Menu
Menu
Menu
Menu
Background: Half of global child deaths occur in sub-Saharan Africa. Understanding child mortality patterns and risk factors will help inform interventions to reduce this heavy toll. The Nanoro Health and Demographic Surveillance System (HDSS), Burkina Faso was described previously, but patterns and potential drivers of heterogeneity in child mortality in the district had not been studied. Similar studies in other districts indicated proximity to health facilities as a risk factor, usually without distinction between facility types. Methods: Using Nanoro HDSS data from 2009 to 2013, we estimated the association between under-5 mortality and proximity to inpatient and outpatient health facilities, seasonality of death, age group, and standard demographic risk factors. Results: Living in homes 40–60 min and > 60 min travel time from an inpatient facility was associated with 1.52 (95% CI: 1.13–2.06) and 1.74 (95% CI: 1.27–2.40) greater hazard of under-5 mortality, respectively, than living in homes < 20 min from an inpatient facility. No such association was found for outpatient facilities. The wet season (July–November) was associated with 1.28 (95% CI: 1.07, 1.53) higher under-5 mortality than the dry season (December–June), likely reflecting the malaria season. Conclusions: Our results emphasize the importance of geographical proximity to health care, distinguish between inpatient and outpatient facilities, and also show a seasonal effect, probably driven by malaria.
Nanoro HDSS site was established in 2009 by the Clinical Research Unit of Nanoro (CRUN), located in the Centre Medicale Saint Camille de Nanoro (CMA), with the goal of evaluating population demography and health living conditions within the health district [22]. Nanoro is located about 85 km from the capital city, Ouagadougou. The Nanoro Demographic Surveillance Area (DSA) lies within the health district of Nanoro and includes 24 villages. All the households within the HDSS area participated in the survey. Initial census started from March to April 2009, and recorded housing, demographic, socio-cultural, and socio-economic characteristics of 54,781 individuals. Since then, census follow-up has been carried out every 4 months. Data collected at the individual level include births, deaths, pregnancies, in/out-migrations (temporary or permanent), and relationships (mother, father, and head of household). Data from 2009 to the end of 2013 were included in this analysis. Nanoro has two main seasons: a rainy season from June to October and a dry season from November to May [22]. In this study, to reflect the malaria mortality seasonality and the potential lag effect of rainy season, the wet season was defined as July to November and the other 7 months were defined as the dry season. There are 16 outpatient health facilities in the Nanoro health district and one inpatient health facility close to the village of Nanoro. There is also an inpatient health facility in Bousse just east of the district, which is the closest inpatient facility for some residents in the DSA, and therefore was included in this study (Fig. 1). Nanoro health district is located in the rural center of Burkina Faso. Green dots represent the HDSS households and red crosses represent the health facilities. The maps of Burkina Faso and the Nanoro health district are our own output using python programming software and publicly available administrative layers to visualize local geographical information system (GIS) data. The source of Africa map on the top right corner of figure is: https://whatsanswer.com/world-map/blank-map-of-africa-large-outline-map-of-africa/ Proximity to both inpatient and outpatient health facilities was measured as Euclidean distance, travel time, and walking travel time. Travel time to the most accessible health facility was calculated using a global “friction surface” provided by the Malaria Atlas Project (MAP) at a resolution of 1 km for 2015, which estimates the travel time through every 1 × 1 km grid square on Earth using the fastest feasible surface travel [19]. A companion algorithm calculates the fastest journey time between any two user-provided points. This index may better capture the opportunity cost of travel than Euclidean or network distance and reflects the information humans use to make transport decisions [19]. We also calculated walking travel time by modifying the friction surface developed by MAP, so that all roads received a fixed walking speed of 5 km per hour [19]. Fastest travel time was the main variable used to describe health-facility access in our models. Hereafter we will refer to this variable simply as “travel time.” Models using the other proximity variables are shown in Supplementary Material. We designed an observational study to identify the associations between various risk factors and child mortality. We estimated the survival probability of children under age five over the study’s nearly 5-year period, as 1 minus the product of average age-specific monthly survival rates from birth through 60 months, multiplied by 1000. Cox proportional hazards regression models [23] were used to estimate the association between under-5 survival and demographic, geographic, and seasonal risk factors. These include physical proximity to health facilities, seasonality of death events during the survey, age groups, gender, maternal education status, ethnicity, multiple birth status and religion. The relationship between each of these factors and mortality risk was assessed one at a time as both categorical and continuous variables (when possible). The final multivariable model adjusted for risk factors that were significant on a univariate model and available for the entire dataset. Among demographic factors, mother education status as well as multiple birth status were missing for children born before the start of HDSS data collection, and were therefore not included in the main model, but the estimates of their effect where not missing is shown in the supplementary material Table S1. For each child, the follow-up time was taken as the time an individual was present within the age group during follow-up, which is the time from the date of first event in the survey, birth or enrollment or in-migration until age 5, out-migration, end of 2013, or death. Village was added as a cluster term to the model to estimate a robust variance. All the analyses and the mapping were performed in R using the survival and ggplot2 packages, respectively [24].
