Meeting Postpartum Women’s Family Planning Needs Through Integrated Family Planning and Immunization Services: Results of a Cluster-Randomized Controlled Trial in Rwanda

Dear Dr. Hodoglugil,

Thank you for your thoughtful comment; you raise an important issue. The answer to your question, however, is not an easy one. As we know, the premise of the randomized control trial is that, through randomization, most potential threats to internal validity (e.g. history, maturation, testing, instrumentation, selection, and regression) are distributed equally across study groups, creating a situation where groups should differ only on exposure to the intervention being tested so that any observed changes can be confidently attributed to that intervention. Of course, we also know in practice for a given study, randomization may not achieve the desired comparability on important confounding factors, particularly when randomizing small numbers of participants/units. It is for this reason that we took the difference of differences approach to analyzing our data and included important covariates in the model.

In this situation, we did investigate, but were unable to identify any specific, systemic factor that could explain the decrease in the control facilities. What we do know is the following: 1) participating health facilities were first randomly selected from a national sampling frame, with no more than one facility per district selected to reduce the likelihood of contamination across study groups, then we randomly allocated the facilities to study groups (intervention or control), resulting in a national distribution of facilities in both groups - no geographic clustering within study groups to influence the outcomes; and 2) there were no known country-wide issues with regard to stock-outs or other National Family Planning Program problems that would have influenced uptake of FP methods during the time frame in control facilities alone.

We also examined pre-/post-intervention changes in FP prevalence by facility within each group. What we found was that in the control group, 3 facilities experienced little to no change between time points in FP prevalence, 1 facility experienced an increase in FP prevalence and 3 facilities experienced decreases in FP prevalence. Among intervention facilities, 6 facilities experienced an increase in FP prevalence and 1 experienced a decrease in FP prevalence. Based on these results, we assume that, in the absence of the intervention, the net change in FP prevalence across intervention facilities would have mirrored that observed in the control facilities, with similar variation across facilities. When looking at the characteristics of the samples between the study groups and time points, we did not observe any differences that could explain the drop in contraceptive uptake in the control group.

However, the question remains, why did FP prevalence, in particular injectable use, decrease between baseline and endline across the health facilities, beyond factors that could be attributable to study or sampling design? One possible influence on FP prevalence, nationwide, could have been the bi-annual (March and November) Mother and Child Week activities during which FP is promoted as are other health interventions targeting maternal and child health. However, both our baseline and endline data collection periods occurred a short time after these outreach activities, which should have led to relatively equal effects at both time points, but this cannot be known for certain.

In summary, the point you make is one that certainly caught our attention and led to considerable searching for an answer. Though we have tried to tease out the reason for the net decrease in FP prevalence in our control facilities, we were not able to identify a reason that is associated with the allocation of facilities within the study groups; therefore, we can only conclude that the reason must be some unmeasured factor, which should have had influence across study groups, and the intervention that we implemented led to the observed differences between study groups. That said, we fully acknowledge that this is but one study with its limitation of having a small number of primary sampling units to randomize to study groups. Despite our limitations, we do believe that the rigor in our design and methods support our conclusion that integrated FP/immunization services appear to lead to better FP method use among postpartum women. It will be important for future research to confirm or refute these findings.

Sincerely,

Lisa Dulli for the study team

Conflict of Interest:

None declared