Accounting for Smoking in Forecasting Mortality and Life Expectancy

Area of Science:

• Demography
• Epidemiology
• Biostatistics

Background:

• Smoking significantly impacts human mortality and life expectancy globally.
• Smoking contributes to nonlinear growth in life expectancy and geographic/sex disparities.
• Accurate mortality forecasts require incorporating the predictable smoking epidemic.

Purpose of the Study:

• To develop a novel Bayesian hierarchical model for forecasting life expectancy at birth.
• To improve life expectancy forecasts by accounting for smoking-related mortality.
• To forecast life expectancy for 69 countries, considering both sexes.

Main Methods:

• Proposed a Bayesian hierarchical model for non-smoking life expectancy at birth.
• Introduced an age-cohort model for the age-specific smoking attributable fraction (ASSAF).
• Converted non-smoking life expectancy forecasts to overall life expectancy forecasts using ASSAF.

Main Results:

• The new method demonstrated improved forecast accuracy compared to four other common forecasting methods.
• Out-of-sample validation confirmed the enhanced performance of the proposed Bayesian model.
• Observed improvements in model calibration, indicating better fit and reliability.

Conclusions:

• The proposed Bayesian hierarchical model effectively forecasts life expectancy at birth.
• Accounting for smoking through ASSAF enhances the accuracy and reliability of life expectancy predictions.
• This approach offers a valuable tool for demographic and epidemiological forecasting.