This study used a unique data source, the maternity record book, to assess whether a referral to antenatal care by a community health worker was associated with IPTp3 uptake. This is of high importance in sub-Saharan Africa, where health systems face both human resource and financial constraints. However, the low uptake of IPTp3 reported in this study, which is similar to that reported in some sub-Saharan African subregions, [11,12,13] indicates that more improvements in coverage must be made in order to reach the Sustainable Development Goal Target 3.3, to end the malaria epidemic (in addition to the AIDS and tuberculosis epidemics). We found that pregnant women who were referred to ANC by a CHW had 1.6 times greater odds of completing IPTp3. Furthermore, there were no differences in referrals by maternal characteristics, indicating that the CHWs applied the referral procedures equally, rather than to selective groups of women. Our findings also suggest that receiving IPTp from a CHW did not disrupt ANC attendance or result in lower ANC attendance. This further strengthens the evidence base for the continued scale-up of community-directed interventions to close the gap between ANC attendance and IPTp3 uptake.
Although they have no medical training, CHWs have potential to improve access to maternal and child health interventions, including IPTp [14]. This study demonstrated that women referred by CHWs were more likely to receive IPTp3 than those who were not referred. Previous studies in Nigeria, Burkina Faso, and Malawi have shown that the implementation of community-based IPTp increased coverage of two or three doses (dependent on study) of IPTp [8, 9, 15, 16]. Similar to the findings in Nigeria and Burkina Faso, C-IPTp did not distract from ANC attendance in this setting; there were no differences in the number of ANC visits between women who received community distributed IPTp and those who only received IPTp at a facility. This is different from the findings in Malawi, where the attendance of at least two ANC visits dropped significantly in the intervention villages (87.3 to 51.5% of women). The authors hypothesized that this could be due to health systems issues (efficiency and professionalism of care), which resulted in women preferentially accessing care through the CHWs rather than at the facilities. It is possible we did not see the same impact on ANC attendance due to the different setting and health systems, which did not present the same deterrents to care. Likewise, a critical component of the approach in TIPTOP was a focus on ANC, including refresher trainings for facility-based providers, engagement of facilities in supporting and mentoring CHWs, and strengthening linkages between communities and facilities. These parts of our strategy likely contributed to C-IPTp not having a negative impact on ANC attendance. A qualitative study in Burkina Faso found that with proper training and supervision, the delivery of IPTp by CHWs was both feasible and acceptable as reported by facility-based health workers and CHWs [17]. Taken together, the quantitative and qualitative evidence supports continued scaling up of the C-IPTp model.
There were significant differences in IPTp3 coverage by health facility, where the proportion of women receiving IPTp3 ranged from 0 to 65%. Nearly 15% of the variation in the IPTp3 outcome was attributed to unobserved facility characteristics. Other studies have outlined numerous health-system level barriers to IPTp uptake, including stock outs, increased fees for SP doses, distance to the facility, long waiting times at the facility, unclear IPTp protocols and insufficient provider training [18,19,20,21]. It’s likely that these barriers affected the 25 facilities in our sample to varying degrees, though unfortunately we were unable to assess their impact in our study. Further research that elucidates whether and how the community distribution of IPTp model mitigates these barriers would be beneficial for future programming and policy efforts across sub-Saharan Africa.
In our study, timing of the first ANC visit was not significantly associated with IPTp3 receipt. Unlike other studies that have shown that earlier attendance results in a higher likelihood of IPTp receipt [18, 22, 23], the trend in our population showed slightly increased odds of IPTp3 for women who presented in the second trimester compared to the first, though these were not statistically significant. A possible explanation is that because IPTp is initiated as early as possible in the second trimester, if a woman that initiates ANC in the first trimester but does not return for the appropriate number of visits, she may in fact have fewer opportunities to receive SP than a woman who presented in the second trimester and attended the same number of visits. In our study, only 18.38% of women attended four or more antenatal care visits, which may have played a role in the trend described above. This proportion is far lower than the most recent Demographic and Health Survey estimate of 57% of women attending four or more ANC visits [4]. A possible explanation for the discrepancy is the DHS collects antenatal care data by maternal report of her most recent live birth in the last 5 years, which introduces the possibility of recall and social desirability bias. This may result in an overestimation of the number of visits, if the woman believes that reported a greater number of visits would be viewed as a favorable response by the interviewer [24]. This potential bias would not apply to our data source, the MRB, as it is completed by health care workers in real time. An alternate explanation, given the missing data for other variables in the analysis, is that the MRB data is incomplete, resulting in an underestimation of the number of ANC visits a woman attended. We also found that women who received malaria-related education during ANC were more likely to receive IPTp3, but this was only significant in the bivariate model. This aligns well with other studies that have demonstrated that knowledge about IPTp and the risk of malaria during pregnancy is associated with increased uptake of IPTp [18, 25]. Unlike other studies the present study was unable to explore the effect of socioeconomic status, education level and marital status on uptake of IPTp due to missing data.
An aspect of the intervention that was not captured by the MRB was the counseling on the benefits of ANC by the CHW to the pregnant woman prior to administering the referral, which likely was a key motivator for ANC attendance. A recent meta-analysis reported that CHW interventions, including counseling, increased knowledge and ANC utilization amongst pregnant women [26]. For future research, it would be beneficial to capture the counseling component in addition to the referral, to examine the role of counseling by CHWs and whether there it results in a difference in ANC attendance and IPTp uptake.
A strength of this study is the use of the data source, the maternal record booklet, which allowed for individual-level data that covered the course of the woman’s pregnancy. To our knowledge, this is the first time MRB data were used outside of clinical care. This allowed for the analysis of individual-level, regularly collected data, which is not available in more standard HMIS data sources where the data are only in aggregate and that avoids the potential limitations and biases from household surveys, as referenced earlier. In line with the global recognition of the importance of strengthening routine health information systems, the MRB source must be strengthened as well for future research and programmatic efforts, as the high missingness of maternal characteristic data indicates there are some quality issues. A second strength of this study is that we used a multi-level model to account for facility-level differences, which produces more robust effect estimates. A limitation of the analysis is that although the outcome and primary covariate data recorded in the MRB were complete, as mentioned earlier, a substantial number of observations had missing information for maternal characteristics (e.g., education, occupation) that did not allow for their inclusion in the analysis. A second limitation is that the MRB data source did not allow for the inclusion of facility characteristics in the model, such as availability of trained providers or facility readiness, which could have an impact on IPTp3. A final limitation is that there may be some selection bias, as we could not visit all sites initially considered due to instability issues, which could mean that those women and facilities included in the analysis are systematically different from those excluded, limiting the generalizability of the results.