Modeling causes of death: an integrated approach using CODEm.

Area of Science:

• Epidemiology
• Biostatistics
• Public Health

Background:

• Accurate cause of death data is vital for public health decision-making and priority setting.
• Existing cause of death data often suffers from unavailability and comparability issues.
• Developing robust models for cause of death estimation is crucial for understanding health trends.

Purpose of the Study:

• To propose general principles for developing, validating, and reporting cause of death models.
• To introduce and detail the Cause of Death Ensemble model (CODEm) as an implementation of these principles.
• To improve the accuracy and reliability of cause of death trend estimation.

Main Methods:

• Developed the Cause of Death Ensemble model (CODEm), an analytical tool for estimating cause of death trends.
• Employed a covariate selection algorithm to identify plausible model combinations.
• Utilized four model classes, including mixed effects linear models and spatial-temporal Gaussian Process Regression.
• Assessed models using out-of-sample predictive validity and combined them into an ensemble for optimal performance.

Main Results:

• Ensemble models, including CODEm, demonstrated superior performance over single component models.
• CODEm excelled in root mean square error, predicting correct temporal trends, and achieving 95% prediction interval coverage.
• Detailed results were presented for maternal mortality, cardiovascular disease, and several cancers.

Conclusions:

• CODEm provides enhanced estimates of cause of death trends compared to previous methodologies.
• The ensemble approach reduces susceptibility to bias in model specification.
• CODEm proves effective for estimating trends in major causes of death, supporting public health initiatives.