ment of intracoronary stents, other new devices, and recently developed adjunctive pharmacologic therapies. No generally accepted methodology exists that might allow comparison of coronary interventional mortality rates between widely disparate laboratories and hospi- tals. Validation of models that allow multivariate case- mix adjustment would facilitate comparison of proce- dural results between hospitals and operators. As such, it was the purpose of this study to determine whether a risk equation based on patient-related variables and derived from an entirely independent dataset 11 could accurately predict procedural death after percutaneous intervention at our institution. Methods Mayo Clinic Angioplasty Registry A prospective interventional database has been kept at the Mayo Clinic under the auspices of an Institutional Review Board–approved protocol since percutaneous transluminal coronary angioplasty (PTCA) was first performed in 1979 at the Mayo Clinic and includes demographic, clinical, angio- graphic, and procedural data. Angiographic characteristics are coded at the time of the interventional procedure, immediate and in-hospital events are recorded, and all patients are subse- quently contacted by telephone at 6 and 12 months and yearly thereafter. The current study was performed with the approval of the Institutional Review Board of the Mayo Foundation. Percutaneous coronary interventional (PCI) proce- dures are widely performed in the United States and around the world. Regional variation in results and hospital and operator volumes has focused attention on credentialing and quality. Comparison of results after PCI procedures between hospitals and operators is inherently extremely difficult because of differences in case mix, referral patterns, and the rapid evolution of procedural techniques. Although an empiric relation between procedural volumes and case-fatality rates has been demonstrated for several cardiac and noncardiac surgical procedures, 1-4 other factors that may contribute to death have not been as well studied. An inverse volume-outcome relation after balloon angioplasty has been reported in some 5-7 but not all 8-11 series. A recent report extended these observations to patients undergoing stent deploy- ment. 12 However, most reports antedate the develop- From the Division of Cardiovascular Diseases and Internal Medicine and the Sec- tion of Biostatistics, Mayo Clinic and Mayo Foundation. Submitted August 2, 1998; accepted December 28, 1999. Reprint requests: C. Rihal, MD, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Copyright © 2000 by Mosby, Inc. 0002-8703/2000/$12.00 + 0 4/1/105299 doi:10.1067/mhj.2000.105299 Interventional Cardiology Prediction of death after percutaneous coronary interventional procedures Charanjit S. Rihal, MD, Diane E. Grill, MS, Malcolm R. Bell, MD, Peter B. Berger, MD, Kirk N. Garratt, MD, and David R. Holmes Jr, MD Rochester, Minn Background The prediction and comparison of procedural death after percutaneous coronary interventional proce- dures is inherently difficult because of variations in case mix and practice patterns. The impact of modern, expanded patient selection criteria, and newer technologic approaches is unknown. Our objective was to determine whether a risk equation based on patient-related variables and derived from an independent data set can accurately predict procedural death after percutaneous coronary intervention in the current era. Methods and Results An analysis was made of the Mayo Clinic Coronary Interventional Database January 1, 1995, to October 31, 1997. Expected mortality rate was calculated with the use of the New York State multivariate risk score. In 3387 patients, 3830 procedures (55.1% stents) were performed, with an expected mortality rate of 2.32% and observed mortality rate of 2.38% (P = not significant). The risk score derived from the New York multivariate model was highly predictive of death (chi-square = 213.8; P < .0001). The presence of a high-risk lesion characteristic such as calcium, thrombus, or type C lesion was modestly associated with death. Conclusions The New York State multivariate model accurately predicted procedural death in our database. (Am Heart J 2000;139:1032-8.)