Abstract
OBJECTIVE: To estimate the association between oncologists' receipt of payments from the pharmaceutical industry and delivery of non-recommended or low value interventions among their patients. DESIGN: Cohort study. SETTING: Fee-for-service Medicare claims. PARTICIPANTS: Medicare beneficiaries with a diagnosis of incident cancer (new occurrence of a cancer diagnosis code in proximity to claims for cancer treatment, and no such diagnosis codes during a ≥1 year washout period) during 2014-19, who met additional requirements identifying them as at risk for one of four non-recommended or low value interventions: denosumab for castration sensitive prostate cancer, granulocyte colony stimulating factors (GCSF) for patients at low risk for neutropenic fever, nab-paclitaxel for cancers with no evidence of superiority over paclitaxel, and a branded drug in settings where a generic or biosimilar version was available. MAIN OUTCOME MEASURES: Receipt of the non-recommended or low value drug for which the patient was at risk. The primary association of interest was the assigned oncologist's receipt of any general payments from the manufacturer of the corresponding non-recommended or low value drug (measured in Open Payments) within 365 days before the patient's index cancer date. The two modeling approaches used were general linear model controlling for patients' characteristics and calendar year, and general linear model with physician level indicator variables. RESULTS: Oncologists were in receipt of industry payments for 2962 of 9799 patients (30.2%) at risk for non-recommended denosumab (median $63), 76 747 of 271 485 patients (28.3%) at risk for GCSF (median $60); 18 491 of 86 394 patients (21.4%) at risk for nab-paclitaxel (median $89), and 4170 of 13 386 patients (31.2%) at risk for branded drugs (median $156). The unadjusted proportion of patients who received non-recommended denosumab was 31.4% for those whose oncologist had not received payment and 49.5% for those whose oncologist had (prevalence difference 18.0%); the corresponding values for GCSF were 26.6% v 32.1% (5.5%), for nab-paclitaxel were 7.3% v 15.1% (7.8%), and for branded drugs were 88.3% v 83.5% (-4.8%). Controlling for patients' characteristics and calendar year, payments from industry were associated with increased use of denosumab (17.5% (95% confidence interval 15.3% to 19.7%)), GCSF (5.8% (5.4% to 6.1%)), and nab-paclitaxel (7.6% (7.1% to 8.1%)), but lower use of branded drugs (-4.6% (-5.8% to -3.3%)). In physician level indicator models, payments from industry were associated with increased use of denosumab (7.4% (2.5% to 12.2%)) and nab-paclitaxel (1.7% (0.9% to 2.5%)), but not with GCSF (0.4% (-0.3% to 1.1%)) or branded drugs (1.2% (-6.0 to 8.5%)). CONCLUSIONS: Within some clinical scenarios, industry payments to physicians are associated with non-recommended and low value drugs. These findings raise quality of care concerns about the financial relationships between physicians and industry.