Impact of AIS 2015 versus 1998 on injury severity scoring and mortality prediction - single centre retrospective comparison study

Abstract

BackgroundA key component of trauma system evaluation is the Injury Severity Score (ISS). The ISS is dependent on the AIS, and as AIS versions are updated this effects the number of patients within a health system which are considered severely injured (ISS >15). This study aims to analyse the changes comparing AIS1998 and AIS2015, and its impact on injury severity scoring and survival prediction model in a major trauma centre. MethodsThis retrospective study reviewed all blunt trauma admissions from 1 January 2020 to 31 December 2020 from the trauma registry of Prince of Wales Hospital, Hong Kong. Patients were manually double coded with AIS1998 and AIS2015 by the same experienced trauma nurse who have completed both AIS 1998 and AIS 2015 Courses. AIS patterns and Injury Severity Scores (ISS) derived from AIS 1998 and 2015 were compared using the Wilcoxon Signed Rank Test. The area under the receiving operator curve (AUROC) was compared based on the Trauma and Injury Severity Score (TRISS) model using AIS 1998 and AIS 2015. Results739 patients were included. There were 34 deaths within 30 days (30-day mortality rate 4.6%). Patients coded with AIS2015 compared with AIS1998 had significant reductions in the classification of serious, severe and critical categories of AIS, with a substantial increase in the mild and moderate categories. The largest reduction was observed in the head and neck region (Z = −11.018, p < 0.001), followed by the chest (Z = −6.110, p < 0.001), abdomen (Z = −4.221, p < 0.001) and extremity regions (Z = −4.252, p < 0.001). There was a 27% reduction in number of cases with ISS >15 in AIS2015 compared with AIS1998. Rates of 30-day mortality, ICU admission, emergency operation and trauma team activation of ISS > 15 using AIS 1998 were similar to the cut off for New Injury Severity Score (NISS) >12 using AIS 2015. The AUROC from the TRISS (AIS2015) was 0.942, and not different from the AUROC for TRISS (AIS1998) of 0.936. The sensitivity and specificity were 93.9% and 82.1% for TRISS (AIS2015), and 93.9% and 76.0% for TRISS (AIS1998). ConclusionTrauma centres should be aware of the impact of the AIS2015 update on the benchmarking of trauma care, and consider the need for updating the ISS cut off for major trauma definitions.