BACKGROUND. Use of case-mix reimbursement in psychiatric inpatients has bee
n limited as a result of a lack of systems which effectively group patients
according to required resource needs. In recognition of the fact that many
patient factors, in addition to diagnosis influence delivery of care in ps
ychiatry, new measures of patient need are emerging.
OBJECTIVE. This study compared improvement realized by using a multidimensi
onal measure of patient severity, the Computerized Severity Index (CSI), to
predict length of stay (LOS) in psychiatric inpatients over that achieved
by using patient variables routinely collected in the discharge abstract.
METHOD. Through retrospective chart review, severity ratings were made on 3
55 psychiatric discharges with primary diagnoses of psychotic or major depr
essive disorders. Those ratings were combined with demographic and diagnost
ic data available in discharge abstracts and were then entered into multiva
riate regression analyses to model LOS.
RESULT. CSI ratings significantly contributed to prediction models, which a
ccounted for an additional 9% to 11% of variation in LOS over discharge abs
tract data. Among patients with psychotic disorders, maximum severity durin
g hospitalization was the best predictor of LOS, whereas among patients wit
h depressive disorders, it was an increase in severity following admission.
CONCLUSION. Severity ratings, based on chart review, improved prediction of
LOS over discharge abstract variables for psychiatric inpatients in two di
agnostic groups. Further research is needed to estimate the impact of incor
porating severity ratings into a grouping system for all psychiatric inpati
ents. Estimation of predictive accuracy is important to determine the amoun
t of risk passed on to providers in a payment system based on psychiatric c
ase mix.