Impact of the Umoyo mother-infant pair model on HIV-positive mothers’ social support, perceived stigma and 12-month retention of their HIV-exposed infants in PMTCT care: evidence from a cluster randomized controlled trial in Zambia

The goal of eliminating vertical human immunodeficiency virus (HIV) transmission is embedded within the Joint United Nations Programme on HIV/AIDS (UNAIDS) 90–90-90 targets which are to be achieved by 2020 [1]. Taking antiretroviral therapy (ART) during pregnancy and breastfeeding can significantly suppress the HIV viral load of a pregnant woman, reducing the risks of mother-to-child transmission of HIV to negligible levels [2]. However, more than half of mother-to-child transmissions occur during breastfeeding, signalling the need for more attention on postpartum follow-up and retention in care during breastfeeding [3].

HIV stigma and discrimination are two important detractors to the elimination of mother-to-child transmission of HIV as they are strong barriers against accessing HIV prevention, testing, and treatment services [4‐7]. Abundant evidence shows that people living with HIV/AIDS are better able to cope with the negative health stressors of discrimination and stigma in the presence of social support as it has been associated with positive health outcomes for such as self-efficacy, adherence to ART, quality of life, physical health, and mental health [8‐13]. Indeed, the WHO has recognized that people living with HIV/AIDS have important psychosocial needs that need to be integrated into HIV care such as prevention and treatment of mental health disorders, assistance on coping with discrimination and stigma and financial forms of assistance [14].

While the vast majority of research linking ART adherence to social support has been conducted on adults, little is known about the relationship between social support for mothers and health outcomes of their HIV-exposed infants (HEIs). Interventions focused on structured peer-support models such as the Mother2mother mentor program have been implemented in selected African countries though evidence on the impact of social support on retention of HEIs along the Prevention of Mother-to-Child Transmission (PMTCT) cascade is promising but mixed [15‐17].

In resource-constrained settings like Zambia, significant strides have been made in increasing PMTCT service utilization and coverage with the national roll out of WHO Option B+ in 2013. However, more needs to be done to achieve elimination of mother-to-child transmission. Nationwide data on the retention rates of mother-infant pair (MIP) cohorts within PMTCT are not available but overall, HIV-positivity rates for early infant diagnostic tests are increasing in infant age from those that test at 6 weeks up to 24 months [19]. Efforts to eliminate mother-to-child transmission of HIV are constrained by a plethora of challenges including but not limited to limited human resources for health; prolonged turnaround times for early infant diagnosis test results; poor 24-month retention, weak cohort-monitoring systems for tracking MIPs; and limited community support systems [20]. Additionally, the standard of PMTCT care in Zambia is a potential barrier to improved retention because mothers of HEIs attend monthly under-five clinics together with HIV-negative mothers. Anecdotally, mothers of HEIs receiving additional services (e.g., prophylaxis and counseling) beyond what HIV-negative mothers are receiving can betray the HIV statuses of the different mothers. Further, to avoid speculation on HIV statuses by offering different services to mothers depending on their status, health workers sometimes request mothers of HEIs to wait until all HIV-negative mothers have been attended to, in order to receive their HIV-specific services. However, the fact that these positive mothers stay longer at the clinic for a routine under-five clinic is also a fuel for additional speculation from their HIV-negative peer mothers who are released early.

Within the context of constrained budgets and competing health needs, promising interventions, such as the Mother2mother program, are unlikely to be prioritized for support by the Zambian Ministry of Health (MOH) due to the financial costs. Alternative approaches that can be feasibly implemented by public health systems are needed. In Zambia, there is an MIP clinic model being implemented in selected public health facilities in two districts (Kitwe in Copperbelt Province and Chipata in Eastern Province). This MIP clinic approach, called Umoyo which is translated as “Clinic of Life” from a local Zambian dialect, is being sustained by a lean staff of health workers supported by few volunteers without external financing or material support outside what is received from the government. In Umoyo clinics, under-five clinics are scheduled in such a way that HIV-positive mothers and their HEIs receive integrated HIV and routine post-natal care services on a day separate from HIV-negative mothers. Therefore, a Umoyo MIP clinic ensures that MIPs receive the services that they need and fosters the development of informal mother-to-mother peer-support systems.

The Zambian MOH was interested in evidence on whether Umoyo clinics could improve retention of HEIs in care for possible scale-up. However, only anecdotal evidence existed in support of this model and, therefore, a rigorous evaluation of this clinic model was necessary to inform policy-makers. Therefore, this study was commissioned to answer the question: “What is the impact of Umoyo MIP clinics on retention of HIV-exposed infants in PMTCT care at 12 months after birth?”

Figure 1 below illustrates the final sample of facilities obtained for the study.

Of the original 124 public facilities that were supported by our implementing partner (CIDRZ), eight were removed because of ongoing interventions similar to the Umoyo program. Among the remaining 116 facilities, facilities were ranked on their catchment population, and the 28 facilities were selected (across the three provinces and eight districts) that had the largest catchment populations and thus likely the largest HIV client population. Within the original 28 selected, five were replaced due to logistical considerations. Specifically, to reduce the budget and logistical challenges, districts where at least one other facility had been selected were included in the sample; these original five facilities were from districts where only one facility was selected. The final sample was 28 facilities across two provinces (Eastern and Lusaka) and seven districts (Fig. 1).

Table 5 displays the breakdown of our study sample for the facilities, children, and mothers.

Of the 28 facilities, not all facilities had facility-characteristic data collected at both time points. Two of the pre-implementation facility surveys, while collected, were not entered into the system at the time. Overall, no facilities reported a stock out in the year before Umoyo, but there were reports of stock outs of dried-blood-spot test kits after Umoyo implementation with 10% of control facilities and 20% of intervention facilities reporting a stock out in the 12 months before the endline survey. The affected facilities which experienced stock outs are found in a similar geographical region which was had overall challenges with stock supply for a short period.

For the child-level analysis, we report the statistics only among the facilities used in the final analysis (14 control and 11 intervention): two facilities did not have ART services in 2016 (pre implementation), and one facility did not have any eligible children for the study in 2017 (post intervention). During the pre-implementation phase, the average number of children in the control arm was 11.9 children per facility (95% CI 2.6, 21.1) while the average for the intervention arm was 37.7 per facility (95% CI − 1.6, 77.1). There was large variation in the number of children per facility as evidenced by the relatively large CIs for facilities in both the control and intervention arms. There was no evidence of a difference in the change of eligible children per facility between the two arms at an alpha of 0.01 (p value: 0.1). During the pre-implementation phase, eligible infants in the control arm were roughly 52 days old (95% CI 47.2, 56.7) when they received their 6-week HIV test while eligible children in the intervention arm were 50 days old (95% CI 46.2, 53.9); there was no evidence of a difference in the change of mean age at testing between the two arms (p value 0.2).

In Table 6, we report the child-level outcomes for the outcomes of interest.

We found that the Umoyo program did not improve the proportion of children who were retained in care at 12 months (− 11%; 99% CI −40.1%, 17.2%) or continually from 6 to 12 months (− 6.5; 99% CI − 24.7%, 11.7%). In logistic GEE models, evidence even suggested that the Umoyo program may have had a significantly negative impact on retention as the treatment effect over time (i.e., interaction term) showed a reduction of 0.2 lower odds of retention at 12 months from baseline to endline (99% CI 0.1, 0.4) as well as a 0.3 lower odds of retention at 12 months (with 6- and 9-month retention) from baseline to endline for children enrolled in an Umoyo facility as compared to the change for children in control facilities (Table 7).

Table 8 shows distribution of the mothers’ outcomes on stigma and social support for mother exit interviews. Over time, according to an alpha of < 0.01, we did not see evidence that the Umoyo program improved the Social Support Scores (mean 0.2, 99% CI − 0.4, 0.8), enacted stigma (mean − 0.1, 99% CI − 0.3, 0.01), HCW stigma (mean − 0.2, 99% CI − 0.5, 0.03), or the internalized stigma (mean 0.01, 99% CI − 0.3, 0.3).

With the linear GEE models, however, we did find that the treatment effect over time; (i.e., the interaction term) showed that mothers in the intervention arm from baseline to endline increased their Social Support Score 0.31 units more as compared to the change in Social Support Score for mothers in control facilities (99% CI 0.08, 0.54). We also found that the HCW stigma decreased over time to a larger degree from baseline to endline for mothers in the intervention facilities as compared to the change in reported HCW stigma for mothers in control facilities (mean difference -0.27; 99% CI -0.46, -0.08) (Table 9).

The findings of this study show that the Umoyo MIP clinic did not have a significant impact on the 12-month postpartum retention of HEIs in PMTCT care. Rather, GEE results suggest that the 12-month retention outcomes over time for HEIs in control facilities were better than those observed in the intervention sites. At the same time, the Umoyo MIP clinic may have had a statistically significant impact on improving social support over time and reducing HCW stigma but did not have any effect on reducing internalized stigma nor enacted stigma.

The finding that the Umoyo MIP clinics did not have a significant effect on retention of HEIs has been found in cluster randomized controlled trials testing the impact of Umoyo MIP clinic-like interventions in similar settings such as Malawi and Zimbabwe. In Malawi, an MIP clinic offering integrated services to HIV-positive mothers and their HEIs on the same day including sending SMS reminders to mothers on follow-up appointments found no statistically significant impact on retention of infants in PMTCT care. The lack of impact was largely attributed to partial implementation fidelity where about 42% of the mothers were exposed to the intervention [25]. In Zimbabwe, a cluster randomized controlled trial found that mother support groups at health facilities had no statistically significant impact on retention of HEIs at 12 months [26]. However, compared to our work these studies found a higher proportion of retention in the intervention than in the control, though insignificant, while our study suggested a lower rate of retention due to the program.

Among others, such as lower power due to facility exclusion and larger than anticipated variation in facility sizes and/or that the intervention may not have had an impact, two possible additional reasons for the potentially null or negative impact observed in this study are described next. First, the evidence on exposure to the intervention was limited but shown to be relatively low at roughly 50%. Based on conversations with the facility staff and field visits, it was obvious that the intervention was implemented in non-standard ways, pointing to weak monitoring of the intervention. Even in places where strong Umoyo MIP clinic champions were present at facility level at the outset, staff turn-over led to variable content and quality of Umoyo MIP clinics. The second possible reason for the negative impact of Umoyo MIP clinics on retention of HEIs is that facility staff reported that the intervention may have had an adverse effect on data completeness of facility registers. Anecdotally, facility staff in intervention sites mentioned that they were the busiest on Umoyo clinic days due to the huge volume of services to be provided, a fact that was compounded by constrained human resources. Therefore, it is possible that staff may have de-prioritized the need to update the facility registers even though PMTCT-related services had been provided. In turn, this missing data in facility registers, may have limited our ability to examine the impact of the Umoyo model. Future versions of the model should work to include more comprehensive data management and monitoring systems. Additionally, the study team did collect additional data from secondary sources such as the TB Register, Isoniazid Register and the Cotrimoxazole Dispensing Register, these secondary registers were not uniformly used across facilities and are not reported anywhere in this manuscript as they would serve to increase bias.

The indicative positive results of the intervention on social support and HCW stigma are similar to other peer-to-peer mentor support studies because the Umoyo MIP clinics are founded on the idea that mothers interact with the mothers much more than would be the case without them [17, 18]. Additionally, since the running of Umoyo MIP clinics is HCW-driven, it is possible that a reduction in HCW stigma was observed. However, there is no literature on studies that have evaluated the evolution of stigma using the HASI-P instrument in order to compare with our study findings for non-improvement in internalized and enacted stigma.

Several lessons have been learned in the course implementing this research that other researchers may benefit from. Firstly, if resources permit, it would be valuable to validate the final sample generated from a sampling frame especially if health facility characteristics are an important part of the selection criteria and randomization. Secondly, having formal measures of implementation fidelity would have given more insights into whether the intervention was really not helpful or that the intervention was poorly implemented. Thirdly, the reliance on data from the public health system remains a challenge and while we had deliberate reasons for doing so, researchers should continue to find innovative way of collecting information that is already contained in formal registers to avoid duplication.

Our study has several limitations. One limitation is that three facilities in the intervention arm and one in the control arm were excluded from the difference-in-difference analysis due to lack of implementation. Thus, power was lower than anticipated and the variation in facility sizes was large; thus, our minimal effect size was lower than estimated. Moreover, the original sampling frame obtained for randomization and sample selection was not as accurate as we had anticpated; thus, we were unable to determine the risk to power ahead of time. Another limitation is that though there was no statistical difference between the two arms and examined covariates (save for the patient load per facility), there may have been residual imbalance at the time the intervention began and we were not able to collect individual-level data about other potential confounders.

This study aimed to build on existing health-system protocols and documentation without creating additional requirements or resources; however, it should be noted that completeness of the data registers was an issue. We sometimes found registers other than the primary register as alternative sources of the required information. As these secondary registers were not present in all facilities, the extent to which they had complete data compared to the primary register is unknown. Therefore, in facilities where other registers were preferred by facility staff as the primary register for recording information, then the outcomes may have been underestimated. The study also did not account for the variability in the use of the primary register across all the facilities nor did the study compare the data completeness in these secondary registers to the primary register. We do not believe, however, that this would have differed between the study arms. More worrying, however, is the anecdotal point above: that those intervention facilities may have de-prioritized accuracy of the records during busy Umoyo clinic days. Future models would benefit by creating innovative solutions to record keeping given the limited staff.

The other limitation of the study relates to the overall fidelity of the intervention. Some facilities, usually the small facilities, had near perfect fidelity to the intervention while the large-volume sites had poor fidelity. Furthermore, the study had no formal measurements of implementation fidelity. Finally, regarding the mothers’ stigma and peer-support outcomes, the limitation with these results is that they were not performed on the same mothers before and after the intervention. Therefore, their applicability is subject to the limitations of the methodology employed, i.e., confounding factors may be able to explain the presence or absence of the impact seen.