Developing an optimal short-form of the PTSD Checklist for DSM-5 (PCL-5)

Area of Science:

• Psychiatry
• Psychological assessment
• Machine learning in healthcare

Background:

• Existing Posttraumatic Stress Disorder (PTSD) Checklist (PCL) short forms were developed using heuristic methods.
• There is a need for an optimal short-form PCL for DSM-5 (PCL-5) developed with rigorous statistical and machine learning techniques.

Purpose of the Study:

• To create an optimal 4-item integer-scored short-form PCL-5 using machine learning and conventional scale development methods.
• To evaluate the performance of the developed short-form scale against the full PCL-5 using independent datasets.

Main Methods:

• Utilized independent datasets from the Army Study to Assess Risk and Resilience among Service members for training (n=8,917) and validation (n=11,728).
• Employed exploratory factor analysis, stepwise logistic regression, and a novel machine learning method to develop short-form scales (1-8 items).
• Selected the optimal short-form scale based on area under curve (AUC), calibration, sensitivity, specificity, and net benefit.

Main Results:

• A 4-item integer-scored short-form scale, selected via stepwise regression, demonstrated strong performance.
• The optimal short-form scale achieved comparable calibration (Brier scores 0.015-0.032) and high AUC (0.984-0.994) in the validation sample.
• This scale exhibited the highest net benefit across a range of decision thresholds.

Conclusions:

• The recommended 4-item integer-scored short-form PCL-5 provides diagnoses that closely parallel the full PCL-5.
• This optimal short-form scale is well-suited for screening purposes in clinical and research settings.