Costs of Hospitalizations with a Primary Diagnosis of Acute Myocardial Infarction Among Patients Aged 18-64 Years in the United States

Acute Myocardial Infarction (AMI) is both a common and deadly type of cardiac event in the United States. Although the age-adjusted hospitalization rate for AMI and its in-hospital case fatality rates have both declined since the mid-1990s, there were still 634,000 inpatient admissions in 2009 for which AMI was listed as the primary diagnosis [1, 2]. Moreover, Americans suffered an estimated 610,000 first-time AMIs and 325,000 recurrent attacks, and 133,958 deaths in 2008 [2]. Because the declines in hospitalization and in-hospital mortality rates have been associated with more aggressive therapeutic interventions [1], it is impor‐ tant to evaluate the cost-effectiveness of these interventions.


Introduction
Acute Myocardial Infarction (AMI) is both a common and deadly type of cardiac event in the United States. Although the age-adjusted hospitalization rate for AMI and its in-hospital case fatality rates have both declined since the mid-1990s, there were still 634,000 inpatient admissions in 2009 for which AMI was listed as the primary diagnosis [1,2]. Moreover, Americans suffered an estimated 610,000 first-time AMIs and 325,000 recurrent attacks, and 133,958 deaths in 2008 [2]. Because the declines in hospitalization and in-hospital mortality rates have been associated with more aggressive therapeutic interventions [1], it is important to evaluate the cost-effectiveness of these interventions.
To evaluate specifically the cost-effectiveness of various interventions against AMI, direct cost estimates of AMI are required [3][4][5]. Surprisingly, however, these cost estimates have not been comprehensively examined in the U.S. Many studies have investigated the economic burden of AMI, but all had some limitations [6][7][8][9][10][11][12][13][14][15][16][17]. Furthermore, in part because of limitations in available studies, the costs of coronary heart disease (CHD) were used in one study to represent the costs for AMI [6], albeit this is inappropriate. For example, a previous study of insured adults aged 18-64 years found that only about 30% of CHD cases represented AMI [9]. Moreover, the American Heart Association recently estimated that the total prevalence of CHD among persons aged ≥20 years was 7% but the AMI prevalence of AMI in this group was 3.1% [2]. In addition, in 2005, hospitalization costs for AMI admissions among adults aged 18-64 years were about $5000 more than those for CHD admissions of non-AMI [9]. Clearly, information on costs that does not clearly distinguish between AMI and non-AMI admissions is of little use in evaluating the cost-effectiveness of interventions to treat AMI [18].
In the present study we estimated AMI-specific costs by exploring the hospitalization costs of AMI while incorporating the impacts on costs of percutaneous coronary intervention (PCI), coronary artery bypass graft (CABG) surgery, comorbidities, complications, ST-elevation status, and length of stay (LOS) while controlling for age, sex, geographic regions, and urban versus non-urban location. Because PCI, CABG surgery, and LOS are likely to be the most influential factors on the costs and relevant factors for evaluating cost-effectiveness of AMI interventions, we also conducted multivariate logistic regressions to identify the factors predicting PCI, CABG surgery, and LOS.

Data source
The 2006-2008 MarketScan Commercial Claims and Encounter inpatient database was used for this study; this database contains information on patients up to age 64 years from approximately 40 privately insured employers, including state governments, with an average of nearly 21 million covered lives per year. In 2006-2008 the database had more than 2.4 billion service records representing commercially insured employees, qualified retirees and dependents from over 100 geographically diverse health insurance plans in all 50 U.S. states and the District of Columbia. The advantages of using the MarketScan database for economic studies include the large sample, detailed diagnosis codes for medical services, and hospitalization costs that are based on payment to providers [19]. Many researchers have used the MarketScan database to investigate medical costs associated with cardiovascular disease [9,20,21] Using the International Classification of Diseases, 9 th revision (ICD-9) codes, we identified hospitalizations with a primary diagnosis of AMI among patients aged 18-64 years who were enrolled in non-capitated health insurance plans. We further separated the hospitalizations into ST-elevated myocardial infarction (STEMI) and non-ST-elevated myocardial infarction (NSTEMI) cases. Based on secondary diagnosis codes, we identified major comorbidities, complications, and procedures for these hospitalizations (Table 1).
We excluded patients younger than 18 years because AMI is very uncommon in that group. We did not include patients in capitated health insurance plans because their costs of hospitalization would not reflect the medical services provided to them. We excluded hospitalizations with a LOS greater than 30 days because we determined that these hospitalizations (n=131, figure 1) would skew our results. To further limit the influence of extreme values on the cost estimates, we excluded all hospitalizations with a cost in the lowest or highest 1% of values ( Figure  1). The costs in our study included all those for physician services, diagnostic tests, therapeutics, supplies, and room fees during the hospitalizations. These costs, as noted above, represented total payment to providers rather than hospital charges. Accordingly, we did not need to adjust charges into payments to reflect the true economic burden of hospitalizations, nor did we use unit cost per bed day or an expert panel's suggested cost as in many other studies [5,11,7,22,23]. We expressed the costs in 2008 dollars by adjusting the 2006 and 2007 value by the consumer price index (CPI) provided by the Bureau of Labor Statistics [24].

Statistical analysis
After deriving the sample means of the costs for different population groups, AMI types, comorbidities, complications, and procedures, we specified various versions of multivariate regression models to examine the factors influencing the costs while controlling for demographic variables and Charlson comorbidity index (CCI) [25]. We used CCI as a comprehensive measure of disease severity. It measures the likelihood of death or serious disability in the subsequent year by diagnosis codes of up to 18 different diseases. In addition to estimating the various versions of regression for the whole study sample, we ran a regression on the costs for STEMI and NSTEMI patients separately. Because PCI, CABG surgery and LOS were major factors determining the costs, we used logistic regression to investigate the predictors of these three factors. For the regression estimation, we used mixed-effects models with a repeated measures approach to account for the fact that a single patient might have multiple admissions during the 3-year period. All tests of statistical significance were 2-tailed, and a p<0.001 was considered significant. All statistical analyses were performed using SAS version 9.1 [26].      CABG: Coronary artery bypass graft. Table 3. Coefficient estimates of hospitalization costs for patients with acute myocardial infarction PCI and CABG surgery increased the cost for both the STEMI and NSTEMI groups, with both procedures increasing the cost more for the NSTEMI group than for STEMI. LOS, in contrast, increased the cost more for the STEMI than the NSTEMI group, while living in an urban area (MSA in Table 3) decreased cost by $1496 for STEMI and $903 for NSTEMI.
Logistic regression indicated that patients aged 18-44 years were less likely than those aged 55-64 to undergo PCI or to have CABG surgery, and they were more likely to have a shorter LOS (i.e., <5 days) ( Table 4). Patients in urban area were more likely to have PCI, but less likely to have CABG. Men were more likely to undergo PCI or to have CABG surgery than were women, but their odd of a short LOS was greater. Versus patients who did not live in urban areas, urban patients were more likely to have PCI, but they were less likely to undergo CABG surgery. Compared with patients in the West, patients in other regions were more likely to have a long LOS (i.e., ≥5 days), but they were usually less likely to have PCI and CABG surgery, with PCI in the North Central region the exception. STEMI patients were more likely than NSTEMI patients to undergo PCI and CABG surgery, and they were more likely to have a long LOS. Patients with comorbidities or complications were more likely to have a long LOS, but they were less likely to have PCI or CABG surgery. Patients undergoing PCI were more likely to have a short LOS, while patients undergoing CABG surgery were far more likely to have a long LOS.

Discussion
The large number of hospitalizations in our economic study of inpatients who had suffered an AMI enabled us to explore a variety of factors that influenced their costs. The results suggest that CABG and PCI are the biggest drivers of hospital costs for AMI patients, adding, respectively, $12,546 and $28,406 to the cost of a stay. The cost effects of PCI and CABG in our study were comparable to the $15,089 and $28,974 additional costs, respectively, found in a Medicare population [7]. Another study reported similar costs for PCI and CABG [17].
In an earlier study using MarketScan data from 2003 to 2006, Zhao and Winget found that the total hospitalization costs of PCI and CABG surgery patient costs were, respectively, $31,379 and $63,909 [10]. Unfortunately, Zhao and Winget did not explore the effects of PCI and CABG on the costs of stay, as we did in our study. Such information is needed to evaluate the cost-effectiveness of AMI interventions [4].
Two other significant drivers of cost in our study were complications and LOS. Having one or more complications increased the cost by over $4600, and LOS increased the cost by over $2900 per day. LOS was highly correlated with CABG surgery and with complications, as indicated in our logistic models (Table 4). Thus, interventions aiming to prevent or better manage the complications of AMI patients might be cost-effective in reducing the hospitalization costs of this group.
Hospitalizations with STEMI had, on average, higher costs than NSTEMI hospitalizations, but after including PCI and CABG surgery as well as complications, comorbidities, and LOS in the regression model, the magnitude of the effect became much smaller. This may be because of differences in treatment approaches and in complications between the two kinds of hospitalizations. For example, over 80% of STEMI hospitalizations had a PCI while only about 51% in the NSTEMI group did. However, the NSTEMI group had a higher rate of CABG surgery than did STEMI (12% vs. 8%) (not shown in tables). On the other hand, compared with NSTEMI cases, the STEMI group had a higher rate of cardiogenic shock, ventricular tachycardia, and ventricular fibrillation, but it had a lower rate of heart failure, atrial tachycardia, and atrial fibrillation. All of these factors would affect the cost differences between STEMI and NSTEMI hospitalizations. The fact that STEMI cost more than NSTEMI was consistent with the literature; in Mexico, for example, STEMI cost nearly $2800 more than NSTEMI [11].
The predictors of PCI, CABG surgery, and LOS that we set forward in this study provide important information for secondary cost-effectiveness analyses of AMI interventions. We found that male patients were more likely than females to have PCI and CABG surgery, but their odds of a shorter LOS (<5 days) were greater. STEMI status greatly increased the probability of having PCI (coefficient estimate of 4.514) and significantly increased the probability of CABG surgery (coefficient estimate of 1.337), and it was associated with greater odds of a longer LOS (≥5 days). Patients with comorbidities and complications were relatively less likely to undergo PCI and CABG surgery, but they were more likely to have a longer LOS. All of these results could be used as inputs in cost-effectiveness evaluations of AMI interventions.
The numerous strengths of this study notwithstanding, several limitations should be considered when interpreting our results. First, all of our patients were covered by non-capitated private insurance plans. Although the costs of these patients accurately reflect the true economic burden imposed by their hospitalizations, the special population may have limited the generalizability of our results to the broader U.S. population. Second, all of our patients were 18-64 years old. The elderly population (aged >64 years) has much higher incidence and prevalence of AMI and its related comorbidities and complications [1,2]; as a consequence, the total costs of AMI should be higher in this population than among those 18-64. Although many studies have focused on the cost of AMI among the elderly [4,5,8], new estimation methods are needed along with high-quality data to develop better cost estimates for this population. Unfortunately, our data would not be appropriate for an analysis of costs among the elderly population for AMI hospitalization. A third limitation is that we estimated the costs of hospitalizations only. With survival rates increasing because of advances in technology [1], AMI patients are living longer. Correspondingly, the lifetime costs of outpatient care and medications for afflicted patients should be increasing. Additionally, productivity losses from the morbidity and premature mortality associated with AMI are also high [10] and should be considered in any comprehensive economic evaluations.
Given all of these factors, the hospitalization costs presented in our report should be treated as a conservative estimate of the economic burden associated with AMI. Moreover, we should note the limitation that we analyzed the costs of hospitalizations with AMI as a primary diagnosis. Although this decision let us cover the majority of AMI cases, there may be substantial additional hospitalizations in which AMI is a secondary diagnosis [9]. These hospitalizations should certainly be included in any complete analysis of the costs of hospitalizations of AMI patients. Because examining the costs of AMI as a secondary diagnosis would require a different analytical framework from the one we used, it would have been beyond the scope of our analysis.

Conclusion
Using a large set of claims data, we estimated the hospitalization costs of patients with a primary diagnosis of AMI and identified the main cost drivers of this important problem. Because most previous studies did not provide any information on the predictors of the costs of AMI hospitalizations [27], we hope that the present study has to some degree filled this gap in the literature. The high costs of AMI could be an economic justification for policy makers to support efforts to prevent AMI. In addition, the detailed information presented herein about the impact of various factors on the costs, procedures, and LOS associated with hospitalizations having a primary diagnosis of AMI can be used to evaluate and support health economic research such as studies on the cost-effectiveness of interventions to control this problem.