Predicting posttraumatic stress disorder following a natural disaster

Area of Science:

• Psychiatry and Psychology
• Data Science and Machine Learning
• Disaster Mental Health

Background:

• Earthquakes are significant natural disasters with a high incidence of survivors developing posttraumatic stress disorder (PTSD).
• Existing research has identified risk factors for PTSD but lacks an optimized prediction tool for post-earthquake mental health triage.
• Effective triage is crucial for allocating limited mental health resources to survivors most in need.

Purpose of the Study:

• To develop and validate an optimized post-earthquake PTSD prediction tool using advanced machine learning methods.
• To identify key risk factors that can be assessed early after a disaster to predict PTSD development.
• To improve the accuracy of PTSD risk assessment compared to existing methods for disaster survivors.

Main Methods:

• Utilized a two-wave survey dataset of 23,907 Chilean individuals exposed to the 2010 earthquake.
• Applied super learning, an ensemble machine learning technique, to a set of 67 risk factors assessed within one week post-event.
• Assessed probable post-earthquake PTSD using the Davidson Trauma Scale and evaluated model performance using cross-validation and ROC curves.

Main Results:

• The super learner algorithm demonstrated superior predictive performance compared to 39 individual algorithms, including logistic regression.
• Achieved an area under the receiver operating characteristic curve of 0.79, outperforming existing post-disaster PTSD risk tools.
• The top 5%, 10%, and 20% of individuals by predicted risk score accounted for a significant proportion (17.5%, 32.2%, 51.4%) of all probable PTSD cases.

Conclusions:

• Developed a highly accurate and optimized post-earthquake PTSD risk score using super learning, suitable for near-term implementation.
• The study validates the utility of super learning for creating robust prediction models for mental health outcomes in disaster contexts.
• This tool can significantly enhance the triage of earthquake survivors to appropriate mental health interventions.