Exploring Multilevel Determinants of Stillbirth: A Comprehensive Analysis Across Sub-Saharan African Countries

Study design and overview

Hypothesis: Individual-level risk factors for stillbirth, such as maternal age and parity, interact with contextual factors, such as access to healthcare services, socioeconomic status, and household dynamics.

The study drew from a retrospective dataset from the DHS, collected every five years using pretested validated quantitative tools and structured methodologies.¹⁵ The DHS’s two-stage sampling procedure consists of a primary survey unit; participants are randomly selected from clusters in each country.¹⁶ The DHS sampling methodology has been published in the literature.¹⁷ The survey method’s consistency over time and across countries made a multi-country analysis possible. DHS datasets used in this study are publicly available on the DHS website and can be downloaded for free upon request via https://dhsprogram.com/data/available-datasets.cfm.

Countries lacking the variable (v-cal) for stillbirth calculation, including Cameroon, Congo, Cote d’Ivoire, Chad, and Togo, were excluded from the analysis (Figure 1). We extracted this study’s data from the most recent nationally representative DHS iterations of 29 SSA countries with information on SSA birth histories between 2010 and 2019 (Figure 2). The sample represents 59% of SSA countries (29/49). We pooled the data from individual recode files in each country that reported at least one birth in the five years before the surveys and included a live-term birth or stillbirth. Our total sample was 197,329; samples for specific countries are outlined in Table 1. Our reporting is guided by the Strengthening Reporting of Observational Studies in Epidemiology (STROBE) guidelines.¹⁸

• Download figure
• Open in new tab

Figure 1. Study eligibility criteria.

• Download figure
• Open in new tab

Figure 2. Factors influencing stillbirth at individual and contextual levels of the socioecological model

Outcome variable

The primary outcome was stillbirth, defined as pregnancies lasting at least 28 weeks that ended in a fetal death^.1 Birth outcome refers to whether a woman had a stillbirth or a live birth in her most recent pregnancy. We input all births within the prior five years by looping through the calendar, summing births, non-live pregnancies, and stillbirths from the births reported in the DHS. A loop iterates the entire (vcal_1) variable (calendar information string) within a five-year window. Within the loop, if the current position substring matches the code for stillbirth (“TPPPPPP”), the ‘stillbirths’ count was incremented by 1 (https://www.dhsprogram.com/data/calendar-tutorial/). We restricted our sample to women who gave birth within the last five years (v208), with stillbirth as the outcome variable. Stillbirth was dichotomized: 0 (“no stillbirth”) for absence of stillbirth, and 1 (“stillbirth”) for presence of stillbirth.

The observed stillbirth rate (SBR) was calculated as the ratio of reported stillbirths to total births (including both stillbirths and live births): SBR = sb/ (sb + lb) *1000 live births within a given period. Here, “sb” refers to the number of stillbirths ≥ 28 weeks of gestational age, and “lb” refers to the number of live births regardless of gestational age or birthweight.¹⁹

Explanatory variables

We examined two groups of independent variables, considering the hierarchical structure of DHS data, where women are nested within clusters or communities. Following a comprehensive literature review, we selected individual — and contextual-level explanatory variables available in the DHS datasets. ^20,21 We included eleven variables in the analysis and categorized them into individual- and contextual-level factors.

Individual factors considered were age, marital status, education, parity, and working status. We categorized age range as 15-19, 20-24, 25-29, 30-34, 35-39, 40-44, or 45-49. Education level was ‘no formal education,’ ‘primary,’ ‘secondary,’ or ‘higher.’ Another essential individual-level factor included in this study was parity, with four categories: number of children (1, 2, 3, or ≥4) ²²; marital status (‘never married,’ ‘married,’ ‘cohabitation,’ ‘widowed,’ ‘divorced,’ or ‘separated’); current working status (‘no’ or ‘yes’). These categories were pre-coded in the DHS dataset.

Contextual factors included mass media exposure, including frequency of watching television, reading newspapers/magazines, and listening to the radio. Each mass media variable was categorized as ‘no exposure’ (not at all), ‘low exposure’ (less than once a week or at least once a week), or ‘high exposure’ (almost every day). ²³ The household wealth index was based on the DHS measure of items available in each household. We used principal component analysis (PCA) ²⁴ to group the available items into ‘poorest,” poorer,’ ‘middle,’ ‘richer,’ or ‘richest,’ representing the household wealth index.²³ Place of residence (rural or urban) and travel to the health facility (‘big problem getting to the facility’ or ‘not a problem getting to the facility’) were pre-coded in the DHS dataset. Community-level female literacy was categorized into low, middle, or high literacy levels.²⁵ Drawing on the Socioecological Model,²⁶ Figure 2 illustrates how complex interactions between individual and contextual factors at various levels affect stillbirth.

Statistical analyses

We weighted the dataset per DHS guidelines (v005/1 000 000) to account for the complex sampling structure of the data and used the survey command (svy) in Stata.¹⁷ Sample weight equalization gave equal weight to each survey, so if one survey had a large sample, it did not predominate the pooled results. This process ensured that the sample weights aligned with each country’s clusters and strata. We used STATA software, version 18·0 SE (Stata Corporation, TX, USA) for weighted data analysis.²⁷ We appended all data from SSA DHS after extracting the essential variables. In our analyses, we included participants with complete information on all covariates but excluded a “missing” category.

We sampled sociodemographic variables to generate represented percentages for each country. The weighted datasets for the 29 countries were merged to develop the overall prevalence of stillbirth among women across the study. We used a map to visually represent the varying prevalence of stillbirth by country (Figure 3). We used bivariate analysis to examine the relationship between stillbirth and the explanatory variables. The Pearson chi-square test of independence provided information on the distribution of stillbirth rates across the different variables and indicated any significant associations. We checked for multicollinearity among the explanatory variables using the variance inflation factor (VIF). The results showed no evidence of high collinearity (maximum VIF is 2·07, the minimum VIF is 1·03, mean VIF is 1· 38, and condition number is 30·02).

• Download figure
• Open in new tab

Figure 3. Prevalence of stillbirth in Sub-Saharan African countries

Finally, we analyzed multilevel binary logistic regression using four models (Model O-III). Model O was an empty model without any explanatory variables, serving as a baseline. Model I included only individual-level variables, Model II had only contextual variables, and Model III, the complete model, incorporated individual and contextual-level variables. The regression analysis results were presented as adjusted odds ratios (aOR) with corresponding 95% confidence intervals (CI), and statistical significance was set at p < 0·001 and p < 0·005, indicating associations between the explanatory variables and stillbirth. The analysis accounted for non-response and under-sampling by applying the survey sample weight. We used the Stata command ‘melogit’ to fit these models, as well as the log-likelihood ratio and Akaike’s information criterion (AIC) tests for model comparison.

Ethical considerations

This study was based on a secondary dataset with no identified participant information. The authors obtained and were approved to use the dataset by MEASURE DHS.

Prevalence of stillbirth in SSA

The sample was 197,329 DHS participants from 29 SSA countries (Table 1). The prevalence of stillbirth among women who had given birth within the last five years preceding the interview in SSA was 319·06 per 1000 live births. Burundi recorded the highest prevalence, 21·4/1000 live births, whereas Comoros recorded the lowest, 3·6/1000 live births (Figure 3).

Individual-level predictors of stillbirth among women in sub-Saharan Africa

Age, education, parity, and marital status were significantly associated with a stillbirth at p<0· 001 (Table 2). Several factors were associated with an increased risk of stillbirth among women in SSA countries (Table 3). There was no significant difference in the odds of stillbirth among women in age categories 15-19 and 20-24 compared to those in the 45-49 age category. However, the odds of stillbirth were higher in those aged 25-29, 30-34, 35-39, and 40-44 (aOR=1·98, 95%CI: 1·26 to 3·11), (aOR=2·14, 95%CI: 1·37 to 3·33), (aOR=2·29, 95%CI: 1·45 to 3·60), and (aOR=1·99, 95%CI: 1·24 to 3·50) respectively (Figure 4). Maternal education was negatively associated with the risk of stillbirth. Compared to women with no education, higher maternal education levels were associated significantly with decreased risk of stillbirth with coefficients ranging from 64% to 90% (aOR=0· 90 95% CI: 0·80 to 1·00, and aOR=0·64 95% CI: 0·45 to 0·90 respectively) (Table 3). While single (never in a union) women had significantly lower odds of stillbirth (aOR=0·38, 95% CI: 0·25 to 0·58), married or cohabitating women had higher odds of stillbirth (aOR=1·40, 95% CI: 1·04 to 1·88) and (aOR=1·39, 95% CI: 1· 02 to 1· 90) respectively. Women with higher parity had significantly higher odds of experiencing stillbirth in women with two births (aOR= 0·38,95% % CI of 0·32 to 0·46) or three births (aOR= 0·31,95%CI of 0·26 to 0· 59), compared to women with one birth (Table 3). Women with a job had increased risk (aOR 1· 20 95%CI:1· 08 to 1· 35) compared to women without a job.

• Download figure
• Open in new tab

Figure 4. Individual-level predictors of stillbirth

Contextual (household and community-level) predictors of stillbirth

Place of residence, sex of household head, and household wealth index were significantly associated with a stillbirth at p<0· 001 (Table 2). Living in a rural residential area was a significant risk factor (aOR=1·22 95%CI:1·06 to1·41). Exposure to mass media did not significantly affect stillbirth (aOR 1·01, 95% CI 0· 85to 1·22). When the household head was female, the odds of stillbirth were significantly lower (aOR=0·86, 95% CI: 0·75 to 0·99) compared to households headed by males. Community literacy level was not a significant predictor of stillbirth (Table 3).

Random effects (measures of variation) results

The random effects model (Table 3) indicated a statistically significant variation in determinants of stillbirth across the clusters. In the empty model, there were substantial variations in the likelihood of determinant of stillbirth across the clustering of the primary sampling units (PSUs;1·21(95% CI 1·14 to 1·28). The empty model’s intraclass correlation (ICC) value showed 0·055 of the variation in stillbirth rates attributed to the between-cluster variations of the characteristics. The between-cluster difference dropped to 0·055 in Model I, to 0·054 in Model II, and maintained its value 0·055 in Model III.

These ICC results suggested that the variations in the likelihood of experiencing stillbirth could be attributed to the variances across the clusters. The AIC values exhibited a similar U-shaped pattern as the ICC values, reaching their lowest point in Model III. Therefore, we chose Model III as the most suitable model to analyze the determinants of stillbirth (AIC; 24593·13).

Implications for practice

The study’s findings highlight the importance of interventions to reduce stillbirth rates, particularly among older women, women with lower levels of education, and women living in rural areas. Interventions could include providing education about the signs of stillbirth, improving access to quality healthcare, and promoting safe birth practices. At the community level, awareness campaigns and educational programs can empower women with knowledge about prenatal care and maternal health. In addition, the study’s findings support broader national and international strategies to improve maternal and child health to achieve Sustainable Development Goal 3, which focuses on ensuring healthy lives and promoting well-being for all ages.

Strengths and limitations

The major strengths of this study were the use of recent nationally representative DHS data from 29 SSA countries and the adoption of a multilevel approach to the analyses. This approach allowed for examining associations between stillbirth and individual and contextual-level factors, providing a more comprehensive understanding of the complex determinants of stillbirth across different geographic contexts. Notably, this study serves as the first comprehensive analysis of stillbirth risk factors pooling data from such a large and diverse group of SSA countries.

We acknowledge some limitations. The data was cross-sectional; therefore, no causality could be inferred. While the DHS dataset is large and nationally representative, the stillbirth sample size was relatively small, particularly in specific SSA countries. This limited the statistical power to detect associations between potential determinants and stillbirth. Further research and longitudinal studies are needed to establish a more definitive understanding of the determinants of stillbirth in this context.

Conclusion

Our results indicate that individual characteristics and some social and environmental factors contribute to stillbirth risk. Our findings can inform the development of interventions to reduce stillbirth rates in this region. Policy, community, and health system modifications are imperative to enhance birth outcomes among women in SSA. Health systems should prioritize the development of comprehensive, culturally sensitive maternal health programs that address the unique challenges faced by women in SSA, thereby contributing to improved birth outcomes and reducing stillbirths.