We conducted a cross-sectional study using data from the NCDR’s PINNACLE Registry and the OPP general payments database between January 1, 2017, and December 31, 2017. The PINNACLE Registry, established in 2008, was the world’s largest outpatient cardiovascular registry and has been described previously. ^,,, It captures data on real-world management of 4 common cardiovascular conditions: heart failure (HF), coronary artery disease (CAD), atrial fibrillation (AF), and hypertension. The diagnoses in the PINNACLE registry were abstracted from physician notes using a standardized collection tool, rather than relying on diagnostic codes, and thus are likely to accurately reflect clinical conditions. However, medication data were based on patient-reported use during clinic visits and was not verified through prescriptions, refills, or pharmacy dispensation records. To ensure data quality, the NCDR team employed rigorous data definitions and periodic audits. All analyses were performed at the Baim Institute for Clinical Research, Boston.

Details about the OPP have also been previously reported. ^,,, The Centers for Medicare & Medicaid Services (CMS) mandated that manufacturers and group provider organizations report all individual payments of $10 or more or aggregate payments exceeding $100 per year (in 2014 USD). Industry payments to physicians were broadly categorized in the OPP as general payments, research payments, and ownership payments. This study focused exclusively on “general payments,” which include consulting fees, travel expenses, speaking at promotional talks, and food and beverage. Each transaction contains information about the associated manufacturer, the product, the amount and nature of the payment. In this study, we restricted our analysis to general payment data, as prior studies demonstrated that only 1% of physicians held ownership interests and only 2% of total research payments were made to individual physicians. ^,

Using PINNACLE Registry data from January 1, 2017 to December 31, 2017, we defined 3 distinct study cohorts: atherosclerotic cardiovascular disease (ASCVD) and/or dyslipidemia, heart failure with reduced ejection fraction (HFrEF), and nonvalvular atrial fibrillation (NVAF). Henceforth, we refer to patients with diagnosis of dyslipidemia and/or atherosclerotic cardiovascular disease as the ASCVD cohort. Patients whose physicians had a missing national provider identifier (NPI) and/or a missing first or last name were excluded. For the NVAF cohort, patients with valvular atrial fibrillation and patients with additional conditions requiring oral anticoagulation (eg, venous or systemic embolism) were excluded. We linked the treating physician with industry payments in OPP using NPIs.

For each cohort, patients were stratified into two groups based on receipt of payments for the target drugs: (1) patients whose prescribing physicians received payments related to PCSK9is (alirocumab and evolocumab) in ASCVD, ARNis (sacubitril/valsartan) in HFrEF, or DOACs (dabigatran, apixaban, and rivaroxaban) in NVAF cohort, and (2) patients whose prescribing physicians did not receive any such payments. We further stratified the NVAF and ASCVD cohorts–drug classes with multiple options–based on the specific drug for which a physician received the largest proportion of the total payments. For example, a physician who received $100 payment for dabigatran and $150 payment for rivaroxaban was assigned to the rivaroxaban group in the NVAF cohort. In our analysis, only payments related to the target drugs in each cohort were included; for example, in the NVAF cohort, we included payments for DOACs (dabigatran, apixaban, and rivaroxaban) and excluded payments for unrelated drugs (eg, PCSK9 inhibitors or ARNI) or devices (eg, stents). The primary exposure was receipt of payments related to the target medications by the prescribing physician.

The primary outcome was patient-level prescription of a target medication in each study cohort: PCSK9 inhibitors in the ASCVD cohort; ARNi in the HFrEF cohort, and DOAC in the NVAF cohort. Secondary outcomes included the proportion of patients receiving specific medications within in each drug class—for example, dabigatran, apixaban, or rivaroxaban in the NVAF cohort; and alirocumab or evolocumab in the ASCVD cohort.

Descriptive analyses of physician payments were performed at the physician-level ( Table 1 ) and all other analyses were conducted at the patient level. For each cohort, we analyzed the total number of physicians, total annual payment amount, and annual median payment amount per physician (interquartile range [IQR]). Physicians were grouped into 4 predefined categoriesby annual total payment: $1-<$100, $101-$1,000, and ≥$1,001. We also calculated the proportion of the total payment value by the nature of the payments: consulting fee, speaker fee, food and beverages, travel and lodging, and “other”—which includes gifts, grants, honoraria, education, entertainment, charitable contribution, and ownership.

We compared the clinical and demographic characteristics of patients based on whether their physician received industry payments. To assess medication use, we conducted frequency table analysises—for the ASCVD cohort: statin vs PCSK9i vs no therapy; for the HFrEF cohort: ARNi vs ACEi/ARB; for the AF cohort: warfarin vs DOAC vs no therapy. For drug classes with multiple options (eg, DOACs and PCSK9is), we examined concordance between the specific drug prescribed and the payments made for that drug (vs, physician receiving payments for rivaroxaban and physician’s prescription of rivaroxaban).

For each cohort, we developed multivariable hierarchical logistic regression models to identify covariates associated with prescription of DOACs, PCSK9is, or ARNis. Physician-level variables included receipt of payments, practice location (urban vs rural), and subspecialty (general cardiology, interventional cardiology, electrophysiology, or “other”, which includes nurse practitioners and physician assistants). Previous analyses showed that in entire PINNACLE registry, 78% of providers were MDs/Dos and 20% were nurse practitioners or physician assistants; thus, the “other” category may not reflect a specific cardiology subspecialty designation. ¹⁰ Receipt of industry payments was modeled both as a binary variable (yes vs no) and as an ordinal variable with payment categories: none, $1-<$100, $100-<$1,000, and ≥$1,000. Due to small proportion of physicians receiving >$1,000 for PCSK9is (2.4%), only 3 payment categories were used for the ASCVD cohort: none, <$100, and ≥$100. For NVAF and HFrEF cohorts, we used 3-level hierarchical models (patients nested within physicians and physicians nested within practice). Fixed effects included receipt of payments, physician specialty, and practice location; random effects were included at the physician and practice levels. For the ASCVD cohort, we used two-level hierarchical models due to the low prescription rate of PCSK9 inhibitors. We also conducted separate models for individual durgs within each class (, dabigatran, apixaban, or rivaroxaban in the NVAF cohort; alirocumab or evolocumab in the ASCVD cohort). All analyses were repeated using 2016 data for validation.

For the PINNACLE Registry data, we excluded variables with >50% missing data were excluded from multivariable models. For the remaining variables, missing values were imputed using the group median for continuous variables and the mode for categorical variables, following prior PINNACLE Registry methodology. ^, For the OPP data, physicians with missing NPI number in either 2016 or 2017 were included only for the year in which their NPI was present. All tests for statistical significance were 2-tailed and evaluated at a significance level of 0.05. The data was accessed for research purposes between Januray 07, 2020 and June 30, 2022. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, North Carolina). The data used for this study was deidenitified, and informed consent was waived. The study was approeved by the Advarra Institutional Review Board approved use of a limited dataset from the PINNACLE Registry for research purposes. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.