Application of International Classification Injury Severity Score to National Surgical Quality Improvement Program Defines Pediatric Trauma Performance Standards and Drives Performance Improvement

Background: The American College of Surgeons National Surgical Quality Improvement Program is becoming a core methodology to define performance as a ratio of observed to expected events. We hypothesized that application of this rising International Classification of Injury Severity Score (ICISS) for individual patient risk stratification to a group of hospitals contributing data to the National Pediatric Trauma Registry (NPTR) would apply objective evidence of actual injuries to define an expected standard and identify performance outliers. Methods: Using a blinded code, children entered into phase III of the NPTR were aggregated by treating hospital. Individual patient ICISS survival probability (P,) were calculated using survival risk ratios (SRR) derived from the phase II NPTR dataset (n = 53,253). For each center, sample size, observed mortality, and ICISS P, were calculated. Probability of mortality (P-m) was computed as 1 - P-s. Logistic regression was used to develop a predictive model for mortality. Logit transformation of P-m was performed to adjust for the skew of minor injury in children and reduce overestimation of low P-m, fatalities. Mean P-m was computed for each center and multiplied by its volume to determine expected frequency. Observed to expected ratio (O/E) and 95% confidence interval were calculated to define expected performance and outliers above or below 1 SD of the mean O/E. Results: Patients treated at 30 pediatric trauma centers (mean volume 451 +/- 258/patients per center) were evaluated. Mean O/E was 1.001 with SD = 0.404. Twenty-two centers fell within the reference range; O/E of 12 centers exceeded 1, suggesting performance below expectation. Trauma center volume, as reflected by sample, did not correlate to O/E performance. Conclusions: Application of ICISS P. from a national pediatric benchmark population simplifies determination of expected mortality necessary to compute the expected component of National Surgical Quality improvement Program. Analysis of these ratios of expected to observed mortality demonstrates variance among centers, defines performance against peers using the same benchmarks, and can drive performance improvement based oil the objective evidence of injury diagnoses actually encountered.