Elucidating policies that may affect health outcomes such as life expectancy and child mortality across the globe is of particular interest to many researchers and decision makers. The present study was built on and extended beyond existing literature in the range of conditions considered as well as in the statistical procedures used to derive explanatory models. It includes 161 countries representing all possible economic and demographic conditions in the world.
Although data from several reliable sources were gathered, large amounts of missing data were still encountered. We dealt with this problem by using multiple imputation techniques that exploit the clustered structure of the data. Our methodology was based on Bayes hierarchical multivariate linear mixed models. Imputations derived with regularized Bayes multivariate hierarchical means models are efficient for these types of data.
A main finding was the discovery of three well-delimited groups in health outcomes space, indicating varying demographic conditions and health transitions in the world. The results showed the relevance of education and income distribution in high-birth-rate countries, and of investment in health systems in the remaining countries. The findings indicate the significance of some variables and complex design and analysis that have not been included in previous studies.
This work was done is collaboration with Sue Thomas Hegyvary, and Devon M. Berry. Sue Thomas Hegyvary is Professor and Dean Emeritus, School of Nursing, and Adjunt Professor, School of Public Health and Community Medicine, University of Washington. Devon M. Berry is Assistant Professor, Cedarville University, Ohio.