Presenter: Ashwin Nathan, MD, MS, of the University of Pennsylvania, presented these results on behalf of his colleagues from his institution, Beth Israel Deaconess Medical Center in Boston, and Duke Clinical Research Institute in Durham, North Carolina(1).

PCI outcomes are submitted to the NCDR. Quarterly reports produced for hospitals and individual physicians are increasingly shared publicly to compare the performance of physicians with each other and help patients better understand the quality of the hospitals and physicians caring for them. The NCDR uses risk adjustment to allow for fairer comparisons of hospitals and physicians. Risk adjustment accounts for the baseline condition of the patient, allowing for a fair comparison of hospitals taking care of sicker patients with hospitals taking care of less-sick patients. Nathan said this technique works for post-PCI mortality. He presented data that showed the observed mortality of patients after PCI matching fairly well with the predicted mortality based on their pre-PCI condition. Even so, he said, most physicians view risk adjustment with suspicion. Reporting of outcomes is associated with an increase in “risk avoidant” behavior, Nathan said, despite risk adjustment’s attempt to level the playing field.

One potential strategy to improve outcomes is for a hospital to avoid all high-risk cases. Nathan and colleagues sought to test the hypothesis that this strategy would not necessarily improve outcomes for hospitals. Nathan and colleagues used all hospitals submitting data to the NCDR in 2017 and reassessed their outcomes using three risk-avoidant strategies. The researchers evaluated the change in hospitals’ risk-adjusted outcome rate and in their ranking based on each of the strategies. Two of the strategies are not possible to perform in the real world because they would require hospitals and physicians to know beforehand which patients would live or die in the days following PCI. But one of the assessed strategies, drop all cases in the highest 10% of risk – “blanket risk avoidance” – is possible, Nathan said.

The study included 1,634 hospitals with 742,293 procedures. Among the top tertile ranking of hospitals, the observed mortality was consistently lower than predicted mortality across the spectrum of patient risk. As Nathan said, “Good hospitals perform well on all patients.” For the middle tertile of hospitals, observed mortality was comparable to predicted mortality across low-, medium-, and high-risk patients. Observed mortality was consistently higher than predicted mortality across all patient risk levels at poor-performing hospitals. Importantly, Nathan said, predicted mortality rates of patients in the same risk level were similar for good, medium and poor-performing hospitals. The blanket risk-avoidance strategy, Nathan said, produced mixed results when applied to the NCDR data, with 883 (56.4%) hospitals reducing their mortality rate but 610 (39%) increasing it. This was also true of the effect on rankings, with some hospitals seeing their performance rank rise and while others’ ranks fell. Furthermore, Nathan said, the analysis showed that there were no hospital or patient characteristics that could predict the effect on hospital performance of removing cases in the top 10% risk level. “Most importantly, risk-avoidant behavior may ultimately harm patients that may need this procedure and are not receiving it because of perceived performance,” he said. “And it is irrational to practice this to improve your perceived performance, and we would not recommend practicing systematic risk avoidance.”

All Authors: Giorgio Medranda, MD; Brian C. Case, MD; Jason P. Wermers, BS; Natalie Morrison, BA(Hons); Ron Waksman, MD, MSCAI. 

References

1. Hospital-Level Percutaneous Coronary Intervention Performance with Simulated Risk Avoidance Strategies