FormalPara Key Points For Decision Makers

Substantial post-acute myocardial infarction costs are incurred through the 90-day post-discharge period irrespective of the care pathway, suggesting that moving from traditional 30-day payment models to 90-day payment models may increase financial risk to hospital providers.

Costs of post-acute care are substantially higher for medically managed patients than for those receiving percutaneous coronary intervention, highlighting an unmet need for therapies that reduce recurrent cardiovascular events and readmissions in these patients.

Heterogeneous post-discharge costs of care must be considered when implementing the Centers for Medicare & Medicaid Services Bundled Payment for Care Improvement and similar programs.

1 Introduction

In the United States, there are an estimated 805,000 incidences of acute myocardial infarction (AMI) per year, and with an average age of 65.6 years for males and 72.0 for females for first AMIs [1], a substantial proportion of AMI patients are eligible for federally sponsored health insurance under Medicare (i.e., ≥ 65 years). In fee-for-service Medicare, the Centers for Medicare & Medicaid Services (CMS) reimburses hospitals for inpatient services based on Medicare Severity Diagnosis Related Groups (MS-DRGs) [2]. These MS-DRG payments are typically standardized and designed to cover all charges associated with inpatient stay. Reimbursements under Medicare to hospitals for acute care episodes, such as AMI, have been typically linked to services used during a 30-day period [3]. However, in 2013, CMS introduced a voluntary bundled payment system, the Bundled Payment for Care Improvement (BPCI) Program [4], which involves a single payment for services used during an episode of care and extends the period up to 90 days after discharge [5]. The current model in place up to December 31, 2023 is BPCI advanced (year 4) [6]. Under the current model, CMS reimburses episodes based on pre-defined, yearly fixed target prices. Target prices are compared with actual episode costs at the end of each year, and providers must pay the difference if actual costs exceed the target price [6].

Because hospitals are often the accountable provider for managing bundled payments, extended 90-day reimbursement programs such as BPCI present implications for providers when applied to AMI as an episode of care. First, treatment of AMI is complex and usually includes follow-up appointments that extend beyond discharge. Second, patients with AMI are at high risk of subsequent cardiovascular (CV) events, such as recurrent AMI, stroke, and CV death [1]. Of the 805,000 AMIs that occur in the United States per year, it is estimated that a quarter are recurrent, with recurrent AMIs associated with a twofold increase in 5-year mortality versus index AMIs [1, 7]. Costs associated with CV-related hospitalizations have been found to be $20,000 higher than non-CV related hospitalizations (all costs reported in this article are in US dollars) [8], and Medicare expenditures for AMI in the 30-day period have been found to be higher than for other CV-related conditions such as heart failure [9]. Moreover, there is growing evidence that readmission burden for AMI extends beyond the typical 30-day period [9, 10]. A recent study found that approximately a quarter of AMI episodes in the USA are associated with readmission within 90 days, an increase of approximately 65% on the proportion readmitted within 30 days [11]. There are, therefore, concerns that programs such as BPCI may increase financial risk to enrolled providers.

A key factor to consider relating to provider burden early after AMI is the treatment pathway of the index event, specifically with non-surgical interventions whereby patients are treated by either percutaneous coronary intervention (PCI) or purely managed pharmacologically (i.e., medical management [MM]). As PCI is associated with higher cost of acute care for AMI, CMS segregates payments for AMI between those who are medically managed and those who undergo PCI [5]. Depending on the clinical presentation and treatment prognosis, primary PCI may be the preferred pathway of care for patients with AMI [12,13,14,15] and data show that between 7.9 and 27.4% of AMI events are medically managed [16]. However, the need and suitability for PCI depends upon many clinical factors, including if the event was associated with ST-segment elevation, and other factors including individual anatomy and comorbidities [12,13,14,15, 17]. Further, there are numerous patient factors that increase clinical burden following AMI, including age, sex, and comorbidities such as diabetes, peripheral arterial disease (PAD), and multivessel disease (MVD) [11, 18,19,20].

To investigate the potential extent of risk faced by hospitals enrolled in an extended 90-day bundled payment system, in a Medicare claims analysis, we evaluated payments and healthcare utilization associated with AMI as an episode of care up to 90 days post-discharge and the impact of treatment pathway (PCI or MM). A further aim was to evaluate the association of pathway of care and patient comorbidities, with particular focus on MVD, on post-discharge healthcare utilization and costs.

2 Methods

The present study was a retrospective analysis of Medicare claims (October 1, 2015 to September 30, 2016).

2.1 Data Sources

The analytic file was derived from the CMS 100% Standard Analytic Files (SAFs) for AMI hospitalizations. These contain adjudicated medical claims submitted by inpatient hospital providers for payment for services provided, including complete information on diagnoses, procedures, MS-DRG, dates of service, reimbursement amounts, provider, and beneficiary demographics. These claims data are compiled into a limited dataset, which only contains one data element of protected health information (date of service) and are standardized to remove geographic variation.

The CMS March 2018 Provider of Services File was utilized to obtain information on whether the hospitals have cardiac catheterization rooms and cardiac surgical units. In addition, the Inpatient Prospective Payment System (IPPS) Final Rule Impact File was used to obtain information relating to community size (number of residents: < 50,000, 50,000 to < 1 million, and ≥ 1 million), hospital size (number of beds: < 250, 250–499, and ≥ 500), readmission adjustment factor (the adjustment made by CMS to yearly payments because of readmission performance measures, with a high of 1 [100%] and a low of 0.97), and percentage of hospital days paid by Medicare as a percentage of total days.

2.2 Sample

The sample included Medicare beneficiaries (continuously enrolled in Medicare Part A and B for 9 months prior to admission through 3 months after the discharge date) with a primary diagnosis for AMI (10th revision of the International Statistical Classification of Diseases and Related Health Problems [ICD-10] code I21.x) on an inpatient claims record from an IPPS hospital and discharged with an MS-DRG of 280–282 or 246–251. The identification and selection of eligible study participants, including the exclusion criteria applied, is detailed in Figure 1. Patients for whom PCI would be usually contraindicated (e.g., those with kidney disease, sepsis) or not be performed (e.g., patients undergoing coronary artery bypass grafting [CABG] surgery) were excluded from the analysis to minimize confounders in the assessment of economic costs associated with pathway of care. In addition, 5735 visits for patients who were not treated at hospitals that participate in CMS’ IPPS were excluded. This exclusion was applied because the data on facility characteristics was not available for those hospitals.

Fig. 1
figure 1

Flow diagram to identify population eligible for analysis. AMI acute myocardial infarction, CABG coronary artery bypass grafting, FFS fee for service, IPPS Inpatient Prospective Payment System, MI myocardial infarction, MS-DRG Medicare Severity Diagnosis Related Groups. *Patients may have multiple hospital stays if they suffer ≥ 1 AMI during the 1-year evaluation period. †Death prior to the 90-day post-acute window does not exclude patients. ‡Patients who undergo a CABG during or immediately after the inpatient stay (within 45 days) are excluded from the study; however, a history of CABG qualifies a patient as having multivessel disease and does not exclude a patient from the study

Full size image

2.3 Measures

Distribution of acute and post-discharge costs across facilities was assessed. Each hospitalization was assigned to a pathway of care: MM without referral to catheterization lab, MM following referral to catheterization lab, PCI, or CABG (Table 1). Patients were assigned a diagnosis of multivessel coronary artery disease (MVD) if multiple vessels were treated with a PCI or atherectomy, or if the patient had previously had a CABG. Other comorbidities diagnosed during the inpatient stay were captured.

Table 1 Eligible MS-DRGs associated with AMIs included in the analysis
Full size table

2.4 Analyses

Patient demographics were compared between groups using chi-square tests for categorical variables and t tests for continuous variables. Medicare payments were standardized to remove geographical variation and separated into reimbursements for services during hospitalization and from discharge to 90 days post-discharge. Results were stratified by MS-DRG individually, and mean payments are reported. Costs presented represent actual costs during the study time frame and are not updated to current year dollars. The mean for utilizers is the total costs for a service (e.g., emergency department [ED] visits) divided by the total number of utilizers of a service (e.g., MS-DRG 280 discharged patients who visit an ED within 90 days of discharge). Median and interquartile range (IQR) payments are also reported.

Payments were separately estimated for the two bundles under study (i.e., PCI or MM), which are described in Table 1. Chi-square tests were used to evaluate the significance of relationships between categorical variables, and analysis of variance (ANOVA) for categorical and ratio level variables.

Excess hospital readmissions are penalized by CMS, and, therefore, may be a primary source of financial risk for hospitals. Post-acute care (in the form of skilled nursing facilities and, to a lesser extent, home health, long-term care hospitals, and inpatient rehabilitation hospitals) is another potential source of cost and risk. Because pathway of care (i.e., PCI vs MM) is related to the utilization of post-acute nursing care, which is in turn related to likelihood of rehospitalization, logistic regression was used to evaluate the risk-adjusted likelihood of utilization of post-acute nursing care and all-cause readmission.

All analyses were conducted in SAS version 9.4.

3 Results

3.1 Study Population

In total, 96,546 unique patient stays were identified (Fig. 1). The study population had a mean age of 74 years, were predominantly white, and more were male (Table 2). Characteristics including age, sex, race, and MVD status were significantly different between MM and PCI treated patients (all p < 0.001; Table 2). Additional observations of note were that female patients were significantly older than male patients (75.3 vs 72.0 years, p < 0.0001; Table 2), patients treated with PCI were significantly younger than MM patients (71.5 vs 76.1 years, p < 0.0001; Table 2), a significantly greater proportion of white patients were treated with PCI than African Americans (48,986/84,428 [58.0%] vs 3393/7067 [48.0%], p < 0.0001; Table 2), and more male than female patients were treated with PCI (34,536/53,051 [65.1%] vs 20,741/43,495 [47.7%], p < 0.0001; Table 2).

Table 2 Characteristics of patient population
Full size table

3.2 Payments for Healthcare Services for AMI Patients

The overall mean cost for an AMI episode of care (weighted by the number of individuals discharged under each DRG) was $22,034 (median $17,561, IQR $13,065) from hospital admission through 90 days post-discharge (Table 3). Within both PCI and MM MS-DRG families, inpatient (acute) and outpatient (post-acute) payments varied substantially but were highest amongst patients with major complications or comorbidities. The largest total mean payment was associated with MS-DRG 248 (PCI with non-drug eluting stent with major complication/comorbidity or ≥ 4 vessels/stents; mean $32,714, median [IQR] $27,007 [$15,486]), followed by MS-DRG 250 (PCI without coronary artery stent without major complication/comorbidity; mean $31,963, median [IQR] $24,478 [$16,669]), MS-DRG 246 (PCI with drug-eluting stent with major complication/comorbidity or ≥ 4 vessels/stents; mean $30,879, median [IQR] $26,287 [$11,433]), and MS-DRG 280 (AMI, discharged alive with major complication/comorbidity; mean $29,164, median [IQR] $19,983 [$22,047]). The lowest mean costs were associated with MS-DRG 282 (AMI, discharged alive without complication/comorbidity or major complication/comorbidity; mean $17,752, median [IQR] $9910 [$15,240]).

Table 3 Distribution of total payments (in US dollars) across inpatient and outpatient settings of care
Full size table

Payments for acute services were considerably higher for patients who had undergone PCI than for those who were medically managed ($16,230 vs $8536). However, post-acute services represented a higher proportion of the total expenditure in MM patients (55–62%) compared with PCI patients (26–41%) (Figure 2).

Fig. 2
figure 2

Total healthcare spending and percentage incurred in post-acute period, by MS-DRG. MM medical management, MS-DRG Medicare Severity Diagnosis Related Group, PCI percutaneous coronary intervention

Full size image

3.3 Utilization of Post-discharge Services

Utilization of post-discharge services by MS-DRGs is shown in Supplementary Table 1 (see the electronic supplementary material). Non-ED post-discharge care was the most frequently used classification of post-discharge service, and the least utilized services were long-term care facilities. Compared to patients receiving PCI, MM patients were significantly more likely to be readmitted (30.5% vs 15.4%, p < 0.0001), to require home health agency services (16.7% vs 8.2%, p < 0.0001), to be discharged to a skilled nursing facility (16.5% vs 5.0%, p < 0.0001), to receive rehabilitation services (23.7% vs 21.8%; p < 0.0001), and to require an additional ED visit (1.2% vs 0.6%; p < 0.0001). In contrast, PCI-managed patients used significantly more outpatient (non-ED) services than MM patients (65.8% vs 58.3%, p < 0.0001).

3.4 Factors Associated with Utilization of Post-acute Nursing Care

Apart from dementia, the strongest predictor of the requirement for post-acute nursing care was MM without catherization versus PCI, with these patients around twice as likely to be referred than those who received PCI (odds ratio [OR] 2.01, 95% confidence interval [CI] 1.92–2.11; Table 4). Multiple comorbidities predicted need for post-acute care, the strongest of which being dementia, followed by malnutrition, and sepsis. Comorbidities that typically correlate with MVD (e.g., diabetes with/without complications, treatment for diabetes, PAD, coronary atherosclerosis, and a history of AMI) were more likely to be referred to post-acute care (Table 4). Referral rates to post-acute care were higher in hospitals without open heart surgical units, located in more populous areas (over a million residents), and in which Medicare paid for more than 45% of days (Table 4). Female gender and older age were also significantly associated with an increased likelihood of referral to post-acute care.

Table 4 Predictors of post-acute nursing care and all-cause readmission
Full size table

3.5 Predictors of All-Cause Readmissions

The strongest predictor of 90-day hospital readmissions was utilization of post-acute nursing care, which was associated with more than a fourfold increase in the likelihood of 90-day readmission (OR 4.38, 95% CI 4.21–4.55; Table 4). There was also an influence of hospital size/type on all-cause readmission rates, with readmissions more likely from hospitals with fewer beds, hospitals without open heart surgical units, and Medicare-dependent hospitals (Table 4). Other predictors of readmission were generally similar to predictors of post-acute nursing care. Again, MM was associated with greater utilization than PCI, with MM patients treated without catheterization being the most likely to be readmitted. Notable exceptions were that MVD was associated with greater odds of readmission, the lack of association between sex and readmissions, and that increasing age was associated with a decrease in the chance of 90-day readmission.

4 Discussion

In this Medicare claims analysis, we found the 90-day period after AMI was associated with substantial economic burden irrespective of the pathway of care (MM or PCI), and that post-acute care spending (i.e., up to 90 days post-discharge) encompassed a significant proportion of total healthcare spending for AMI as an episode of care. Interestingly, the pathway of care influenced when the bulk of these costs are incurred; MM incurred up to approximately two-thirds of costs in the post-acute period, whereas in PCI-managed patients, the majority of spending was for services delivered in hospital. Importantly, our results suggest that this was not purely due to the higher cost of PCI treatment relative to MM, as costs for post-acute care settings such as inpatient readmissions were higher in MM patients than in PCI patients. This is also supported by our finding that MM was associated with a twofold increase in the likelihood to require post-acute nursing care and be readmitted to hospitals.

We found that comorbidities such as MVD, diabetes, PAD, and prior myocardial infarction (MI) were associated with a higher risk of 90-day readmission and, similarly, smaller hospitals with less surgical cardiac expertise were associated with higher rates of 90-day readmission. Conversely, increasing age was weakly associated with a decreased likelihood of 90-day readmission, although this should be treated with caution, since age is correlated with other comorbidities used in our model and thus may be subject to collinearity. These results are in line with a prior Medicare analysis by Culler et al. [5], which also reported that MM is associated with higher readmissions within 90 days versus PCI and higher rates of comorbidities, and showed that PAD, diabetes, and prior MI are associated with higher frequency of readmissions. Taken together with our cost analysis, our findings build upon this by demonstrating that this high clinical risk is associated with a substantial economic burden.

Prior analyses of hospitals participating in the BPCI for various episodes of care, which included AMI, have reported that despite covering costs incurred over an extended 90-day period, the BPCI is associated with unchanged or lower payments for acute care episodes compared to 30-day payment programs [21,22,23,24]. This is important given recent evidence of high hospitalization burden in the 90 days post-discharge, with readmission rates ranging between 24–28% [5, 11]. Moreover, 38% of major adverse CV event readmissions after AMI in the United States occur between 30 and 90 days after discharge [25], and the 90-day post-AMI period has been reported to be associated with an approximate 65% increase in AMI readmissions versus the 30-day period [11]. Thus, if reimbursements for acute care episodes in 90-day models are comparable to those of 30-day payment programs [21,22,23,24], moving from a typical 30-day reimbursement model to a 90-day model such as BPCI may increase the financial risk incurred by hospitals since they must pay the difference if the actual cost of care exceeds the fixed reimbursement given by the payment program [6]. This is particularly relevant to medically managed AMI patients, for whom we found around two-thirds of costs are incurred in the post-acute period up to 90 days after discharge.

Interestingly, we found that higher PCI use and centers with cardiac units were associated with lower post-discharge expenditure; this is supported by Pandey et al. [26], who found these factors to be associated with higher post-discharge ‘home time,’ which may be a better metric of hospital performance. Moreover, although the MM MS-DRGs incurred the highest post-discharge expenditures, they have the lowest reimbursements (and presumably, lowest costs of inpatient care). This may result in a situation where hospitals become financially incentivized to perform PCI, despite it not always being clinically appropriate/possible, e.g., due to high bleeding risk [27]. Despite the BCPI aiming to reduce costs and improve quality, when applied to CV care in general, it has been shown that the program has not significantly improved quality of care or reduced spending [28].

Our analysis supports that MVD is associated with higher rates of readmission within 90 days. This is interesting given the increasing focus on MVD as a key driver of poor outcomes following AMI, with the risk of recurrent AMI events increasing in a stepwise manner with the increasing number of non-culprit plaque/vessels [29, 30]. There is growing evidence that complete revascularization, i.e., treatment of culprit as well as non-culprit lesions in a single or staged procedure, improves clinical outcomes, and consequently, standard of care is moving towards complete revascularization where clinically appropriate [17, 31, 32]. Nonetheless, the increased risk of 90-day readmissions among MVD patients supports that higher atherosclerotic burden in these patients translates into higher risk of 90-day readmission.

The finding that post-acute nursing care utilization was more likely in more populous areas suggests that availability was lower in rural areas, as higher post-acute care utilization did not translate into higher 90-day readmissions in more populous areas. The finding that smaller, presumably rural, hospitals showed worse 90-day readmissions may be related to the provision of standard of care medications, which has been shown to be less comprehensive by rural hospitals [33]. In this regard, pharmacist-led transition of care programs have been shown to significantly decrease readmissions [34]. Similarly, we found that a lack of specialized cardiac units was associated with increased readmissions; in addition to this, there could be variation between individual surgeons. However, as previously surmised [5], Medicare claims data highlight an overall unmet need in preventive therapeutics in the high-risk 90-day period post-AMI, and that new strategies are required. Currently, therapies that reduce the pro-atherogenic lipoprotein low-density lipoprotein cholesterol (i.e., statins, ezetimibe, and proprotein convertase subtilisin/kexin type 9 inhibitors) take several months to years to have a pronounced effect on CV events [35, 36]. Pharmacological interventions targeting atherosclerotic plaque may be needed for those who do not receive PCI and have comorbidities placing them at high risk, particularly MVD. This may in turn reduce the financial risk on providers in a 90-day bundled payment system.

This study had some limitations. First, since there is currently no ICD-10 diagnosis code for MVD, the approach used, as described in the methods, may have been unable to capture all true MVD patients. Second, the exclusion criteria used to minimize confounders in the assessment of economic costs associated with the pathway of care may have introduced bias in the reported clinical outcomes. Third, although the CMS data files used were the most recent available at study conception, future studies using currently available CMS data are warranted to incorporate changes in clinical practice and implementation of new treatment modalities. Fourth, the analysis did not include non-fee-for-service Medicare populations, e.g., Medicare Advantage and commercially insured or Medicaid populations. These populations may have different post-discharge risks and costs. Finally, AMI episodes of care were only included where AMI was the primary diagnosis; patients treated for AMI in combination with other conditions may also have had different post-discharge risks and costs.

5 Conclusions

The 90-day post-AMI period is associated with significant post-acute care costs, particularly among patients who are medically managed. Care pathway impacts the timing that costs are incurred, with medically managed patients incurring higher costs in the post-acute period in relation to some care settings compared to those undergoing PCI. As the CMS BPCI and similar programs are implemented, there will be a need to account for these heterogeneous post-discharge costs of care. Centering payments for AMI around a 90-day model further highlights that the clinical and economic burden of AMI goes beyond the immediate discharge period, representing a current unmet need in preventative cardiology. There is a particular need for additional preventative measures in those who are medically managed, both in a clinical sense, and to reduce financial risk to hospitals/providers.