Estimating Child Mortality Risk: Application and Validation of Gaussian Process Regression

A long standing priority in global health has been meeting the health needs of the youngest population, children under the age of 5. Efforts to improve the wellbeing of the population under 5 years of age requires accurate and timely assessment of the current levels and trends of mortality risk in children under the age of 5. Here we present an updated methodological strategy using Spatio-temporal and Gaussian process regressions to synthesize disparate mortality data sources into a coherent time series of 5q0 estimates with 95% uncertainty bounds. We anticipate that in rigorous predictive validity tests this updated modeling strategy will again outperform other synthesis modeling options including LOESS and spline-regression.