Objective: To assess the predictive ability of preillness and illness variables, impact of care, and discharge variables on the post-intensive care mortality.
Setting and patients: 5,805 patients treated with high intensity of care in 89 ICUs in 12 European countries (EURICUS-I study) surviving ICU stay.
Methods: Case-mix was split in training sample (logistic regression model for post-ICU mortality: discrimination assessed by area under ROC curve) and in testing sample. Time to death was studied by Cox regression model validated with bootstrap sampling on the unsplit case-mix.
Results: There were 5,805 high-intensity patients discharged to ward and 423 who died in hospital. Significant odds ratios were observed for source of admission, medical/surgical unscheduled admission, each year age, each SAPSII point, each consecutive day in high-intensity treatment, and each NEMS point on the last ICU day. Time to death in ward was significantly shortened by different source of admission; age over 78 years, medical/unscheduled surgical admission; SAPSII score without age, comorbidity and type of admission over 16 points; more than 2 days in high-intensity treatment; all days spent in high treatment; respiratory, cardiovascular, and renal support at discharge; and last ICU day NEMS higher than 27 points
Conclusions: Worse outcome is associated with the physiological reserve before admission in the ICU, type of illness, intensity of care required, and the clinical stability and/or the grade of nursing dependence at discharge.