Multicenter external validation of the Geriatric Trauma Outcome Score: A study by the Prognostic Assessment of Life and Limitations after Trauma in the Elderly (PALLIATE) consortium

Allyson C. Cook, Bellal Joseph, Kenji Inaba, Paul A. Nakonezny, Brandon R. Bruns, Jeff D. Kerby, Karen J. Brasel, Steve E. Wolf, Joe Cuschieri, M. Elizabeth Paulk, Ramona L. Rhodes, Scott C. Brakenridge, Herb A. Phelan

Research output: Contribution to journalArticlepeer-review

50 Scopus citations

Abstract

Background A prognostic tool for geriatric mortality after injury called the Geriatric Trauma Outcome Score (GTOS), where GTOS = [age] + [ISS × 2.5] + [22 if transfused any PRBCs by 24 hours after admission], was previously developed based on 13 years of data from geriatric trauma patients admitted to Parkland Hospital. We sought to validate this model. Methods Four Level I centers identified infjects who are 65 years or older for the period of the original study. The GTOS model was first specified using the formula [GTOS = age + (ISS × 2.5) + 22 (if given PRBC by 24 hours)] developed from the Parkland sample and then used as the sole predictor in a logistic mixed model estimating probability of mortality in the validation sample, accounting for site as a random effect. We estimated the misclassification (error) rate, Brier score, Tjur R2, and the area under the curve in evaluating the predictive performance of the GTOS model. Results The original Parkland sample (n = 3,841) had a mean (SD) age of 76.6 (8.1) years, mean (SD) ISS of 12.4 (9.9), mortality of 10.8%, and 11.9% receiving PRBCs at 24 hours. The validation sample (n = 18,282) had a mean (SD) age of 77.0 (8.1) years, mean (SD) ISS of 12.3 (10.6), mortality of 11.0%, and 14.1% receiving PRBCs at 24 hours. Fitting the GTOS model to the validation sample revealed that the parameter estimates from the validation sample were similar to those of fitting it to the Parkland sample with highly overlapping 95% confidence limits. The misclassification (error) rate for the GTOS logistic model applied to the validation sample was 9.97%, similar to that of the Parkland sample (9.79%). Brier score, Tjur R2, and the area under the curve for the GTOS logistic model when applied to the validation sample were 0.07, 0.25, and 0.86, respectively, compared with 0.08, 0.20, and 0.82, respectively, for the Parkland sample. Conclusion With the use of the data available at 24 hours after injury, the GTOS accurately predicts in-hospital mortality for the injured elderly. Level of Evidence Prognostic study, level III.

Original languageEnglish (US)
Pages (from-to)204-209
Number of pages6
JournalJournal of Trauma and Acute Care Surgery
Volume80
Issue number2
DOIs
StatePublished - Feb 1 2016

Keywords

  • Geriatric
  • elderly
  • prognosis
  • score
  • trauma

ASJC Scopus subject areas

  • Surgery
  • Critical Care and Intensive Care Medicine

Fingerprint

Dive into the research topics of 'Multicenter external validation of the Geriatric Trauma Outcome Score: A study by the Prognostic Assessment of Life and Limitations after Trauma in the Elderly (PALLIATE) consortium'. Together they form a unique fingerprint.

Cite this