Rondinelli R D, Murphy J R, Wilson D H, Miller C C
Department of PM&R, Rose Medical Center, Denver, CO 80220.
Arch Phys Med Rehabil. 1991 Jun;72(7):447-53.
Consecutive patients (n = 1,289) discharged from two inpatient rehabilitation facilities were prospectively examined to determine the extent to which rehabilitative outcomes, functionally based progress, and associated resource utilization (in terms of rehabilitation length of stay) can be concomitantly predicted using the Tufts/New England Medical Center functional assessment tool and bivariate and multivariate statistical comparisons. A high percentage (greater than 50%) of statistically significant associations between the predictor variables and seven outcome measures were seen. The R2s corresponding to these associations were generally small and resistant to enhancement by commonly accepted statistical manipulations. Consequently, they are, in general, of limited predictive value for determining functional prognosis or resource utilization among rehabilitation inpatients. Although functionally based predictors appear to be the best predictors of functional progress, their effectiveness as predictors of other domains (eg, discharge outcome and rehabilitation length of stay) is variable. The implication is that a prospective payment system using such an array of predictors and directed primarily at cost containment is likely to overlook potentially important gains in functional progress and patient outcome. Furthermore, the functionally based predictors, taken individually, varied in effectiveness as predictors among facilities, and, taken collectively, they varied in effectiveness as predictors for different diagnostic groupings within a facility. The inconsistency of predictions according to the various domains appears to further limit their application to a prospective payment model.
