Predicting Suicide Attempt Trends in Youth: A Machine Learning Analysis Using Google Trends and Historical Data

Area of Science:

• Computational psychiatry
• Digital epidemiology
• Machine learning in public health

Background:

• Suicide is a major cause of death in young people, with limited predictive tools available.
• Predicting suicide attempts (SA) in individuals under 18 remains a significant challenge.
• This study explores the use of Google Trends data for SA prediction in paediatric populations.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting SA in youth.
• To identify reliable predictors of SA using online search data.
• To assess the feasibility of using Google Trends for real-time suicide risk monitoring.

Main Methods:

• Analysis of Relative Search Volumes (RSVs) from Google Trends for suicide risk-related terms.
• Identification of terms strongly correlated with SA rates using Pearson Correlation Coefficients (PCC).
• Development and evaluation of ML models including Random Forest Regression, Support Vector Regression (SVR), XGBoost, and Linear Regression, assessed by PCC, MAE, MSE, RMSE, and MAPE.

Main Results:

• Terms like 'psychiatrist' and 'anxiety disorder' showed strong correlations with SA rates (PCC ≥ 0.90).
• Random Forest Regression performed best (PCC = 0.953), identifying 'burnout,' 'anxiety disorder,' 'antidepressants,' and 'psychiatrist' as key predictors.
• Other models showed varying performance: XGBoost (PCC = 0.446), SVR (PCC = 0.833), and Linear Regression (PCC = 0.947).

Conclusions:

• ML models utilizing Google Trends data show potential for short-term prediction of youth SA.
• Online search data can be a valuable tool for identifying real-time suicide risk in paediatric populations.
• Further research is warranted to refine these predictive models and integrate them into public health strategies.