Explaining sleep disorders in children after earthquakes using SHAP modelling: Container City modelling

Area of Science:

• Pediatric Psychology
• Computational Psychiatry
• Disaster Medicine

Background:

• Sleep disturbances and trauma are prevalent in children post-event.
• Container cities present unique challenges for child well-being after disasters.

Purpose of the Study:

• To predict factors influencing post-traumatic sleep disorders in children using machine learning.
• To identify key predictors for sleep problems in children residing in post-disaster container settlements.

Main Methods:

• A descriptive, cross-sectional study involving 1085 children in container cities.
• Utilized logistic regression, random forest, and XGBoost models for prediction.
• Assessed sleep disorders in 55.3% of the participants.

Main Results:

• The XGBoost model demonstrated superior performance with 0.89 accuracy.
• Achieved an F1 score of 0.87, precision of 0.90, recall of 0.85, and NPV of 0.89.
• 'Sleep problem situation' was the most significant predictor.

Conclusions:

• Machine learning effectively predicts childhood sleep disorders post-trauma.
• Findings inform rehabilitation strategies and highlight the need for integrated psychiatric and pediatric nursing interventions.
• Shelter problems during earthquakes were less influential on sleep disorders.