Given persistent mother-to-child HIV transmission in resource-limited settings with high HIV burden, identifying effective implementation strategies that can be scaled through routine management structures is critical. The SAIA-SCALE program tested an approach for delivering a multi-component system engineering strategy through district-level PMTCT management systems in order to evaluate a model for further scale-up. Reach, adoption, implementation, and maintenance are determinants of whether this strategy is viable and should be considered for wider scale and long-term adoption in Mozambique and elsewhere.
Targeting the highest volume facilities in each district allowed for high coverage and geographic coverage throughout a province of over two million inhabitants while working in a relatively low proportion of public facilities. The 29% of public facilities in Manica province included in the program covered 53% of first antenatal care visits and 52% of institutional deliveries during the study period. High utilization of public sector health services enabled a high proportion of total pregnancies and births in the province to be reached. The sampling strategy involved selecting the three highest volume health facilities from each district, which resulted in SAIA-SCALE reaching rural populations as some districts in Manica province are predominantly rural. However, our sampling strategy did not reach the smallest and most remote facilities. Sampling the highest volume facilities in the province as a whole (rather than by district) would have resulted in reaching an even higher proportion of pregnancies and births, but at the cost of not reaching rural populations throughout the province. Focusing on higher volume facilities also resulted in covering higher density areas where HIV prevalence is typically higher, which aligns well with the goal of targeting a PMTCT program to where it is most needed. In order to scale this approach province-wide, additional resources would be required that would likely make the program less efficient, where district supervisor would need to travel to smaller, more remote facilities, which would increase costs and the burden on their time. Remote mentorship via audio or video calls is one potential strategy for reaching more remote facilities that has shown promise in previous studies,[38,39] and is a priority area for further research.
Organizational readiness survey results indicate that facility staff felt highly committed and confident in their peers’ ability to implement the SAIA strategy. This aligns with adoption indicators – as 100% of districts and facilities included in the study adopted the SAIA intervention. Given 100% adoption, change commitment and change efficacy cannot be assessed as determinants on adoption in this study. However, our assessment of change commitment and change efficacy demonstrated little variation across study facilities. Given that our study targeted district health systems leadership as disseminating agents, included buy-in from provincial leadership, and included support from an external non-governmental organization that has worked in Manica province since 1987, it is logical that both district and facility staff universally adopted the intervention. In Mozambique, health policy is set centrally and disseminated throughout the public healthcare system (including provincial, district, and facility levels). It is likely that this in such hierarchical systems the primary driver of adoption is buy-in from higher levels rather than healthcare workers’ commitment or motivation. Thus, in Mozambique the success of further scale-out efforts may be dependent on initially targeting buy-in at provincial and district levels.
Fidelity—defined as mentorship and improvement cycles being carried out—was high during both the intensive phase and during the maintenance phase with limited external support. Mentorship visits, workplan development, and implementation of proposed workplan tasks were relatively stable from the intensive implementation to the maintenance phase, and adoption of micro-interventions into routine practice decreased only slightly from 70% to 62%. Mentorship visits during intensive implementation slightly exceeded the target of one visit per facility-month, then decreased to one visit per facility-month during the maintenance phase. The number of micro-interventions proposed per workplan decreased from the intensive to maintenance phase. This could indicate facilities being less ambitious when external staff were not present at workplanning meetings, or may reflect an increase in experience over the course of the study and a corresponding shift toward more feasible and efficient workplanning. There was a slight decrease in workplans being implemented, and slight decrease in micro-interventions being adopted into routine practice (six and eight percentage points, respectively), which could indicate a decay in motivation over time due to the long duration of exposure to SAIA and/or due to external research staff not being present at monthly SAIA cycles. An alternative explanation is that some level of saturation had been reached in which the largest issues had been addressed or the issues remaining were more difficult to fix and thus ultimately proposed solutions were abandoned. However, as workplans were fully implemented and micro-interventions adopted the majority of the time in both implementation and maintenance phases, and given the minor decrease observed from implementation to maintenance phase, it is unlikely that this decrease reflects a meaningful decrease in motivation. Rather, this indicates that the reduced level of support during the maintenance phase was sufficient to support continued implementation. Staff turnover at the district level would likely halt implementation during the maintenance phase, in which case a booster intervention and additional training would be required.
Approximately half of the micro-interventions tested related to educating patients (including calling new mothers to encourage retention in care). This could indicate that micro-interventions were selected to address one of the main barriers to successful PMTCT scale up, which is postpartum retention in care.[40] However, it could also indicate focusing on individual patient responsibility rather than more costly and difficult to address systemic issues. Focusing on patient education over systemic issues is not unique to our study. A review of published knowledge translation interventions found that patient education was the most common type of intervention, with 58% of interventions being at the patient level (versus health system or providers).[41] The second most common category of micro-interventions in SAIA-SCALE was service organization (including service reorganization such as task shifting). This could reflect the success in other settings of using alternative entry points to engage women, such as community-groups and mother-to-mother support groups.[5]
Transitioning from an efficacy trial to a scaled-up implementation strategy and effectiveness trial is challenging and not always successful. A recent study in Uganda found that the intensive “demonstration” phase facilities had sustained outcomes, while scale-up facilities had outcomes decline after implementation ended.[42] In our study, minimal sustained funding led to sustained program outputs, which may increase the likelihood of sustained PMTCT cascade outcomes. Maintenance of the SAIA-SCALE program up to 24-months post intensive support was high. A systematic review found that the most commonly reported barrier to program sustainment was funding ending.[32] Our study is an example of maintenance happening despite a large reduction in external support, while maintaining funding for key input (transportation). There are other factors that could have led to high maintenance. A recent review found that progress monitoring, stakeholder participation, integration with existing policies, and training can contribute to sustainability,[43] all of which were present in the SAIA-SCALE program. Future research could study maintenance for longer than 24 months, and compare different levels of external support to find the most efficient sustainment strategy as well as determinants of sustainment beyond just financial support – all of which are recognized research priorities.[44,45]
Limitations
While the stepped wedge cluster randomized trial design enabled measuring maintenance for 24 months for wave 1 facilities, it also restricted measuring maintenance to 12 months for wave 2 facilities, and to no maintenance measurement for wave 3 facilities. Full adoption is a positive outcome for the program, but limited our ability to assess predictors of adoption. Two implementation and maintenance measures were only available as self-reported data: Workplan implementation and micro-intervention adoption into routine practice; self-reported data is subject to recall and desirability bias.