Using claims data to attribute patients with breast, lung, or colorectal cancer to prescribing oncologists.

Background: Alternative payment models frequently require attribution of patients to individual physicians to assign cost and quality outcomes. Our objective was to examine the performance of three methods for attributing a patient with cancer to the likeliest physician prescriber of anticancer drugs for that patient using administrative claims data.

Methods: We used the HealthCore Integrated Research Environment to identify patients who had claims for anticancer medication along with diagnosis codes for breast, lung, or colorectal lung cancer between July 2013 and September 2017. The index date was the first date with a record for anticancer medication and cancer diagnosis code. Included patients had continuous medical coverage from 6 months before index to at least 7 days after index. Patients who received anticancer drugs during the 6 months prior to index were excluded. The three methods attributed each patient to the physician with whom the patient had the most evaluation and management (E&M) visits within a 90-day window around the index date (Method 1); the most E&M visits with no time window (Method 2); or the E&M visit nearest in time to the index date (Method 3). We assessed the performance of the methods using the percentage of the study cohort successfully attributed to a physician, and the positive predictive value (PPV) relative to available physician-reported data on patient(s) they treat.

Results: In total, 70,641 patients were available for attribution to physicians. Percentages of the study cohort attributed to a physician were: Method 1, 92.6%; Method 2, 96.9%; and Method 3, 96.9%. PPVs for each method were 84.4%, 80.6%, and 75.8%, respectively.

Conclusion: We found that a claims-based algorithm - specifically, a plurality method with a 90-day time window - correctly attributed nearly 85% of patients to a prescribing physician. Claims data can reliably identify prescribing physicians in oncology.