Background and Importance Real-world data sources can facilitate essential understanding of the epidemiological features of anaphylaxis. However, the accuracy of case-identifying definitions based on diagnosis codes for anaphylaxis in healthcare databases remains understudied.

Aim and Objectives To evaluate the accuracy of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes to identify anaphylaxis in claims data from the largest healthcare system in Taiwan.

Material and Methods We conducted a cross-sectional study analysing claims data from the largest multi-institutional healthcare system in Taiwan from 2017 to 2021. We included patients with incident anaphylaxis identified by either ICD-10-CM codes for anaphylaxis (Group 1) or ICD-10-CM codes for severe allergic or drug adverse events and additional modifier codes for acute allergy events (e.g., epinephrine, intramuscular or intravascular injection) (Group 2). We randomly selected 20% of the cases to determine the positive predictive value (PPV) of anaphylaxis case-identifying definitions in Groups 1 and 2 after independent review of electronic medical records by two physicians. The clinical criteria for anaphylaxis, proposed at the Second National Institute of Allergy and Infectious Disease/Food Allergy and Anaphylaxis Network, served as the gold standard to confirm anaphylaxis diagnosis (Groups 1 and 2).

Results From the original cohort (n=2,176), we randomly selected 433 patients (20%) with either a diagnosis of anaphylaxis (Group 1), or a diagnosis of severe allergic and drug adverse events with additional modifier codes for acute allergy events (Group 2). In Group 1, we judged 135/170 patients as true anaphylaxis cases (median age: 47 years; female: 46.5%), giving a PPV of 79.4% (95% CI: 73.3–85.5). In Group 2, we judged 47/263 patients as true anaphylaxis cases (median age: 48 years; female: 54.0%), giving a PPV of 17.9% (95% CI: 13.3–22.5). The underlying causes for false-positive anaphylaxis identification in Group 2 were urticaria (76.7%) and angioedema (23.4%).

Conclusion and Relevance Acceptable PPVs were observed when anaphylaxis cases were identified by ICD-10-CM codes for anaphylaxis, but not by ICD-10-CM codes for severe allergic or drug adverse event with additional modifier codes for acute allergy events. Our multi-institutional findings could serve as a fundamental reference for further studies of anaphylaxis based on real-world healthcare databases.

Conflict of Interest No conflict of interest.