# Revisiting the social determinants of health with explainable AI: a cross-country perspective

## TL;DR

This study uses explainable AI to analyze how various social and health factors predict mortality across different countries.

## Contribution

The study introduces a self-devised algorithm to uncover consistent and context-specific mortality predictors using cross-country data.

## Key findings

- Demographic and socioeconomic factors are consistently significant predictors of mortality across datasets.
- Individual risk factors show notable differences, highlighting the context-specific nature of predictors.
- The algorithm integrates explanation and prediction to reveal domain-level patterns in mortality.

## Abstract

In social science and epidemiological research, individual risk factors for mortality are often examined in isolation, while approaches that consider multiple risk factors simultaneously remain less common. Using the Health and Retirement Study in the United States, the Survey of Health, Ageing and Retirement in Europe, and the English Longitudinal Study of Ageing in the UK, we explore the predictability of death with machine learning and explainable AI algorithms, which integrate explanation and prediction simultaneously. Specifically, we extract information from all datasets in 7 health-related domains, including demographic, socioeconomic, psychology, social connections, childhood adversity, adulthood adversity, and health behaviors. Our self-devised algorithm reveals consistent domain-level patterns across datasets, with demography and socioeconomic factors being the most significant. However, at the individual risk-factor level, notable differences emerge, emphasizing the context-specific nature of certain predictors.

This article is part of a Special Collection on Cross-National Gerontology.