Abstract
Objective
To develop a population-derived, parsimonious, and objective risk stratification model for dogs following trauma and compare its predictive performance to the animal trauma triage (ATT) score.
Design
Observational cohort study using data from the American College of Veterinary Emergency and Critical Care Veterinary Committee on Trauma (VetCOT) trauma registry acquired between September 2013 and October 2017.
Setting
Nine Level I and Level II veterinary trauma centers.
Animals
Nine hundred eighty-four dogs assessed within 24 h of traumatic injury.
Interventions
None.
Measurements and Main Results
Patient mortality was 10.8%. The VetCOT model was constructed based on 4 variables: plasma lactate and ionized calcium obtained within 6 h of admission, and presence or absence of clinical signs consistent with either head or spinal trauma. The VetCOT score had good discriminatory performance (AUROC = 0.87, 95% CI = 0.83-0.91) comparable to that of the 6 variable ATT score for the same population (area under the receiver operator characteristic [AUROC] = 0.87; 95% CI, 0.84–0.90). No statistical difference in discriminatory performance between the 2 scores was identified (P = 0.98). The VetCOT score showed good calibration on this population (Hosmer–Lemeshow test P = 0.93), whereas the ATT score failed to calibrate (P = 0.02) due to overprediction of mortality at low scores. Sensitivity and specificity for outcome of the VetCOT score at a risk probability cutoff of 0.5 for this population were 28.97% and 97.95%, respectively.
Conclusions
The VetCOT score is a more parsimonious model with comparable discriminatory performance and superior calibration to the ATT score for risk stratification in dogs following trauma. Further prospective validation studies are required to confirm the discriminatory performance of the VetCOT score.
Journal of Veterinary Emergency and Critical Care, EarlyView.Wiley: Journal of Veterinary Emergency and Critical Care: Table of Contents