Accuracy and Usability of Emergency Department Diagnostic Codes: A Multicentre Observational Study

Le résumé fourni par la source

Objectives: To evaluate the accuracy and usability of emergency department (ED) clinician-assigned diagnostic codes across all four Tasmanian public hospitals. Methods: We conducted a retrospective study of 300 ED presentations from all four public EDs in Tasmania. Two reviewers, a specialist emergency physician and a clinical research fellow, classified the accuracy of diagnostic codes across five levels which were then further classified into usable (levels 1–2) and unusable (levels 3–5). Rater agreement was evaluated. For accuracy and usability evaluation, conflicting reviewer classifications were removed. Associations between usability and patient, presentation, and site characteristics were evaluated using chi-square and Fisher’s exact tests. Results: Overall rater agreement was moderate for both accuracy (69.3%, κ = 0.52) and usability (81.7%, κ = 0.58). Among 208 eligible presentations, 139 (66.8%) diagnostic codes were correct, 153 (73.6%) were acceptable for secondary use, 9 (4.3%) were clearly incorrect and 13 (6.3%) had insufficient documentation. Coding performance varied across hospitals, acuity, arrival mode, and representation intervals, with hospital-level usability ranging from 63.6% to 88.9%. Conclusions: Across all four Tasmanian public EDs, 66.8% of diagnostic codes were accurate, 73.6% were appropriate for secondary use, and 4.3% were clearly incorrect, with moderate rater agreement for accuracy and usability (κ = 0.52 and κ = 0.58). These findings, together with marked between-hospital variation, underscore the context dependence and subjectivity of ED diagnostic coding. They support a cautious approach to using ED diagnostic codes for health service planning, performance reporting, and observational research, while highlighting the need to address barriers and opportunities to strengthen coding performance.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.