Abstract 4020: A Modified Method for Estimating Volume-Outcome Relationships for Physicians and Hospitals Performing Percutaneous Coronary Intervention
Background: Prior studies have documented a consistent inverse relationship between volume of percutaneous coronary intervention (PCI) procedures by physicians and/or hospitals and adverse outcomes. However these studies utilized methodologies to quantify physician and hospital volume that potentially created bias and imprecision in the results.
Methods: We used New York State PCI data from 1997–1999 to calculate adjusted volume-outcome relationships for physicians and hospitals. For each PCI, physician volume was calculated using the volume of procedures performed by that physician in the previous year and hospital volumes were calculated after excluding the operator’s volume in the previous year from the hospital volume. The relation between volume and outcome was also calculated using the conventional methodology where physician volume is derived from the total number of procedures performed in the same year as the procedure of interest and hospital volume includes the volume of the treating physician. The results of both methods were compared.
Results: PCI was performed on 98,561 patients with a mortality rate of 0.85%. By the conventional approach there was no relationship between physician volume and mortality. However, with the modified method, increasing physician volume was associated with a reduction in the adjusted odds of death. Similarly, with the conventional method, no relationship was found between hospital volume and mortality. But with the modified method, mortality was significantly reduced in the highest volume hospitals (OR, 0.77, 95% CI, 0.62– 0.95). When analyzing the interaction of hospital and physician volume on mortality, the conventional method did not detect a significant interaction. In contrast, the modified method demonstrated a significant interaction between physician and hospital volume with a reduced risk of death when PCI was performed in the highest volume hospitals by the highest volume physicians (OR, 0.643, 95% CI, 0.44 – 0.94).
Conclusion: A more precise measure of physician and hospital volumes leads to different estimates of volume-outcome relationships than conventional methods.