Estimation of mortality rates for disease simulation models using Bayesian evidence synthesis

Area of Science:

• Biostatistics
• Epidemiology
• Health Simulation Modeling

Background:

• Estimating competing risks is vital for disease simulation models, especially for diseases causing significant all-cause mortality.
• Risk factors often influence both disease-specific and other-cause mortality, necessitating integrated estimation approaches.

Purpose of the Study:

• Propose a Bayesian approach for estimating competing risks for disease simulation models.
• Apply this method to model lung cancer, considering its impact on all-cause mortality and shared risk factors.

Main Methods:

• Employed Bayesian evidence synthesis to estimate other-cause mortality stratified by smoking status.
• Utilized US National Health Interview Survey (NHIS) data linked with death registries.
• Fitted cause-specific hazard models for lung cancer, heart disease, and other causes, controlling for demographic variables and smoking status.

Main Results:

• Estimated annual mortality rates for lung cancer and other causes, stratified by age, race, gender, and smoking status (1987-1995).
• Identified black current smokers as having the highest mortality rates.
• Successfully synthesized NHIS data with national vital statistics, addressing data inconsistencies.

Conclusions:

• Bayesian evidence synthesis provides an effective framework for estimating cause-specific mortality rates.
• The method allows for stratification by key demographic factors, enhancing the accuracy of disease simulation inputs.