Predicting hospital mortality in adult patients with prolonged stay (>14 days) in surgical intensive care unit

Background. The aim of this paper was to identify the factors at surgical intensive care unit (SICU) admission and during the following SICU course that influence hospital mortality of patients with prolonged SICU stay (> 14 days). Methods. This prospectively-planned study enrolled 1661 patients over 16 years old with prolonged SICU stay in a tertiary-care teaching hospital over a 4-year period. Data at SICU admission, physiologic factors on the 14 ^th SICU day and the indications of prolonged ICU stay were collected. A multivariate logistic regression model with a least absolute shrinkage and selection operator technique was adopted to identify factors associated with hospital mortality in prolonged-stay patients at the 14th SICU day. Results. Prolonged-stay patients accounted for 9.7% of the total SICU admissions, but consumed 51.7% of total SICU days. The hospital mortality of these patients was 34%. For predicting the hospital mortality in prolonged SICU stay patients, the predictors at ICU admission included gender, longer pre-ICU days, higher Charlson comorbidity index, and not admitted from emergency. Predictors on the 14^th SICU day included lower Glasgow coma scale, lower mean arterial pressure, higher dosage of inotropes required, higher serum lactate level, higher serum bilirubin level, lower platelet count, and the use of renal replacement therapy. Among the indications for prolonged SICU stay, predictors included the need for mechanical circulatory support, worsening acute encephalopathy with altered mental status, hemodynamic instability due to bleeding, and sepsis with unstable vital signs. Conclusion. This validated predictive model reached clinically accurate discriminatory power, and may serve to improve patient care and resource utilization in the SICU.