Development, Calibration, and Clinical Validation of a Deployable PROMIS-Based Adaptive Assessment Tool (PROMIS-CAT) for Real-Time Patient-Reported Outcomes in Spine Care

Study Design. Instrument development & validation Objective. To develop a Patient-Reported Outcomes Measurement Information System-computer adaptive test (PROMIS-CAT) capable of estimating PROMIS T-scores using fewer questions by applying adaptive item selection & machine-learning–based calibration. Summary of Background Data. Patient-reported outcomes capture aspects of health that cannot be measured through imaging or examination. PROMIS is widely used in clinical research and practice, but PROMIS Short Forms require patients to answer 8–10 items per domain creating cognitive burden, increasing survey abandonment, and slowing workflow. Current PROMIS CATs require an external server connection, limiting integration into custom applications, EMRs, or offline environments. Methods. Three cohorts were used: development (n=2,000), calibration (n=900), and validation (n=887). PROMIS-CAT system adaptively assessed six domains: Physical Function, Pain Interference, Fatigue, Anxiety, Depression, and Ability to Participate in Social Roles. After each response, system estimates ability (θ) using a graded response model and selects the item with maximal information at θ. The cycle continues until measurement precision meets PROMIS CAT standards. To convert θ to PROMIS T-scores, linear calibration models were applied to four domains (Physical Function, Pain Interference, Fatigue, Social Roles). Anxiety and Depression required isotonic regression to correct bias. Results. In the 887-patient validation cohort, the PROMIS-CAT engine showed high accuracy compared with PROMIS Short Forms. Physical Function, Pain Interference, Fatigue, and Social Roles met all predefined interchangeability benchmarks, achieving mean absolute errors (MAEs) <3 T-score points with minimal bias. Anxiety and Depression were initially inaccurate (MAE ~12) but improved markedly after isotonic calibration (MAE ~4), eliminating bias. Conclusions. PROMIS-CAT engine accurately estimates PROMIS T-scores with far fewer questions and without reliance on PROMIS servers, enabling real-time scoring within mobile apps, patient portals, and EMRs. This tool supports efficient, precise, and clinically meaningful outcome measurement in routine spine care.