Abstract:
Background Globally, droughts are occurring more regularly and are having negative effects on population health,
particularly in countries such as Kenya, where agriculture is a primary driver of the economy and a source of
subsistence for many communities. Children are particularly susceptible to weather-related shocks. Previous research
has shown an association between drought and cross-sectional indicators of malnutrition, such as stunting. In this
study, we explored various longitudinal and spatial analysis approaches to evaluating the effect of drought on height-
for-age z-scores (HAZ) over time and space among young children in Kenya.
Methods Using anthropometric data from three georeferenced Kenyan Demographic and Health Surveys (KDHS)
and the self-calibrated Palmer Drought Severity Index, we developed multivariate linear regression and spatial lag and
error models (with Moran’s I calculations) to investigate the association between drought severity and HAZ in
children aged 0–5 years. Initial covariates included age of the child, sex, maternal age, height and education, wealth
index, urban or rural location, and size at birth. We then did multilevel and geographically weighted regression
modelling using frequentist or Bayesian methods and with inclusion of household-level covariates, such as livelihood
zones. To assess the effect of changes in drought severity on child HAZ over time, KDHS data from 2003, 2008–09,
and 2014 were analysed with spatiotemporal modelling.
Findings Preliminary results from the multivariate linear model showed a negative, non-significant association
between drought severity and HAZ among Kenyan children in 2014 (β=0·033, p=0·101); however, there was a
significant interaction between drought severity and age (β=–0·002, p<0·0007). The spatial lag model gave similar
results. Other variables associated with HAZ included wealth index, age, sex, maternal education, and maternal
height. Global Moran’s I calculations indicated that there was a slight positive spatial autocorrelation across child
HAZ (I=0·047, p<0·0001).
Interpretation Increased drought severity was associated with a non-significant decrease in HAZ among Kenyan
children. However, a significant interaction between age and drought severity indicates that the effect of drought on
HAZ varies by age. Findings from this study will help to inform the development of methodological approaches for
improving understanding of the effects of climate change on child health. Expanding these analyses to other east-
African countries will also contribute to the development of national adaptation strategies and planning in anticipation
of increased climate variability.
Funding Canadian Institutes of Health Research (Institute of Public and Population Health).