Adult patients in the National Trauma Data Bank (NTDB version 7.0) were analyzed, and the following physiologic injury parameters were compared: RTS, systolic blood pressure (SBP), shock (SBP ≤ 90 mm Hg), Glasgow coma scale-total (GCS-T), and GCS-motor (GCS-M). Areas under the receiver-operating characteristic curves (AUROC) were calculated for unadjusted and multivariate regression models to predict mortality after trauma.
There were 1,484,648 patients who met inclusion criteria. In unadjusted analyses, RTS had the highest proportion of missing data (21 % ) and was highly predictive of mortality (AUROC = 0.85). SBP and shock had a much lower AUROC of 0.67 and 0.66, respectively, but had many fewer missing cases. The combination parameters of GCS-M with SBP or GCS-M with shock showed AUROC comparable to RTS (0.85) with approximately 80,000 fewer missing cases.
The discriminatory power of RTS is significantly better than SBP, shock, or GCS alone. Given the limitation of missing data associated with RTS, the combination of SBP and GCS-M is a more reliable and equally effective method of assessing physiologic injury severity in studying trauma outcomes.