Quality of life item banks for type 2 diabetes: psychometric evaluation and computerized adaptive testing simulations

Abstract Purpose We evaluated the psychometric properties of type 2 diabetes (T2DM) quality of life (QoL) item banks (IBs), and assessed their efficiency for computerized adaptive testing (CAT) using simulations. Methods In this cross-sectional, clinical study, patients with T2DM answered 256 items across seven IBs: Activity Limitation (AL); Symptoms (SY); Diet (DT); Emotional (EM); Concerns (CN); DM Management (DMn); and Work (WK), referred to collectively as “DiabCAT”. The psychometric properties of each IB (e.g. measurement precision; item “fit”; differential item functioning) were assessed using Rasch analysis. The mean number of items required for ‘high’ (0.3 standard error [SE]) and ‘moderate’ (0.387 SE) precision was determined using CAT simulations. Results The 298 patients (57.4% male) had a mean (SD) age of 60.7 (10.8) years and DM duration of 15.3 (9.7) years. Overall, SY, DT, WK, EM, and CN displayed acceptable measurement performance after deleting 15 problematic items, with some limitations in targeting at higher levels of the latent trait, particularly for EM. The DMn domain was multidimensional and subsequently split into two unidimensional domains (DMn-Convenience [DMn-CV] and DMn-Concerns [DMn-CN]). Owing to unresolvable psychometric issues, AL was dropped. In CAT simulations, the mean number of items required per IB ranged from 9 (CN, DMn-CV, WK) to 11 (EM) for moderate, and 14 (DT) to 18 (EM) for high measurement precision. Conclusions Comprising 213 items within seven psychometrically valid and efficient IBs, DiabCAT enables researchers and clinicians to comprehensively assess the QoL impact of T2DM and its management with relatively low administration burden. Further assessment of its reliability, validity and responsiveness is warranted.