Public health data is often highly aggregated in time and space. The consequences of temporal aggregation for modeling in support of policy decisions have largely been overlooked. We examine the effects of changing temporal scale on spatial regression models of pediatric diarrhea mortality patterns, mortality rates and mortality peak timing, in Mexico. We compare annual and decadal level univariate models that incorporate known risk factors. Based on normalized sums of squared differences we compare between annual and decadal coefficients for variables that were significant in decadal models. We observed that spurious relationships might be created through aggregating time scales; obscuring interannual variation and resulting in inflated model diagnostics. In fact, variable selection and coefficient values can vary with changing temporal aggregation. Some variables that were significant at the decadal level were not significant at the annual level. Implications of such aggregation should be part of risk communication to policy makers.
Bibliographical noteFunding Information:
The Etiology, Risk Factors and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development Study (MAL-ED) is carried out as a collaborative study supported by the Bill and Melinda Gates Foundation, the Foundation for the NIH and the National Institutes of Health/Fogarty International Center. S.L. and J.R.N. were also funded by interagency personnel agreements between the Fogarty International Center (NIH) and the University of Colorado at Boulder and Colorado State University, respectively. The authors thank two anonymous reviewers for their constructive critics.
- Disease modeling
- Spatial models
- Temporal aggregation