Geographic Variation in the Treatment of Acute Myocardial Infarction: Title and subTitle BreakThe Cooperative Cardiovascular Project FREE

Each year approximately 1.5 million people in the United States experience acute myocardial infarction (AMI), and approximately one third of these patients die in the acute phase of the AMI. The annual economic burden of AMI is in excess of $60 billion.¹ Due to its frequency and severity, AMI has been the topic of intense scientific and clinical interest. Randomized trials have confirmed the efficacy of some therapies resulting in evidence-based treatment guidelines.^{2 - 3} These guidelines make it possible to evaluate the processes of care in a meaningful manner and to identify areas for the improvement of the care of patients with AMI.

Since 1992 the Health Care Financing Administration has implemented a continuous quality improvement approach to ensuring the quality of care for its Medicare beneficiaries through its Health Care Quality Improvement Initiative.⁴ The first national project of this program is the Cooperative Cardiovascular Project (CCP), focusing on the treatment of patients with AMI. Quality indicators, based on clinical practice guidelines, were developed by a steering committee.³ Data on more than 200,000 patients admitted for treatment of AMI were abstracted from clinical records. Patients were classified as eligible or ideal for the specific therapies described by the quality indicators.

During this same time period, studies were conducted to describe the geographic variation of health care in the United States.⁵ These studies used Medicare claims data to empirically derive hospital service areas (HSAs), which are local market areas for health care, and hospital referral regions (HRRs). Combining these HRRs with the national sample of data collected by CCP allowed us to study geographic variation in the treatment of AMI in the United States.

The sampling approach used bills submitted by acute care hospitals (UB-92 claims form data) and contained in the Medicare National Claims History File to identify all Medicare discharges with an International Classification of Diseases, Ninth Revision, Clinical Modification principal diagnosis of 410 (myocardial infarction), excluding fifth digits of 2, which indicate a "subsequent episode of care."⁶ Data were collected during 8-month periods, staggered so that all discharge dates were between February 1994 through July 1995. Approximately .003% of the bills submitted could not be linked to medical records. The National Claims History File does not necessarily include bills for all of the approximately 12% of patients treated under Medicare managed care risk contracts. The primary sampling relies on the billing principal diagnosis of AMI. The accuracy of billing diagnoses has been questioned frequently.^{7 - 8} The criteria for a confirmed AMI required elevated levels of creatine kinase isoenzyme MB or lactate dehydrogenase with LD₁ levels greater than LD₂ levels or 2 of the following: chest pain, 2-fold elevation of the creatine kinase level, or electrocardiographic evidence of an AMI. This data set contains the 186,800 patients from the 50 states with AMI confirmed by these criteria.

There were also 3 special circumstances in the sampling scheme. The pilot states (Alabama, Connecticut, Iowa, and Wisconsin) were sampled initially between June 1 and December 31, 1992, to test the indicator and abstraction methods and then remeasured between August 1 and November 30, 1995. Sampling in Minnesota was delayed until April through November 1995 so that it would not interfere with a local study of AMI; only about a 60% sample of AMI diagnoses was included in the CCP study. One hospital that initially ignored record requests was sampled from July 1995 through February 1996. We believe that the sampling strategy used in this study is highly representative of the Medicare fee-for-service patient population in the United States.

Quality indicators are measurable aspects of care that are presumed on the basis of evidence or consensus to be related to the quality of care delivered and to favorable outcomes. They were developed through an intensive review process, beginning with the establishment of a joint Health Care Financing Administration and American Medical Association Steering Committee for the CCP. Indicators were based on treatment guidelines for AMI published by the American College of Cardiology, the American Heart Association, and expert consensus.³ Representatives from the American College of Cardiology, the American Academy of Family Practice, and the American College of Physicians reviewed and refined these indicators. The peer review organizations in the 4 pilot states further refined the quality indicators and developed data collection instruments and computer algorithms.^{9 - 10}

For each indicator, specific criteria determined whether a patient could have been a candidate for a particular intervention. Potential candidates were divided into 2 subgroups. Ideal candidates are patients for whom the treatment would almost always be indicated, and less-than-ideal candidates are patients for whom the therapy was controversial (eg, the use of β-blocking agents for patients with congestive heart failure), for whom it was contraindicated (eg, thrombolytics for patients with a recent hemorrhagic stroke), or for whom critical data for determining the appropriateness of the intervention were missing (eg, data on thrombolytic timing were missing). Detailed information on the variable definitions, the indicator algorithms, and the data abstraction software are available at the CCP World Wide Web home page (www.usccp.org).

Data were obtained from the medical record by trained abstractors employed by independent contractors. Quality was monitored and maintained by random reabstractions and the percentage of agreement was generally high (range, 94.8%-98.3%).¹¹

All Medicare records of hospitalizations of patients residing in the 50 states for 1995 and 1996 were used to assign each ZIP code in the United States to 1 of 3436 geographically distinct HSAs based on where the plurality of patients traveled for hospital care. Based on patterns of care for major cardiovascular surgery, HSAs were aggregated into 306 HRRs.⁵ The HRRs were empirically derived regional health care markets. The CCP data were mapped to these markets using the ZIP code of the institutional provider of care for each episode of myocardial infarction. Details of the geographic analysis methods have been published.⁵ Data from sparsely populated areas were suppressed in these analyses. We also suppressed rates of adherence to specific indicators for any HRR that had less than 11 ideal cases for the indicator.

Quality indicators were evaluated only for patients during their first confirmed AMI during the sampling time frame and who were judged ideal candidates for a specific treatment (Table 1). Rates of adherence to performance indicators (ie, the proportion of ideal patients with confirmed AMI who received the specified treatment) were calculated for each of 306 HHRs. These proportions were adjusted for age and sex using logistic regression analysis.¹² The 20th and 80th percentiles were contrasted as a measure of variability.¹³ Standard statistical methods were used for the calculation of the correlation coefficient.¹⁴

The data set used in these analyses consists of 186,800 patients with confirmed AMI. Figure 1 and Figure 2 show the geographic variation in performance for each of the 7 CCP quality indicators. The white lines on the map are the state boundaries, and the black lines indicate the boundaries of the HRRs. Five levels of color intensity indicate the percentage of ideal patients with confirmed AMI receiving the therapy described by the quality indicator, ranging from white (0%-20% of patients received the therapy) to red (80%-100%). The gray coloring indicates sparsely populated areas or those in which there are insufficient data to calculate meaningful rates. The overall rates of adherence to these quality indicators are summarized in Table 2.

More than 90,000 patients were judged ideal for aspirin therapy, and it was widely prescribed during hospitalization (Figure 1, A). The mean rate was 86.2% (range, 67.8%-100.0%) with 20th and 80th percentiles of 82.6% to 90.1%. No HRRs had rates less than 60%, and the majority were greater than 80%. Most of the HRRs with rates of 60% to 79% were in the south central and southeast regions. Aspirin was prescribed at hospital discharge in 77.8% of the approximately 60,000 patients judged ideal for this therapy (Figure 1, B). The range was between 52.1% and 96.0% with 20th and 80th percentiles of 72.5% to 83.9%. Approximately half of the HRRs had rates more than 80%, and only a few were less than 60%. Highest rates were observed in the northeast, north central, and mountain regions.

The quality indicator for reperfusion includes treatment with thrombolytic agents or primary percutaneous transluminal coronary angioplasty during the first 12 hours of hospitalization. The mean rate of reperfusion among 17,071 ideal patients was 67.2% (Figure 1, C). Substantial variation was found with a range from 33.0% to 93.3% and 20th and 80th percentiles of 59.8% to 75.1%. Lowest rates were in the south central and mid Atlantic states with a few HRRs in the west and northwest. Rates greater than 80% were observed in 23 widely scattered HRRs.

There were many exclusions for the use of β-blocking agents, thus only 14,839 (7.9%) of the 186,800 patients with confirmed AMI were judged ideal for this treatment. The large number of HRRs in Minnesota with insufficient data is a consequence of the smaller proportion of AMI patients sampled in that state. Approximately half (49.5%) of all patients judged ideal received β-blocking agents (Figure 2, A). The range was 0%-92.7% with 20th and 80th percentiles of 35.8% to 61.5%. Rates of use of β-blocking agents were highest in the northeast and north central regions and lowest in the south central and south east regions.

Angiotensin-converting enzyme (ACE) inhibitors were prescribed to 59.3% of the approximately 18,000 patients judged ideal for this treatment (Figure 2, B). The range of rates is wide (6.7%-100%) with 20th and 80th percentiles of 49.2% to 69.2%. Rates were somewhat higher in the north central region, but there was no clear geographic pattern of ACE inhibitor use.

The quality indicator for calcium channel blocking agents specifies that they be withheld from patients with impaired left ventricular function (ejection fraction <40%). Since only 4.8% of the patients had this degree of impairment of ventricular function, approximately 25% of HRRs had insufficient data on which to base adherence rates for this quality indicator. Overall, calcium channel blocking agents were withheld from 81.9% of patients for whom their use was contraindicated. There was little variability in rates with 20th and 80th percentiles of 73.6% to 90.8% and no clear geographic pattern (Figure 2, C).

Of the 186,800 patients with confirmed AMI, 22,024 (11.8%) were identified in the clinical record as current cigarette smokers. Of these, 41.9% have documented evidence of smoking cessation advice. The rates of smoking cessation advice varied widely (7.3%-81.7%) with 20th and 80th percentiles of 32.8% to 51.3% (Figure 2, D). Rates were highest in Alaska and the mountain states and lowest in the south central and southeast regions.

To explore the effects of chance, we stratified the 306 HRRs into tertiles based on the observed sample size. We calculated the mean and the 20th and the 80th percentiles for small, medium, and large HRRs. These results (not shown) were similar in the small, medium, and large HRRs for every quality indicator. Furthermore, there were only relatively small differences in the 20th and the 80th percentiles of the estimates for the quality indicators.

We also examined the correlation between the various quality indicators by HRR. There was a significant positive correlation between the indicators for treatment with aspirin, β-blocking agents, and ACE inhibitors (r=0.26-0.59; all P values <.001). Smoking cessation advice was also positively correlated with these indicators (r=0.16-0.27; all P values <.01). The withholding of calcium channel blocking agents among patients with impaired left ventricular function was weakly and inconsistently related to the other quality indicators. Reperfusion was not significantly correlated with any of the other quality indicators.

In these analyses, based on 186,800 patients treated for a confirmed AMI and judged "ideal" for the specific treatment, we found considerable variability in the rates of adherence to the CCP quality indicators. Aspirin was used frequently both during hospitalization (86.2%) and at discharge (77.8%), and calcium channel blocking agents were withheld from most patients with impaired left ventricular function (81.9%). Lower rates were seen in the use of ACE inhibitors at discharge (59.3%), reperfusion using thrombolytic therapy or coronary angioplasty (67.2%), prescription of β-blocking agents at discharge (49.5%), and for documented smoking cessation advice (41.9%). Quality indicators for aspirin, β-blocking agents, ACE inhibitors, and smoking cessation advice were positively correlated within HRRs. There was substantial geographic variation in the treatment of AMI. A wide range of treatment rates by HRR was found for most of the quality indicators. This data set does not have information on any health system characteristics, but the observed geographic variability observed does not support the existence of systematic effects. High and low rates of quality indicators can be found within the same state or region. Although some rather striking regional variations exist, much of the observed variability seems to relate to local clinical practices. It is unlikely that this variation between the care described by the quality indicators and the kind of care actually provided is a consequence of chance or of differences in patient case mix in this large population of patients homogeneous with respect to age, sex, and access to health care.

There are 2 limitations to these data and the inferences that can be drawn from this study. First, these data rely on information contained in the medical record. Medical records are sometimes illegible or poorly organized, and the task of medical record abstraction is quite difficult. Reliability testing of the medical record abstraction in the CCP study has shown excellent interrater reliability. While it is possible that variations noted were due to care that was provided and not documented, this would suggest a significant departure from established documentation policies. In all hospitals, it is required that physician orders be recorded in the medical record for specific therapies to be administered, and documentation of discharge medications is required by the Joint Commission on Healthcare Organizations. Thus, we believe that the CCP data abstraction system yields an overall accurate representation of the care of these patients. Second, the patterns of geographic variation in the treatment of AMI observed in this national study may accurately represent care during this study period but may not represent long-term patterns of care.

Prior studies have shown variability in use of medical and surgical services.^{15 - 16} Studies of treatment of AMI have evaluated variability in the use of diagnostic and therapeutic procedures among specific patient subgroups, by payers, or by location.^{17 - 20} Studies of specific medical therapies have been far less frequent. Pilote et al²¹ studied regional variation in the management of AMI in the United States among 21,772 patients and found substantial variation in the use of medications and in invasive procedures. Rogers et al²² reported observations from the National Registry of Myocardial Infarction and concluded that there was substantial underuse of thrombolytic therapy and β-blocking agents and substantial overuse of calcium channel blockers.

Reports from randomized trials or voluntary registries may not represent the care provided for the entire United States. The CCP results describe AMI care for patients with AMI participating in the Medicare fee-for-service program. These findings are in general agreement with the conclusions reached by Pilote et al²¹ and Rogers et al²² and extend the findings of these studies by identifying patients ideal for the specific therapy and demonstrating variation in compliance to guidelines throughout the United States. This study shows the opportunity for improvement of the care of patients with AMI using existing knowledge. Aspirin, β-blocking agents, ACE inhibitors, and thrombolytics have been shown in randomized trials to reduce mortality due to AMI.^{23 - 25} Reperfusion using percutaneous transluminal coronary angioplasty has been shown to have a protective effect on short-term mortality after AMI.²⁶ The American College of Cardiology and American Heart Association guidelines²⁷ have listed primary percutaneous transluminal coronary angioplasty as an alternative to thrombolytic therapy. In these CCP data, substantial numbers of patients judged ideal for these treatments did not receive them. Approximately 14% of such patients did not receive aspirin during hospitalization, 22% did not receive aspirin at discharge, 50% did not receive β-blocking agents, 41% did not receive ACE inhibitors, and 33% did not receive reperfusion. Calcium channel blocking agents were administered to 18.1% of patients with impaired left ventricular function. Current guidelines generally recommend against the use of calcium channel blocking agents in the acute or long-term care of patients with AMI.²⁷ There was no documentation that 58% of patients who smoke received smoking cessation advice.

The CCP quality indicators are practice guidelines based largely on evidence from randomized clinical trials and were not controversial at the time this study was conducted. They describe practices of care that will optimize patient outcomes and reduce the use of possibly harmful therapies.²⁸ Why then does this gap between the care provided and the quality indicators exist? Physicians may be unaware of the guidelines or disagree with their content²⁹ ; or they may be resistant to guidelines because they perceive the loss of professional autonomy.^{30 - 31} Alternatively, changes in behavior lag behind knowledge.³² It is likely that all of these factors contribute to the lack of uniformity of treatment observed in this study. Improvement of clinical care is possible. Feedback of the CCP data to hospitals improved clinical care in the 4 pilot test states³³ ; local medical opinion leaders are successful in accelerating the rate of improvement of AMI management³⁴ ; and techniques of continuous quality improvement are useful to change current clinical practices.^{35 - 37}

These gaps between knowledge and practice have important consequences. Acute myocardial infarction is a common and serious condition, and the evidence from clinical trials strongly suggests that adherence to these established guidelines will result in better patient outcomes. It is undoubtedly true that some AMI patients experience unnecessary morbidity or mortality because they receive substandard medical care. This study finds that there is currently unrealized potential for more effective care of patients with AMI. Using the lessons learned from studies such as the CCP to improve continually the outcomes of cardiovascular care is a goal that should be shared by society, payers, and clinicians. Because these goals are primarily the responsibility of clinicians, it is essential for the public good that clinicians shoulder that responsibility by improving the quality of care.