Comprehensive Summary
This study, published in the American Journal of Epidemiology, examines how social determinants of health influence mortality among older adults across the United States, the United Kingdom, and Europe, using machine learning and explainable AI. Drawing on three large longitudinal datasets, the Health and Retirement Study (HRS), the English Longitudinal Study of Ageing (ELSA), and the Survey of Health, Ageing and Retirement in Europe (SHARE) researchers analyzed predictors of death from seven domains: demography, socioeconomic status, psychology, social connections, childhood adversity, adulthood adversity, and health behaviors. Across all countries, mortality was found to be highly predictable, with demographic and socioeconomic factors consistently ranking as the strongest contributors. However, the relative importance of other risk factors, such as smoking, physical activity, and occupational status, varied by context. For example, current smoking was a key predictor in continental Europe, while lower education and occupational status carried more weight in the U.K. Age and gender remained universally important, but the influence of health behaviors diminished in very old age.
Outcomes and Implications
This research underscores the central role of socioeconomic inequality in shaping health and survival, supporting policies that address disparities in income, education, and access to resources as key levers for reducing mortality risk. The findings also highlight the need for context-specific health policies, since risk factors differ across countries and cultural settings. Methodologically, the study demonstrates the value of explainable AI in epidemiology, showing how machine learning can provide both high predictive accuracy and interpretable insights for policymakers and clinicians. Practically, the results suggest that preventive interventions may be most effective when targeted earlier in life, before the predictive power of behaviors like smoking and physical activity declines with age. Future research should expand to low- and middle-income countries and incorporate biological data, offering a more comprehensive global perspective on how social and biological factors interact to drive health inequalities.