The principle of parsimony: Glasgow Coma Scale score predicts mortality as well as the APACHE II score for stroke patients.

Abstract

Although the development and use of severity-of-illness measures has gained widespread enthusiasm, uncertainty remains as to the optimal measure for stroke patients. The Health Care Financing Administration recently derived a severity-of-illness measure based on the APACHE II system to explain differences in Medicare mortality rates among hospitals treating stroke patients. We hypothesized that the Glasgow Coma Scale score provides prognostic information of accuracy comparable to that of the APACHE II score for stroke patients, yet is simpler and cheaper to abstract from the medical record. We therefore studied 246 patients hospitalized with stroke, including 49 oversampled mortalities. The Glasgow Coma Scale score was as accurate as the APACHE II score in predicting stroke mortality both before (r = -0.50 and r = 0.50, respectively) and after (r = -0.40 and r = 0.38, respectively) the oversampled mortalities were excluded. The APACHE II score required abstraction of 16 variables from the medical record compared with three for the Glasgow Coma Scale score and required more than three times the time to abstract from the medical record. Therefore, in the interest of parsimonious data collection, the Glasgow Coma Scale may be a preferable severity-of-illness measure for patients with stroke.