Background
Since the establishment of the Millennium Development Goals in 1990, there has been substantial progress in reducing child mortality globally, from 93 deaths in 1990 to 39 deaths in 2017 per 1000 live births. Nonetheless, an estimated 5.4 million children under age five died in 2017, out of which 2.5 million died during the first month of their life [
1]. About half of child deaths occurred in sub-Saharan Africa [
2]. In 2015, the Sustainable Development Goals (SDGs) were defined, aiming to reduce under-five mortality in every country to below 25 per 1000 live births by 2030 [
3]. To achieve these targets, urgent action in sub-Saharan Africa is needed and higher-quality information to guide this action [
4]. Among sub-Saharan countries, Burkina Faso, where our study area is situated, has made great progress in reducing under-5 mortality by about 58% from 201 to 84.6 deaths per 1000 live births between 1990 and 2016, but this rate is still much higher than the SDGs [
1].
To track progress towards child survival goals and to plan effective interventions for child health, identifying the major drivers of child mortality as well as data-driven estimates of child mortality are necessary [
4]. However, countries with the highest child mortality burden lack civil registration and vital statistics (CRVS) systems accounting for all births, deaths and causes of death. In these countries, the location and timing of child deaths and the overall death rates, are highly uncertain. What we know about these crucial public-health questions is informed mostly by nationally representative surveys such as the Demographic and Health Surveys (DHS), conducted every several years.
A Health and Demographic Surveillance System (HDSS) is a local CRVS system that routinely monitors the health and demographic characteristics of a population living in a specific area. HDSS data facilitate detailed local studies of public health in general, and child mortality in particular. As of 2020, 49 HDSS sites participate in the International Network for the Demographic Evaluation of Populations and Their Health in Developing Countries (INDEPTH), recording the life events of over 3 million people in 17 African and Asian countries [
5]. Several studies have investigated spatial [
6‐
10], temporal [
11,
12] and demographic [
11,
13,
14] factors affecting child mortality in HDSSs. However, no study to date has analyzed such patterns in the relatively new Nanoro HDSS in rural north-central Burkina Faso.
Risk of child mortality varies over space and time, and it is important to identify the areas at the highest risk in order to focus intervention-based efforts in those areas. One source of heterogeneity is proximity to health facilities [
15,
16]. Poor access to health care remains a concern in many low-income countries [
17]. A growing number of studies have estimated the effect of distance from a health facility upon child mortality. The first meta-analysis of such studies was published in 2012 [
16] and was updated more recently [
18]. They found that living
> 5 km away from a facility is associated with 62% higher neonatal mortality based on 4 studies, and 57% higher under-5 mortality based on 9 studies; both effects were deemed highly significant. In addition, a study aggregating 29 DHSs from 21 countries found that living
> 10 km from a facility was strongly associated with 27% higher odds of neonatal mortality. Both the meta-analyses and the DHS-based study did not distinguish between smaller and larger facilities. Most above mentioned studies used simple Euclidean distance, or local expert opinion about distance or travel time, as the exposure variable. More sophisticated approaches to estimate real-life travel distance or time [
19] have been published only rarely in this context.
Mortality also varies over time as a result of changes in health care-seeking, age and season of birth and death [
11,
12], and environmental conditions [
20]. In the Nouna HDSS, Burkina Faso, infants born during the rainy season were associated with higher mortality risk compared with those born during the dry season [
11]. During the rainy season, flooded roads limit access to health care, especially in rural region. In most of West Africa, the rainy season also coincides with food shortage until the harvest arrives [
11]. Seasonality also drives cause-specific mortality patterns due to malaria, pneumonia, and diarrhea, which were the leading causes of child mortality in Burkina Faso in 2010 [
21]. Previous studies have also found associations between demographic factors such as birth spacing, twin births, ethnicity, maternal age, and child mortalities [
11,
12]. Twin status in the Nouna HDSS, Burkina Faso was strongly associated with infant mortalities [
11]. Children of young mothers also were at higher risk of mortality than older mothers [
12].
Against this background, Nanoro presents some unique research opportunities. It is relatively new (our study begins with its inception in 2009), and thus less susceptible to potential participation (Hawthorne) effects seen in longer-standing HDSSs that had carried out many surveillance and research projects over the years. It is also unique in being completely rural yet hosting a strong tertiary health center in its main village. In addition, recent progress on global proximity estimates provides new tools for quantifying local patterns of access and inequality. Our study presents an attempt to leverage these opportunities, focusing on drivers of heterogeneity in child mortality risk within the Nanoro HDSS.
Discussion
Our study provides insight into child mortality patterns in the Nanoro health district, Burkina Faso by linking it to various demographic, spatial and temporal risk factors. One distinction of our study is the evaluation of proximity to both inpatient and outpatient health facilities. In the recent meta-analysis by Rojas-Gualdrand and Caicedo-Velazquez [
18], the majority of studies included in its under-5 mortality endpoint estimate measured distance from any health center with no distinction between inpatient and outpatient. There were also inconsistencies regarding the effect of proximity to health care on child and neonatal mortalities. In Malawi, DHS data showed no association between distance to delivery care and early neonatal mortality, and in Zambia, early neonatal survival was higher with increasing distance [
25]. On the other hand, analysis of DHS data in Madagascar showed a higher risk of infant mortality among those who lived further from a health facility [
26]. In rural western Burkina Faso, rural Ethiopia and Tanzania, proximity to health facilities was found to be a major risk factor for infant, child and overall under-5 mortality [
15,
16,
27]. Our analysis is in agreement with the latter studies, and indicates that impeded access to an inpatient health facility might be a major risk factor for child mortality. Our study also suggests that proximity to outpatient health facilities does not drive the pattern of child mortality in the study area. We speculate that outpatient health facilities do not provide the level of care children need in a life or death situation. We note the confounding factor that inpatient health facilities are usually located in towns and major villages, with better food, water, and other living conditions for residents, as well as generally higher education and socioeconomic status. Another distinction of our study is the use of the recently developed global proximity map that accounts for the spatial locations and properties of roads, railroads, rivers, water bodies, topographical characteristics, land cover, and national borders [
19]. Accounting for these features leads to a more accurate measurement of proximity than Euclidean or network distance that has been commonly used in previous studies.
There was a significant association between seasonality of death and under-5 mortality, with the wet season having a higher mortality rate, reflecting the malaria mortality pattern. The higher rate of out-migration during the dry season also highlights its potential effect on the child mortality pattern.
Some of the limitations of our work are other risk factors that we have not accounted for and may be important to our outcomes, such as family wealth status, family health-seeking behavior, sanitation and hygiene information, and effects of flooding. Wealth status of households could highlight the distribution of resources and health services in the district. Furthermore, in families with access to improved sanitation facilities, children are less exposed to infectious diseases. Variations in health-seeking frequency between households can also affect child mortality risk. Flooding could also inundate road and therefore limit access to care. Adjusting for each of these factors may help better explain the heterogeneities in the data, and alter the relationship between the proximity to the health facilities and child mortality. Also, the travel time index we used in this study is based on the assumption that everyone might use the fastest travel method possible. However, our analysis using walking travel time showed a similar association with under-5 mortality. An additional limitation is that the friction surface developed by MAP does not account for seasonal variations, which can affect travel time to health facilities. Last but not least, this is an observational study, and therefore any association is subject to potential confounding factors, as discussed above for inpatient facilities.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.