Mortality Factors in Geriatric Blunt Trauma Patients: Creation of a Highly Predictive Statistical Model for Mortality Using 50,765 Consecutive Elderly Trauma Admissions from the National Sample Project

Abstract

Elderly patients are at high risk for mortality after injury. We hypothesized that trauma benchmarking efforts would benefit from development of a geriatric-specific model for risk-adjusted analyses of trauma center outcomes. A total of 57,973 records of elderly patients (age older than 65 years), which met our selection criteria, were submitted to the National Trauma Database and included within the National Sample Project between 2003 and 2006. These cases were used to construct a multivariable logistic regression model, which was compared with the American College of Surgeons Committee on Trauma's Trauma Quality Improvement Project's (TQIP) existing model. Additional spline regression analyses were performed to further objectively quantify the physiologic differences between geriatric patients and their younger counterparts. The geriatric-specific and TQIP mortality models shared several covariates: age, Injury Severity Score, motor component of the Glasgow Coma Scale, and systolic blood pressure. Our model additionally used temperature and the presence of mechanical ventilation. Our geriatric-specific regression mode generated a superior c-statistic as compared with the TQIP approximation (0.85 vs 0.77; P = 0.048). Spline analyses demonstrated that elderly patients appear to be less likely to tolerate relative hypotension with higher observed mortality at initial systolic blood pressures of 90 to 130 mmHg. Although the TQIP model includes a single age component, these data suggest that each variable needs to be adjusted for age to more accurately predict mortality in the elderly. Clearly, a separate geriatric model for predicting outcomes is not only warranted, but necessary.