A machine learning-based screening tool for classifying post-traumatic stress disorder in firefighters: Firefighter...

Area of Science:

• Psychiatry
• Computer Science
• Occupational Health

Background:

• Post-traumatic stress disorder (PTSD) is under-reported in South Korean firefighters due to stigma.
• A machine learning model was developed to address this under-reporting.

Purpose of the Study:

• To develop and validate the Firefighter PTSD Latent Risk Evaluation (FLARE) model.
• To identify firefighters at risk of probable PTSD using demographic and psychosocial features.

Main Methods:

• Utilized data from 52,428 firefighters, including those with and without probable PTSD.
• Applied an extreme gradient boosting (XGBoost) algorithm with selected clinical and occupational factors.
• Evaluated model performance using the area under the receiver operating characteristic curve (AUROC).

Main Results:

• The FLARE model achieved an AUROC of 0.90 for classifying probable PTSD.
• Key features included concentration difficulties, guilt, sleep disturbances, and perceived work pressure.
• Temporal validation confirmed generalizability with an AUROC of 0.91.

Conclusions:

• The FLARE model effectively identifies firefighters at risk for PTSD.
• It utilizes a concise set of clinical and psychosocial indicators.
• This tool can improve PTSD screening and early intervention in stigmatized populations.