We seek to extensively advance the technology for measuring health-related quality of life (HRQOL) for adults with diabetes using item response theory (IRT) and computerized adaptive testing (CAT) methods. HRQOL, increasingly accepted as an important outcome in clinical research, is also becoming appreciated as an important predictor (risk factor) and outcome in disease management. If more readily available, such information could be useful in informing treatment choices and enhancing patient-caregiver communication in diabetes care. Among the major factors limiting their widespread routine use is the impracticality (respondent burden) of today's HRQOL measures. Short-forms have proven more practical and acceptable to patients; however, """"""""ceiling"""""""" and """"""""floor"""""""" effects typically result from measuring a limited range within each health domain. Further, short forms (e.g., the SF-36) lack the precision necessary to detect changes in individual patient scores. To address these deficiencies, the principal aims of Phase I are to use preliminary IRT calibrations applied to diabetes-specific questionnaire items covering the major domains of disease impact (body pain, physical functioning, social and role participation, psychological distress and well-being, vitality) to: (1) build a prototype web-based diabetes impact CAT; (2) estimate actual reductions in respondent burden, range of levels measured, and score accuracy of DIABETES-CAT in comparison with scores for the full-length (""""""""static"""""""") survey in a pilot field test of patients with diabetes (N=100) in a clinical setting; (3) evaluate patient and clinician acceptance of actual CAT administrations relative to full-length questionnaires; and (4) evaluate the discriminant validity of the disease-specific and generic CAT and static tools in relation to differences in glycemic control. The product of Phase I will be a prototype version of a DIABETES-CAT with preliminary evidence regarding feasibility, acceptability and empirical performance. In Phase II, we hope to """"""""marry"""""""" the new diabetes-specific CAT and the expanded item pools for the eight generic CAT forms (based on SF-36 and other widely-used measures) and create a single system. Items will be evaluated rigorously and those fitting an IRT model will be calibrated on a common metric. A substantial advantage of our approach is the ability to utilize IRT models to cross-calibrate scores with current widely used diabetes-impact tools and to meaningfully compare results. DIABETES-CAT will be evaluated in terms of reliability, validity (discriminant and responsiveness) and precision across score levels to create a fully operational system that will facilitate clinical research and improve assessment methods for use in clinical practice. By greatly lowering data collection costs, reducing respondent burden, eliminating """"""""ceiling"""""""" and """"""""floor"""""""" effects and increasing the precision of individual patient scores, routine monitoring of disease-specific and generic HRQOL may become feasible as a clinical tool and lead to acceptance of CAT among patients with diabetes and other chronic illnesses.
Schwartz, Carolyn; Welch, Garry; Santiago-Kelley, Paula et al. (2006) Computerized adaptive testing of diabetes impact: a feasibility study of Hispanics and non-Hispanics in an active clinic population. Qual Life Res 15:1503-18 |