Machine learning methods to predict child posttraumatic stress: a proof of concept study.

Area of Science:

• Computational psychiatry
• Pediatric psychology
• Machine learning applications in healthcare

Background:

• Accurate prediction of childhood Posttraumatic Stress Disorder (PTSD) is crucial for effective intervention.
• Machine Learning (ML) has shown promise in disease prediction but has not been applied to childhood PTSD.
• This study explores the feasibility of using ML for predicting childhood PTSD.

Purpose of the Study:

• To determine if ML methods can create accurate predictive classification models for childhood PTSD.
• To identify specific variables with potential causal links to PTSD from these models.

Main Methods:

• Applied ML predictive classification with causal discovery feature selection to a dataset of 163 children hospitalized with injuries.
• Collected 105 biopsychosocial risk factor variables at hospitalization.
• Assessed PTSD symptoms three months post-discharge.

Main Results:

• A predictive classification model for childhood PTSD was successfully developed with significant accuracy.
• Models using subsets of causally relevant features performed comparably to models using all variables.
• Causal Discovery identified 58 variables, with 10 being the most stable predictors.

Conclusions:

• This proof-of-concept study demonstrates ML's capability in predicting childhood PTSD and identifying causal variables.
• The applied techniques offer a valuable addition to the methodological tools for PTSD research.
• Future research should focus on replicating, refining, and expanding these findings.