Do early-life circumstances predict late-life suicidal ideation? Evidence from SHARE data using machine learning

Area of Science:

• Gerontology
• Psychiatry
• Machine Learning in Health

Background:

• Suicidal ideation in late life is linked to early-life circumstances, but their predictive power requires further clarification.
• Existing research suggests a connection, yet a comprehensive evaluation of specific early-life factors is lacking.

Purpose of the Study:

• To employ a machine learning approach to assess the predictive importance of 32 early-life circumstances across six domains on suicidal ideation in old age.
• To identify key early-life factors that significantly influence the risk of suicidal ideation in later life.

Main Methods:

• Utilized data from the Survey of Health, Aging and Retirement in Europe (SHARE), a cross-national longitudinal survey.
• Applied the XGBoost machine learning model to analyze recalled early-life circumstances (SHARE wave 7) and reported suicidal ideation (SHARE wave 8) in adults over 50.
• Evaluated 32 circumstances spanning socioeconomic status, health, healthcare, and relationships.

Main Results:

• The XGBoost model demonstrated strong predictive performance (AUC=0.80, accuracy=0.77).
• Key predictors of late-life suicidal ideation included childhood relationship quality, socioeconomic status, health, and healthcare experiences.
• Specifically, having a supportive peer group in childhood emerged as a critical influencing factor.

Conclusions:

• Early-life circumstances demonstrate a modest but significant predictive capacity for suicidal ideation in later life.
• Findings underscore the potential for early intervention and preventive strategies targeting childhood experiences to mitigate risks in older adults.