Abstract
OBJECTIVE: To evaluate the validity of the Utah statewide All-Payer Claims Database (APCD), we compared breast cancer-specific treatments and dosages with gold-standard abstraction of medical records. STUDY DESIGN: In this pilot study, breast cancer treatments were abstracted by a certified tumor registrar at the Utah Cancer Registry (UCR) for patients diagnosed in 2013 with breast cancer. The abstraction of medical records was the gold standard for comparison with treatments identified in the APCD. The reliability and agreement between the treatment identified in the APCD and abstraction data were measured with sensitivity and specificity. Dose consistency was measured with the intraclass correlation coefficients (ICC). RESULTS: Compared with the 186 abstractions, the sensitivity of the APCD to identify chemotherapy agents was high: 89% for any agent, 91% for carboplatin, 83% for docetaxel, 82% for doxorubicin, or 94.7% for biologic therapy. The consistency between the chemotherapy dosage identified in the claims and the abstraction varied from 63% to 76%. For radiotherapy, the sensitivity of the claims to identify the completed radiotherapy regimen was 66%. The ICC between radiotherapy doses identified in the claims and the abstraction was 54% (95% confidence interval [CI], 48%, 67%). CONCLUSIONS: Employing these novel methods, the claims were highly reliable in identifying cancer treatment agents overall, namely carboplatin, docetaxel, and trastuzumab. The claims were of moderate utility in capturing the treatment dose information. In addition to the APCD, the use of multiple data sources improved the completeness of cancer treatment information.