Development of symptom assessments utilising item response theory and computer-adaptive testing—A practical method based on a systematic review

Abstract 

Assessment of individual patients’ distress is a cornerstone of clinical care for advanced cancer. Patients’ ability to fill out lengthy questionnaires is compromised by many factors. Computer-adaptive tests (CAT) offer a promising approach to developing tailored instruments, that administer only items relevant to the individual patient. A systematic review of the literature about CATs in medical databases was conducted. Based on the results, a method for developing a CAT was designed that requires nine steps: (1) build an item pool; (2) administer the items to a predefined sample in a calibration study; (3) eliminate inappropriate items; (4) examine whether all items are influenced by a single dominant trait; (5) calibrate the items to the best-fitting item response theory (IRT) model; (6) evaluate items’ parameter equivalence across subgroups; (7) build an item bank with the calibrated items; (8) develop the CAT; and (9) pilot test the developed CAT. CAT offers the chance to extend the usefulness of patient-reported outcome (PRO) measurements from clinical studies to daily clinical practice.

Keywords: Oncology, Patient-reported outcomes, Item response theory, Computer-adaptive testing