Development and calibration data for the Medication Adherence Item Bank: a new computer adaptive test for persons with type 2 diabetes mellitus

Purpose To develop a new computer adaptive test that evaluates important aspects of medication adherence for persons with type 2 diabetes mellitus. Methods Two hundred and twenty-five people with type 2 diabetes mellitus completed 41 items related to medication adherence. Results Exploratory analysis supported the essential unidimensionality of the initial item pool. Five items were deleted due to low item-adjusted total score correlations (resulting in 36 items). Confirmatory factor analysis supported the retention of 27 items. A graded response model identified no items for exclusion, based on misfit. No items were flagged for meaningful differential item functioning (DIF). The final item bank is comprised of 27 items; an associated 6-item short form was constructed that balanced both psychometric factors (e.g., item information values) and clinical input. Initial analysis of the simulated CAT and static short form supported both the reliability (i.e., internal consistency and test–retest reliability) and validity (i.e., convergent, discriminant, and known groups) of both administration formats. Conclusions The new medication adherence item bank provides a reliable and valid assessment of the ability to take medications accurately among people with T2DM; it will be available in early 2023 through healthmeasures.net.