The importance of validating intracranial bleeding diagnoses in The Health Improvement Network, United Kingdom: Misclassification of onset and its impact on the risk associated with low‐dose aspirin therapy

Purpose To evaluate misclassification of intracranial bleeding (ICB) onset in The Health Improvement Network and assess its impact on risk associated with low‐dose aspirin preventive therapy. Methods A total of 199 049 new users of low‐dose aspirin and 1:1 matched non‐users were followed to identify incident ICB cases with validation involving manual review of patient records and linked hospital data. The index date was the date of the recorded diagnosis (initial cases). After a second manual review including free text comments, the index date was backdated to the first symptom date; prevalent cases were excluded. Two nested case‐control analyses were undertaken: using initial cases and using final cases (with backdated index date). Current, recent, and past low‐dose aspirin use was defined as 0 to 7, 8 to 90, and 91 to 365 days before the index date, respectively. Results Among 1635 initial and 1611 final ICB cases, there were, respectively, 37% vs 38% current users of low‐dose aspirin, 10% vs 8% recent users, and 7% vs 7% past users. Current users with duration <3 months accounted for 6% of initial cases and 4% of final cases. Compared with never use, odds ratios (95% confidence intervals) using initial and final cases, respectively, were current use 0.97 (0.84‐1.12) vs 0.97 (0.84‐1.12); <3 months current use, 1.65 (1.28‐2.13) vs 1.13 (0.85‐1.50); recent use 1.52 (1.24‐1.88) vs 1.10 (0.88‐1.37); and past use 1.30 (1.03‐1.64) vs 1.17 (0.92‐1.48). Conclusions Misclassifying ICB onset in The Health Improvement Network marginally impacts the categorization of low‐dose aspirin use yet notably affects estimates of associated risk.