Purpose: To investigate agreement between "indirect" and "direct" methods for measuring outcome of inpatient rehabilitation programmes.
Method: A consecutive sample of 610 patients with musculoskeletal, cardiopulmonary and other diseases completed questionnaires 2 weeks before and 4 to 6 weeks after inpatient rehabilitation treatment. Overall response rate was 80%. Patient outcome variables were measured at baseline and follow-up. These included 14 items covering general health, pain, functional limitations, sleep, energy, mood, and social contacts. In the follow-up questionnaire patients were also asked to recall their baseline situation and to answer the same items according to this recalled condition and, furthermore, to report directly the perceived change on a 3-point scale ("deteriorated", "unchanged", "improved"). Thus, three different methods for measuring change could be compared: changes observed prospectively between baseline and follow-up, change as the difference of recalled baseline status and observed follow-up status, and perceived change.
Results: Spearman correlation coefficients r(S) and Cohen' Kappa statistic were used to analyse association and chance-corrected agreement, respectively, between these three different methods. Correlation and agreement between the different indicators of change was only moderate to fair with r(S) varying between 0.10 and 0.65, Cohen's Kappa ranged between 0.07 and 0.62. Results of principal components analysis indicated that items representing perceived change did not reflect the substantive aspects of change and were assigned indiscriminately to one single component.
Conclusion: Prospective and retrospective methods for measuring change are not functionally equivalent. Whenever feasible, prospective methods using observed baseline and follow-up data ("indirect" measurement of change) should be used in the evaluation of patient outcome in rehabilitation.