G308 Comparing paediatric trauma standardised mortality using tarn and picanet benchmarks

Abstract

Introduction and aims Illness and injury severity scoring systems enable assessment of quality of care and benchmarking between units by correcting for case acuity. UK trauma network members are assessed by Trauma Audit and Research Network (TARN) calculations of Probability of Survival(TPS). This is based on Injury Severity Score(ISS), age, gender, conscious state and (adult specific) modified Charlson Comorbidity Index. This is not validated for children resulting in wide confidence intervals and limited clinical utility. Paediatric Intensive Care Units are benchmarked by the validated Paediatric Index of Mortality (PIM) calculated from physiological variables at time of admission and risk stratified diagnosis groups. We observed a clinical disparity between the scores of these methods and sought to determine if we could identify patterns of discrepancies. Methods TARN and PICANet data submissions for our regional trauma centre and PICU were retrospectively analysed for 2014–2017, comparing PIM3 (PIM2 prior to 2016) and TARN mortality (TM calculated as 1-TPS) predictions by age and injury type. All cases with discordant predicted mortality >50% were reviewed, categorising the key influencing factors for their scores. Results 110 trauma patients were admitted to PICU with an actual mortality of 10.9% and mean predicted mortality of PIM 10.3% vs TPS 13.8% (p=0.5). These included all trauma deaths in our Major Trauma Centre during this period. There was minimal correlation between methods (Pearson coefficient 0.4). Examining discordant outliers, children with a high TM (63%) but low PIM (7%) had high ISS (mean 44) with a 60% risk of death. The group with low TM (9%) but high PIM (78%) were predominantly infants with abusive head trauma presenting with out of hospital cardiac arrest and fixed dilated pupils. They had lower ISS (32) but a mortality of 75%. Discussion and conclusion We have found significant discrepancy between the two standard benchmarking systems for critically injured children in the UK. The current use of TARN adult trauma score does not accurately account for variation in paediatric case-mix with variation particularly driven by known predictors of poor neurological outcomes. We suggest simple modifications to allow the derivation of a paediatric specific method to enable meaningful benchmarking of paediatric trauma.