PREDICTORS OF SURVIVAL IN PEDIATRIC TRAUMA: A COMPARISON OF THE INTERNATIONAL CLASSIFICATION INJURY SEVERITY SCORE (ICISS) TO EXISTING INJURY SCORING TOOLS Casey J. Allen, MD, Jonathan P. Meizoso, MD, William M Hannay, BS, Juliet J Nissan, MD, Carl I. Schulman, MD, PhD, Alan S. Livingstone MD, Juan E. Sola, MD, Nicholas Namias, MD, and Kenneth G. Proctor, PhD Dewitt-Daughtry Department of Surgery, Divisions of Trauma and Surgical Critical Care and University of Miami Miller School of Medicine, Miami, FL 33136, USA Supported in part by: Grants #N140610670 from the Office of Naval Research and #09078015 from U.S. Army Medical Research & Materiel Command None of the authors have declared conflicts of interest. BACKGROUND Traditional injury scoring systems for adults are not as reliable in pediatrics due to altering physiologies, disproprotionate organ sizes, and unique injury patterns. The International Classification Injury Severity Score (ICISS) predicts survival as calculated from the product of survival risk ratios (SRRs) for a patient's 3 worst injuries. SRRs are derived as the proportion of fatalities for every International Classification of Diseases, Ninth Edition (ICD-9) in a "benchmark" population. In adults, previous studies have demonstrated that ICISS predicts outcome more accurately than ISS. In pediatrics, the value of ICISS has been less well defined. One study validated ICISS as objective metrics for determining performance using the National Surgical Quality Improvement Program (NSQIP) methodology. Another study demonstrated that ICISS simplified determination of expected mortality necessary to compute the expected component of NSQIP. Analysis of expected to observed mortality demonstrates variance among centers, defines performance against peers using the same benchmarks, and can drive performance improvement based on the objective evidence of injury diagnoses actually encountered. To our knowledge, the ICISS has never been compared to existing injury scoring tools in predicting survival in pediatric trauma at a single institution. OBJECTIVES To assess the utility of the ICISS in predicting survival compared to existing injury scoring tools in pediatric trauma at a single institution, we provide a demographic analysis and survival analysis from over a decade of experience of a level 1 pediatric trauma center. We evaluate the use of ICISS based on probabilities of survival by ICD-9 coding from the Florida Agency for Health Care Administration (AHCA) discharge dataset from 1991 to 2010. We hypothesize that the ICISS survival prediction model derived from the Florida AHCA dataset will accurately predict survival in our independent sample, and that this algorithm better predicts survival than existing injury scoring tools at a single level 1 pediatric trauma center. Ryder Trauma Center, 1800 NW 12th Avenue, T-242, Miami, FL 33136 | (305) 585-1280 c.allen7@med.miami.edu METHODS From January 2000 to December 2012, consecutive pediatric admissions (≤17y) at a Level I pediatric trauma center were retrospectively reviewed for demographics, mechanisms of injury (MOI), Injury Severity Score (ISS), Revised Trauma Score (RTS), Trauma Score – Injury Severity Score (TRISS), length of stay (LOS), and survival. The ICISS survival probabilities were calculated for all admitted patients using SRRs computed from all pediatric admissions from 1991 to 2010 in the Florida AHCA dataset (236,317 total pediatric injury codes, with 6,907 mortalities). Receiver operator characteristic (ROC) analysis was used to independently assess and compare the predictive capability of ICISS, ISS, RTC, and TRISS. Data are represented as mean standard deviation if normally distributed, or median (interquartile range) if not. Student’s t, Fisher’s exact, or non-parametric tests were used as appropriate. Significance was assessed at p=0.05. RESULTS The study population was comprised of 1,928 pediatric trauma admissions (Table 1) with an ISS of 13±12, RTS of 7.84(0.29), TRISS of 0.981(0.011), and ICISS of 0.981(0.011). Subsequent overall survival was 96.6%. The independent predictive capabilities of ICISS, ISS, RTC, and TRISS are detailed in Table 2. ICISS accurately predicts survival (area under ROC: 0.970) in the overall population, however the existing scoring tools are better predictors even in separate age groups. CONCLUSIONS ICISS is an accurate predictor of survival in patients at this trauma center, but it is inferior when compared to existing scoring tools. This may reveal the poor utility of a tool derived from outcomes at various levels of trauma centers across the nation/state in predicting outcomes in those treated at a large volume level 1 pediatric trauma center with inherently different systems and capabalities. Table 1: Population. n=1,928 Male 70% Sex Female 30% Age, years MOI Blunt Penetrating ISS RTS TRISS ICISS LOS, days Survival 11±6 76% 24% 13±12 7.84(0.29) 0.981(0.011) 0.903(0.168) 3(7) 96.6% Table 2: AUROC of injury scoring models on survival, separate age groups. n=1,928 0-5y 6-11y 12-17y All pediatric ISS 0.978 0.939 0.940 0.992 RTS 0.972 0.970 0.889 0.982 TRISS 0.999 0.982 0.992 0.992 ICISS 0.892 0.850 0.878 0.970