Prehospital end-tidal carbon dioxide is predictive of death and massive transfusion in injured patients: An Eastern Association for Surgery of Trauma multicenter trial.

BACKGROUND: Prehospital identification of the injured patient likely to require emergent care remains a challenge. End-tidal carbon dioxide (ETCO2) has been used in the prehospital setting to monitor respiratory physiology and confirmation of endotracheal tube placement. Low levels of ETCO2 have been demonstrated to correlate with injury severity and mortality in a number of in-hospital studies. We hypothesized that prehospital ETCO2 values would be predictive of mortality and need for massive transfusion (MT) in intubated patients. METHODS: This was a retrospective multicenter trial with 24 participating centers. Prehospital, emergency department, and hospital values were collected. Receiver operating characteristic curves were created and compared. Massive transfusion defined as >10 U of blood in 6 hours or death in 6 hours with at least 1 U of blood transfused. RESULTS: A total of 1,324 patients were enrolled. ETCO2 (area under the receiver operating characteristic curve [AUROC], 0.67; confidence interval [CI], 0.63-0.71) was better in predicting mortality than shock index (SI) (AUROC, 0.55; CI, 0.50-0.60) and systolic blood pressure (SBP) (AUROC, 0.58; CI, 0.53-0.62) (p < 0.0005). Prehospital lowest ETCO2 (AUROC, 0.69; CI, 0.64-0.75), SBP (AUROC, 0.75; CI, 0.70-0.81), and SI (AUROC, 0.74; CI, 0.68-0.79) were all predictive of MT. Analysis of patients with normotension demonstrated lowest prehospital ETCO2 (AUROC, 0.66; CI, 0.61-0.71), which was more predictive of mortality than SBP (AUROC, 0.52; CI, 0.47-0.58) or SI (AUROC, 0.56; CI, 0.50-0.62) (p < 0.001). Lowest prehospital ETCO2 (AUROC, 0.75; CI, 0.65-0.84), SBP (AUROC, 0.63; CI, 0.54-0.74), and SI (AUROC, 0.64; CI, 0.54-0.75) were predictive of MT in normotensive patients. ETCO2 cutoff for MT was 26 mm Hg. The positive predictive value was 16.1%, and negative predictive value was high at 98.1%. CONCLUSION: Prehospital ETCO2 is predictive of mortality and MT. ETCO2 outperformed traditional measures such as SBP and SI in the prediction of mortality. ETCO2 may outperform traditional measures in predicting need for transfusion in occult shock. LEVEL OF EVIDENCE: Diagnostic test, level III.

Beyond Mortality: Does Trauma-related Injury Severity Score Predict Complications or Lengths of Stay Using a Large Administrative Dataset.

INTRODUCTION: Despite its shortcomings, trauma-related injury severity score (TRISS) correlates well with mortality in large trauma datasets. The aim of this study was to determine if TRISS correlates with morbidity and hospital lengths of stay using data from an institutional registry at a Level I Trauma Center. We hypothesized that higher TRISS correlates with increased complications and longer hospital stays. METHODS: A retrospective review of our institutional registry was performed, examining all trauma admissions between January 1999 and June 30, 2015. Out of a total of 32,026 patient records, TRISS data were available in 23,205 cases. Abstracted data included patient age, gender, ISS, TRISS, presence of complication, Glasgow Coma Scale (GCS), hospital length of stay, intensive care unit LOS, step-down unit LOS, functional independence measure, and 30-day mortality. RESULTS: TRISS was highly predictive of mortality, with the AUC value of 0.95 (95% confidence interval 0.936-0.954, P < 0.01) compared to ISS (AUC 0.794), GCS (AUC 0.827), and age (AUC 0.650). TRISS also performed better than the other variables in terms of the ability to predict morbidity events (AUC 0.813). TRISS was comparable to ISS in terms of prediction of ICU admission (AUC 0.801 versus 0.811, respectively). After correcting for patient age and gender, higher TRISS significantly correlated with longer hospital stays . CONCLUSIONS: Despite previous criticisms, we found that TRISS is superior to ISS for mortality and morbidity prediction. TRISS correlated significantly with a hospital, step down, and ICU lengths of stay using a large administrative dataset.