APACHE system consistently under-predicts ICU length of stay for persons with severe sepsis

INTRODUCTION: Severe sepsis is a major worldwide cause of morbidity and mortality - is an important cause of admission in the intensive care units (ICU); the leading cause of death in non coronary intensive care units; and the 10th leading cause of death overall. Accurate prediction of length of stay (LOS) of severe sepsis patients in ICUs is critical to ICU resource management. APACHE-IV model offers the best predictive accuracy. OBJECTIVE: To assess the ICU-LOS predictability of APACHE-IV system for severe sepsis patients. METHODS: After institutional ethical committee clearance, ICU data [June, 2006 – August, 2008] from AMRI hospitals, Kolkata India for APACHE-IV predicted ICU-LOS of severe sepsis patients with complete data were compared with actual observed ICU-LOS, days on mechanical ventilation and other clinically important factor impacting ICU-LOS, employing t-test, correlation coefficients and ANOVA of suitably transformed variables where needed. RESULTS: Out of 3949 ICU admissions, 218 severe sepsis admissions were identified [168 unique admissions] where 134 patients [80%] had complete usable data: 59% men; median age: 63. Years [IQR:24]; 58% did not have dialysis; 84% were on mechanical ventilation [MV]. Mean ICU-LOS [9.3 days (SD:6.7)] was significantly greater than APACHE-predicted ICU-LOS [6.3 days (4.3) - paired t-test; p=0.0017]. ICU-LOS was very strongly correlated with days on MV [r=0.9]. Mean ICU-LOS was significantly greater for those receiving blood transfusion [p<0.0001]; on MV [p=0.0018]; having surgery [p<0.0001]. CONCLUSIONS: APACHE-IV model underestimates the actual length of stay in ICUs by about 3-days (33% underestimation).