Validity of applying TRISS analysis to paediatric blunt trauma patients managed in a French paediatric level I trauma centre

Abstract

Objective: Using a weighted combination of the Revised Trauma Score (RTS), the Injury Severity Score (ISS), the type of injury (blunt or penetrating) and patient age, the TRISS method is used to calculate the probability of survival (ps) in trauma patients. The goal of this study was to compare the ability of the American Major Trauma Outcome Study (MTOS) norm for adult blunt trauma patients (ADULT) and the specific norm for paediatric patients (PED) to estimate the ps of injured children using TRISS methodology.

Design: Retrospective analysis using a paediatric trauma patient database.

Setting: A French level 1 paediatric trauma centre.

Patients: Four hundred seven consecutive paediatric blunt trauma patients, treated over a 3-year period.

Measurements: The observed and expected survivals were compared, using the M, W and Z scores, with both ADULT and PED. The W score is the number of survivors more or less than expected from the MTOS predictions for 100 patients. A Z score, which measures the significance of W, between -1.96 and +1.96, indicates no significant difference between observed and expected survivors. A value of M less than 0.88 indicates a disparity in the severity match between the study group and the MTOS group. We calculated the standardised W score (Ws), which represents the W score that would have been observed if the case mix of severity was identical to that of the MTOS group. Accordingly, a standardised Z score (Zs) was also calculated. In addition, we calculated the area under the receiver operating curve (aROC) using both norms, while calibration was also assessed by calculation of the Hosmer-Lemeshow goodness-of-fit tests.

Results: Using PED, the number of actual survivors (n = 364) was not significantly different from the MTOS (n = 358). The value of M, 0.65, indicated a disparity in the severity match between the study group and the MTOS group, due to a higher proportion of patients with lower ps (TRISS < 0.95, 52 vs 27%). We was +1.06% (95% confidence interval -0.34 to 2.08) and Zs was 1.48, indicating no significant difference from the MTOS. Using ADULT, the number of observed survivors (n = 364) was significantly higher than that expected (n = 354), with a W score of +2.70% (Z = +1.98, p < 0.05). There was a disparity in the severity match (M = 0.67) between the study group and the MTOS group, due to a higher proportion of patients with lower ps. Ws was +1.32% (95% confidence interval -0.12 to 2.37) and Zs = +1.79 (NS), indicating no significant difference from the MTOS. The Hosmer-Lemeshow statistics indicated that ADULT (Cg = 7.24, p = 0.51; Hg = 4.45, p = 0.81) and PED (Cg = 6.08, p = 0.64; Hg = 3.55, p = 0.90) provided sufficient goodness-of-fit. There was no significant difference in the aROC of the TRISS between the two norms (0.935 +/- 0.050 vs 0.936 +/- 0.050; NS).

Conclusion: Both adult and paediatric norms were equally good predictors of the probability of survival of injured children, provided that Ws and Zs are used when there is a disparity in the severity match between the study group and the MTOS group.