Interventions to improve adherence to antenatal and postnatal care regimens among pregnant women in sub-Saharan Africa: a systematic review

Following the PRISMA guidelines [10], Fig. 1 illustrates the process for selecting studies that were included in this systematic review. The initial search identified 186 abstracts. After screening these abstracts, 19 studies were accepted for full-text review based on our PICO search strategy. Subsequently, 5 of these studies were still eligible after full text review and data extraction. The other 14 studies were rejected because they did not report treatment effects on ANC or PNC adherence as an outcome of interest.

Table 1 provides a high-level overview of findings from this systematic review. Overall, three ANC/PNC home visit interventions had no effect on ANC attendance, with two out of the three studies judged to have an “unclear risk” of bias and the third considered to have a “high risk” of bias. On the contrary, two behavioral interventions increased ANC uptake and PNC utilization, respectively, with one study assessing ANC judged to have an “unclear risk” of bias and the other assessing PNC considered to have a “high risk” of bias.

Table 2 summarizes the study design, intervention characteristics, and findings for each of the five studies included in this review [17‐21]. Because of the small sample size of publications included and the differences across studies in interventions and adherence measures, there was little commonality to quantify differences between groups or calculate effect sizes that would allow comparisons of findings across studies. Thus, the results of the studies presented in Table 1 and discussed below, outline whether there were statistically significant differences in adherence to ANC/PNC between the treatment arms being compared within each individual study.

All of the studies were cluster randomized controlled trials of interventions to improve maternal and infant health that took place within the past six years. Three were conducted in Uganda [17, 18, 21], one in Ghana [19], and one in Tanzania [20]. Although ANC or PNC adherence was measured in each study, the primary outcome of interest varied across studies, with two focusing on delivery in a health facility [17, 20], one on coverage of key essential newborn care behaviors, such as breastfeeding, thermal care, and cord care [21], one on neonatal mortality and newborn care behaviors [19], and one on ANC attendance explicitly [18].

Three studies implemented a home visit intervention in which trained community health workers visited the homes of identified pregnant women and provided educational counseling on preventive and promotive care during pregnancy [17, 19, 21]. Two of the three included two prenatal and three postnatal home visits [19, 21], while the remaining study included two prenatal home visits and one postnatal home visit [17]. A variety of topics related to preventive and promotive care were covered in the visits. Prenatal visits focused on danger signs in pregnancy, birth preparation, and clean delivery practices promoting the health of the newborn, including hygienic cord care, proper wrapping, early/exclusive breastfeeding, and delayed bathing. Postnatal visits focused on screening for and counseling on maternal and newborn danger signs, as well as encouragement and reinforcement of breastfeeding, skin-to-skin contact, newborn immunization, and prompt care-seeking. In each study, there was no statistically significant difference in ANC attendance between the treatment group that received the home visit intervention and the standard care control group (see Table 2). Importantly, ANC adherence to the WHO recommendations of four ANC visits during pregnancy remained fairly low across treatment arms in the two studies that measured this particular outcome, with rates of 74% adherence amongst the control and 76% amongst the treatment group in the Kirkwood et al. study[19], and rates of 44% adherence amongst the control and 47% amongst the treatment group in the Waiswa et al. study [21].

The remaining two studies implemented behavioral interventions that used planning and incentive schemas, respectively, to promote ANC/PNC uptake among pregnant women. In the Magoma et al. study, ANC healthcare providers helped pregnant women in the treatment group to develop a birth plan prior to delivery [20]. Developing a birth plan involved discussions on place of delivery, importance of skilled delivery care, transport arrangements, funding, possible blood donors should an emergency occur, birth companions, and home support. As a result, PNC utilization in the first month after delivery was higher amongst the intervention group compared to the standard care control group (62% vs. 32%, respectively; 95% CI 15.4 - 47.2, p = 0.0009). Further, women in the treatment group sought PNC approximately three times sooner than those in the control group (see Table 2).

In the final included study [18], the treatment group was exposed to advertisement for ANC, and informed about the availability of a portable ultrasound (pOBU), in three separate conditions: 1) word of mouth advertisement of ANC and pOBU; 2) word of mouth advertisement of ANC and pOBU plus radio advertisement of only ANC; and 3) word of mouth advertisement of ANC and pOBU plus radio advertisement of both ANC and pOBU. ANC uptake was significantly higher among those subjected to word of mouth plus radio advertisement of ANC and pOBU compared to the control group that received word of mouth advertisment of ANC only with no mention of pOBU (65% vs. 11%, respectively, rate ratio 5.86, 95% CI 2.6 - 13.0, p < 0.0001; see Table 2). There were no differences in rate ratio attending ANC among the three variants of the intervention arm, and when comparing each of the first two intervention variants to the control group.

Thus, overall, two of the five studies [18, 20] included in this review demonstrated the effectiveness of behavioral interventions to increase ANC/PNC attendance. The remaining three studies, which implemented a home visit intervention, did not seem to have an impact on adherence to recommended ANC regimens [17, 19, 21].

Table 3 summarizes the results from the Cochrane risk of bias assessment (see Additional file 3 for comprehensive evidence supporting each judgment). Risk of bias was low in most categories across studies. All five studies received “low risk” ratings for selection bias by documenting the use of random sequence generation and adequate treatment allocation concealment. All five studies were also judged to be “low risk” for attrition bias based on the extent of missing data from each group, the reasons provided, and the type of analysis conducted. When evaluating detection bias via blinding of outcome assessors from knowledge of which intervention a participant received, all studies received “unclear risk” ratings due to lack of information provided. The greatest variation across studies was in the performance bias domain, assessing blinding of participants and personnel to treatment allocation, with one study rated as “low risk” because the researchers intentionally masked the presence of pOBU when initial consent was obtained in the control arm [18]; three as “unclear risk” because they do not mention or use a procedure to blind participants and study personnel to treatment allocation [17, 19, 21]; and one as “high risk” because it did not allow blinding of birth plan providers or pregnant women who participated in the study to the treatment allocation [20]. To examine reporting bias, each protocol was checked for discrepancies between outcomes of interest study authors said they would measure and those they report on; three of the five studies [17‐19] received “low risk” ratings in selective reporting, however, the Waiswa et al. study [21] was rated as “high risk” in this category due to the presence of pre-specified intermediate outcomes that appear to be excluded in the published paper. Finally, sources of other bias were “low risk” in four out of five studies [18‐21], with one study judged to have “unclear risk” due to baseline imbalance between treatment arms [17].

Synthesizing the overall level of evidence across domains, three of the five studies [17‐19] included in this review were considered to have an “unclear risk” of bias, while the remaining two studies [20, 21] had a “high risk” of bias.