Relationship of physician ratings of severity of illness and difficulty of clinical management to length of stay.

Abstract In a concurrent prospective study, medical and surgical residents rated the severity of illness and difficulty of clinical management of each of their patients within 24 hours of admission, and on a daily basis throughout the patient's stay. Data were collected on consecutive admissions resulting in 661 admissions with complete data for analysis. Results indicate that difficulty and severity are correlated, each explaining variations in length of stay (LOS), and together explaining up to 44 percent. Four alternative measures are tested, first-day values, average values over the stay, peak or highest values, and a zero-one measure indicating whether or not the severity or difficulty fluctuated over the stay. First-day and average measures of severity and difficulty explain little variation in LOS; peak and fluctuating measures are highly explanatory. After adjusting for diagnosis-related groups (DRGs), fluctuating severity adds 34 percent, and adjusting for both DRGs and severity, fluctuating difficulty adds 10 percent for a total of 53 percent variance explained. In comparable results, peak severity adds 21 percent, and peak difficulty 4 percent, for a total of 34 percent variance explained. Findings indicate that difficulty had independent value as a predictor, and the high explanatory power of the fluctuating measures suggests that a third dimension, instability, may be as important as severity and difficulty in explaining LOS.