Anders Davidson, MD1, Judson Janak, PhD2, Jeff Howard, PhD2, Ian Stewart, MD1. 1David Grant Medical Center, 2U.S. Army Institute of Surgical Research
Background: Mortality prediction scores are widely utilized to assess prognoses in injured populations. These scores are helpful in care-oriented discussions and in resource allocation. We sought to develop and internally validate a mortality prediction score for injured combat casualties.
Methods: Data were obtained from the Department of Defense Trauma Registry, the Theater Medical Data Store, the Composite Health Care System and the Armed Forces Medical Examiner System. United States service members that were critically wounded in Iraq or Afghanistan from 1 Feb 2002 to 1 Feb 2011 who survived until evacuation to Landstuhl Regional Medical Center, Germany were analyzed. Variables analyzed included presenting vital signs to Role I-III: mean arterial blood pressure, heart rate, temperature, demographic variables, injury severity score (ISS), location of injury, mechanism of injury, presence of traumatic brain injury, and presence of acute kidney injury (AKI) ascertained within the first 7 days using the KDIGO serum creatinine criteria. The outcome examined was 90 day mortality.
Results: Of 6,011 records, 3,807 were included for analysis after excluding patients with missing data. Based on prior work, variables known to be associated with mortality were added stepwise into logistic regression models. To compare the predictive ability of the models, receiver operating characteristic area under the curves (ROCAUC) were evaluated. The ROCAUC for ISS alone was 0.862. The only variable examined that significantly increased this ROCAUC was AKI stage (ROCAUC 0.905, p<0.001). Given the relatively small number of subjects in the cohort bootstrapping techniques were used to internally validate the combined ISS and AKI model (95% confidence interval for the AUC 0.880-0.930). Utilizing the coefficients and intercept of the model, we developed a formula that estimates the risk of death for this cohort. For example, an ISS of 25 without AKI carries a mortality risk of 1.7%. The presence of AKI increases the risk to 4.8%, 11.1%, and 35.3% for stages 1-3 respectively.
Conclusion: Mortality prediction is an important instrument in injured populations. Utilization of ISS and AKI stage has the potential to be a simple and accurate approach for mortality prediction following combat injury in patients that survive to be evacuated from theater. Further work must be done in other populations to externally validate this prediction score.
Presented at the SAGES 2017 Annual Meeting in Houston, TX.
Abstract ID: 77702
Program Number: MSS23
Presentation Session: Full-Day Military Surgical Symposium – Trauma/Critical Care Presentations
Presentation Type: MSSPodium