Introduction: Excessive response time remains a barrier to using patient-reported outcome measures in Taiwan’s clinical settings. Using the Rasch model, we developed a computerized adaptive test (CAT) version of the Patient-Reported Outcomes Measurement Information System (PROMIS) Depression scale to enhance clinical efficiency.
Objective: We aimed to assess the measurement efficiency and validity of the PROMIS Depression CAT Taiwan version by examining its correlation with full-scale responses and the influence of demographic factors.
Methods: 409 psychiatric patients from a northern Taiwan medical center and community were included. The 28-item PROMIS® Depression Scale (5-point Likert) was used for assessment. The sample was split, with one-half training the CAT model and the other validating it. The PROMIS Depression CAT Taiwan version was developed using the Rating Scale Model (RSM) and the Expected A Posteriori (EAP) method to iteratively update ability estimates (θ), with Maximum Fisher Information (MFI) guiding item selection. The R program and ChatGPT were used to generate CAT parameters and modes. Item difficulty parameters were obtained from Winsteps, and CAT-estimated θ values were analyzed using Pearson correlation and demographic comparisons in SPSS.
Results: CAT-estimated θ values strongly correlated with full-scale responses (r = 0.958, p < 0.001), achieving 85% confidence (SE = 0.38). Compared to the 28-item full scale, the CAT achieved comparable accuracy with an average of 10.7 items, which could be completed in 3–4 minutes. Among demographics, only gender significantly influenced θ values (t = -2.862, p = 0.005), with males scoring lower than females.
Conclusion: The PROMIS Depression CAT Taiwan version maintains precise measurement while reducing test length, making it a promising tool for clinical and research applications in Taiwan.