Exploring the Healthcare Value of Percutaneous Coronary Intervention: Appropriateness, Outcomes and Costs in Michigan Hospitals

Abstract

Background:

Assessments of healthcare value have largely focused on measuring outcomes of care at a given level of cost with less attention paid to appropriateness. However, understanding how appropriateness relates to outcomes and costs is essential to determining healthcare value.

Methods:

In a retrospective cohort study design, administrative data from fee-for-service Medicare patients undergoing PCI in Michigan hospitals between 6/30/2010 and 12/31/2014 were linked with clinical data from a statewide PCI registry to calculate hospital-level measures of: 1) appropriate use criteria (AUC) scores, 2) 90-day risk-standardized readmission and mortality rates, and 3) 90-day risk-standardized episode costs. We then used Spearman’s correlation coefficients to assess the relationship between these measures.

Results:

A total of 29,839 PCI’s were performed at 33 PCI hospitals during the study period. A total of 13.3% were for STEMI, 25.0% for NSTEMI, 47.1% for unstable angina, 9.8% for stable angina, and 4.7% for other. The overall hospital-level mean AUC score was 8.4 ± 0.2. Ninety-day risk-standardized readmission occurred in 23.7% ± 3.7% of cases, 90-day risk-standardized mortality in 4.3% ± 0.6%, and mean risk-standardized episode costs were $26,159 ± $1,074. Hospital-level AUC scores did not correlate with 90-day readmission, mortality or episode costs.

Conclusion:

Among Medicare patients undergoing PCI in Michigan, we found hospital-level AUC scores did not correlate with 90-day readmission, mortality or episode costs. This finding suggests that a comprehensive understanding of healthcare value requires multidimensional consideration of appropriateness, outcomes and costs.

Introduction

The challenge of understanding value in healthcare is increasingly important for hospitals and providers pursuing patient-centered, high-quality care. Percutaneous coronary intervention (PCI) is one of the most commonly performed and highly scrutinized medical procedures in the United States, responsible for an estimated $25 billion in costs each year to the healthcare system^{1, 2}. As a result, PCI has become the focus of numerous healthcare value initiatives. The Center for Medicare and Medicaid Services (CMS) has included components of PCI in its hospital value-based purchasing program for acute myocardial infarction (AMI) as well as its recent bundled payments for care improvement initiative³. Corresponding initiatives in the private sector are also being enacted^4–7.

Despite growing enthusiasm for these initiatives, most programs focused on procedures like PCI are currently limited by fundamental challenges with measurement of healthcare value. For instance, original definitions have considered healthcare value as “outcomes achieved at a given level of cost”⁸. This definition, however, does not explicitly account for whether the procedure provided would result in an expected benefit to the patient – a concept captured by measuring the appropriateness of a procedure for a given clinical scenario. This raises several potential concerns for a procedure like PCI that has variable appropriateness from its use in AMI where it can be lifesaving to its use in asymptomatic ischemia where its clinical benefits are substantially lower. Therefore, to truly understand the healthcare value of PCI, one most consider the appropriateness of the procedure for a given patient, in addition to costs and patient outcomes. Yet prior work in this area has understudied appropriateness in its assessments. In a large population of patients undergoing PCI in the Veterans Affairs (VA) Healthcare System, investigators reported variability in outcomes and costs across hospitals, but did not assess appropriateness^{9, 10}. In an analysis of the National Cardiovascular Data Registry (NCDR) CathPCI registry, appropriateness of PCI was found to be uncorrelated with procedural complications in the hospital; however, no examination of post-discharge outcomes or costs was performed¹¹. In an Ontario based study, the relationship between long-term outcomes and appropriateness in stable CAD patients was analyzed in patients managed with revascularization and medications, but healthcare costs were not assessed¹². Understanding the relationship of appropriateness with outcomes and costs is essential because a PCI that is of less appropriateness is of lower value to the patient, regardless of the outcome or cost.

Accordingly, we simultaneously examined the interrelationships between these three domains of PCI care – appropriateness, outcomes, and costs – in a large fee-for-service (FFS) Medicare population in Michigan. We accomplished this by linking clinical data from a unique statewide registry to comprehensive administrative and payment data¹³. We calculated hospital-level measures of appropriateness, 90-day outcomes, and costs for PCI and then correlated these measures to develop a more comprehensive view of healthcare value. Our findings have implications for policy-makers and payers seeking to introduce healthcare value initiatives for PCI and similar procedures.

Methods

Data Sources

We utilized two data sources that were part of the Michigan Value Partnerships, a regional quality improvement partnership between Blue Cross Blue Shield of Michigan and practitioners, physician organizations, and acute care hospitals in the state. The investigators were given complete access to these data sources. Due to agreements within our quality collaborative, the data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.

The first data source was the Blue Cross Blue Shield Cardiovascular Consortium (BMC2), a registry of all non-federal PCI-capable hospitals in the state of Michigan with 47 participating centers¹⁴. Captured data include detailed patient and hospital characteristics, procedural findings, interventions, and outcomes based on pre-specified data elements defined by the collaborative and the NCDR. Numerous prior publications have utilized the BMC2 PCI registry, including reports documenting the assessment of appropriateness and outcomes. A careful and detailed audit of this data source has been maintained for nearly two decades^{15, 16}.

The second data source involved the Michigan Value Collaborative (MVC), a statewide quality improvement collaborative also funded by Blue Cross Blue Shield of Michigan. MVC’s overall aim is to help hospitals and providers in Michigan provide high quality care at the lowest reasonable cost by working with its member hospitals to advance value-based initiatives. MVC maintains a claims-based registry that provides detailed information pertaining to standardized payments and utilization relating to an episode of care that includes data on FFS Medicare beneficiaries and Blue Cross preferred provider organization (PPO) patients¹⁷. It was a resource for longitudinal risk adjusted episode costs and outcomes that are not available in the BMC2 registry. Several publications have utilized data from MVC^{18, 19}. Our particular analysis used administrative data from FFS Medicare patients 65 years or older from 2010 through 2014 to generate estimates of 90-day outcomes and costs for PCI that complemented data from the BMC2. We linked the following CMS files on Medicare beneficiaries to the BMC2 registry: Master Beneficiary Summary File (MBSF), Medicare Provider Analysis Review (MedPAR), Outpatient (OP), Carrier, Home Health Agency (HHA), Hospice, Inpatient, and Skilled Nursing Facility (SNF).

Study Population

We employed iterative deterministic matching to link BMC2 and MVC data to construct a cohort of all Medicare FFS patients (29,839 patients) undergoing PCI from 2010 to 2014 in the State of Michigan (Figure 1). We used patient date of birth, gender, procedure dates, and hospital and physician national provider identifiers as the characteristics for indirect matching. MVC identified PCI index cases at hospitals based on hospital diagnosis related groups (DRGs) on a facility claim or professional claims with CPT codes for PCI. This comprehensive approach allowed us to capture both inpatient and outpatient PCIs. Once the PCI index case was identified, all clinically relevant claims within 90 days of the procedure date were included in the episode. Overall, 84% of MVC administrative data could be linked to a specific patient within the BMC2 registry. We excluded 3,114 PCI episodes for which data were insufficient for calculating appropriate use criteria (AUC) or risk-adjusted mortality. This limitation is owed largely to the fact that 2012 AUC are more comprehensive for classifying acute indications for PCI as compared with stable disease. Future iterations of the AUC may continue to reduce the number of uncertain cases over time but this is far from certain²⁰. We also excluded PCIs performed at 14 hospitals without on-site cardiac surgical backup due to the unique nature of these hospitals and their patients (i.e. all PCIs at these hospitals were for emergent indications during this time period).

Figure 1.

Description of Cohort Construction.

Measures: Appropriateness, Outcomes and Costs

We sought to understand the interrelationships between appropriateness, outcomes, and episode costs based on a novel conceptual framework for healthcare value (Figure 2). In this framework we consider value to be positively related to better outcomes and more appropriate care at a given level of cost. We performed this analysis at the hospital level to further understand the concept of value as it relates to hospital performance assessments.

Figure 2.

Original and proposed conceptual frameworks for value.

For this analysis, we used BMC2 clinical data to calculate appropriateness using the AUC developed by the American College of Cardiology as follows: appropriate (7–9), uncertain (4–6), or inappropriate (1–3)^{21, 22}. We used an algorithm previously developed by members of the Clinical Outcomes and Assessment Program and the Northern New England Cardiovascular Disease Study Group to automate calculation of the AUC based on clinical data elements within the BMC2 registry using the 2012 criteria. Details of this algorithm have been previously published²³. Example scores for 6 different patient scenarios are included in the supplementary appendix.

Outcomes for 90-day readmission were defined as any admission occurring within 90 days after being discharged alive from the index PCI hospitalization or procedure. We defined 90-day mortality as a death that occurred anytime in the 90 days after index PCI. Longitudinal outcomes were assessed using MVC administrative claims data^{24, 25}.

Costs were obtained by aggregating insurance payments across an episode of care, which was defined as the time period encompassing the index PCI case and the subsequent 90 days that included all testing, inpatient, and outpatient services delivered during this window. Individual episode payments were classified into one of four payment categories: 1) index hospitalization, 2) professional services, 3) readmissions and 4) post-acute care (e.g., skilled nursing facility, rehabilitation, home health, emergency department visits). Payments were price standardized to eliminate the impact of hospital specific adjustments (i.e. for geographic location, medical education, health information technology adoption) allowing us to measure resource utilization. Price standardization was performed by assigning a standardized payment to each PCI, reflecting the average Medicare payment in Michigan. Episode payments were risk standardized using the following variables: age, gender, 70 comorbidities, and high spending for the prior six months as assessed by ICD-9 codes at presentation²⁶.

Statistical Analysis

We hypothesized that appropriateness is an important dimension of healthcare value that is independent of risk-standardized outcomes and costs.

We calculated and reported AUC scores at the hospital-level. AUC scores do not require adjustment since they are specifically designed to create scenarios where clinical risk and benefits are equally evaluated by participating clinicians. Risk-standardized outcomes at the hospital-level were obtained using previously developed prediction models where available with patient-level characteristics incorporated as fixed effects in hierarchical mixed effects regression models with a hospital-level random intercept²⁷.

For the 90-day post-PCI mortality, patient-level risk estimates were obtained from the BMC2 random forest mortality model and included as a fixed effect term in a hierarchical mixed effects regression model with a hospital-level random intercept^28–30. For the 90-day readmission measure, initial patient-level risk estimates were obtained using the coefficients previously identified by CMS and similarly included in the hierarchical mixed effects regression model³¹. Risk-standardized measures at the hospital-level were then obtained by multiplying the overall outcome rate observed in the cohort by the ratio of mean predicted to mean expected rate estimates for each site from the fitted mixed effects model, where the expected rate is calculated using the fixed effect fitted coefficients and the predicted rate incorporates both the fixed effect contribution and the posterior mode of the random intercept estimated using an empirical Bayes algorithm. This allowed for adjustment of both risk and reliability at hospitals with lower PCI volumes.

For standardized episode costs, a hierarchical regression model was fitted adjusting for the index hospitalization DRG, along with comorbidity flags generated from ICD-9 diagnostic codes present on the index admission billing record.

We then examined correlations between hospital-level AUC scores with risk-standardized outcomes and costs. Spearman’s correlation coefficients and scatterplots with superimposed locally weighted scatterplot smoothing curves were utilized to assess within site correlation between AUC scores, risk-standardized outcomes, and risk-standardized episode costs. In addition to our overall analyses, we also examined correlations after stratifying patients as to whether or not PCI was performed in the setting of an AMI, as previous work has suggested that 98.6% of acute indications for PCI are rated as appropriate³².

Results

Overall Cohort

From June 30, 2010 to December 31, 2014, 32,953 episodes of PCI were identified in fee-for-service Medicare beneficiaries at 33 hospitals in Michigan. Of these, 29,839 (90.5%) PCIs were successfully mapped to the AUC. Excluded patients (n = 3,114) were more likely to have presented with stable angina (42.7% vs. 9.8%) and a PCI indication of “other” (63.5% vs 4.7%) (i.e., not the defined categories for primary PCI, other ST-elevation AMI presentation or high risk Non-ST elevation AMI or unstable angina). The mean number of PCI’s performed per hospital was 904 and the median number was 696 over the study period. The baseline characteristics of the entire cohort and stratified groups based on the presence of AMI are summarized in Table 1.

Table 1.

Cohort Baseline Characteristics.

		Overall (n = 29,839 )	Non-AMI (n = 18,415)	AMI (n = 11,424)
Demographics
	Age	72.2 ± 9.7	72.10 ± 9.3	72.42 ± 10.4
	Male	17,417 (58.4%)	10,859 (59.0%)	6,558 (57.4%)
	White	26,069 (87.4%)	16,166 (87.8%)	9,903 (86.7%)
	African American	2,970 (10.0%)	1,721 (9.3%)	1,249 (10.9%)
	Other	665 (2.2%)	400 (2.1%)	255 (2.2%)
Comorbidities
	Current/Recent Smoker (w/in 1 year)	5,889 (19.7%)	3,128 (17.0%)	2,761 (24.2%)
	Hypertension	26,879 (90.1%)	17,115 (93.0%)	9,764 (85.5%)
	Dyslipidemia	25,375 (85.1%)	16,617 (90.3%)	8,759 (76.8%)
	Family History of Premature CAD	4,379 (14.7%)	2,887 (15.7%)	1,492 (13.1%)
	Prior MI	10,363 (34.7%)	6,662 (36.2%)	3,701 (32.4%)
	Heart Failure	5,718 (19.2%)	3,642 (19.8%)	2,076 (18.2%)
	PCI	13,886 (46.5%)	9,935 (54.0%)	3,951 (34.6%)
	CABG	6,472 (21.7%)	4,221 (22.9%)	2,251 (19.7%)
	Diabetes Mellitus	12,334 (41.3%)	7,827 (42.5%)	4,507 (39.5%)
CAD Presentation
	Non-STEMI	7,448 (25.0%)	NA	7,448 (65.2%)
	Stable angina	2,930 (9.8%)	2,930 (15.9%)	NA
	ST-Elevation MI (STEMI) or equivalent	3,976 (13.3%)	NA	3,976 (34.8%)
	Unstable angina	14,063 (47.1%)	14,063 (76.4%)	NA
	Other	1,420 (4.7%)	1,420 (4.7%)	NA
Cardiogenic Shock at Start of PCI		503 (1.7%)	33 (0.2%)	470 (4.1%)

Hospital Level Outcomes

At the hospital-level, the overall mean AUC score was 8.4 ± 0.2 (range, 7.97 to 8.79) with the mean proportion of cases identified as appropriate at the hospitals calculated at 91% ± 3.7%. The mean hospital-level risk-standardized 90-day readmission rate (RSRR) was 23.7% ± 3.7% (range 16.1% – 30.5%). Common reasons for readmission within 90-days based on the readmission diagnosis related group included repeat PCI (20.6%), heart failure (11.9%), gastrointestinal hemorrhage with or without complications (5.7%), and AMI (4.9%). The mean hospital-level risk-standardized 90-day mortality rate (RSMR) was 4.3% (range 2.9% – 5.5%). The mean hospital-level risk-standardized episode costs (RSEC) were $26,159 ± 1,074 (range $24,287 – $29,161). Neither hospital-level AUC scores nor proportions of appropriate cases were associated with RSRR, RSMR, or RSEC. (Figures 3, 4, 5). Higher RSRR were strongly associated with higher RSEC (R 0.73, p < 0.0001).

Figure 3.

The chart demonstrates the mean hospital appropriate use criteria (AUC) scores at each of the 33 Michigan hospitals vs. 90-day risk-standardized mortality rates (RSMR).

Figure 4.

The chart demonstrates the mean hospital appropriate use criteria (AUC) scores at each of the 33 Michigan hospitals vs. 90-day risk-standardized episode costs (RSEC).

Figure 5.

The chart demonstrates the mean hospital appropriate use criteria (AUC) scores at each of the 33 Michigan hospitals vs. 90-day risk-standardized readmission rates (RSRR).

Figure 6 is a 3-dimensional plot that displays the interrelationships between mean hospital-level AUC scores, RSMR and RSEC. The plot demonstrates the lack of relationship between hospital-level AUC scores and the other domains of PCI care.

Figure 6 (Central Illustration).

A 3-dimensional plot that displays the interrelationships between mean hospital-level AUC scores, RSMR and RSEC. The three red hospitals highlighted on this figure suggest facilities with poor performances in these categories and the three green dots identify high performers.

Non-AMI Cohort

We stratified our analysis by the presence of AMI given the important role of this diagnosis in driving appropriateness in the AUC. During the study period, 18,415 patients (61.6%) underwent PCI for a non-AMI indication. In this cohort, the mean hospital-level AUC score was 8.1 ± 0.3, mean RSRR was 23.0% ± 2.7 (range 17.5% – 28.5%), RSMR was 1.7% ± 0.3 (range 1.3% – 2.5%), and mean RSEC were $23,918 ± 1,048 (range $22,434 – $25,934). Higher RSRR were associated with higher RSMR (R 0.50; p = 0.0038). Higher RSEC were also associated with higher RSRR (R 0.77, p<0.0001) and RSMR (R 0.51, p=0.0027). Neither hospital AUC scores nor proportions of appropriate cases were associated with RSRR, RSMR, or RSEC.

AMI Cohort

In this subgroup, 11,424 patients underwent PCI for AMI. The mean AUC score at the hospital-level was 8.9 ± 0.1, the RSRR was 24.9% ± 4.2 (range 17.6% – 33.8%), the RSMR was 8.3% ± 0.9 (range 6.2% – 11.0%), and RSEC were $29,726 ± 1,362 (range $27,330 – $33,809). Higher RSEC were associated with higher RSRR (R 0.62, p = 0.0002) but not higher RSMR (R 0.24, p = 0.16). Neither mean hospital-level AUC scores nor proportions of appropriate cases were associated with RSRR, RSMR, or RSEC.

Discussion

In this study, we examined the key relationship of appropriateness with outcomes and costs of PCI. Our most important finding was that average AUC scores did not correlate with either 90-day outcomes or episode costs at the hospital-level. In contrast and not surprisingly, we found that poor outcomes were associated with higher costs. Taken together, these findings suggest that appropriateness is an independent measure of healthcare value for PCI that needs to be considered separately along with outcomes and costs when assessing the overall performance of the procedure. We are unaware of prior work directly examining AUC scores with these other dimensions of value simultaneously.

Value in healthcare was originally defined as achieved outcomes per unit of cost⁸. However, the morbidity, mortality, and cost burden of inappropriate procedures adds no value to health care delivery. Performing a PCI in an asymptomatic patient with a non-obstructive lesion can be done at low cost and mortality, but with no expected clinical benefit to the patient. Traditional outcome measures are increasingly a major focus of quality programs and some may argue that comprehensive outcome measurement may reduce the role of the AUC by encompassing appropriateness. However, we believe that appropriateness remains a distinct concept under most current views that attempts to quantify the potential benefit of a particular procedure and the decision-making that led to care taking place. In the future, this may change if additional outcomes measurement begins to incorporate the reason behind the procedure into its assessments. For example, if we are able to more comprehensively and reliably measure angina in the future then these measurements can inform both the appropriateness of the procedure and assessment of its outcomes.

Although improving appropriateness has been the focus of quality efforts, its relationship with episode costs and outcomes has been understudied. Overall, the appropriate use of PCI has been improving nationally, and our findings are consistent with this trend³³. We found similarly low and declining rates of stable angina consistent with recent contemporary reports. Whether these findings represent improvements in care, changes in documentation, or intentional up-coding of the AUC toward ACS presentations is unknown. Hospital level variations even at high levels of appropriateness suggest significant variations in care delivery. Appropriateness has been associated with better quality of life and freedom from angina in PCI³⁰. Ko and colleagues also found that appropriate use of PCI in patients with stable angina was associated with better outcomes when compared with medical management alone in Ontario. This highlights that improving patient selection in the cases of both underuse and overuse of PCI may improve outcomes¹². Appropriateness has been examined in the setting of procedural complications within the NCDR. In that study, procedural complications did not correlate with appropriateness suggesting that the latter represented a distinct aspect of PCI care³⁴. Like this prior report from the NCDR, we found no relationship between appropriateness and 90-day outcomes of readmission and mortality.

However, our results differ from prior investigations in the VA Healthcare System (8,9). Unlike these studies, we found an association between higher risk-adjusted costs and higher readmissions and mortality. Care delivery at private institutions is subject to significantly different payment structures and incentives when compared with the VA Healthcare System. This may have played a role in this important discrepancy. For example, in the hospitals we studied, poor outcomes due to complications may have led to higher downstream utilization of services and raised costs for payers, in contrast to the VA’s integrated delivery system where these incentives do not exist. Thus, it appears that quality improvement aimed at improving readmissions and longitudinal mortality may have a significant impact on costs in many private hospitals.

Early papers describing value-based healthcare defined it over specific cycles of care (24). Going a step further and adding the dimension of appropriateness to the definition of value identifies downstream episode costs that could be altogether avoided. In our investigation, we chose to define this cycle as 90 days post-procedure because we felt this was a long enough period to capture outcomes and utilization related to the coronary disease presentation treated with PCI. Prior investigations have defined post-PCI cycles of care as short as one month and as long as one year^{9, 10}. These investigations found the majority of PCI cost variation occurs during the index procedure, supporting our decision for using this cycle. Defining the ideal cycle of care in the context of appropriateness has important implications for how best to design bundled payment in PCI and will require additional investigations³⁵. It is important to pursue a strategy that encourages appropriate patient selection, post-procedure care coordination, and judicious post-procedure testing.

What are the possible implications of our findings? One key conclusion is that we need to reconsider PCI value in a multidimensional space that includes all relevant components of value (Figure 6). Including appropriateness along with outcomes and costs allows us to consider the extent to which a PCI was of expected clinical benefit when the decision was made to pursue it. For example, when one examines all three of these dimensions simultaneously the benefits of performing high value PCI becomes more readily apparent. This makes intuitive sense as some hospitals may perform PCI with low mortality and low costs but in less appropriate patients. The three orange hospitals highlighted on this figure (Figure 6) suggest such facilities where specific improvement efforts may be targeted, and the three green hospitals identify high performers where best practices can be investigated. This demonstrates the potential to create a PCI “value index” that may be used to compare the value of PCI across institutions and can also be used to internally monitor quality improvement efforts.

Our study has important limitations. First, it is restricted to fee-for-service Medicare patients in a single state; conclusions about this study population should be extrapolated with caution to national cohorts or other states with different case-mixes of patients. Our results may not reflect the care of patients undergoing PCI by commercial payers both in the elderly (e.g. Medicare Advantage programs) and non-elderly. Second, hospitals in this cohort also have been actively participating in a statewide quality collaborative aimed at improving outcomes for over a decade. This could be a reason for the high overall performance including high rates of appropriateness and low rates of adverse outcomes. A high proportion of the PCI’s in our cohort were performed for acute indications, which contributes to overall high hospital-level AUC scores. This high degree of appropriateness limits our ability to evaluate how this characteristic may interact with outcomes and costs at levels of lower performance. It also highlights the potential need for better and more refined AUC to better explore this dimension of value. Third, important outcomes that are also critical to the clinical evaluation of non-acute PCI were not assessed, such as angina burden, functional status, or health related quality of life. Unfortunately, these outcomes are not collected post-procedurally in our registry and are poorly collected in most registries despite their critical role in assessing healthcare value. Future work will need to incorporate this information. Fourth, our dataset only contains patients who underwent PCI. We do not have an assessment of other patients during this time period who were either managed medically or underwent surgical revascularization. Greater detail on revascularization patterns at hospitals may be useful to assessing healthcare value. Finally, we chose to analyze our cohort at the hospital-level as we were most interested in understanding PCI value as a metric for assessing healthcare value at a facility. Extrapolation of our findings to individual patients is not recommended as our analysis was not designed to make such conclusions.

Despite these issues, we believe our investigation adds significantly to the discussion of healthcare value in PCI. Our findings support the argument that a comprehensive evaluation of healthcare value for PCI must take into account appropriateness. The lack of a relationship between hospital-level appropriateness and outcomes or costs is an important finding which demonstrates the importance of a new multidimensional conceptual framework for value that can potentially serve as the basis for novel metrics moving forward. By factoring appropriateness into the definition of healthcare value, policymakers and payers may better transition to reimbursing high-value care.

Abbreviations:

PCI

Percutaneous Coronary Intervention

CMS

Center for Medicare and Medicaid Services

AMI

Acute Myocardial Infarction

VA

Veterans Affairs

NCDR

Healthcare System, National Cardiovascular Data Registry

FFS

fee-for-service

BMC2

Blue Cross Blue Shield Cardiovascular Consortium

MVC

Michigan Value Collaborative

PPO

preferred provider organization

DRGs

diagnosis related groups

AUC

appropriate use criteria

RSMR

Risk-standardized 90-day mortality rate

RSRR

Risk-standardized 90-day readmission rate

RSEC

Risk-standardized episode costs