Improving Accuracy of KABCO Injury Severity Assessment by Law Enforcement Officers
Abstract
ABSTRACT
Injury severity as assessed by law enforcement officers is one of the metrics used when allotting safety funds by transportation agencies. The Crash Outcome Data Evaluation System (CODES) is a database that contains injury severity assessments by law enforcement officers and the actual health outcomes as measured by medical practitioners. Crashes from 2008 through 2012 were analyzed to determine reasons for discrepancies between law enforcement officer's assessment of injury severity and that of a medical practitioner. Comparing the two assessments of injury severity shows law enforcement officers only rate serious injury severity "A" (Incapacitating Injuries) accurately for 33% of crash victims. Overall accuracy of injury severity is only 51% for all crash victims. Crashes where injury severity was inaccurately assessed were analyzed. Factors such as alcohol, gender, vehicle type, and lighting conditions contributed to law enforcement officer's inaccurate injury severity assessment. It was also found law enforcement officers have difficulty assessing injuries to almost every body region.
Crash victims with accurate injury severity assessments were also analyzed. Injuries and other crash data noted by a law enforcement officer in the crash report was collected to build a logistic regression model and classification trees to estimate injury severity. For both logistic regression and classification trees injury severity estimation improved from law enforcement officers current accuracy, particularly for "Incapacitating" (injury severity "A") and "Non-Incapacitating" (injury severity "B") injuries. Logistic regression was found to be the best method to improve injury severity. Injury severity "A" estimation increased to 77% from the 33% accuracy of law enforcement officers currently. Injury severity "B" estimation increased to 51% from 18% currently. Overall, accuracy was improved from 51% to 70%.