Injury coding in a national trauma registry: a one-year validation audit in a level 1 trauma centre

Hospitals must improve patient safety and quality continuously. Clinical quality registries can drive such improvement. Trauma registries code injuries according to the Abbreviated Injury Scale (AIS) and benchmark outcomes based on the Injury Severity Score (ISS) and New ISS (NISS). The primary aim of this study was to validate the injury codes and severities registered in a national trauma registry. Secondarily, we aimed to examine causes for missing and discordant codes, to guide improvement of registry data quality.

We conducted an audit and established an expert coder group injury reference standard for patients met with trauma team activation in 2015 in a Level 1 trauma centre. Injuries were coded according to the AIS. The audit included review of all data in the electronic health records (EHR), and new interpretation of all images in the picture archiving system. Validated injury codes were compared with the codes registered in the registry. The expert coder group's interpretations of reasons for discrepancies were categorised and registered. Inter-rater agreement between registry data and the reference standard was tested with Bland-Altman analysis.

We validated injury data from 144 patients (male sex 79.2%) with median age 31 (inter quartile range 19-49) years. The total number of registered AIS codes was 582 in the registry and 766 in the reference standard. All injuries were concordantly coded in 62 (43.1%) patients. Most non-registered codes (n = 166 in 71 (49.3%) patients) were AIS 1, and information in the EHR overlooked by registrars was the dominating cause. Discordant coding of head injuries and extremity fractures were the most common causes for 157 discordant AIS codes in 74 (51.4%) patients. Median ISS (9) and NISS (12) for the total population did not differ between the registry and the reference standard.

Concordance between the codes registered in the trauma registry and the reference standard was moderate, influencing individual patients' injury codes validity and ISS/NISS reliability. Nevertheless, aggregated median group ISS/NISS reliability was acceptable.