Background: Hospital charges associated with percutaneous transluminal coronary revascularization (PTCR) in the United States exceeded $6 billion in 1994 and are likely to be constrained in some manner in the near future. Despite this high cost to the public, little is known about the major determinants and sources of variability of PTCR.
Methods and results: From a consecutive series of 1258 procedures with attempted PTCR at a single tertiary referral center, we analyzed 65 clinical, angiographic, physician, and outcome variables as potential correlates of total (hospital and physician) cost. Direct and indirect costs, both hospital and physician, were determined on the basis of resource utilization using "top-down" methodology and were available for 1237 procedures (1086 patients) (98.3%). Mean (+/- SD) patient age was 62 +/- 11 years, 76% were male, 3% had acute myocardial infarction, 71% had unstable angina, 58% had multivessel disease, left ventricular ejection fraction was 54 +/- 12%, 26% had use of at least one nonballoon revascularization device, and median length of stay was 4.4 days. Procedural success was obtained in 89%, and major complications (death, bypass surgery, or Q-wave myocardial infarction) occurred in 3.8%. The median cost was $9176, but it was asymmetrically distributed, and the interquartile and total ranges were wide ($7333 to $13,845 and $3422 to $193,474, respectively). Analyses of independent correlates of cost and loge(cost) were performed using multivariate linear regression in training and test populations. Modeling found 15 independent preprocedural correlates of loge(cost) (R2 = .37) and 23 overall correlates (R2 = .65), excluding length of stay per se. Additional of length of stay to the model increased the explanatory power of the model to R2 = .82. Preprocedural variables most predictive of loge(cost) included presentation with acute myocardial infarction, decision delay (> 48 hours between admission and diagnostic angiography and/or > 24 hours between angiography and intervention), weekend delay, use of intra-aortic balloon counterpulsation, intention to stent, creatinine > or = 2.0 mg%, and lesion complexity (modified American College of Cardiology/American Heart Association score) (all P < .001). In the model that included postprocedural variables as well, length of stay, noncardiac death, urgent bypass surgery, use of the Rotablator, Q-wave myocardial infarction, rise in creatinine > or = 1.0%, and blood product transfusion were all strong independent correlates of loge(cost) (P < .001).
Conclusions: The range of total hospital costs associated with percutaneous intervention is extraordinarily wide. Baseline patient characteristics account for nearly half of the explained variance, but procedural complications and system delays account for much of the remainder. Quantification of the determinants of cost may promote more economically efficient care in the future.