Assessing the accuracy of ICD-10 codes for identifying acute thromboembolic events among patients receiving anticoagulation therapy.

International classification of disease (ICD) codes can improve the efficiency of epidemiological research provided the codes accurately identify outcomes of interest. The purpose of this retrospective cross-sectional study is to evaluate the accuracy of ICD-10 codes for identifying thromboembolic events occurring during anticoagulation therapy. Medical charts of patients hospitalized for any reason while receiving anticoagulant therapy between September 1, 2017 and December 31, 2017 were reviewed by two reviewers blinded to ICD-10 code status. Following identification of confirmed acute thromboembolic events, ICD-10 codes were unblinded and sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) along with 95% confidence intervals (CI) were calculated for coding in any diagnosis position ("principal" or "other"). There were 661 hospitalizations identified among 487 anticoagulated patients. There were 27 thromboembolic events identified during chart review. Stroke and venous thromboembolism were the most common thromboembolic event types. Overall thromboembolic ICD-10 coding sensitivity was 100.0% (95% CI 87.2-100.0); specificity was 79.3% (75.9-82.4). PPV was 17.1% (11.6-23.9%), and NPV was 100% (99.3-100.0). ICD-10 codes can reliably be used for ruling out hospitalizations for thromboembolic events in patients receiving anticoagulation therapy but should not be used for identifying thromboembolic complications without confirmatory chart review.