Diagnostic Accuracy of International Classification of Diseases, Tenth Revision (ICD-10) Codes for Vertebral Artery Dissection and Its Complications in Japan: A Single-Center Validation Study.

INTRODUCTION

Vertebral artery dissection (VAD) is a leading cause of cerebrovascular disease in young and middle-aged adults. Although medical claims databases facilitate large-scale observational studies, validated case-identification algorithms for unruptured VAD and its complications remain limited. We evaluated the diagnostic accuracy of International Classification of Diseases, Tenth Revision (ICD-10) codes for identifying unruptured VAD and its complications in Japanese medical claims data.

METHODS

We conducted a single-center cross-sectional validation study using electronic medical records from April 2018 to December 2025. Three ICD-10-based case definitions were evaluated: I72.6 for unruptured VAD, I60.5 for VAD-related subarachnoid hemorrhage (SAH), and the combination of I72.6 with I63.0-I63.9 for unruptured VAD with cerebral infarction. Two independent clinicians determined reference standards based on imaging findings and documented clinical diagnoses. The primary outcome was a positive predictive value (PPV), with 95% confidence intervals (CI) calculated using the Wilson method.

RESULTS

A total of 104 patients met the ICD-10-based case definitions (64 with I72.6, 12 with I60.5, and 28 with combined I72.6 and I63.0-I63.9). The PPV was 0.81 (95% CI, 0.70-0.89) for I72.6 alone, 0.82 (95% CI, 0.55-0.95) for I60.5, and 1.00 (95% CI, 0.88-1.00) for the combined I72.6 and I63.0-I63.9. When I72.6 was restricted to codes assigned on the hospital admission date, the PPV increased to 0.93 (95% CI, 0.81-0.98).

CONCLUSION

The combination of ICD-10 codes for unruptured VAD and cerebral infarction yielded a high PPV and may be suitable as a pragmatic initial case identification algorithm for pharmacoepidemiologic studies using Japanese medical claims databases.