Dissatisfaction with Medicare's current system of paying for rehabilit
ation care has led to proposals for a rehabilitation prospective payme
nt system, but first a classification system for rehabilitation patien
ts must be created. Data for 36,980 patients admitted to and discharge
d from 125 rehabilitation facilities between January 1, 1990, and Apri
l 19, 1991, were provided by the Uniform Data System for Medical Rehab
ilitation. Classification rules were formed using clinical judgment an
d a recursive partitioning algorithm. The Functional Independence Meas
ure version of the Function Related Groups (FIM-FRGs) uses four predic
tor variables: diagnosis leading to disability, admission scores for m
otor and cognitive functional status subscales as measured by the Func
tional Independence Measure, and patient age. The system contains 53 F
RGs and explains 31.3% of the variance in the natural logarithm length
of stay for patients in a validation sample. The FIM-FRG classificati
on system is conceptually simple and stable when tested on a validatio
n sample. The classification system contains a manageable number of gr
oups, and may represent a solution to the problem of classifying medic
al rehabilitation patients for payment, facility planning, and researc
h on the outcomes, quality, and cost of rehabilitation.