It takes a village: A pilot cross-randomized trial to enhance pregnancy care and support in northern Ghana

Abstract

Background: Maternal mortality remains a challenge in Ghana, where 263 women per 100 000 live births die during pregnancy or childbirth. Barriers to reaching the recommended antenatal care (ANC) include poor access to quality health care, cultural factors, and lack of support for pregnant women.

Methods: We piloted two cross-randomized interventions: durbars, or local community meetings that incorporated education about ANC and supporting pregnant women, and an enhanced ANC model that added phone calls and a home visit to standard care. The study took place in 30 villages in the Upper East Region of Ghana between August 2021 and November 2022.

Results: We tracked 277 women during pregnancy, with 120 women randomly assigned to the enhanced ANC intervention and 142 women living in villages randomized to the durbar intervention. Women who were randomized to the enhanced ANC intervention were 13.2 percentage points (pp) (95% confidence interval (CI) = 0.1, 24.3) more likely to have average or above average knowledge of pregnancy danger signs, 22.1 pp (95% CI = 9.1, 36.5) more likely to have a birth plan, 28.2 pp (95% CI = 13.0, 42.4) more likely to prepare the plan with their partners, and 16.4 pp (95% CI = 0.9, 29.3) more likely to pay for delivery with funds set aside in birth plan. They were also 35 pp (95% CI = 16.1, 48.1) more likely to make blood donor arrangements than control women who made birth plans. We found no impact of the durbar intervention on study outcomes.

Conclusions: Evidence from this pilot suggests that interventions that increase interactions between health providers and pregnant women outside of the health facility may substantially improve women's experience during pregnancy and maternal health outcomes. Providing structured ways for men to get engaged in ANC increased their involvement. Although we found no evidence the community meetings improved study outcomes, larger studies with repeated meetings and community-wide surveys are needed to make causal conclusions.

Registration: American Economic Association RCT Registry: 10360; ISCRNT: ISRCTN95961119.

Figure 1

Map of Upper East Region of Ghana showing health facilities and villages included in the study. The map displays health facilities where the intervention was delivered (red cross), villages that were randomly assigned to durbar intervention (squares), and control villages where no durbars were held (triangles).

Figure 2

Topics discussed during enhanced ANC intervention based on a sample of 120 women randomly assigned to the enhanced ANC intervention. Each bar represents the proportion of respondents in the enhanced ANC intervention who said they discussed the topics during phone calls and home visits.

Figure 3

No evidence the durbar intervention affected study outcomes related to first ANC visit based on 277 women participating in the study. Blue bars represent the control group mean. Red bars represent the control group mean plus treatment effect of the durbar intervention. The treatment effect, which can be interpreted as pp difference from the control mean, is also reported as β above the bars. The error bars are 95% CIs around the treatment effects. The outcomes are labeled below bars along the x-axis. Each outcome was estimated using a separate regression model that controlled for primiparity, age and age squared, education, partner's age and education, religion, wealth index quintile, and catchment fixed effects. Standard errors were clustered at the village level and bootstrapped by the wild method. Regression results are also presented in Table S6 in the Online Supplementary Document

Figure 4

Impact of enhanced ANC intervention on study outcomes based on 277 women participating in the study. Blue bars represent the control group mean. Red bars represent the control group mean plus treatment effect of the durbar intervention. The treatment effect, which can be interpreted as pp difference from the control mean, is also reported as β above the bars. The error bars are 95% CIs around the treatment effects. The outcomes are labeled below bars along the x-axis. Each outcome was estimated using a separate regression model that controlled for primiparity, age and age squared, education, partner's age and education, religion, wealth index quintile, and catchment fixed effects. Standard errors were clustered at the village level and bootstrapped by the wild method. Regression results are also presented in Table S7 in the Online Supplementary Document. Panel A. Enhanced ANC improved knowledge of danger signs and participation in medical decisions. Panel B. Enhanced ANC improved birth preparedness.

Figure 5

Components of birth plans for women randomized to enhanced ANC intervention compared to control women. Based on 208 women participating in the study who developed birth plans. Each bar represents the treatment effect of the enhanced ANC intervention, which can be interpreted as pp change from the control mean women. 95% CIs calculated around the treatment effect. Each birth plan component was estimated in a separate regression model that was regressed on a binary indicator for enhanced ANC intervention, conditional on having a birth plan, and controlled for primiparity, age and age squared, education, partner's age and education, religion, wealth index quintile, and catchment fixed effects. Robust standard errors were bootstrapped by the wild method.