Hospital Patterns of Medical Management Strategy Use for Patients With Non–ST-Elevation Myocardial Infarction and 3-Vessel or Left Main Coronary Artery Disease

Ralf E. Harskamp, MD; Tracy Y. Wang, MD, MHS, MSc; Deepak L. Bhatt, MD, MPH; Stephen D. Wiviott, MD; Ezra A. Amsterdam, MD; Shuang Li, MS; Laine Thomas, PhD; Robbert J. de Winter, MD, PhD; Matthew T. Roe, MD, MHS


Am Heart J. 2014;167(3):355-362. 

In This Article


Study Population

Our study population was derived from the ACTION Registry–GWTG, which is a nationwide database containing hundreds of thousands of patient records from more than 700 participating hospitals in the U.S. The registry uses a standardized data set with written definitions, has requirements in place to ensure uniform data entry and transmission, and is subject to data quality checks ( The general methodology of the registry has been previously described.[12]

For the current study, our starting population included 417,828 patients; of these, we selected patients undergoing coronary angiography from hospitals with both PCI and CABG capabilities from January 1, 2007 through March 31, 2012. Then, we identified those patients who had 3VD (lesions ≥50% in each of the three major epicardial coronary arteries) and/or a LMD (≥50% lesion in the left main vessel). As shown in the flowchart (Figure 1), from the starting population of 417,828 patients at 717 hospitals, we excluded patients with a short version of the data collection form that did not include angiographic details, patients with ST-segment elevation myocardial infarction (STEMI), patients with prior CABG surgery, patients without angiography, patients treated in a hospital without CABG capabilities, patients without 3VD/LMD on angiography, and patients transferred to other institutions since in-hospital outcomes could not be ascertained after inter-hospital transfer. Patients with prior CABG were excluded from the analysis because revascularization decisions are often confounded in these patients due to complex technical and angiographic features, such as degenerated vein grafts, and prior CABG is a very strong predictor of MMS use for NSTEMI patients.[11] After applying our exclusion criteria, 42,535 patients with 3VD/LMD from 423 hospitals were included in the primary analysis cohort.

Figure 1.

Study Population. Flow-diagram that displays the applied exclusion criteria that led to the study population used for this analysis. DCF, data collection form.

Statistical Analysis

In order to study temporal trends in the use of a MMS versus revascularization in the primary analysis cohort, we tested the difference in proportional use of MMS during each quarter of the analysis time period (2007–2012) using the Cochran-Armitage trend test. Next, we assessed whether there was variation in the proportion of MMS use versus revascularization (PCI or CABG) among hospitals over the 5-year period using a generalized linear mixed model with hospitals as a random effect and the variance component tested against zero. Since we expected the range of use to vary widely in hospitals that contributed only a small number of patients, we restricted this and subsequent analyses to 316 hospitals that included at least 25 patients during the 5-year period (total number of patients = 41,310), therefore excluding 107 hospitals with <25 patients included during their participation in the ACTION Registry–GWTG (total number of patients excluded = 1,225). To illustrate the range of MMS use versus revascularization, we plotted the data into a figure showing the percentage and 95% confidence intervals for the use of MMS at each hospital. We also performed a sensitivity analysis to evaluate the variation in MMS use among the 69 hospitals that consistently submitted data during each quarter of the analysis time period.

We then explored whether the variation in hospital-level MMS use could be explained by differences in patient case-mix by stratifying the 316 hospitals with adequate volume into tertiles delineated by the proportion of patients who were treated with a MMS at each hospital. Differences in hospital characteristics, annual volume of NSTEMI patients, use of CABG, patient baseline characteristics, unadjusted in-hospital mortality rates, discharge medications, and interventions were compared among these hospital tertiles during the entire time period. Additionally, we calculated the hospital-level predicted mortality risk (using the ACTION Registry–GWTG in-hospital mortality risk score) to determine if the patient risk profile differed among hospitals.[13] Continuous variables were summarized using medians with interquartile ranges and compared with Kruskall-Wallis tests, whereas categorical variables were displayed using percentages and compared with Mantel-Haenszel χ2 tests. All analyses were performed using SAS version 9.2 (SAS Institute, Cary, NC).