Accuracy of ICD-10 codes for patients with acute myocarditis: a retrospective study at a large tertiary centre in London, UK

Abstract

Abstract Background Acute myocarditis (AM) is an inflammatory disease of the myocardium that is associated with heterogenous clinical presentations, which may be non-specific. Furthermore, several cardiac diseases can mimic its clinical phenotype. Research in AM frequently utilises ICD-10 codes from hospital admissions for case identification. We retrospectively confirmed or refuted a diagnosis of AM according to ESC Position Statement criteria in patients with an ICD-10 code for AM. Methods We performed a single-centre retrospective analysis of all unique admissions with ICD-10 codes corresponding to myocarditis or myopericarditis in the first three coding positions at King's College Hospital, London, United Kingdom. The diagnosis was classified as “confirmed” if proven by cardiac magnetic resonance imaging (CMR) or endomyocardial biopsy (EMB) or “suspected” in the absence of CMR or EMB if other ESC Position Statement criteria were met. To identify additional cases, we used an open-source retrieval system for unstructured clinical data (CogStack). We searched hospital and Intensive Care Unit (ICU) discharge summaries for inpatients discharged alive containing the keywords “myocarditis” or “myopericarditis”. We also searched for patients who died during the study period where the keywords “myocarditis” or “myopericarditis” were included on the death notification. Results We identified 308 unique admissions with an ICD-10 code for myocarditis or myopericarditis in this study, presenting between 2008 and 2020 (Figure 1). Overall, 26.0% of patients (n=80/308) could be excluded from a diagnosis of AM on review of the clinical summary. A total of 16.2% of patients (n=50/308) had insufficient evidence of AM, 1.9% (n=6/308) had not had coronary artery disease excluded as a culprit for the presentation, and 10.1% (n=31/308) had an alternative diagnosis. Only 45.8% (n=141/308) of all patients met criteria for suspected or confirmed AM. Of those, 86.5% (n=122/141) of cases were confirmed by CMR. Overall, 39.6% of patients with an ICD-10 code indicating AM had a confirmed diagnosis. An additional 46 suspected and 197 confirmed cases were identified using open-source retrieval from unstructured clinical data. Conclusion AM has heterogenous and sometimes non-specific clinical presentations, which may be compounded by limited access to CMR and EMB. We identified significant misclassification using ICD-10 codes. It is crucial to ensure that studies investigating AM include only patients meeting appropriate diagnostic criteria, thereby ensuring a high-quality evidence base in this disease. Funding Acknowledgement Type of funding sources: None.