ICU physicians are unable to accurately predict length of stay at admission: a prospective study

Abstract

Objective: To evaluate the accuracy of prediction of intensive care unit length of stay made by physicians at patient admission. Design: Prospective cohort study. Setting: Three medical-surgical intensive care units in an oncology hospital. Patients: All patients admitted between January and December 2014. Interventions: None. Main Outcome Measurements: Intensive care unit (ICU) length of stay was estimated by the physicians responsible for patient admission and categorized as < 48 h, 2-5 days or more than 5 days. Agreement between predicted and actual intensive care unit length of stay was calculated. Results: A total of 2955 patients were admitted during the study period. Physicians accurately predicted ICU length of stay in 1557 (52.7%) admissions. ICU length of stay was underestimated in 864 (29.2%) and overestimated in 534 (18.1%) cases. Agreement between predicted and actual intensive care unit length of stay was poor (Kappa = 0.22) and not associated with physician characteristics. Predictions of an intensive care unit length of stay of > 5 days were significantly less accurate than those of < 48 h and of 2-5 days (31.1, 59.8 and 53.1%, respectively, P < 0.001). Conclusions: The intensive care unit length of stay prediction in these oncological intensive care units is inaccurate and, ideally, should not be made at admission.