Vital sign based shock scores are poor at triaging South African trauma patients.

BackgroundTraumatic shock cannot be diagnosed by a single physiological measurement and a number of vital sign based combined shock scores (CSS) have been proposed to identify and triage trauma patients with shock. This audit uses data from a prospectively entered electronic trauma registry to compare the ability of these CSS to predict in-hospital mortality, need for surgery, need for blood transfusion and ICU admission.Materials and methodsThe data used in the study was obtained from the Hybrid Electronic Medical Record (HEMR) in Pietermaritzburg from January 2012-September 2015. The calculated scores (Systolic Blood Pressure [SBP], Mean Arterial Pressure [MAP], Shock Index [SI], Modified Shock Index [MSI] and Shock Index multiplied by Age [SIA]) were plotted against each outcome parameter and the inflection points at which they started to increase, for each parameter, was determined and compared.ResultsA total of 8793 patients met the inclusion criteria. After the datasets with missing data were removed, a total of 7623 patients were available for analyses. There was a slightly higher incidence of blunt trauma (46%) compared to penetrating trauma (43%). Area under the Receiver Operating Curves (AUROC) for prediction of mortality revealed the MSI and SIA performed best, with values of 0.69 and 0.70, respectively. In both the 'need for ICU' prediction as well as the 'need for blood transfusion' prediction, MSI performed best with scores of 0.73 and 0.79, respectively. None of the parameters performed well in the 'need for surgery' prediction. None of the CSS parameters reached a 'good predictor capability' score of 0.8.ConclusionThe currently available vital sign based scores (SBP, MAP, SI, MSI, SIA) used in the prediction of shock severity and triage are not good predictors of mortality, need for ICU, need for theatre or need for blood transfusion in our population where half the trauma is penetrating and there are long pre-hospital delays. Our data suggests that none of the proposed CSS's are capable of reliably and accurately identifying and categorizing shock states in South African trauma patients.