Socio-demographic, not environmental, risk factors explain fine-scale spatial patterns of diarrhoeal disease in Ifanadiana, rural Madagascar.


Evans MV(1)(2), Bonds MH(3)(4)(5), Cordier LF(4)(5), Drake JM(1)(2), Ihantamalala F(3)(4)(5), Haruna J(4)(5), Miller AC(3), Murdock CC(1)(2)(6)(7), Randriamanambtsoa M(8), Raza-Fanomezanjanahary EM(9), Razafinjato BR(4)(5), Garchitorena AC(4)(5)(10).
Author information:
(1)Odum School of Ecology, University of Georgia, Athens, GA, USA.
(2)Center for Ecology of Infectious Diseases, University of Georgia, Athens, GA, USA.
(3)Department of Global Health and Social Medicine, Blavatnik Institute at Harvard Medical School, Boston, MA, USA.
(4)PIVOT, Ranomafana, Madagascar.
(5)PIVOT, Boston, MA, USA.
(6)Department of Infectious Diseases, College of Veterinary Medicine, University of Georgia, Athens, GA, USA.
(7)Department of Entomology, College of Agriculture and Life Sciences, Cornell University, Ithaca, NY, USA.
(8)National Institute of Statistics, Antananarivo, Madagascar.
(9)Ministry of Health, Antananarivo, Madagascar.
(10)MIVEGEC, Univ. Montpellier, CNRS, IRD, Montpellier, France.


Precision health mapping is a technique that uses spatial relationships between socio-ecological variables and disease to map the spatial distribution of disease, particularly for diseases with strong environmental signatures, such as diarrhoeal disease (DD). While some studies use GPS-tagged location data, other precision health mapping efforts rely heavily on data collected at coarse-spatial scales and may not produce operationally relevant predictions at fine enough spatio-temporal scales to inform local health programmes. We use two fine-scale health datasets collected in a rural district of Madagascar to identify socio-ecological covariates associated with childhood DD. We constructed generalized linear mixed models including socio-demographic, climatic and landcover variables and estimated variable importance via multi-model inference. We find that socio-demographic variables, and not environmental variables, are strong predictors of the spatial distribution of disease risk at both individual and commune-level (cluster of villages) spatial scales. Climatic variables predicted strong seasonality in DD, with the highest incidence in colder, drier months, but did not explain spatial patterns. Interestingly, the occurrence of a national holiday was highly predictive of increased DD incidence, highlighting the need for including cultural factors in modelling efforts. Our findings suggest that precision health mapping efforts that do not include socio-demographic covariates may have reduced explanatory power at the local scale. More research is needed to better define the set of conditions under which the application of precision health mapping can be operationally useful to local public health professionals.