LAUSR

Background Socio-economic and demographic determinants of child growth at ages 0–5 years in developing countries are well documented. However, Precision Public Health interventions and population targeting require more finely grained knowledge about the existence and character of temporal changes in child growth associations. Methods We evaluated the temporal stability of associations between height-for-age z-score (HAZ) of children aged 0–59 months and child, parental, household, and community and infrastructure factors by following 25 countries over time (1991–2014) in repeated cross-sections of 91 Demographic and Health Surveys using random effect models and Wald tests. Results We found that child growth displayed relatively more time stable associations with child, parental, and household factors than with community and infrastructure factors. Among the unstable associations, there was no uniform geographical pattern in terms of where they consistently increased or decreased over time. There were differences between countries in the extent of temporal instability but there was no apparent regional grouping or geographic pattern. The instability was positively and significantly correlated with annual changes in HAZ. Conclusions These findings inform about the generalizability of results stemming from cross-sectional studies that do not consider time variation – results regarding effects of child, parental, and household factors on HAZ do not necessarily need to be re-evaluated over time whereas results regarding the effects of infrastructure and community variables need to be monitored more frequently as they are expected to change. In addition, the study may improve the Precision Public Health population targeting of interventions in different regions and times – whereas the temporal dimension seems to be important for precision targeting of community and infrastructure factors, it is not the case for child, parental, and household factors. In general, the existence of temporal instability and the direction of change varies across countries with no apparent regional pattern.