PT - JOURNAL ARTICLE AU - Montoya, Luisa AU - Davie, Gabrielle AU - Kool, Doctor Bridget AU - Dicker, Bridget TI - 84 Comparison of manually and algorithm derived injury severity scores AID - 10.1136/injuryprev-2022-safety2022.30 DP - 2022 Nov 01 TA - Injury Prevention PG - A10--A10 VI - 28 IP - Suppl 2 4099 - http://injuryprevention.bmj.com/content/28/Suppl_2/A10.2.short 4100 - http://injuryprevention.bmj.com/content/28/Suppl_2/A10.2.full SO - Inj Prev2022 Nov 01; 28 AB - Background Various attempts have been made to reduce the administrative burden of manually assigning Abbreviated Injury Severity (AIS) codes to derive Injury Severity Scores (ISS) in trauma registry data. The accuracy of the resulting measures remains unclear, especially in the New Zealand (NZ) context. The aim of this study was to compare ISS derived from hospital discharge International Classification of Diseases (ICD-10-AM) codes with ISS recorded in the NZ Major Trauma Registry (NZ-MTR).Methods Individuals admitted to hospital and enrolled in the NZ-MTR between 1 December 2016 and 30 November 2018 were included. ISS were calculated using a modified ICD/AIS mapping tool (ICD-ISS). The agreement between MTR-ISS and ICD-ISS for raw scores was assessed by Bland-Altman charts and the Intraclass Correlation Coefficient (ICC). Analysis was conducted by gender, age, ethnicity, and mechanism of injury.Results 3,154 patients fulfilled the inclusion criteria. The ICC for agreement between MTR-ISS and ICD-ISS was 0.40 (95% CI: 0.37–0.43), indicating poor agreement. The Bland-Altman plot showed a bias toward lower ISS by the ICD-ISS than the MTR-ISS (average difference was 1.79±10.02). A moderate agreement (ICC=0.50, 95%CI: 0.40–0.60) between MTR-ISS and ICD-ISS was found in the paediatric population (<15 years). Differences by mechanism of injury were observed.Conclusion The overall agreement between MTR-ISS and ICD-ISS was poor. On average, the ICD/AIS mapping tool tended to underestimate ISS, with a few exceptions.Learning Outcomes Further modifications are needed to current algorithms to develop higher alignment with current manual calculations.