The positive predictive value of ICD-10-AM S06.0~ concussion codes for mild traumatic brain injury

Background: Increasing concern regarding long-term consequences of mild traumatic brain injury (mTBI; concussion) highlights the need for an accurate understanding of its epidemiological landscape. Current estimates of mTBI presentations to the emergency department (ED) are based on principal diagnosis International Classification of Diseases, Tenth Revision, Australian Modification (ICD-10-AM) S06.0~ concussive injury codes assigned by the treating ED clinician in the ED information system (EDIS). While mTBI is defined by absence of loss of consciousness (LOC) or LOC < 30 minutes, the S06.0~ codes include injuries associated with longer periods of LOC. To ensure reliable case estimates to inform healthcare resource allocation, it is essential to understand which S06.0~ codes most accurately capture mTBI and to minimise its misclassification. Objective: To identify the positive predictive value (PPV) of principal diagnosis S06.0~ codes for mTBI using information from EDIS and retrospective chart review. Method: All episodes of care assigned principal diagnosis S06.0~ codes between 1 July 2022 and 30 June 2023 at Royal Perth Hospital ED, Western Australia, were identified, and clinical diagnoses were confirmed via medical record review. PPVs were calculated for each S06.0~ code, and logistic regression models implemented (RStudio v4.3.3) to explore the influence of patient characteristics on coding. Results: Code S06.00 was most frequently assigned ( n = 279, 51.7%). No cases were coded S06.04 or S06.05 (>24 hours LOC). For mTBI, the adjusted PPV of S06.0~ was 71.8% (95% confidence interval (CI) 66.85–76.22), increasing to 74.0% (95% CI 68.72–78.69) after excluding S06.03 (LOC 30 minutes to 24 hours). The highest PPV was observed for the four-digit S06.0 code (adjusted PPV = 87.0%; 95% CI 74.36–93.90). PPVs were lower among patients arriving via emergency transport, older adults (>65 years), and cases without a documented injury mechanism. Conclusion: There are limitations to using ICD-10-AM concussive injury codes to identify mTBI in ED settings, and patient characteristics influence coding accuracy. Reconsideration of current ED coding practices, including exploration of alternative models that reduce clinician burden and improve coding accuracy, is recommended. Implications for health information management practice: The introduction of electronic medical records provides an opportunity to standardise documentation of key clinical findings and develop artificial intelligence tools to improve coding of mTBI.