A systematic review of intensive care length of stay indicators

A prolonged ICU length of stay (LOS) has been associated with many medical diagnoses and conditions but it is difficult to predict on admission to the ICU. Prolonged ICU LOS can adversely affect patient outcomes by increasing the risk of complications, and possibly mortality. Identification of such indicators may help to improve ICU resource utilization (e.g. bed triage, ICU staffing).

Objective

To systematically review the literature to determine common 'early' predictors of ICU LOS for adult patients.

Methods

A computerized search of MEDLINE, EMBASE and the Cochrane Database was undertaken from 1966–October 2003 to identify trials evaluating predictors of ICU LOS. Search terms included 'ICU', 'critical care', 'LOS', 'predictors', 'indicators', and 'duration of stay'. Bibliographies of all articles retrieved were searched for relevant articles not identified by the computerized search.

Results

Five studies were identified that were published in full, with a combined total of 16,107 patients. The definition of prolonged LOS varied with the population evaluated (e.g. > 14 days general ICU, > 5 days CVICU) and thus prevented meta-analyses. Approximately 10% of patients had prolonged LOS as defined by the study. Universal positive early indicators were: emergent surgery or admission, trauma, or need for mechanical ventilation in < 24 hours. Abbreviated LOS was associated with coma, DNR orders, and nontrauma surgical reasons for admission. APACHE created as a predictive model for mortality consistently did not predict LOS.

Conclusion

Patients with prolonged stay form a small percentage of ICU patients. There are few early predictors that are universally significant or important when determining prolonged ICU LOS. Many factors appear to be specific to subgroups and thus may limit broad applicability of general ICU scoring systems to specialized populations.