Professor of Medicine and Epidemiology Department of Medicine, Rutgers Robert Wood Johnson Medical School, New Brunswick, NJ New Brunswick, United States
Background: Regulatory agencies in Japan encourage the use of real world data for post-marketing surveillance. While venous thromboembolism (VTE) is one of serious side effects of old and new medications, algorithms to define VTE have not been not validated in Japan.
Objectives: To determine positive predictive values (PPV) of claims-based algorithms for any deep vein thrombosis (DVT), any pulmonary embolism (PE), or any VTE.
Methods: We analyzed claims and electronic medical record (EMR) data from a large tertiary healthcare system. A committee of clinical/epidemiology experts developed claims-based algorithms for VTE. We randomly sampled 166 algorithm-defined any VTE cases (with 3 subgroups: DVT only, PE only, and PE & DVT) - 93 any DVT (with 2 subgroups: DVT only and PE & DVT); 84 any PE (with 2 subgroups: PE only and PE & DVT) -, reviewed EMR to identify gold standards (GSs), conducted descriptive analyses, and calculated PPV of algorithms with 95% confidence interval (CI) for: GS1 - physician diagnosis of VTE in medical records and GS2 - adjudication by experts, considering Japanese VTE guidelines.
Results: Among 686 claims-defined VTE cases (mean age 62 and 62% female), 479 were any DVT (mean age 63 and 65% female) and 257 were any PE (mean age 59 and 55% female). 166 cases randomly sampled, all received ≥1 treatments: 42% warfarin, 66% direct oral anticoagulants, 57% other anticoagulants, 9% relevant procedures, and 23% oxygen administration. Supporting symptoms were observed in 48% of cases with 19% dyspnea, 5% chest pain, 28% leg swelling, 3% syncope, and 4% palpitation. Abnormal was observed in 82%, 71%, and 66% on computerized tomography angiography, compressive ultrasound, and ventilation perfusion, respectively. For prevalent cases, PPV using GS of physician diagnosis were 70% (CI: 57-83), 76% (CI: 62-89), and 74% (CI: 64-83); while using expert adjudication, were 71% (CI: 58-84), 77% (CI: 64-91), and 75% (CI: 66-85) for any DVT, any PE and any VTE, respectively. For incident cases, we excluded patients with chronic thromboembolic pulmonary hypertension (CTEPH) within ±1 month of PE diagnosis. PPV using GS of physician diagnosis were 78% (CI: 65-83) and 74% (CI: 65-85); while using expert adjudication, were 78% (CI: 65-92) and 76% (CI: 66-85) for any PE and any VTE, respectively.
Conclusions: Our validation study for VTE in Japan showed PPVs of 70-80%, similarly to PPVs from US, Canada, and Europe. More validation studies for VTE involving different types of institutions are needed to assess/confirm the generalizability of our results.
