What's in a score:A longitudinal investigation of scores based on item response theory and classical test theory for the Amsterdam Instrumental Activities of Daily Living Questionnaire in cognitively normal and impaired older adults

Objective: We aimed to investigate whether item response theory (IRT)-based scoring allows for a more accurate, responsive, and less biased assessment of everyday functioning than traditional classical test theory (CTT)-based scoring, as measured with the Amsterdam Instrumental Activities of Daily Living Questionnaire. Method: In this longitudinal multicenter study including cognitively normal and impaired individuals, we examined IRT-based and CTT-based score distributions and differences between diagnostic groups using linear regressions, and investigated scale attenuation. We compared change over time between scoring methods using linear mixed models with random intercepts and slopes for time. Results: Two thousand two hundred ninety-four participants were included (66.6 ± 7.7 years, 54% female): n = 2,032 (89%) with normal cognition, n = 93 (4%) with subjective cognitive decline, n = 79 (3%) with mild cognitive impairment, and n = 91 (4%) with dementia. At baseline, IRT-based and CTT-based scores were highly correlated (r = −0.92). IRT-based scores showed less scale attenuation than CTT-based scores. In a subsample of n = 1,145 (62%) who were followed for a mean of 1.3 (SD = 0.6) years, IRT-based scores declined significantly among cognitively normal individuals (unstandardized coefficient [B] = −0.15, 95% confidence interval, 95% CI [−0.28, −0.03], effect size = −0.02), whereas CTT-based scores did not (B = 0.20, 95% CI [−0.02, 0.41], effect size = 0.02). In the other diagnostic groups, effect sizes of change over time were similar. Conclusions: IRT-based scores were less affected by scale attenuation than CTT-based scores. With regard to responsiveness, IRT-based scores showed more signal than CTT-based scores in early disease stages, highlighting the IRT-based scores’ superior suitability for use in preclinical populations.