Purpose: The aim of this study was to build a self-report instrument to measure function and disability among older adults for clinical and public health research using Item Response Theory and Computer Adaptive Testing approaches. We structured the instrument to distinguish between limitations in function (the ability to perform discrete physical tasks and activities) and limitations in participation in society (the ability to perform complex activities and social roles). This abstract describes the development of the Disability Scale of the Late-Life Function and Disability Instrument CAT (LLFDI-CAT), which measures limitations in participation (disability) from the perspective of the older adult.
Methods: We conducted an extensive literature review, clinician and subject focus groups, and cognitive testing to guide development of an item bank consisting of 54 items. A convenience sample of 520 community-dwelling older adults answered all 54 items for the calibration study. Confirmatory factor analyses (CFA) were conducted to assess the dimensionality of the scale by comparing the CFA results across different models: (1) one-factor unidimensional; (2) two-factor multidimensional (MIRT); and (3) bi-factor MIRTs. Item Response Theory methods were employed for item calibration and to assess fit. To test accuracy, we used the dataset to simulate 5, 10, and 15 item CATs and compared the correlation, bias and root mean square error (RMSE) between the CATs and the full item bank across Fisher and posterior expected KL (Kullback-Leibler) information select methods. To understand the precision of the score estimation, we examined the standard error of scores for the 3 simulated CATs.
Results: CFA suggested the bi-factor multidimensional IRT (MIRT) provided best fit; therefore the Overall Disability Scale was calibrated with separate “relationship” and “doing things” sub-factors. Fit statistics were acceptable: CFI=0.94; TLI=0.93; and RMSE=0.06. The 15-item CAT demonstrated accuracy compared with the item bank (r>=0.97; and RMSE <0.23) and better precision in the middle of the scale than at the extremes. Specifically, standard errors for Overall Disability scores for the 15-item CAT were less than 0.3 across the -2 to 0 logit score range and less than 0.4 across the -2 to 0 logit score range for the subscale scores.
Conclusions: The LLFDI-CAT Disability scale demonstrated accuracy and precision and has the potential to advance disability assessment in research with older adults.
See more of: The 33rd Annual Meeting of the Society for Medical Decision Making