Objective: To evaluate LOS in developing a concept of borderline ICU LOS for a realistic reimbursement of intensive care.
Design: Retrospective analysis of LOS and cost data extracted from patients' electronic records.
Setting: Surgical ICU of the University Hospital Göttingen, Germany.
Patients and participants: All adult ICU admissions with LOS >24 h over a 24-month period (1 January 2000 to 31 December 2001; n=1631.)
Interventions: None.
Measurements and results: Cluster analysis partitioned the ICU population into three homogeneous groups based on ICU LOS and total direct costs: cluster 1 (n=1405; mean LOS=2.8; mean cost= Euro 2399); cluster 2 (n=190; mean LOS=13.4; mean cost=Euro 12,754); cluster 3 (n=36; mean LOS=34.9; mean cost= Euro 34,173). Cost distribution between cluster 1 and clusters 2 and 3 combined was 48 vs 52%. Upper 95 percentile LOS of 6.7 allowed cluster 1 to be replaced by an LOS profile population of < or = 7 days population (n=1355; 96% population and 91% total ICU cost overlap with cluster 1) representing 83% of total ICU population and 44% of total ICU costs. Stratification of >7 day population into LOS less than or >20 days (n=220; n=56) were further differentiated by mortality (11 vs 23%) and sepsis incidence (33 vs 79%).
Conclusions: It may be feasible to formulate a LOS-based reimbursement scheme for ICU services in Germany based on the selection of (appropriate) patients' ICU LOS profiles.