Author + information
- Received August 28, 1998
- Revision received April 7, 1999
- Accepted May 16, 1999
- Published online September 1, 1999.
- Gerald T. O’Connor, PhD, DSc, FACC∗,‡,*,
- David J. Malenka, MD, FACC∗,†,
- Hebe Quinton, MS‡,
- John F. Robb, MD, FACC∗,†,
- Mirle A. Kellett Jr., MD, FACC§∥,
- Samuel Shubrooks, MD, FACC¶,
- William A. Bradley, MD, FACC#,
- Michael J. Hearne, MD, FACC#,
- Mathew W. Watkins, MD, FACC∗∗,
- David E. Wennberg, MD§∥,
- Bruce Hettleman, MD, FACC∗,†,
- Daniel J. O’Rourke, MD, MS∗,†,
- Paul D. McGrath, MD§,
- Thomas Ryan Jr., MD, FACC§∥,
- Peter VerLee, MD, FACC††,
- for the Northern New England Cardiovascular Disease Study Group‡‡
- ↵*Reprint requests and correspondence: Dr. Gerald T. O’Connor, Center for the Evaluative Clinical Sciences, Dartmouth Medical School, 7251 Strasenburgh Hall, Room 330, Hanover, New Hampshire 03755-3863
Using recent data, we sought to identify risk factors associated with in-hospital mortality among patients undergoing percutaneous coronary interventions.
The ability to accurately predict the risk of an adverse outcome is important in clinical decision making and for risk adjustment when assessing quality of care. Most clinical prediction rules for percutaneous coronary intervention (PCI) were developed using data collected before the broader use of new interventional devices.
Data were collected on 15,331 consecutive hospital admissions by six clinical centers. Logistic regression analysis was used to predict the risk of in-hospital mortality.
Variables associated with an increased risk of in-hospital mortality included older age, congestive heart failure, peripheral or cerebrovascular disease, increased creatinine levels, lowered ejection fraction, treatment of cardiogenic shock, treatment of an acute myocardial infarction, urgent priority, emergent priority, preprocedure insertion of an intraaortic balloon pump and PCI of a type C lesion. The receiver operating characteristic area for the predicted probability of death was 0.88, indicating a good ability to discriminate. The rule was well calibrated, predicting accurately at all levels of risk. Bootstrapping demonstrated that the estimate was stable and performed well among different patient subsets.
In the current era of interventional cardiology, accurate calculation of the risk of in-hospital mortality after a percutaneous coronary intervention is feasible and may be useful for patient counseling and for quality improvement purposes.
In the 20 years since its use was first reported by Gruntzig (1), percutaneous coronary interventions (PCI) have become a frequently performed therapeutic procedure in patients with coronary artery disease. It has been estimated that in 1994, over 428,000 coronary angioplasty procedures were performed in the U.S. (2). In more recent years, balloon angioplasty has been joined by many nonballoon devices that have enhanced the technique’s capability, extended its indications and, in some cases, lowered the rates of adverse outcome (3). The ultimate benefit of PCI will be judged by its effect on the relief of angina, its prevention of myocardial infarction (MI) and its effects on long-term survival (4). Randomized long-term studies and a recently conducted meta-analysis suggest that PCI may be the treatment of choice for some patients with symptomatic coronary artery disease (5–14). Yet, the anticipated benefits of PCI must be balanced against its risks. In PCI, balloon-induced barotrauma damages the endothelium and often the media and adventitia of the coronary artery (15). Dissection of the arterial wall can be detected in 50% to 80% of patients after PCI (16). Plaque hemorrhage, platelet deposition or clot formation may result in lumen compromise. The major short-term adverse outcomes associated with PCI are the need for urgent or emergent coronary artery bypass graft surgery (CABG), postprocedure MI and in-hospital mortality. Patient counseling, optimal clinical decision making and quality improvement activities require accurate assessment of these risks.
In a previous study (17), we presented a multivariate prediction rule for in-hospital mortality after PCI using regional data from 1990 to 1993. Since then, stents, other nonballoon devices and the newer antiplatelet agents have become more widely available and used. We now have validated outcomes data on consecutive PCI procedures performed in northern New England from 1994 through 1996. This provides an opportunity to identify patient and disease factors that are associated with in-hospital mortality and to develop and validate a multivariate prediction rule in the modern era of PCI.
The Northern New England Cardiovascular Disease Study Group is a voluntary research consortium composed of clinicians, research scientists and hospital administrators at the six regional institutions that are the sole providers of coronary revascularization in northern New England and one Massachusetts-based institution. The intent of the group is to foster continuous improvement in the quality of care of patients with cardiovascular disease in northern New England through the pooling of process and outcome data and the timely feedback of data to clinicians (18,19). Between January 1, 1994 and December 31, 1996, data were collected by 52 interventionists on 15,331 consecutive hospital admissions for a PCI. The capture rate of procedures was assessed by comparing forms submitted to the cardiac catheterization logs at each medical center. The capture rate was 98.6%. The deaths were validated by comparing the status at discharge recorded on the data collection form to that reported on hospital discharge records supplied by the participating medical centers. The capture rate for in-hospital mortality is 100%.
Data were collected on the following variables: Demographic data: patient age, gender, height and weight; Medical history: previous CABG, PCI or MI, family history of premature coronary artery disease, congestive heart failure, hypertension, treated diabetes mellitus, current smoking, hypercholesterolemia, chronic obstructive pulmonary disease, peripheral vascular disease, cerebrovascular disease, renal failure and baseline creatinine; Primary indication for PCI: asymptomatic coronary artery disease, stable angina, unstable angina (including new-onset angina, rest angina, angina of increasing frequency or intensity and angina lasting ≥20 min irrespective of medication, excluding angina occurring within two weeks of a MI, postinfarction angina, postinfarction anatomy, primary therapy for acute MI and cardiogenic shock [systolic blood pressure <80 mm Hg requiring treatment with pressors or inotropes]); Priority at PCI: emergent, urgent and nonurgent, as classified by the cardiologist—emergent: factors dictate that PCI be performed immediately to avoid unnecessary morbidity or death; urgent: medical factors require that the patient stay in the hospital until PCI is performed; nonurgent: factors indicate that the patient could be discharged to return electively for PCI; Therapy before, during and after the procedure: intravenous heparin, intravenous nitroglycerin, thrombolytic therapy (e.g., tissue plasminogen activator, streptokinase, urokinase, anisoylated plasminogen streptokinase activator complex [APSAC]) and insertion of an intraaortic balloon pump (IABP). Cardiac anatomy and function: cardiac catheterizations were performed at the participating or referring institutions using their own standard methods during the course of regular clinical care. Data collected included percent stenosis of the left main coronary artery, number of other diseased (>70% stenosis) native vessels, dominance, number of bypass grafts (distal anastomoses), ejection fraction (EF) and left ventricular end-diastolic pressure (LVEDP); PCI procedure information: location of lesions attempted (using the Coronary Artery Surgery Study map) (20); category of lesion (never before angioplastied, restenosis of lesion angioplastied on previous admission, acute reocclusion, second or subsequent attempt to dilate a lesion); lesion status pre- and postprocedure stenosis using visual estimates or calipers, whatever the local standard; lesion type (21)(A = discrete, <10 mm in length, concentric, readily accessible, <45° bend, smooth contour, little or no calcification, not totally occluded, not ostial, no branch involvement, no thrombus seen; B1 (one of the following) or B2 (≥2 of the following) = tubular, 10 to 20 mm in length, eccentric, moderate tortuosity of proximal segment, 45° to 90° bend, irregular contour, moderate to heavy calcification, total occlusions <3 months in duration, ostial location, bifurcation lesions requiring double guide wires, some thrombus present; C = diffuse, >2 cm in length, excessive proximal tortuosity, extreme angulation [>90°], unable to protect major side branches, total occlusion >3 months in duration, degenerated vein grafts with friable lesions]); collateral vessels (A = feeding distal target vessel beyond stenosis; B = arising from target vessel beyond stenosis; C = no collateral vessels); dissection (flow limiting, spiral dissection of ≥2 vessel diameters); device (balloon, atherectomy, translumen extraction, Rotoblator, laser, stent); Outcome: in-hospital mortality. The number of patients in the data set and their status at hospital discharge were verified using cardiac catheterization laboratory logs and hospital discharge data. Full definitions for all variables and the data collection forms can be obtained at http://www.healthimprov.org/data_forms.htm.
Standard statistical methods were used for the calculation of the odds ratios (OR) with 95% confidence intervals (CI) and p values (22). Logistic regression analysis was performed with Stata (23,24)to assess the relation between patient, disease and treatment characteristics and in-hospital mortality. Among the data elements collected, only EF and creatinine level were missing for a substantial number of patients. Ejection fraction was not available for 49.4% of patients. Missing values were imputed, from age, gender, acuity (emergency, urgent, nonurgent) and previous revascularization, using standard statistical methods (25). Creatinine data were missing for 22.7% of patients, and missing values were coded as <2.0 mg/dl. The likelihood ratio chi-square test (χ2LR) was used to compare nested logistic regression models (26).
The area under the receiver operating characteristic (ROC) curve was used as a measure of model discrimination (27,28). The statistical model was internally validated using the technique of bootstrap resampling (29). This technique is efficient and provides nearly unbiased estimates of the predictive accuracy of the prediction model (30). We developed the multivariate model on the entire data set, then, 100 samples of 70% were drawn at random with replacement. The ROC curve area was calculated for each sample. This allowed the calculation of the standard deviation of the mean ROC area, and from this the standard error of the mean (SD/n0.5) (27). In addition, a series of stratum-specific estimates of the ROC area were calculated to assess the discriminatory ability of the prediction model for specific patient subgroups. Resampling methods (using 100 samples) were used to calculate the standard errors of these estimates. The calibration of the prediction equation was assessed by comparing the observed and expected numbers of deaths by decile of predicted risk. The Hosmer-Lemeshow goodness-of-fit statistic was calculated (31).
During the study period, data were collected on 15,331 consecutive patients who underwent PCI. There were 165 in-hospital deaths (1.08%). Most of the patients (67.8%) undergoing PCIs were men. The mean age was 61.4 years (Table 1). Cardiac risk factors were common. Many patients had a previous revascularization (14.2% CABG, 29% PCI). The most common comorbidity was diabetes (22%), followed by peripheral vascular disease, cerebrovascular disease and chronic obstructive pulmonary disease (COPD). The most frequent indication for the procedure was unstable angina (73.3%). With regard to priority of coronary angioplasty, 8.8% of the procedures were classified as emergent; 55.7% were classified as urgent; and 35.6% were nonurgent. Multivessel coronary artery disease was present in 34.5% of patients (Table 2). Few patients had a depressed EF (9.3% with <40%) or elevated LVEDP (14.8% with >22 mm Hg). Before the procedure, 46.3% of patients were treated with intravenous nitroglycerin and 0.4% received an IABP. Multivessel PCI took place in 7.9% of patients. Of the patients, 17.5% had a PCI in the proximal left anterior descending coronary artery and 5.7% had a graft angioplasty. Type B2 or C lesions were attempted in 27% of patients. The rates of stent use by year were 1.8% in 1994, 15.4% in 1995, and 44.9% in 1996—with an overall rate of 21.7%. The in-hospital mortality rates by year were 1.2% in 1994, 1.1% in 1995 and 1.0% in 1996. Treatment with IIb/IIIa platelet inhibitors (0.63%), directional coronary atherectomy (6.6%) and Rotablator (3.8%) was infrequent.
Risk factors for in-hospital mortality
Univariate assessment of the associations between patient demographic variables, past medical history, patient status and the risk of dying was conducted by calculating ORs and p values (Table 1). Older age was significantly (p trend <0.001) and positively associated with the risk of a fatal adverse outcome. Female gender was significantly associated with a fatal (p < 0.019) adverse outcome. The in-hospital mortality rate for men was 0.96%, whereas that for women was 1.38%. Previous CABG (p < 0.001) was associated with increased risk of in-hospital mortality, but this was not true for a previous PCI. Diabetes (p = 0.007), chronic obstructive pulmonary disease (p < 0.001), peripheral vascular disease (p < 0.001), cerebrovascular disease (p < 0.001) and body surface area (p trend <0.001) were all associated with an increased risk of a postprocedure death. A creatinine level ≥2 mg/dl (OR 6.37, p < 0.001) and a history of congestive heart failure (OR 8.63, p < 0.001) were associated with an especially high risk for in-hospital mortality. One surprising finding in these univariate analyses was that treated hypercholesterolemia was associated with a decreased risk of in-hospital mortality. This unintuitive association may identify patients who are under greater medical surveillance, or it may be a random association. More severe indications for PCI and increased priority of the procedure were also strongly associated (p trend <0.001) with the risk of in-hospital mortality. As expected, those patients requiring an emergency procedure or being treated for an acute MI or cardiogenic shock had the highest rates of death. Although they are treatment variables, the preprocedure use of thrombolytic agents, intravenous nitroglycerin and an IABP may be important proxies for severity of illness and is information available to interventionists before the procedure. All three were significantly (p < 0.001) associated with an increased risk of a fatal adverse outcome.
Angiographic and hemodynamic data are summarized in Table 2. The number of diseased coronary arteries was associated with the risk of a fatal (p trend <0.001) adverse outcome, and those patients with >50% stenosis of the left main coronary artery were 4.26 (p < 0.001) times as likely as those without this lesion to die after the procedure. Both a decreased EF (p trend <0.001) and an increased LVEDP (p trend <0.001) were associated with an increased risk of in-hospital mortality. The risk of in-hospital mortality was not significantly associated with a PCI of the proximal left anterior descending coronary artery (p = 0.504). A PCI in a graft was associated with an increased risk of in-hospital mortality (p < 0.001). Lesion types B1, B2 and C were also associated with increasing in-hospital death (p trend <0.001).
Development and validation of the multivariate prediction rule
All variables that were associated (p < 0.05) with in-hospital mortality in the univariate analyses were included in the initial multivariate prediction rule. Variables not significantly associated with this outcome in the multivariate analysis were systematically dropped. The multivariate model predicting in-hospital mortality was statistically significant (model χ2LR [16 df] = 466.34, p < 0.0001). Variables associated with an increased risk of in-hospital mortality included older age, congestive heart failure, any vascular disease (peripheral or cerebrovascular), increased creatinine level, lower EF, treatment of cardiogenic shock or an MI, urgent or emergency priority, preprocedure insertion of an IABP and PCI of a type C lesion. The multivariate model was tested against the “full” model, which contained all of the significant univariate predictors that had been dropped (χ2LR [20 df] = 26.95, p = 0.1365). The nonsignificant likelihood ratio chi-squared test indicates that these variables did not substantially improve the multivariate prediction model. To examine the ability of the multivariate model to discriminate the ROC curves of 100, 70% samples of the data (with replacement) were calculated. The area under the “average” curve was 0.88, with a standard deviation of 0.019, indicating a good ability to discriminate between patients who died in the hospital and those who did not.
The predicted risks of individual patients were rank-ordered and divided into deciles. Within each decile of estimated risk, the number of deaths predicted was plotted against the actual number of observed deaths (Fig. 1). An identity line (i.e., line of perfect correlation) is shown on the plot. The Hosmer-Lemeshow goodness-of-fit statistic across deciles of risk was not statistically significant (χ2[8 df] = 13.8, p = 0.087), indicating little departure from a perfect fit. The variables used in the multivariate prediction rule, their regression coefficients and the p values associated with each variable are summarized in Table 3. An example of the calculation of predicted risk for an individual patient is also described.
To study the performance of the prediction rule on patient subgroups, the data were stratified according to age (<70 vs. ≥70 years), gender, indications for procedure (stable angina, unstable angina, treatment for MI or cardio-genic shock), acuity (nonurgent vs. urgent vs. emergent), EF (<40%; 40–49%, 50–59%, ≥60%), number of coronary arteries with a lesion ≥70% (1, 2 or 3), worst lesion type (A, B1, B2 or C) and year of procedure. The discriminatory ability of the prediction equation, as measured by the area under the ROC curve, is calculated for each strata (Table 4). They are remarkably consistent. In 21 of 25 strata, the ROC areas were ≥0.80. The equation predicts only slightly less well for the 4.4% of patients undergoing PCI for the treatment of an acute MI (ROC area = 0.77). A lower value was seen among the 0.6% of patients being treated for cardiogenic shock (ROC area = 0.59).
In this contemporaneous, prospective cohort study of in-hospital mortality after PCIs in northern New England in 1994 to 1996, data were collected from 52 interventional cardiologists on 15,331 consecutive hospital admissions for PCI, representing over 98.5% of all patients who underwent a PCI during the study period. These data were used to develop and internally validate a multivariate prediction equation for in-hospital mortality that required only routinely collected data known before the PCI. The discriminatory characteristics of the prediction equation and its calibration were good. This prediction equation may be useful for patient counseling and to assess outcomes of PCI.
Choice of variables
In developing this prediction equation, our choice of variables was guided by face validity and by studies done by other investigators. With respect to face validity, it must be biologically plausible to believe that there would be an association between the specific patient or disease characteristic and the outcome of interest. Another valuable source of insight was obtained from studies done by other investigators. Our initial approach to data collection for the Percutaneous Transluminal Coronary Angioplasty (PTCA) Registry was influenced by the work of Detre et al. (32). In 1991, we developed our first PTCA mortality prediction equation (33). We were participants in the working group of PTCA risk factors convened by Block et al. (34)and were guided by this experience, and we thought that it would be worthwhile to update the prediction equation using contemporaneous data.
In the internal validation of the prediction equation, our approach was to use bootstrap resampling to estimate the standard deviation of the ROC area. This has been shown to be more efficient than split-sample cross validation (35). In addition, we looked at strata of patient groups to see whether the equation discriminates similarly for each. The prediction equation had good performance characteristics overall and within groups.
The regional nature of the data reflects wide experience within this geographic area during a specific period. However, our data set of 15,331 patients had only 165 deaths (1.08%) and may not be sufficient for an accurate prediction in some of the smaller patient subgroups. As the practice of PCI changes, increases in the use of stents and novel antiplatelet agents and further reductions of in-hospital mortality may occur, which could change those risk factors of predictive death. However, in northern New England, there has been very little change over time in the risk of in-hospital mortality after PCI (1.01% in 1990–1993 vs. 1.08% in 1994–1996). Furthermore, this report only considers mortality after PCIs. Mortality rates are required by the Joint Commission on Accreditation of Healthcare Organizations (JCAHO) clinical indicator report, are tracked by hospitals and health care systems and will certainly be reported in the American College of Cardiology National Registry. Mortality rates are important but certainly are not sufficient for the full evaluation of PCI. Nonfatal adverse outcomes are also important, as are patient functional status, patient satisfaction and resource use.
Comparisons with other studies
There have been a number of multivariate analyses of risk factors associated with in-hospital adverse outcomes after PCI. Three studies used data from large registries. Using 1992 data from the Registry of the Society for Cardiac Angiography and Interventions, Kimmel et al. (36)developed a six-variable prediction rule estimating the risk of major complications (death, MI, emergent CABG). The variables included multivessel disease, unstable angina, recent MI, type C lesion or angioplasty of the left main coronary artery, shock, age and aortic valve disease. The prediction rule showed good discriminating ability (ROC area = 0.71) and was tested on independently collected data (ROC area = 0.65). In a series of 5,827 patients undergoing PCI in New York State in 1991, Hannan et al. (37)analyzed risk factors for fatal outcomes. In multivariate analyses, they found that female gender, hemodynamic instability, shock and low EF predicted in-hospital mortality. The logistic regression model had good discriminating ability (ROC area = 0.88); no validation was reported. Using the 1985–1986 National Heart, Lung, and Blood Institute’s Coronary Angioplasty Registry, Kelsey et al. (38)found that female gender, age, congestive heart failure, diabetes and multivessel disease were associated with post-PCI in-hospital mortality. This model was not independently validated.
Other studies have used data from single institutions. Two of these validated their multivariate models. Emphasizing long-term outcome, Mick et al. (39)analyzed data from 5,000 consecutive patients seen at the Cleveland Clinic between 1980 and 1988. Patients undergoing PCI for acute ischemic events were excluded. In multivariate analyses using Cox proportional hazards modeling, they found that age >60 years, extent of coronary artery disease, Canadian Cardiovascular Society functional class, previous PCI, male gender, history of diabetes mellitus, history of hypertension and history of congestive heart failure were significant predictors of event-free survival. This prediction model was developed on 4,000 patients and validated on 1,000 other patients. The ROC area for the model showed relatively good predictive accuracy (ROC = 0.74). From a 1986–1989 case series at Duke University (658 patients), Tenaglia et al. (40)developed a scoring system for predicting abrupt vessel closure associated with PCI from lesion characteristics. Variables included branch location, length of the lesion, presence of thrombus and lesions of the right coronary artery. Bootstrapping was used to provide internal validation of the multivariate model.
The current study is generally in agreement with these reports and with our previous prediction rule for 1990–1993 (17). Still included in the multivariate model are age, procedural priority and indication for and preprocedure use of an IABP. However, gender, history of MI, use of preprocedure intravenous nitroglycerin, LVEDP, number of diseased coronary arteries and an intervention on a proximal left anterior descending coronary artery are no longer multivariate predictors of death. Some of these risk factors are markers for patients who have little cardiac reserve or will develop a large amount of ischemia with an acute complication. It is possible that with the availability of bailout devices like perfusion catheters and stents, some of these risk factors are no longer so important. New risk factors in the model include congestive heart failure, peripheral or cerebrovascular disease, elevated creatinine level, EF and intervention on a type C lesion. The first three risk factors were not previously collected, and there has been improved collection of EF. It is likely that even in the era of stents, type C lesions are difficult to dilate and do not leave much room for technical error.
In this contemporary, regional, prospective study, a number of strong risk factors for in-hospital mortality after PCI were identified and a multivariate prediction equation was developed and internally validated. This equation discriminates well across a broad range of patient risk factors. Using either a programmable calculator or a microcomputer, this equation is easily solved. We conclude that the calculation of context-specific estimate of risk using routinely available preprocedure data is feasible and may be useful for patient counseling and for quality assurance purposes in the modern era of interventional cardiology. External validation remains to be accomplished by using the prediction equation on data other than that from which it was derived. This constitutes a more rigorous test of the multivariate prediction model and should be done both regionally and in other settings.
We thank the data coordinators at each of the participating institutions for their many efforts to ensure that data collection was complete and accurate. We also thank those individuals in Medical Informatics at each institution for their help with the validation study.
Northern New England cardiovascular disease study group
Gerald T. O’Connor, PhD, DSc2, David J. Malenka, MD2, Hebe Quinton, MS2, John F. Robb, MD2, Mirle A. Kellett, Jr., MD5, Samuel Shubrooks, MD1, William A. Bradley, MD7, Michael J. Hearne, MD7, Mathew W. Watkins4, David E. Wennberg, MD6, Bruce Hettleman, MD2, Daniel J. O’Rourke, MD, MS2, Paul D. McGrath6, Thomas Ryan, Jr., MD5, Peter VerLee, MD3
Beth Israel–Deaconess Medical Center, Boston, Massachusetts1
David Brackett, RN, Mary Bogosian, RN, CCP, Christian Campos, MD, Jeannie Fischer, PA, Philip J. Fitzpatrick, MD, Beth Jennings, Robert Johnson, MD, Wendy Kowalker, Patricia Lahey, RN, Stephen J. Lahey, MD, David Leeman, MD, Keith P. Lewis, MD, Stanley Lewis, MD, Maria Lustenberger, RN, Peter R. Maggs, MD, Richard Nesto, MD, Brian O’Connor, CCP, Patty Pawlow, RN, Kathy Peterson, RN, Patricia Rabett, RN, Samuel Shubrooks, MD, Cheryl Sirois, RN, Terri Stokes, RN, MS, Susan Sumner, RN, Paul G. Vivino, MD, Albert Washko, MD, Ronald Weintraub, MD
Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire2
Virginia Beggs, ARNP, MS, John D. Birkmeyer, MD, Nancy J. O. Birkmeyer, PhD, William Burke, RCVT, Edward Catherwood, MD, MS, Mike Chamberlain, RN, Lindsay D’Anna, PA, Lawrence J. Dacey, MD, MS, Gordon Defoe, CCP, Kenneth Dixon-Vestal, RN, Thomas Dodds, MD, Mary Fillinger, MD, Bruce Friedman, MD, Christine Heins, RN, Bruce Hettleman, MD, Douglas James, MD, John E. Jayne, MD, Karen A. Jean, RN, Pamela Jenkins, MD, PhD, Joseph Kasper, ScD, Lori Key, RN, Terry Kneeland, MPH, Judith Kobe, RN, Elizabeth Maislen, ARNP, David J. Malenka, MD, Charles A. S. Marrin, MB, BS, Mary Menduni, RN, Nathaniel Niles, MD, William C. Nugent, MD, Gerald T. O’Connor, PhD, DSc, Elaine M. Olmstead, Daniel O’Rourke, MD, Winthrop Piper, MD, Stephen K. Plume, MD, Hebe B. Quinton, MS, John Robb, MD, Cathy S. Ross, MS, John Sanders, MD, William Schults, William F. Sullivan, MS, Jon Wahrenberger, MD, Beth Wolf
Eastern Maine Medical Center, Bangor, Maine3
Robert Allen, MD, Jim Blum, MS, Deborah Carey-Johnson, RN, MS, Chae C. Choi, MD, Tina Closson, RN, Robert Clough, MD, Donna Dauphinee, Cynthia M. Downs, RN, MSN, Glen D. Garson, MD, Rebecca Henry, RN, Felix Hernandez, Jr., MD, Joseph J. Hessel, MD, Robert M. Hoffman, MD, John H. Jentzer, MD, Edward R. Johnson, MD, Peter Marshall, MD, Helen McKinnon, RN, Cathy Mingo, RN, MS, Craig Pedersen, PA, Wendy Perkins, LPN, Robert Rosenthal, MD, Matthew W. Rowe, MD, Katrina Sargent, M. Theodore Silver, MD, Sherry Sprague, Wolfgang J. T. Spyra, MD, Laurie True, RN, Peter Ver Lee, MD, Paul vom Eigen, MD, Craig Warren, CCP
Fletcher-Allen Health Care, Burlington, Vermont4
Richard G. Brandenburg, PhD, Pamela Brown, Betsy Burns, RN, Mark Capeless, MD, Kevin Carey, MD, Steve Colmanaro, PA, Steve Crumb, RN, CS, Betty Diette, RN, Roy V. Ditchey, MD, Maureen Dwyer, ARNP, Karen Farrell, ANP, Jan Faucett, RN, Sally Gagnon, RN, Susan Geoffrey, RN, Larry Goetschius, Laurie Grenier, Walter D. Gundel, MD, Richard S. Jackson, MD, David Johnson, MD, Charlie Krumholz, CCP, Ann Laramee, RN, Bruce J. Leavitt, MD, Martin Lewinter, MD, Steve Marcus, PA, Karen McKenny, RN, Mitchell Norotsky, MD, Madeline Norse, RN, William C. Paganelli, MD, PhD, Diane Pappalardo, MHSA, Daniel S. Raabe, MD, Melinda Rabideau, RN, Martha Root, RN, Janice Smith, RN, Christopher M. Terrien, Jr., MD, Edward Terrien, MD, Matthew W. Watkins, MD, Jane Wilde, RN, MSN, William Witmer, MD
Maine Medical Assessment Foundation5
Robert B. Keller, MD, David Wennberg, MD, MPH
Maine Medical Center, Portland, Maine6
Lawrence Adrian, PA, Warren D. Alpern, MD, Eric Anderson, Richard A. Anderson, MD, Linda Banister, RN, Claire Berg, RN, Seth Blank, MD, John Braxton, MD, Carl E. Bredenberg, MD, Michael Brennan, PA, David Burkey, MD, Cantwell Clark, MD, Jane Cleaves, RN, Vincent Conti, CEO, Deborah Courtney, RN, MS, Joshua Cutler, MD, Desmond Donegan, MD, Pat Fallo, RN, Rick Forest, CCP, Robert Groom, CCP, Daniel Hanley, MD, Mary Beth Hourihan, MD, Jane Kane, RN, Saul Katz, MD, Mirle A. Kellett, Jr., MD, Robert Kramer, MD, Costas T. Lambrew, MD, F. Stephen Larned, MD, Lee Lucas, Paul D. McGrath, MD, Jeremy R. Morton, MD, Edward R. Nowicki, MD, John R. O’Meara, MD, Sheilia Parker, RN, Patricia Peasley, RN, Cathy Prouty, RN, Reed D. Quinn, MD, Dennis Redfield, Karen Reynolds, MPH, Thomas Ryan, Jr., MD, Jean Saunders, MSN, MPH, Alyce Schultz, RN, PhD, Susan Seekins, RN, Paul W. Sweeney, MD, Karen Tolan, RN, Nancy Tooker, RN, Joan F. Tryzelaar, MD, Paul T. Vaitkus, MD, Kathy Viger, RN, Cynthia Westlund, RN, Wanda Whittet, RN
Catholic Medical Center, Manchester, New Hampshire7
Yvon Baribeau, MD, Ann Becker, RN, Craig C. Berry, MD, Kevin Berry, MD, William A. Bradley, MD, David C. Charlesworth, MD, Susan Cuddy, RN, Robert C. Dewey, MD, Frank Fedele, MD, Louis I. Fink, MD, Erik J. Funk, MD, Alan E. Garstka, MD, Karen Grafton, RN, Dan Halstead, CCP, Michael J. Hearne, MD, J. Beatty Hunter, MD, Dennis Kelly, MD, Mark A. Klinker, MD, Peggy Lambert, RN, Patrick J. Lawrence, MD, Jeffery Lockhart, MD, Christopher T. Maloney, MD, Kathy McNeil, RN, Venkatram Nethala, MD, John Pieroni, CCP, M. Judith Porelle, RN, Joanne Robichaud, RN, Mary Sanford, RN, James Schmitz, MD, Benjamin M. Westbrook, MD, Thomas P. Wharton, MD, Kirke W. Wheeler, MD, Diane White, RN
↵‡‡ The full membership is listed in the .
☆ Funding for this study was provided by the participating institutions of the Northern New England Cardiovascular Disease Study Group.
- coronary artery bypass graft surgery
- confidence interval
- chronic obstructive pulmonary disease
- ejection fraction
- intraaortic balloon pump
- left ventricular end-diastolic pressure
- myocardial infarction
- odds ratio
- percutaneous coronary intervention
- percutaneous transluminal coronary angioplasty
- peripheral vascular disease
- receiver operating characteristic
- likelihood ratio chi-square test
- Received August 28, 1998.
- Revision received April 7, 1999.
- Accepted May 16, 1999.
- American College of Cardiology
- Jollis J.G.,
- Peterson E.D.,
- Nelson C.I.,
- et al.
- Hilborne L.H.,
- Leape L.L.,
- Kahan J.P.,
- Park R.E.,
- Kamberg C.J.,
- Brook R.H.
- Hamm C.W.,
- Reimers J.,
- Ischinger T.,
- Rupprecht H.J.,
- Berger J.,
- Bleifeld W.
- Rodriguez A.,
- Mele E.,
- Peyregne E.,
- et al.
- Potkin B.N.,
- Keren G.,
- Mintz G.S.,
- et al.
- Northern New England Cardiovascular Disease Study Group,
- O’Connor G.T.,
- Malenka D.J.,
- Levy D.G.,
- Disch D.L.,
- Quinton H.B.
- Principal Investigators of CASS and Their Associates
- Ryan T.J.,
- Bauman W.B.,
- Kennedy J.W.,
- et al.
- Hennekens C.H.,
- Buring J.E.
- Cornfield J.,
- Gordon T.,
- Smith W.W.
- (1997) STATA Statistical Software. Release 5.0 (Stata Corp, College Station, Texas).
- Kleinbaum D.G.,
- Kupper L.L.,
- Morgenstern H.
- Swets J.A.
- Hanley J.A.,
- McNeil B.J.
- Hosmer D.W.,
- Lemeshow S.
- Northern New England Cardiovascular Disease Study Group,
- O’Connor G.T.,
- Malenka D.J.,
- Levy D.G.,
- Disch D.L.,
- Quinton H.B.
- Block P.C.,
- Peterson E.C.,
- Krone R.,
- et al.
- Kimmel S.E.,
- Berlin J.A.,
- Strom B.J.,
- Laskey W.K.
- Kelsey S.F.,
- James M.,
- Holubkov A.L.,
- et al.
- Mick M.J.,
- Piedmonte M.A.,
- Arnold A.M.,
- Simpendorfer C.
- Tenaglia A.N.,
- Fortin D.F.,
- Califf R.M.,
- et al.