Predicting Probable Persistent PTSD Following the Sewol Ferry Disaster: Development of an AI Algorithm Based on...

Area of Science:

• Psychiatry and Mental Health
• Artificial Intelligence in Medicine
• Trauma Studies

Background:

• Persistent post-traumatic stress disorder (PTSD) presents significant challenges in early risk identification.
• Developing effective predictive tools is crucial for timely intervention and mitigating long-term impacts.

Purpose of the Study:

• To develop and validate a novel artificial intelligence (AI)-based predictive algorithm for early identification of persistent PTSD.
• Utilize psychological, biological, and psychosocial data for enhanced risk prediction.

Main Methods:

• A cohort of 88 bereaved family members was assessed, categorized into persistent PTSD (n=67) and remitted (n=21) groups after 4 years.
• Linear discriminant analysis (LDA) was employed, with techniques like Borderline Synthetic Minority Oversampling Technique and recursive feature elimination used for model optimization.
• Psychosocial, demographic, and blood test data were collected at initial evaluation.

Main Results:

• The LDA model achieved high performance (AUC = 0.858), with psychosocial factors being the strongest predictors.
• Increased anxiety, depression, insomnia, and rumination were associated with higher PTSD risk.
• Positive resources and social support demonstrated a protective effect, while physiological markers had limited predictive value.

Conclusions:

• The study supports using brief psychosocial measures with an interpretable AI classifier for practical triage of high-risk individuals.
• Further external validation and biomarker analysis are recommended to enhance clinical utility and generalizability.