9 �� 21.1 selleck chemical days (median 38, range 7 to 192 days). ILOS>30 patients comprised only 4% of all ICU patients, but accounted for 8350 bed days (29%) out of a total of 28,771 bed days and 6742 ventilator days (41%) out of a total of 16,335 during this study period.Figure 2Distribution of length of stay of all trauma ICU patients in the study period. X axis = length of stay (days); Y axis = percentage of all trauma intensive care unit (ICU) patients.Demographic and clinical characteristics are shown in Table Table1.1. ILOS>30 patients were significantly older, more severely injured, and had lower Glasgow Coma Scores (GCS) on admission. A modest positive correlation existed between illness severity score (ISS) and ICU LOS (Spearman’s rho = 0.4, P < 0.001).

The LOS>30 patients group also had significantly higher incidences of pre-existing cardiac, renal, pulmonary conditions and diabetes mellitus. Not surprisingly, ILOS>30 patients sustained significantly more complications in the ICU.Table 1Demographic and clinical characteristics for ILOS>30 and ILOS<30 groupsOf the 4920 patients, 3421 (69.5%) were younger than 65 years old compared with 1499 (30.5%) who were 65 years old or older. ICU LOS was significantly associated with patient age (<65 versus >65 years old) when controlled for injury severity except in the least severely injured and most severely injured categories (Table (Table2).2). Age was also significantly associated with mortality. Patients 65 years and older had a mortality rate of 24.4% compared with 6.7% for younger patients (P < 0.001).

When controlled for injury severity, the association of mortality with age was significant for all degrees of injury severity (Table (Table3).3). For the ISS 1-3 patients who died, three had no autopsies (and therefore potential injuries may not have been delineated completely), three suffered anoxic brain injury after hanging and drowning accidents, and one died from necrotizing fasciitis after sustaining minor soft tissue trauma 10 days previously.Table 2Relation between age and intensive care unit length of stayTable 3Relation between age and mortalityUnivariate analysis produced the following predictors of ICU stay of more than 30 days: age over 65 years, ISS > 21 (Receiver operating characterstic curve [ROC] analysis; sensitivity 72% [95% C.I. 65%, 78%], specificity 64% [95% C.I.

63%, 66%]), GCS <12 (ROC curve analysis; sensitivity 43% [95% C.I. 36%, 50%], specificity 73% [95% C.I. 72%, 75%]), pre-existing cardiac, renal, pulmonary or diabetic conditions, and complications that developed during ICU stay (all P GSK-3 < 0.05).Variables with P < 0.2 by univariate analysis were entered into a logistic regression analysis to create a prediction model for ICU LOS of 30 days or longer. The P value was set at 0.2 because some variables may prove to have lower P values in a model or to be important confounders.