Predicting In-Hospital and 1-Year Mortality in Geriatric Trauma Patients Using Geriatric Trauma Outcome Score

• Published in: Journal of the American College of Surgeons Vol. 224; no. 3; pp. 264 - 269
• Main Authors: Ahl, Rebecka, MA(Cantab), MB BChir, Phelan, Herb A., MD, MSCS, FACS, Dogan, Sinan, MD, Cao, Yang, PhD, Cook, Allyson C., MD, Mohseni, Shahin, MD, PhD
• Format: Journal Article
• Language: English
• Published: United States 01-03-2017
• Subjects: Aged; Erythrocyte Transfusion; Female; Geriatric Assessment; Hospital Mortality; Humans; Injury Severity Score; Kirurgi; Logistic Models; Male; Medicin och hälsovetenskap; Outcome Assessment (Health Care); Predictive Value of Tests; Registries; Surgery; Sweden; Wounds and Injuries - mortality; GTOS; ISS; Geriatric Trauma Outcome Score; area under curve; Injury Severity Score; AUC
• ISSN: 1072-7515; 1879-1190; 1879-1190
• DOI: 10.1016/j.jamcollsurg.2016.12.011

Background The Geriatric Trauma Outcome Score (GTOS; [age] + [2.5 × Injury Severity Score] + 22 [if packed RBC transfused within ≤24 hours of admission]), was developed and validated as a prognostic indicator for in-hospital mortality in elderly trauma patients. However, GTOS neither provides information about post-discharge outcomes nor discriminates between patients dying with and without care restrictions. Isolating the latter, GTOS prediction performance was examined during admission and 1-year post discharge in a mature European trauma registry. Study Design All trauma admissions 65 years of age and older in a university hospital during 2007 to 2011 were considered. Data on age, Injury Severity Score, packed RBC transfusion within ≤24 hours, therapy restrictions, discharge disposition, and mortality were collected. In-hospital deaths with therapy restrictions and patients discharged to hospice were excluded. The GTOS was the sole predictor in a logistic regression model estimating mortality probabilities. Performance of the model was assessed by misclassification rate, Brier score, Tjur R2 , and area under the curve. Results The study population was 1,080 patients with a median age of 75 years, mean Injury Severity Score of 10, and packed RBCs transfused in 8.2%. In-hospital mortality was 14.9% and 7.7% after exclusions. Misclassification rate fell from 14% to 6.5% and Brier score from 0.09 to 0.05, and area under the curve increased from 0.87 to 0.88. Equivalent values for the original GTOS sample were 9.8%, 0.07, and 0.82, respectively. One-year mortality follow-up showed a misclassification rate of 17.6% and Brier score of 0.13. Conclusions Excluding patients with care restrictions and discharged to hospice improved GTOS performance for in-hospital mortality prediction. The GTOS is not adept at predicting 1-year mortality.