility.
To calculate sample size, we used the following point estimates for the 12-month maternal retention: MIP arm = 40%, MIP + SMS arm = 60%, and SOC arm = 25%. Infant retention was estimated as follows: MIP arm = 20%, MIP + SMS arm = 40%, and SOC arm = 10%. An equation, which corrects for stratification,21 was used to determine the number of clusters needed for each treatment arm with the following inputs: α = 0.05, β = 80%, m (number of subjects per cluster) = 65, and k (coefficient of variation) = 0.25. Considering only pairwise comparisons between each intervention arm and SOC, the maternal retention outcome yielded the biggest sample size (10 clusters/arm) to detect the expected differences between study arms. As reported previously,17 the 30 health facilities were randomly selected of 61 available facilities in the 2 districts, allowing for buffer zones to avoid contamination.
One year after enrollment, the sample size was revised by adjusting the value of “k” from 0.25 to a more realistic value of 0.20, based on data from a previous study.22 Maintaining 10 clusters/arm and allowing for 15% of unavailable data, the sample size was adjusted to 41 MIPs per cluster or 1230 MIPs across the 3 study arms.
Health care workers (HCWs) trained in good clinical practice requested written informed consent and enrolled eligible HIV+ pregnant women. Basic sociodemographic information was collected. At subsequent visits, HCWs recorded services received by the women and HEIs in health facility registers and patient cards, which served as source documents. Every 4–6 weeks, study research assistants visited the health facilities to extract study-related data from these source documents onto case report forms, assess the quality of implementation of study interventions, and address data quality issues. In addition, a study mentor conducted technical support visits to the health facilities every 3 months. However, more frequent supportive visits were conducted in sites consistently registering challenges with implementing study interventions and data documentation.
Completed case report forms were transferred to the PRIME study office where data entry officers performed electronic double data entry and data cleaning. Routine data quality checks including audits on a random sample were conducted every 3–4 months by study staff. In addition, an independent data quality expert performed data audits twice yearly.
Facility-level, maternal, and infant characteristics were summarized at enrollment by study arm using sample statistics. Primary and secondary outcomes were analyzed by log binomial regression using a generalized estimated equation that accounted for the clustering of health facilities and the stratified sampling design. Although dichotomous on the individual level, retention was a proportion at the study facility level and was presented in bubble plots to show differences between and within study arms. To identify other potential predictors for maternal retention, univariate analyses were performed to calculate unadjusted risk ratios (RRs) and 95% confidence intervals (CIs) expressed against a reference standard along with cluster-adjusted percentages. Subsequently, multivariate analyses were performed that adjusted for stratification and the study interventions. All statistical analyses were conducted using Stata SE 13.
The College of Medicine Research and Ethics Committee and the WHO Ethics Review Committee approved the initial trial protocol and subsequent amendments. Informed consent was sought from all study participants before enrollment.
Enrollment and Follow-up Process
From May 2013 to November 2014, 2536 HIV+ pregnant women were documented to have sought antenatal care services at the 30 study facilities, of whom 1350 (53.2%) enrolled into the trial. The most commonly reported reason for non-enrollment was the inability of the facility-based HCWs to administer the study eligibility checklist because of high workload. Of the enrolled women, 461, 493, and 396 were enrolled in the MIP, MIP + SMS, and SOC arms, respectively. Figure 1 depicts the study flowchart for facility selection and randomization, maternal and infant enrollment, by study arm, including reasons why some women did not remain at the study facilities up to the time of delivery. Overall, 410 enrolled women were not retained in the study at 12 months postpartum, 69.0% of whom dropped out before delivery.
Of the 1350 enrolled women, 1067 (79.0%) had documented live birth deliveries in the health facility registers comprised of 1049 singleton and 18 twin deliveries. Thus, 1085 infants enrolled in the study, distributed as follows: 386, 399, and 300 in the MIP, MIP + SMS, and SOC arms, respectively. The median number (interquartile range) of scheduled visits made by enrolled women from delivery to 12 months postpartum was 6 (2–8), 6 (4–8), and 6 (2–8) in the MIP, MIP + SMS, and SOC arms, respectively. For the HEIs, the median number (interquartile range) of visits were 5 (4–7), 6 (5–6), and 5 (4–5), respectively.
Characteristics of Study Sites and Enrolled Women and Infants
Table 2 displays facility-level and maternal and infant individual-level baseline characteristics, by study arm. Generally, there were no major differences across the study arms in the characteristics of the facilities and enrolled women or infants. However, a slightly lower proportion of women in SOC arm were single and enrolled in the first trimester than those in the MIP + SMS arm. Also, the proportions of women delivering at health facilities and having live births tended to be lower in the SOC than those in the MIP arm. However, the proportion of singleton deliveries tended to be higher in the SOC arm than in the MIP + SMS arm.
Maternal Retention at 12-Month Postpartum
Compared with the 12-month maternal postpartum retention rate in the SOC arm (22.2%), the cluster-adjusted retention rates were similar in the MIP arm [19.3%, RR: 0.85, (95% CI: 0.56 to 1.30), P = 0.46] and in the MIP + SMS arm (24.9%, RR: 1.08, 95% CI: 0.87 to 1.35, P = 0.50). Nevertheless, as shown in Figure 2A, there were wide variations in the retention rates across individual facilities within each of the study arms.
Retention of HEIs at 12-Month Postpartum
Compared with the 12-month cluster-adjusted infant retention rate in the SOC sites (9.8%), the rate was similar in the MIP arm (8.0%, RR: 0.89, 95% CI: 0.31 to 2.58, P = 0.83) but higher in the MIP + SMS arm (19.5%, RR: 1. 40, 95% CI: 0.85 to 2.31), but did not reach statistical significance (P = 0.19). Similar to maternal retention, the 12-month infant retention varied markedly across health facilities within each arm, as shown in Figure 2B.
Secondary Maternal and Infant Outcomes
We also assessed the impact of the trial interventions on maternal and infant retention, using the MoH definitions. Compared with the cluster-adjusted 12-month maternal postpartum retention in the SOC arm (69.1%), retention rates were similar in the MIP arm (72.4%, RR: 1.05, 95% CI: 0.93 to 1.18, P = 0.45) and in the MIP + SMS arm (67.1%, RR: 0.99, 95% CI: 0.93 to 1.05, P = 0.72). Similarly, compared with the 12-month cluster-adjusted infant retention in the SOC arm (78.0%), retention rates were similar in the MIP arm (75.4%, RR: 0.98, 95% CI: 0.89 to 1.09, P = 0.73) and in the MIP plus SMS arm (80.8%, RR: 1.01, 95% CI: 0.96 to 1.07, P = 0.63).
Fidelity in the Implementation of Study Interventions
In view of the lack of impact of our study interventions, data were analyzed to assess the level to which study participants were exposed to the trial interventions. Results have been presented in detail elsewhere.23 Nevertheless, we found suboptimal exposure of women to the MIP service delivery model because only 42% of the mothers' health facility visits occurred on designated MIP clinic dates. We also found inadequate implementation of the study interventions by health workers, as they correctly scheduled only 47% of the mothers' subsequent clinic visits on designated MIP clinic dates. However, when the mothers were scheduled to return on a designated MIP clinic date, 75% of them attended the clinics. Nevertheless, we observed a statistically significant increasing trend in the proportion of maternal visits occurring on designated MIP clinic dates over the span of the study, starting at 23% in at the inception of the trial in 2013 and increasing to a maximum of 59% at the completion of the trial in 2016. A similar increment was noted in the HEIs, from 42% in 2013 to 60% in 2016.
Similarly, we found challenges in the implementation of the SMS-based tracing. When mothers missed scheduled MIP clinics, health workers sent out tracing SMS reminders 43% of the time and documented outcomes 63% of the time. Some of the reported reasons for not initiating SMS reminders were poor documentation of missed visits, poor mobile phone network, uncharged mobile phone handsets, and preference by some HCWs to relay SMS reminders by health surveillance assistants and not community volunteers.
Other Predictors of Maternal Retention
Table 3 shows the effect of other variables on 12-month maternal retention for the whole study cohort. In bivariate analyses, none of the health facility characteristics were significantly associated with 12-month maternal retention. However, older maternal age, higher parity, and participation in the study from 2015 onward were significantly associated with increased maternal retention. Conversely, study enrollment at a later gestational age and first diagnosis of HIV or being ART naive at enrollment were associated with reduced retention. In multivariate analyses, which included study interventions and health facility location (rural/semiurban) variables, only maternal age and participation in the study from 2015 onward were significantly associated with 12-month maternal retention.
Unintended Social Adverse Events
There were 2 incidents at one MIP + SMS facility where participants' husbands expressed concern with community volunteers tracing and requesting their wives to attend health facility without their knowledge and suspected spousal infidelity. One of the husband considered divorcing his wife. Although these incidents were resolved, health facility staff expressed discomfort over lay community volunteers' ability to maintain confidentiality and respect women's rights when tracing defaulting women.
In this cluster randomized trial, which was conducted in a predominantly rural setting in Malawi with low mobile phone ownership, an integrated HIV and MNCH service delivery model and an SMS-based method of tracing clients delivered by a community volunteer, did not impact 12-month postpartum retention of HIV+ mothers and HEIs. We found very low retention rates across the 3 study arms because our definition of maternal and infant retention was more conservative than the routine MoH monitoring and evaluation definition, as it captured not only clinic attendance at 12 months but also compliance with preceding scheduled visits. Nevertheless, regardless of the definition of retention, our study interventions did not improve infant or maternal retention. However, we observed suboptimal exposure of the enrolled women and infants to the trial interventions, which may have blunted the effect of the interventions on retention.
The ineffectiveness of SMS-based tracing to improve retention in our trial is inconsistent with findings from previous randomized controlled trials that showed increased maternal retention at 8–10 weeks postpartum.11,12 This is most likely because of methodological differences among the studies. In contrast to our trial, the previous trials enrolled only women with mobile phones and communicated with them directly, through weekly or biweekly SMS reminders, whereas the PRIME trial sent the reminders to mothers through community volunteers. However, previous studies did not assess the impact of SMS reminders on long-term retention. Nevertheless, our results are consistent with those from one before/after intervention study which found no effect of SMS reminders on 12-month postpartum retention.13 Our findings of ineffectiveness of integrated HIV/MNCH service delivery model on maternal retention mirror those from a previous Kenyan cluster randomized trial.13 Nevertheless, caution must be exercised when comparing these studies because they used different models of integration and definitions of retention.
Our enrolled women had limited exposure to the SMS-based communication system, mostly because of the inability by HCWs to identify mothers who missed visits in timely fashion and send out SMS reminders. Similarly, they had limited exposure to the integrated HIV/MNCH (MIP) clinics mostly because of the inability of HCWs to book their follow-up appointments on designated MIP clinic dates. To alleviate these challenges, project staff intensified health facility mentorship visits. Although study participants' exposure to the intervention improved over time, the challenges persisted, as described elsewhere,23 mostly because of underlying health system challenges, particularly HCW shortages. The study health facilities had a median of 4 HCWs per facility, with most facilities having 1–2 nurses or clinicians. Also, frequent staff turnover at the facilities meant that HCWs could not gain experience in delivering the interventions. HCWs reported experiencing high workloads because they also had to simultaneously deliver other essential maternity and out-patient services. We also found that some health facilities had challenges delivering a “one-stop shop” MIP clinic model because of space limitations. Details of the challenges experienced in implementing the study interventions have been reported elsewhere.23
Our study found that most enrolled women were lost before delivery, regardless of the study arm. This is somewhat consistent with previous studies which showed high early attrition of HIV+ pregnant women newly initiated on ART as part of Option B+ implementation.4,8,10,24 In our multivariate model, we found that older maternal age at enrollment was associated with improved maternal retention. This suggests that to improve maternal retention, interventions must focus on addressing antenatal attrition, particularly among young pregnant women and those newly initiated on ART.
One of the strengths of the design was that the implementation setting reflected reality in most health facilities in Malawi. Actual HCWs and community volunteers and not study staff were in direct control of administering the interventions. Thus, the study findings can be generalized into similar settings. Study limitations included incomplete documentation of services provided to mothers and HEIs in registers and client cards, which served as our source documents. Missed or inaccurate documentation of client visits or services delivered could have led to under-reporting of outcomes. However, it is unlikely that under-reporting differed across study arms.
Our findings cannot be extrapolated to other settings different from Malawi. In this trial, SMS reminders were sent to community health volunteers residing in defaulting women's villages who, in-turn, physically attempted to trace and persuade them to return to the clinic. In settings with high mobile phone ownership, SMS reminders could be sent directly to the defaulting women, hence removing potential concerns of confidentiality, stigma, and discrimination, which may be associated with the involvement of the community-based volunteers. Alternatively, as implemented in other settings,25 trained community-based health workers and not lay community members could be used to receive SMS reminders and attempt to trace defaulting mothers. Similarly, implementation of integrated MIP clinics may be explored in health systems that do not have severe human resource and infrastructure constraints.
In summary, we found that an integrated HIV/MNCH delivery model and SMS-based reminders did not improve 12-month retention. Health system challenges adversely affected optimal implementation of the study interventions. Thus, without adequate long-term health system investments, including skills development and building motivation of HCWs, integration of services in health systems such as in Malawi and other Sub-Saharan African countries seems unrealistic. Suboptimal quality of health care delivery is likely to remain a barrier to long-term retention of mothers and infants.
The authors express their sincere gratitude to Frank Chimbwandira and Michael Eliya for their technical support during implementation; Debbie Jackson, Christiane Horwood, and Sifiso Phakathi for monitoring trial implementation; WHO technical support from Geneva (Nigel Rollins, Nathan Shaffer, April Baller, Nita Bellare, and Shaffiq Essajee) and in the Afro-region (Morkor Owiredu); and in-country WHO team (Ellen Thom) for their advice and support in study operations. The implementation support from the District Health Office and staff in Mangochi and Salima is also highly appreciated.
1. UNAIDS. Prevention Gap Report. Joint United Nations Programme on HIV
/AIDS. 2016. Available at: http://http://www.unaids.org
/sites/default/files/media_asset/2016-prevention-gap-report_en.pdf. Accessed January 27, 2017.
2. WHO. Consolidated Guidelines on the Use of Antiretroviral Drugs for Treating and Preventing HIV
Infection: Recommendations for a Public Health Approach-Second Edition. 2016. Available at: http://apps.who.int/iris/bitstream/10665/198064/1/9789241509893_eng.pdf?ua=1. Accessed January 27, 2017.
3. Centers for Disease C, Prevention. Impact of an innovative approach to prevent mother-to-child transmission of HIV
, July 2011–September 2012. MMWR Morb Mortal Wkly Rep. 2013;62:148–151.
4. Tenthani L, Haas AD, Tweya H, et al. Retention
in care under universal antiretroviral therapy for HIV
-infected pregnant and breastfeeding women (“Option B+”) in Malawi
. AIDS. 2014;28:589–598.
5. Haas AD, Tenthani L, Msukwa MT, et al. Retention
in care during the first 3 years of antiretroviral therapy for women in Malawi
's option B+ programme: an observational cohort study. Lancet HIV
6. Katirayi L, Namadingo H, Phiri M, et al. HIV
-positive pregnant and postpartum women's perspectives about Option B+ in Malawi
: a qualitative study. J Int AIDS Soc. 2016;19:20919.
7. Helova A, Akama E, Bukusi EA, et al. Health facility challenges to the provision of Option B+ in western Kenya: a qualitative study. Health Policy Plan. 2017;32(2):283–291.
8. Woelk GB, Ndatimana D, Behan S, et al. Retention
of mothers and infants in the prevention of mother-to-child transmission of HIV
programme is associated with individual and facility-level factors in Rwanda. J Int AIDS Soc. 2016;19(suppl 4):20837.
9. Kim MH, Zhou A, Mazenga A, et al. Why did I stop? Barriers and facilitators to uptake and adherence to ART in option B+ HIV
care in Lilongwe, Malawi
. PLoS One. 2016;11:e0149527.
10. Cataldo F, Chiwaula L, Nkhata M, et al. Exploring the experiences of women and health care workers in the context of PMTCT Option B plus in Malawi
. J Acquir Immune Defic Syndr. 2017;74:517–522.
11. Odeny TA, Bukusi EA, Cohen CR, et al. Texting improves testing: a randomized trial of two-way SMS
to increase postpartum prevention of mother-to-child transmission retention
and infant HIV
testing. AIDS. 2014;28:2307–2312.
12. Kebaya LN, Wamalwa R, Kariuki D, et al. Efficacy of mobile phone use on adherence to nevirapine prophylaxis and retention
in care among HIV
-exposed infants in PMTCT: a randomized controlled trial. Arch Dis Child. 2014;99:A1–A620.
13. Schwartz SR, Clouse K, Yende N, et al. Acceptability and Feasibility of a mobile phone-based case management intervention to retain mothers and infants from an Option B+ program in postpartum HIV
care. Matern Child Health J. 2015;19:2029–2037.
14. van Lettow M, Bedell R, Mayuni I, et al. Towards elimination of mother-to-child transmission of HIV
: performance of different models of care for initiating lifelong antiretroviral therapy for pregnant women in Malawi
(Option B+). J Int AIDS Soc. 2014;17:18994.
15. Weigel R, Hosseinipour MC, Feldacker C, et al. Ensuring HIV
-infected pregnant women start antiretroviral treatment: an operational cohort study from Lilongwe. Malawi
Trop Med Int Health. 2012;17:751–759.
16. Turan JM, Onono M, Steinfeld RL, et al. Implementation and operational research: effects of antenatal care and HIV
on Elements of the PMTCT cascade: results from the SHAIP cluster-randomized controlled trial in Kenya. J Acquir Immune Defic Syndr. 2015;69:e172–e181.
17. Mwapasa V, Pro G, Chinkhumba J, et al. Mother-infant pair clinic and SMS
messaging as innovative strategies for improving access to and retention
in eMTCT care and Option B+ in Malawi
: a cluster randomized control trial (the PRIME study). J Acquir Immune Defic Syndr. 2014;67(suppl 2):S120–S124.
Demographic and Health Survey 2010. 2011. Available at: http://http://www.nsomalawi.mw
/images/stories/data_on_line/demography/MDHS2010/MDHS2010.report.pdf. Accessed February 13, 2017.
19. MACRA. National Survey on Access to and Usage of ICT Services in Malawi
. 2015. Available at: http://http://www.macra.org.mw
.pdf. Accessed February 13, 2017.
MoH. Clinical Management of HIV
in Children and Adults. 2011. Available at: http://http://www.hiv.health.gov.mw
/cha_repository/file.php/82/82.pdf. Accessed February 13, 2017.
21. Hayes J, Moulton L. Cluster Randomized Trials. Boca Raton, FL: Chapman & Hall/CRC; 2009.
22. Adams G, Gulliford MC, Ukoumunne OC, et al. Patterns of intra-cluster correlation from primary care research to inform study design and analysis. J Clin Epidemiol. 2004;57:785–794.
23. Gunda A, Jousset A, Tchereni T, et al. Integrating HIV
and maternal, neonatal and child health services in rural Malawi
: an evaluation of the implementation processes and challenge. J Acquir Immune Defic Syndr. 2017;75(suppl 2):e132–e139.
24. Tweya H, Gugsa S, Hosseinipour M, et al. Understanding factors, outcomes and reasons for loss to follow-up among women in Option B+ PMTCT programme in Lilongwe. Malawi
Trop Med Int Health. 2014;19:1360–1366.
25. Hirsch-Moverman Y, Daftary A, Yuengling KA, et al. Using mHealth for HIV
/TB treatment support in Lesotho: enhancing patient-provider communication in the START study. J Acquir Immune Defic Syndr. 2017;74(suppl 1):S37–S43.
Keywords:Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.
HIV; retention; SMS; integration; implementation research; Malawi