Each letter (A-O) represents a separate outlier physician.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Yang N, Groeneveld PW, Khatana SAM, Giri J, Fanaroff AC, Nathan AS. Variability in Reported Percutaneous Coronary Intervention Mortality Among Physicians Practicing at Multiple Sites in New York State. JAMA Cardiol. 2021;6(4):477–478. doi:10.1001/jamacardio.2020.6717
New York State (NYS) publishes reports of percutaneous coronary intervention (PCI) mortality at the physician level to inform patients; however, most physicians in states with public reporting describe pressure to avoid high-risk cases owing to concerns about poor reported performance.1-3 Prior studies have suggested that mortality inadequately assesses PCI quality because it occurs rarely and physician-level risk-adjusted mortality rates are unstable year to year.4 Further, in addition to overall physician-level outcomes, physicians’ outcomes are also published by site. For physicians who practice at multiple sites, these estimates may be unstable because their PCI volume performed at each site may be low. While a categorical outlier analysis based on confidence intervals is provided in the reports, this may be more difficult for patients to interpret than numerical mortality rates, which they may use to compare physicians in a given site. We examined variation in site-specific risk-adjusted mortality rates (RAMR) for physicians practicing at multiple sites within NYS.
We used publicly reported 30-day RAMR for PCI performed in NYS between 2014 and 2016 by each physician from the NYS Department of Health website. Institutional review board approval was not sought given the use of publicly available data in published reports on the NYS Department of Health website. We obtained published site-specific 30-day RAMR (ssRAMR) at each hospital where physicians performed PCI from a hospital-level presentation of outcomes. We obtained the overall mean RAMR (mRAMR) during this period from an aggregate presentation of outcomes for physicians practicing at multiple sites. We defined physicians as outliers if either the mRAMR or maximum ssRAMR point estimate was greater than the ninety-fifth percentile for RAMR values for all physicians. All statistical testing was 2-tailed, with P values less than .05 designated statistically significant.
Between 2014 and 2016, 142 853 PCI were performed by 373 physicians at 61 hospitals. A total of 207 physicians (55.5%) practiced at multiple hospitals, accounting for 82 075 procedures (57.5%), with no difference in PCI volume among physicians practicing at 1 or more than 1 site (P = .44). Physicians practicing at more than 1 site performed PCI at a median of 2 (IQR, 2-3) hospitals, with a median of 56 (IQR, 12-215.25) PCI performed per hospital. Among these physicians, the median mRAMR was 1.11% (IQR, 0.66%-1.60%; range, 0-5.33%) and the median ssRAMR was 0.52% (IQR, 0-1.53%; range, 0-47.69%). Overall, 15 physicians practicing at multiple sites were classified as outliers using reported maximum ssRAMR values, but only 4 of these physicians were also classified as an outlier using mRAMR (Figure).
Public reporting of PCI quality focuses on mortality, a rare complication. We found that individual ssRAMRs reported for physicians practicing at multiple hospitals are highly variable, and mRAMR and ssRAMR outlier status are inconsistent with each other. Only 4 physicians who had an outlier for ssRAMR also were outliers for an aggregate mRAMR across all of the sites where they practiced. The extreme variability in ssRAMR is in part owing to the Law of Large Numbers, which states that the accuracy of empirical statistics improves with the number of observations.5 Because physicians’ PCI volumes at individual centers may be small, a single mortality event at a hospital where a physician performs a few cases could result in an extremely high reported RAMR owing to the stochastic nature of mortality following PCI, even with perfect risk adjustment. For physicians practicing at multiple hospitals, mRAMR may be a more stable and accurate estimate of PCI quality than ssRAMR owing to the larger sample of patients included. Thus, given its unreliability, public reporting of ssRAMR in NYS does not adequately reflect the quality of care delivered by physicians performing PCI, may be misleading to patients, and should not be reported for physicians practicing at multiple sites. These findings may more broadly inform quality improvement initiatives beyond PCI in NYS.
Corresponding Author: Ashwin S. Nathan, MD, Cardiovascular Medicine Division, Hospital of the University of Pennsylvania, Perelman Center, South Tower, 11th Floor, 3400 Civic Center Blvd, Philadelphia, PA 19104 (firstname.lastname@example.org).
Accepted for Publication: November 5, 2020.
Published Online: December 30, 2020. doi:10.1001/jamacardio.2020.6717
Author Contributions: Dr Nathan and Ms Yang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: All authors.
Acquisition, analysis, or interpretation of data: Yang, Fanaroff, Nathan.
Drafting of the manuscript: Yang, Nathan.
Critical revision of the manuscript for important intellectual content: Groeneveld, Khatana, Giri, Fanaroff.
Statistical analysis: Yang, Groeneveld, Nathan.
Administrative, technical, or material support: Giri, Fanaroff.
Supervision: Groeneveld, Giri, Fanaroff, Nathan.
Conflict of Interest Disclosures: Dr Giri has served on an advisory board for AstraZeneca and received research support to the institution from Recor Medical and St. Jude Medical. Dr Fanaroff receives research support from the American Heart Association and Boston Scientific and honoraria from the American Heart Association. No other disclosures were reported.