Author + information
- Received November 3, 1998
- Revision received August 16, 1999
- Accepted September 21, 1999
- Published online February 1, 2000.
- William J Rogers, MD, FACC∗,* (, )
- John G Canto, MD, MSPH, FACC∗,
- Hal V Barron, MD, FACC†,‡,
- Joseph A Boscarino, PhD, MPH§,
- David A Shoultz, PhD∥,¶,
- Nathan R Every, MD, MPH, FACC∥,
- for the Investigators in the National Registry of Myocardial Infarction 2#
- ↵*Reprint requests and correspondence: Dr. William J. Rogers, 334 LHR Building, UAB Medical Center, Birmingham, Alabama 35294
We sought to determine the extent to which the capability of a hospital to perform invasive cardiovascular procedures influences treatment and outcome of patients admitted with acute myocardial infarction (AMI).
Patients with AMI are usually transported to the closest hospital. However, relatively few hospitals have the capability for immediate coronary arteriography, percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass graft surgery (CABG), should these interventions be needed.
The 1,506 hospitals participating in the National Registry of Myocardial Infarction 2 were classified according to their highest level of invasive capability: 1) none (noninvasive, 28.1%); 2) coronary arteriography (cath-capable, 25.2%); 3) coronary angioplasty (PTCA-capable, 7.4%); and 4) bypass surgery (CABG-capable, 39.2%). Treatment and in-hospital outcomes were assessed for 305,812 patients admitted from June 1994 through October 1996. Follow-up through 90 days was ascertained in a subset of 30,402 patients enrolled simultaneously in both the National Registry of Myocardial Infarction (NRMI) 2 and the Cooperative Cardiovascular Project (CCP).
The proportion of patients receiving initial reperfusion intervention was only slightly higher at the more invasive hospitals (noninvasive 32.5%, cath-capable 31.2%, PTCA-capable 32.9% and CABG-capable 35.9%, p < 0.001 by chi-square statistic). Among thrombolytic recipients, median door-to-drug time interval differed little among hospital types and ranged from 42 to 45 minutes. At cath-capable, PTCA-capable and CABG-capable hospitals, coronary arteriography was performed in 32.9%, 37.4% and 64.9%, respectively, and PTCA in 0.0%, 5.1% and 31.4%, both p < 0.001 by chi-square statistic. The proportion of patients transferred out to other facilities was 51.0%, 42.2%, 39.9% and 4.4% (p < 0.0001) among noninvasive, cath-capable, PTCA-capable and CABG-capable hospitals, respectively. Among patients in the combined NRMI and CCP data set, mortality at 90 days postinfarction was similar among patients initially admitted to each of the four hospital types.
Although patients with AMI admitted to hospitals without invasive cardiac facilities have a high likelihood of subsequent transfer to other facilities, their likelihood of receiving a reperfusion intervention at the first hospital, their door to thrombolytic drug intervals and their 90-day survival rates are similar to those of patients initially admitted to more invasively equipped hospitals. These data suggest that a policy of initial treatment of myocardial infarction at the closest medical facility is appropriate medical practice.
Over the past two decades, a spectrum of interventions has evolved for the treatment of acute myocardial infarction (AMI), ranging from simple coronary care unit observation with thrombolytic therapy for eligible patients, to more complex interventions like immediate cardiac catheterization followed by percutaneous or surgical coronary revascu
larization. Although all acute care hospitals should be able to provide immediate thrombolytic therapy, less than 20% are equipped to perform immediate coronary arteriography, and even fewer have capability for immediate percutaneous transluminal coronary angioplasty (PTCA) or coronary artery bypass graft surgery (CABG), if needed (1). If the outcome of myocardial infarction (MI) were superior in hospitals with invasive facilities, there might be rationale for the regionalization of acute infarct care, similar to that currently employed for victims of severe trauma. This would entail the immediate transport of all patients with AMI to regional care centers, rather than to the closest hospital, as is currently done.
The National Registry of Myocardial Infarction (NRMI) is an ongoing U.S. registry of patients with AMI and involves hospitals of all levels of invasive capability for cardiac care. The purpose of this investigation was to determine whether, among patients enrolled in NRMI, outcome is significantly influenced by the ability of a hospital to perform invasive cardiovascular diagnostic testing and interventions on patients admitted with AMI.
The NRMI is a voluntary nationwide U.S. registry whose purpose is to describe the characteristics of patients admitted with AMI at participating hospitals, the treatment interventions utilized and the in-hospital outcomes. The second phase of the registry, NRMI 2, began in 1994 and involved 1,506 hospitals, or 26% of all acute-care hospitals in the U.S. (2). The NRMI methodology has been previously described in detail (3). Briefly, upon joining the registry, participating hospitals are required to complete an enrollment form characterizing their local facilities and services. Data from consecutive patients with AMI are then recorded onto two-page case report forms and forwarded to a central data collection and tabulation center, STATPROBE, Inc. (Lexington, Kentucky). Confidential quarterly data reports are provided to participating hospitals, summarizing the hospital’s local registry data in parallel with data from its state, the nation and data from hospitals of like size and invasive capabilities. Local registry data may then be utilized to assess local quality of care of patients with AMI, and national data summaries can be used to assess registry-wide practice patterns. To ensure quality control of registry data, local registry coordinators are trained to complete the data collection forms utilizing a standardized manual of instructions and definitions. Case report forms must pass systematic range and internal consistency checks at the central data collection center before being entered into the national database. Participating hospitals obtain approval for the registry data collection process as dictated by local policy. All patients enrolled in NRMI 2 from June 1994 through October 1996 were included in this study, except for those transferred into the registry hospitals from other institutions.
Because survival ascertainment in NRMI is limited to the initial hospitalization, additional data on survival to 90 days postentry were obtained in a subset of patients enrolled in both the NRMI 2 and the national Cooperative Cardiovascular Project (CCP), sponsored by the Health Care Financing Administration (HCFA).
Registry hospitals were stratified into four mutually exclusive categories based on data furnished by the hospitals when they joined the registry:
1. noninvasive: no capability to perform coronary arteriography (cath), PTCA or CABG;
2. cath-capable: coronary arteriography could be performed, but not PTCA or CABG;
3. PTCA-capable: coronary arteriography and PTCA could be done, but not CABG;
4. CABG-capable: coronary arteriography, PTCA and CABG could be performed.
The study protocol defined AMI as 1) history and presentation suggestive of MI accompanied by either total creatine kinase or creatine kinase-MB ≥ twice upper limit of normal for the hospital’s laboratory or ECG evidence of MI, or, in the absence of the above, alternative enzymatic, scintigraphic, echocardiographic or autopsy evidence of MI, or 2) an International Classification of Diseases, 9th revision, clinical modification, discharge diagnosis code for MI, 410.01 through 410.91.
Hospitals were classified as “urban” if they were located in a county with at least one city having a population of >50,000 or two cities with a combined population of >50,000. Teaching hospitals were those that were members of the Council of Teaching Hospitals. States were grouped into the U.S. census regions as previously described (4). Patients were classified as “not low risk” if there was evidence of ST-segment elevation on their first 12-lead ECG and one or more of the following: age ≥70 years, history of prior MI, initial systolic blood pressure <100 mm Hg and pulse >100 beats/min, evidence of heart failure upon admission or anterior/septal infarct location. Treatment intervals for administration of thrombolytic therapy were defined according to the National Heart Attack Alert Program as follows: “door” indicates the time of patient presentation to the hospital; “data,” the time of acquisition of the diagnostic ECG; “decision,” the time it was decided to administer thrombolytic therapy; and “drug,” the time thrombolytic therapy was initiated (5). Major bleeding included hemorrhagic episodes, other than intracranial bleeding, resulting in substantial hemodynamic compromise.
In comparing hospital categories, basic proportions were compared using the chi-square test (6). Means of continuous variables were compared using one-way analysis of variance (7), and medians were compared using a nonparametric test of medians available in SAS 6.12 (NPAR1WAY procedure) (8). Multivariate logistic regression models were fit to identify independent predictors of hospital mortality (9). These models were fit using a forward stepwise procedure where the indicator of hospital type was forced into the model at the first step. In order to evaluate the potential effect of hospital type and other important covariates on 90-day survival among patients in the NRMI-2 database, a subset of the data was linked to the HCFA-sponsored CCP database. This subset was comprised of 30,402 patients whose data were matched using an iterative algorithm that has been validated and described in detail elsewhere (10). Following the linkage of these two related data sets, Cox regression models were used to estimate the effect of hospital type and other covariates of interest on 90-day survival. Hazard ratios were derived for each variable in these models, and 95% confidence limits were fit around the point estimates of the hazard ratios. Analyses were performed using a commercially available statistical package (SAS 6.12, SAS Institute, Cary, North Carolina). This report is based on data processed by the central data collection center as of October 31, 1996.
Among the 1,506 participating registry hospitals, 423 (28.1%) were classified as noninvasive, 380 (25.2%) as cath-capable, 112 (7.4%) as PTCA-capable and 591 (39.2%) as CABG-capable (Fig. 1). The numbers of staffed beds in the four hospital types were 133 ± 4, 215 ± 5, 211 ± 11 and 380 ± 8 (mean ± standard error of the mean), respectively (p < 0.001). The proportions with urban location were 58%, 75%, 84% and 95%, respectively (p < 0.001), and the proportions that were classified as teaching hospitals were 0.2%, 2.9%, 8.9% and 21.4%, respectively (all p < 0.001). In the South Atlantic census region, cath-capable hospitals were most prevalent (37%), and in the New England and Mid-Atlantic regions, noninvasive hospitals predominated (prevalence 45% and 47%, respectively). In the six remaining U.S. census regions, CABG-capable hospitals were the most common type with prevalence ranging from 40% to 68%.
Among the 305,812 patients satisfying the study criteria, 57,252 (18.7%) were enrolled at noninvasive hospitals, 76,956 (25.2%) at cath-capable hospitals, 24,251 (8.0%) at PTCA-capable hospitals and 147,153 (48.1%) at CABG-capable hospitals. Although minor differences in clinical characteristics were observed among patients enrolled at different types of hospitals (Table 1), none was of sufficient magnitude to appear clinically significant. Due to the large number of patients in each group, achievement of statistical significance did not necessarily imply clinical significance for the variables shown in Table 1or in the other tables.
Among the four hospital types, the proportion of patients receiving an initial reperfusion intervention varied only minimally—from 31.2% in cath-capable hospitals to 35.9% in CABG-capable hospitals (Table 2). When reperfusion therapy was employed, it was always intravenous lytic therapy in patients enrolled at noninvasive hospitals (Fig. 2)and predominantly intravenous lytic therapy at the other types of hospitals. At cath-capable hospitals 0.3% of reperfusion attempted was by intracoronary (IC) lytic therapy. At PTCA-capable hospitals, 10.2% of reperfusion attempted was by primary PTCA and 0.5% by IC lytic therapy. At CABG-capable hospitals, 34.2% of reperfusions attempted were by primary PTCA, 1.0% by IC lytic therapy and 3.0% by immediate CABG. Reperfusion attempted by primary PTCA was significantly more common at CABG-capable hospitals than at PTCA-capable hospitals (34.2% vs. 10.2%, p < 0.001).
The physician ordering lytic therapy, when it was employed, was much less likely to be a primary care physician and much more likely to be either an emergency department physician or a cardiologist at the three invasive categories of hospital (Table 2). The median thrombolytic door-to-drug time was 0.72 hours (43 min) and did not differ substantially among patients enrolled at the four different hospital categories. Among patients undergoing direct angioplasty, the median door-to-balloon time was 2.2 hours.
The likelihood of obtaining a consultation before initiating lytic therapy increased progressively from noninvasive to CABG-capable hospitals. During the first 24 h after admission, the utilization of IV heparin, aspirin and IV nitroglycerin was higher in the more invasive hospitals. During the entire hospitalization, coronary arteriography was performed far more commonly in the CABG-capable hospitals (64.9%) than either the PTCA-capable (37.4%) or cath-capable (32.9%) hospitals (p < 0.001). The use of pacemakers, mechanical ventilators and intraaortic balloon pumps was also higher in the CABG-capable hospitals, even among those patients not having CABG, in whom the procedural frequencies were 4.2%, 9.7% and 4.6%, respectively (all p < 0.001 vs. the less invasive hospitals).
Median length of stay in the coronary care unit was longest in the CABG-capable hospitals (3.0 vs. 2.0 days, p < 0.001) (Table 3). Total hospital stay was also longest in CABG-capable hospitals (6.1 days overall), partially explained by longer duration of hospitalization among the patients undergoing CABG (median 10.7 days vs. 5.6 days without CABG, p < 0.001) and also by the higher proportion of patients transferred out to other facilities among the other hospital types (noninvasive: 51.0%, cath-capable: 42.4%, PTCA-capable: 39.9% vs. 4.4% in CABG-capable, p < 0.0001). Occurring more frequently in the CABG-capable hospitals were stroke (1.6%) and major bleeding (3%), both of which were more common among the subset of patients having CABG (N = 21,272) (stroke 2.3% vs. 1.5% and major bleeding 5.3% vs. 2.6%). Hospital mortality was similar among noninvasive, cath-capable and PTCA-capable hospitals but was higher in CABG-capable hospitals (Table 3)and remained higher after adjustment for baseline covariates, including all the variables shown in Table 1and the U.S. census locations for each hospital (Table 4).
Among the 30,402 patients in the linked NRMI 2 and CCP data sets, the numbers enrolled at noninvasive, cath-capable, PTCA-capable and CABG-capable hospitals were 6,196; 8,089; 2,545 and 13,572, respectively. In-hospital mortality was 13.7%, 13.9%, 13.4% and 15.7%, respectively (p < 0.001), but mortality at 90 days was 23.5%, 22.9%, 22.4% and 22.2%, respectively (p = NS). Relative risk for 90-day mortality, adjusted for baseline covariates, was also similar among the four hospital types (Table 4). Independent predictors of 90-day mortality included age, history of heart failure, stroke, diabetes, no prior PTCA, admission blood pressure and pulse, evidence of heart failure upon admission, an ECG at discharge showing anterior infarct location or Q-waves and East South Central location.
This observational analysis from the NRMI 2 shows that a hospital’s ability to perform coronary arteriography, PTCA and CABG influences the initial management of patients presenting with AMI. More invasively equipped hospitals employ reperfusion therapy in a slightly higher proportion of patients, are more prone to use primary angioplasty rather than thrombolytic therapy and are much less apt to transfer patients out to other facilities. However, mortality at 90 days is similar among the four hospital types.
Distribution of hospitals
The large NRMI 2 hospital population contains a substantial representation of hospitals both with and without invasive capability (Fig. 1). However, in comparison with all U.S. hospitals, noninvasive hospitals are under-represented in NRMI 2 (28.1% vs. 67.4%) in all U.S. hospitals (2). Interestingly, in NRMI 2, noninvasive hospitals do predominate in New England, known for its conservative management of AMI (11).
In theory, hospitals with the ability to perform immediate coronary arteriography and mechanical revascularization might have considerable advantage over hospitals with lesser invasive capability in attempting reperfusion for AMI, because they could offer primary PTCA to patients with contraindications to thrombolytic therapy. Surprisingly, the NRMI 2 data show that the rates of attempting reperfusion are only marginally higher in the CABG-capable hospitals (35.9%) than in those with lesser invasive capabilities (range 31.2% to 32.9%). Intravenous thrombolytic therapy is the most commonly employed reperfusion modality in all categories of hospitals (Fig. 2). These data are consistent with the notion that primary PTCA is usually employed in patients who are also thrombolytic candidates, rather than in patients who have exclusions for thrombolytic therapy (12). Primary PTCA is used three to four times more commonly at hospitals with CABG-capability than at those without CABG-capability, likely reflecting adherence to previous guidelines recommending that backup CABG be present whenever acute PTCA is attempted (13).
Emergency department practices differed somewhat among hospital types. Interphysician consultation before administering thrombolytic therapy, a process known to increase door-to-drug time and hospital mortality (14), occurred in the majority of all hospital types with a prevalence ranging from 57% in the noninvasive hospitals to 65% in the CABG-capable hospitals. The time intervals from hospital presentation to administration of thrombolytic therapy were remarkably similar among the four hospital types. We anticipated that hospitals with PTCA capability might have longer thrombolytic treatment times due to emergency department indecision about whether primary PTCA or lytic therapy was the appropriate treatment option. However, we found that median door-to-drug time was actually shortest in the PTCA- and CABG-capable hospitals (42 min) but still in excess of the recommended goal of 30 min (5).
Other medications known to improve survival of AMI when administered within the first 24 h include aspirin, beta-adrenergic blocking agents and ACE inhibitors (15,16). Although there were minor differences in the use of these therapies among the hospital types, the magnitude of difference was probably not clinically important. More striking were differences in the use of invasive procedures during hospitalization. Coronary arteriography was employed only about half as commonly as cath-capable and PTCA-capable hospitals as it was at CABG-capable hospitals (33%, 37% and 65%, respectively). Furthermore, the frequency of performing PTCA during hospitalization was only 5.1% at PTCA-capable hospitals versus 31% at CABG-capable hospitals. These differences in the use of invasive procedures may relate to a preference for performing sequential or combined coronary arteriography and PTCA at hospitals with CABG backup.
We found that in-hospital mortality rates are similar among noninvasive, Cath-capable and PTCA-capable hospitals but higher in CABG-capable hospitals after adjustment for baseline covariates. However, adjusted 90-day mortality rates were similar in the subset of 30,402 NRMI 2 patients who had extended follow-up in the HCFA CCP database. The relatively higher in-hospital mortality rates among patients initially admitted to the CABG-capable hospitals was likely related to a combination of factors, including their longer duration of hospitalization, their lower likelihood of being transferred out and the small, but finite, risk of the invasive procedures being performed. The similar 90-day mortality among the four hospital types is consistent with findings from randomized trials comparing acute invasive with conservative management post-MI. These trials uniformly failed to show an advantage of the routine invasive approach (17–20).
Our data are also consistent with previous reports showing equally good outcome for patients with AMI cared for at community hospitals rather than at tertiary care hospitals, despite greater utilization of invasive procedures at the latter (21,22). Furthermore, the data support the current policy of initially treating patients with AMI at the closest medical facility, rather than routing them to regional centers with specialized facilities, as is the case for trauma victims. Patients with evidence of continuing ischemia or hemodynamic instability can be transferred from noninvasive to invasive facilities for specialized treatment, according to preestablished interhospital pathways.
Limitations of the NRMI have previously been presented (3). The Registry does not include all U.S. hospitals and is not a random sample of all U.S. hospitals. There is the potential for nonconsecutive patient enrollment, and there is no independent on-site verification of the accuracy of registry data. Conversely, over one-fourth of all U.S. acute care hospitals participate in NRMI 2, and it has grown to be the largest voluntary registry of patients with MI in the world. Hospitals collect registry data primarily to assess the quality of their local practice patterns for MI care, so there is great incentive to accurately record data on consecutive patients. A recent study has shown very high agreement between data collected independently by NRMI 2 coordinators and data collected independently by chart abstractors for the CCP in Medicare-eligible patients at hospitals participating in both studies (10). Finally, NRMI does not collect follow-up data following hospital discharge. Data presented on survival at 90 days for the subset of patients enrolled in both NRMI and the CCP may not be representative of all NRMI 2 patients since the CCP targeted an older population. Furthermore, 90-day survival data may not be sufficient to detect an advantage, if it exists, of invasive over conservative therapy.
Within the limitations of this observational analysis, patients with AMI would appear to receive equivalent immediate care at U.S. hospitals participating in NRMI 2, regardless of the invasive cardiac capability of the hospitals. Although hospitals with invasive capabilities have more options for initial reperfusion, the proportion of patients receiving an initial reperfusion intervention is only minimally greater than it is in hospitals without invasive capabilities. Mortality rates at 90 days follow-up, adjusted for baseline differences, are similar among patients admitted to the four types of hospitals. These data support the present policy of caring for patients with AMI at the closest hospital, rather than routinely and emergently transporting them to regional invasive revascularization centers.
☆ The National Registry of Myocardial Infarction is supported by Genentech, Inc., South San Francisco, California.
- angiotensin converting enzyme
- acute myocardial infarction
- coronary artery bypass graft surgery
- coronary arteriography
- Cooperative Cardiovascular Project
- Health Care Financing Administration
- myocardial infarction
- National Registry of Myocardial Infarction
- percutaneous transluminal coronary angioplasty
- Received November 3, 1998.
- Revision received August 16, 1999.
- Accepted September 21, 1999.
- American College of Cardiology
- Ryan T.J.,
- Anderson J.L.,
- Antman E.M.,
- et al.
- ↵The 1994 Hospital Market Database. Chicago, Illinois: SMG Marketing Group, Inc., 1995.
- Participants in the National Registry of Myocardial Infarction,
- Rogers W.J.,
- Bowlby L.J.,
- Chandra N.C.,
- et al.
- ↵U.S. Bureau of the Census. Statistical Abstract of the United States. Washington, D.C.: U.S. Department of Commerce Bureau of Census 1996.
- Fleiss J.L.
- Fisher L.D.,
- Van Belle G.
- ↵SAS/STAT Version 6.12 Edition. Cary, North Carolina: SAS, Inc., 1996.
- Breslow N.E.,
- Day N.E.
- Every N.R.,
- Frederick P.D.,
- Robinson M.,
- Sugarman J.,
- Bowlby L.,
- Barron H.V.
- Ryan T.J.,
- Bauman W.B.,
- Kennedy J.W.,
- et al.
- ↵Al-Mubarak N, Rogers WJ, Lambrew CT, Bowlby LJ, French WJ. Consultation before thrombolytic therapy in acute myocardial infarction. Second National Registry of Myocardial Infarction (NRMI) Investigators. Am J Cardiol 1998;83:89–93.
- ACE Inhibitor Myocardial Infarction Collaborative Group
- Thrombolysis and Angioplasty in Myocardial Infarction Study Group,
- Topol E.J.,
- Califf R.M.,
- George B.S.,
- et al.
- TIMI II-A Investigators,
- Rogers W.J.,
- Baim D.S.,
- Gore J.M.,
- et al.
- TIMI IIIB Investigators
- Veterans Affairs Non-Q-Wave Infarction Strategies in Hospital (VANQWISH) Trial Investigators,
- Boden W.E.,
- O’Rourke R.A.,
- Crawford M.H.,
- et al.
- TIMI Research Group,
- Feit F.,
- Mueller H.S.,
- Braunwald E.,
- et al.
- RESCATE Investigators,
- Marrugat J.,
- Sanz G.,
- Masia R.,
- et al.