Modeling causes of death: an integrated approach using CODEm

摘要: Data on causes of death by age and sex are a critical input into health decision-making. Priority setting in public should be informed not only the current magnitude problems but trends them. However, cause data often available or subject to substantial comparability. We propose five general principles for model development, validation, reporting. detail specific implementation these that is embodied an analytical tool - Cause Death Ensemble (CODEm) which explores large variety possible models estimate death. Possible identified using covariate selection algorithm yields many plausible combinations covariates, then run through four classes. The classes include mixed effects linear spatial-temporal Gaussian Process Regression fractions rates. All each assessed out-of-sample predictive validity combined ensemble with optimal performance. estimation outperform any single component tests root mean square error, frequency predicting correct temporal trends, achieving 95% coverage prediction interval. present detailed results CODEm applied maternal mortality summary several other death, including cardiovascular disease cancers. produces better estimates than previous methods less susceptible bias specification. demonstrate utility major