TY - JOUR T1 - Item Response Theory, Computerized Adaptive Testing, and PROMIS: Assessment of Physical Function JF - The Journal of Rheumatology JO - J Rheumatol SP - 153 LP - 158 DO - 10.3899/jrheum.130813 VL - 41 IS - 1 AU - James F. Fries AU - James Witter AU - Matthias Rose AU - David Cella AU - Dinesh Khanna AU - Esi Morgan-DeWitt Y1 - 2014/01/01 UR - http://www.jrheum.org/content/41/1/153.abstract N2 - Objective. Patient-reported outcome (PRO) questionnaires record health information directly from research participants because observers may not accurately represent the patient perspective. Patient-reported Outcomes Measurement Information System (PROMIS) is a US National Institutes of Health cooperative group charged with bringing PRO to a new level of precision and standardization across diseases by item development and use of item response theory (IRT). Methods. With IRT methods, improved items are calibrated on an underlying concept to form an item bank for a “domain” such as physical function (PF). The most informative items can be combined to construct efficient “instruments” such as 10-item or 20-item PF static forms. Each item is calibrated on the basis of the probability that a given person will respond at a given level, and the ability of the item to discriminate people from one another. Tailored forms may cover any desired level of the domain being measured. Computerized adaptive testing (CAT) selects the best items to sharpen the estimate of a person’s functional ability, based on prior responses to earlier questions. PROMIS item banks have been improved with experience from several thousand items, and are calibrated on over 21,000 respondents. Results. In areas tested to date, PROMIS PF instruments are superior or equal to Health Assessment Questionnaire and Medical Outcome Study Short Form-36 Survey legacy instruments in clarity, translatability, patient importance, reliability, and sensitivity to change. Conclusion. Precise measures, such as PROMIS, efficiently incorporate patient self-report of health into research, potentially reducing research cost by lowering sample size requirements. The advent of routine IRT applications has the potential to transform PRO measurement. ER -