Objective To develop a computer adaptive testing (CAT) version of the self-reported depression in patients with type 2 diabetes Mellites (T2DM) and evaluate its measurement performance.Methods Convenience sampling was employed to select T2DM patients who met the inclusion criteria from Shanghai Pudong New Area Gongli Hospital and four community health service centers in Jinqiao Medical Federation, and face-to-face paper questionnaire surveys were conducted, with 2,139 questionnaires distributed.The measurement characteristics of the Chinese version of the PROMIS depression item bank were evaluated among T2DM patients, and items were selected accordingly.Then, a CAT version of the self-reported depression for T2DM patients was constructed using the FlexCAT online platform. Post-hoc simulation and Monte Carlo simulation were used to evaluate from three aspects:measurement efficiency, measurement accuracy, and measurement precision.Results A total of 2,062 patients were investigated.The Chinese version of the PROMIS depression item bank met the assumptions of the graded response model, the difficulty parameters was ranging from -0.50 to 3.40 and discrimination parameters from 2.05 to 4.11, which indicated the reasonable range. Significant item function differences were not observed among different genders, ages, and educational levels.Based on item quality analysis, a CAT version of self-reported depression for patients was constructed on the FlexCAT platform.Testers could complete personalized assessments online, and researchers could obtain evaluation results in the background.Post-hoc simulation and Monte Carlo simulation showed that the CAT version had 4-12 response items, with mean values of 4.77 and 4.57, and more than 90% of the testers could complete the assessment by responding to 4-6 items.The Pearson correlation coefficients between the simulated latent depression trait levels and the full item bank estimates were 0.942 and 0.953, with Cohen′s d effect size of 0.03.Conclusion The items of the Chinese version of the PROMIS depression item bank have good measurement characteristics and can be used for the construction of the CAT version.The constructed CAT version of self-reported depression for T2DM patients performed well in terms of measurement efficiency, precision, and accuracy, with personalized assessment characteristics, which can reduce the burden of patient responses, alleviate the work pressure of medical staff, and improve clinical work efficiency.