Epidemiology and Predictors of Long-Stays in Medical ICU: A Retrospective Cohort Study

Introduction: Prolonged stays in ICU have been associated with overconsumption of resources but little is known about their epidemiology. We aimed to identify predictors and prognostic factors of extended stays, studying a long-stay population. Methods: We present a retrospective cohort study between July 2000 and December 2013 comparing patients hospitalized in a medical ICU for >= 30 days (long-stay patients-LSP) with patients hospitalized for <30 days (short-stay patients-SSP). Admission characteristics were collected from the local database for every patient and evolution during the ICU stay was retrieved from LSP files. Results: Among 8906 patients hospitalized in the ICU, 417 (4.7%) were LSP. At admission, male sex (adjusted odds-ratio (aOR) 1.4 [1.1; 1.7]), inpatient (aOR 2.0 [1.6; 2.4]) and in-ICU hospitalizations for respiratory (aOR 2.9 [1.6; 3.5]) or infectious diseases (aOR 1.6 [1.1; 2.5]) were all independently associated with a long stay in the ICU, while hospitalizations for metabolic (aOR 0.2 [0.1; 0.5]) or cardiovascular diseases (aOR 0.3 [0.2; 0.5]) were in favor of a short stay. In-ICU and in-hospital LSP mortality were 38.8% and 48.2%. Age (aOR 1.02 [1.00-1.04]), catecholamines (aOR 3.9 [1.9; 8.5]), renal replacement therapy (aOR 2.4 [1.3; 4.3]), primary disease-related complications (aOR 2.5 [1.4; 4.6]) and nosocomial infections (aOR 4.1 [1.8; 10.1]) were independently associated with mortality in LSP. Conclusion: LSP were highly comorbid patients mainly hospitalized for respiratory diseases. Their mortality was mostly related to nosocomial infections but the majority were discharged alive from the hospital.