UVA_Trauma_NISS_Data

Raw data for publication of: <br> Emily J. Larkin, Marieke K. Jones, Steven D. Young, Jeffrey S. Young. 2022. Interest of the MGAP score on in-hospital trauma patients: Comparison with TRISS, ISS and NISS scores. Injury <br> https://doi.org/10.1016/j.injury.2022.05.024. <strong>Patients and Methods</strong> The study design was a retrospective, single-center trauma registry study. Patients 18+ years old arriving at the Level I UVA Trauma Center between 2003-2015 who suffered blunt or penetrating trauma injuries formed the cohort for study (n = 18692). Patients with injuries other than blunt and/or penetrating injuries or data records missing components necessary to calculate MGAP, ISS/NISS, and TRISS were excluded from analysis, resulting in 16265 complete records. To determine if registry data missing on our variables of interest would result in a different predicted mortality rate, the full cohort (n = 18692) was examined for missingness (Figure 1) and then used in sensitivity analysis to robustly determine the effect of missingness on results. Independent variables included mechanism of injury (blunt or penetrating), admitted vital signs needed for trauma scoring calculations, GCS, ISS, NISS, and the retrospectively calculated MGAP and TRISS scores. Metrics from which the ISS, NISS, MGAP, and TRISS scores are derived were all collected upon hospital admission. Thus, although the MGAP was originally calculated using pre-hospital metrics4, in-hospital parameters were used for its calculation in this study. The dependent variables were hospital discharge status (dead / alive) and disposition from the emergency room to either the hospital floor, considered a low acuity disposition (n = 8847), or the ICU, operating room, or morgue, considered a high acuity disposition (n = 7418). No patients in the registry were discharged directly home from the emergency room. Mortality was defined as mortality during hospital admission and death occurring after discharge was not considered. Data were analyzed using R (R Foundation for Statistical Computing, Vienna, Austria, version 4.0.2). To compare the ability of MGAP, ISS, NISS, and TRISS to predict mortality and indicators of morbidity, AUROC was calculated.28 These curves plot sensitivity versus 1-specificity, with an area under the curve (AUC) of 1 indicating a “perfect” test and an AUC of 0.5 indicating the test is no better than chance (i.e. a coin flip).29,30 Differences in AUC were statistically determined using DeLong’s test for two correlated ROC curves.28,31 To assess if MGAP, ISS, NISS, and TRISS accomplished the under-triage rates under ACSCOT guidelines, sensitivity was calculated at the accepted cutoffs for increased risk of mortality, MGAP&lt;23, ISS/NISS&gt;15, and TRISS&lt;0.912,4. Optimizing the sensitivity of trauma predictive models was prioritized in this study over specificity to minimize the number of patients experiencing major trauma not classified as such.4,15 Measurements of specificity, and positive and negative predictive value were also calculated.32