Comparison of observed adherence to 4 classes of evidence-based medications after acute myocardial infarction (angiotensin-converting enzymes [ACE] inhibitors or angiotensin receptor blockers [ARBs], β-blockers, statins, and P2Y12 inhibitors). All dual-eligible patients received the LIS, but of the Medicare-only patients, only 11% received the LIS. After statistical adjustment, all differences were significant in comparisons of dual-eligible vs Medicare-only patients without LIS and Medicare-only patients with LIS vs Medicare-only patients without LIS. There were no significant differences between dual-eligible patients and Medicare-only patients with LIS. For all supporting data, see eTable 3 in the Supplement.
eTable 1. 1-year mortality risk by dual eligibility, within subgroups
eTable 2. Post-discharge outcomes by dual eligibility status, among patients discharged to home
eTable 3. Medication adherence in the first year by dual eligibility and LIS status
Customize your JAMA Network experience by selecting one or more topics from the list below.
Doll JA, Hellkamp AS, Goyal A, Sutton NR, Peterson ED, Wang TY. Treatment, Outcomes, and Adherence to Medication Regimens Among Dual Medicare-Medicaid–Eligible Adults With Myocardial Infarction. JAMA Cardiol. 2016;1(7):787–794. doi:10.1001/jamacardio.2016.2724
Copyright 2016 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
How do the outcomes and medication adherence of dual Medicare-Medicaid–eligible older adults compare with those of patients with Medicare coverage only after a myocardial infarction?
In this observational study of 17 419 patients from a national myocardial infarction registry, dual Medicare-Medicaid–eligible patients were less likely to receive invasive therapies during the index hospitalization and had worse short- and long-term outcomes than patients with Medicare coverage alone. These disparities were not driven by medication regimen nonadherence; in fact, rates of adherence at 1 year were greater for the dual-eligible patients.
Despite the additional support provided by Medicaid coverage, dual-eligible older adults are a vulnerable population that may benefit from interventions to optimize post–myocardial infarction outcomes, including efforts to improve use of evidence-based therapies in the hospital and at the time of discharge.
Patients with dual Medicare-Medicaid eligibility have a higher burden of chronic disease conditions and increased health care utilization compared with patients with Medicare coverage alone, but the treatment patterns and outcomes of dual-eligible patients with myocardial infarction (MI) are unknown.
To examine the association of dual-eligible status with clinical outcomes and adherence to medication regimens (hereinafter “medication adherence”) among older adults after MI.
Design, Setting, and Participants
In this retrospective study conducted from February 2015 to April 2016, we linked patients 65 years or older enrolled in a national myocardial infarction registry (the Acute Coronary Treatment Intervention Outcomes Network Registry–Get With the Guidelines [ACTION Registry-GWTG]) from July 1, 2007, to December 31, 2009, to Medicare claims data to obtain 1-year follow-up and medication adherence data. The ACTION Registry-GWTG is the largest quality-improvement registry of patients with MI in the United States. Included patients were all 65 years or older; had Medicare Parts A, B, and D; presented with MI; and survived to hospital discharge.
Dual Medicare and Medicaid eligibility.
Main Outcomes and Measures
Death, readmission, major adverse cardiovascular events (death, recurrent MI, stroke), and medication adherence at 1 year.
Of 17 419 Medicare patients discharged alive after MI, 4674 (27%) were dual eligible. Dual-eligible patients were more likely to be female (64% vs 49%) and nonwhite (29% vs 6%), with a higher prevalence of comorbid conditions and more frequent presentation with non-ST elevation MI (non-STEMI) (75% vs 69%). Dual-eligible patients were less likely to receive primary percutaneous coronary intervention for STEMI (77% vs 81%), revascularization for non-STEMI (58% vs 65%), and prescription of secondary prevention medications at discharge. After multivariable adjustment, dual eligibility status was associated with a higher risk of readmission at 30 days (hazard ratio [HR], 1.16; 95% CI, 1.06-1.26), death at 1 year (HR, 1.24; 95% CI, 1.14-1.36), and major adverse cardiac events at 1 year (HR, 1.21; 95% CI, 1.12-1.31). Dual-eligible patients had higher 1-year adherence to medications prescribed at discharge (HR, 1.55; 95% CI, 1.39-1.74) than Medicare-only patients.
Conclusions and Relevance
Compared with Medicare-only patients, older adults with dual Medicare-Medicaid eligibility presenting with MI have superior rates of medication adherence but higher rates of postdischarge readmission and adverse cardiovascular outcomes.
Nearly 6 million adults 65 years or older in the United States are eligible for both Medicare and Medicaid.1 Patients with dual eligibility often have a higher burden of chronic disease conditions and higher health care utilization than individuals with Medicare coverage alone.2-4 For this reason, the Centers for Medicare and Medicaid Services (CMS) and the Affordable Care Act (ACA) prioritize improved health outcomes and reduced cost of care for dual-eligible patients.2,5,6
It is unknown if disparities exist in the treatment and outcomes of dual-eligible patients with acute myocardial infarction (MI). Poverty is one of the major requirements of Medicaid eligibility, and thus dual-eligible patients are the poorest subset of the Medicare population. Prior studies have shown that low socioeconomic status (SES) is associated with increased incidence of MI,7 lower utilization of cardiovascular procedures,8,9 higher rates of post-MI mortality and rehospitalization,10-12 and lower adherence to medication regimens (hereinafter “medication adherence”).13 However, dual-eligibility status is not just a marker of low SES, since Medicaid coverage provides additional financial support for medical care and prescription drugs that may affect medication adherence and health outcomes.1,14 Whether disparities in post-MI outcomes and medication adherence persist despite this nearly universal insurance coverage is unknown, but if they do, there may be significant policy implications.
Therefore, we sought to characterize dual-eligible older adults by linking the largest MI registry in the United States to Medicare claims data. Our objectives were to (1) describe the clinical characteristics and inpatient treatment patterns of dual-eligible older adults compared with Medicare-only patients with Part A, B, and D coverage; (2) examine the association of dual-eligible status with readmission, mortality, and major adverse cardiac events (MACE); and (3) examine the association of dual eligibility and the Low Income Subsidy (LIS) with postdischarge medication adherence.
The National Cardiovascular Data Registry (NCDR) Acute Coronary Treatment Intervention Outcomes Network Registry–Get With the Guidelines (ACTION Registry-GWTG) is the largest ongoing US registry of patients with MI. Hospitals submit inpatient data including demographics, baseline clinical characteristics, in-hospital treatment and outcomes, and discharge medications using a standardized set of data elements and definitions.15 This registry was either approved by an institutional review board or considered quality assurance data and not subject to institutional review board approval based on individual site determinations. The present analysis was approved by the Duke institutional review board. Patients 65 years or older in the registry were linked to the Medicare fee-for-service database using 5 indirect identifiers in combination (date of birth, sex, hospital ID, date of admission, and date of discharge) to obtain longitudinal outcomes and medication adherence data.16
From July 1, 2007, to December 31, 2009, there were a total of 44 927 admissions of patients 65 years or older with Medicare Part A and B enrollment in the ACTION Registry-GWTG. To ensure a reasonable comparison between dual-eligible patients (all of whom are provided Part D coverage) and Medicare-only patients, we excluded all patients without Part D coverage (n = 23 248). We required at least 3 months of Part D coverage prior to the index hospitalization to accurately characterize post-MI adherence to medications that may have been prescribed prior to the index MI and continued at discharge. We also excluded patients who died during the index hospitalization (n = 1620), were transferred out to another acute care facility (n = 950), were discharged to hospice, comfort care, or against medical advice (n = 808), or were discharged without any evidence-based medications (n = 256). Finally, only 1 admission per patient was analyzed, so 626 nonindex admissions were excluded. Our primary sample included 17 419 patients treated at 361 hospitals.
Dual-eligible status was determined using the Dual Eligible Status Code from the Beneficiary Summary File. Patient LIS status was obtained from the Part D Cost Share Group.17
Detailed definitions for data elements of the ACTION Registry-GWTG are available at https://www.ncdr.com/webncdr/action/home/datacollection. Comorbidities for the Charlson index were generated from ACTION data elements when available; otherwise, they were obtained from CMS Parts A and B claims from the index admission and any admission within the prior 6 months.18 Community socioeconomic variables were drawn from the Area Health Resource File, which is collected at the county level, and matched to patients by zip code. For each patient, we calculated a SES score that incorporated information from income, education level, and employment rate.19
We examined 3 clinical outcomes after discharge from the index MI: (1) all-cause readmission at 30 days and 1 year, (2) death at 1 year, and (3) MACE (defined as the composite of death, readmission for MI, or stroke) at 1 year. Readmission was defined as the first hospitalization after index discharge. Admissions for possible staged revascularization procedures (percutaneous coronary intervention [PCI] or coronary artery bypass grafting [CABG]) that occurred within 30 days of discharge were not counted, unless the principal discharge diagnoses included indications not consistent with elective readmission (heart failure, MI, unstable angina, arrhythmia, and cardiac arrest). Mortality was ascertained from the Medicare denominator files. Readmissions for MI and stroke were ascertained from the primary diagnosis code for subsequent hospitalizations in the linked Medicare claims file. Myocardial infarction and stroke were defined using International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes 410.x1 (MI), and 430.x, 431.x, 432.x, 433.x1, 434.x1, 436.x, 437.1, 437.9, 997.02 (stroke).
Medication adherence was assessed using Part D prescription filling data as the proportion of days covered of medications prescribed at discharge. An 80% proportion of days covered during the first year after discharge was considered adherent,20 after accounting for any rehospitalization days. We assessed 4 medication classes that are indicated for post-MI treatment: angiotensin-converting enzymes (ACE) inhibitors or angiotensin receptor blockers (ARBs), β-blockers, statins, and P2Y12 inhibitors. Aspirin was not included because it is often obtained without a prescription.
Patients were grouped by dual-eligibility status. Baseline characteristics and treatment variables were compared using Pearson χ2 tests for all categorical variables and Wilcoxon rank sum tests for all continuous variables.
We assessed postdischarge outcomes using the cumulative incidence (readmissions) or the Kaplan-Meier method (death, MACE). Multivariable adjustment was performed using Cox proportional hazards models with robust sandwich covariance estimators to account for patient clustering within hospitals.
Covariates in the model were adapted from a previously published long-term mortality risk model21 with additional variables selected a priori as potentially unbalanced between the groups. Covariates included age, sex, race, medical history (current or recent smoker, dialysis status, prior congestive heart failure, prior PCI, prior CABG, prior stroke, hypertension, dyslipidemia, diabetes, peripheral artery disease, prior MI), presentation (STEMI vs non-STEMI [NSTEMI]), signs of congestive heart failure on arrival, shock on arrival, heart rate on admission, systolic blood pressure on admission, BMI, home medications (aspirin, P2Y12 inhibitor, warfarin, β-blocker, ACE inhibitor or ARB, aldosterone blocker, statin), in-hospital treatments (diagnostic catheterization, PCI, drug-eluting stent, CABG), laboratory values and measurements (left ventricular ejection fraction [LVEF], initial creatinine level, initial hemoglobin level, peak troponin level), discharge medications (ACE inhibitor or ARB, β-blocker, statin, P2Y12 inhibitor), site characteristics (hospital region, type of hospital, teaching hospital, total hospital beds, rate of defect-free discharge care), and median household income in patient’s zip code.
Missing values of continuous variables were imputed using the sex- and/or STEMI -specific median for patient-level covariates and the overall median for hospital- or zip code–level covariates. The percentage of missing data was less than 2% for all variables with the exception of LVEF (11%). For categorical variables, missing demographic and medical history variables were imputed to the most common value; missing medications and procedures were imputed as “no.” Continuous variables were transformed using restricted cubic splines where necessary to accommodate nonlinearity. All models began at the time of discharge. For assessment of mortality by subgroups including age, sex, race, and MI type, a separate model was used for each subgroup with the addition of interaction terms.
Since a substantial portion of dual-eligible patients reside in long-term care facilities, which could bias our analysis of postdischarge outcomes, we also performed a sensitivity analysis by excluding all patients who were not discharged to home from the index admission (n = 3472). This sample of home-dwelling older adults included 13 947 patients.
Analysis of medication adherence was limited to patients discharged to home and surviving at 1 year after discharge with Part D prescription filling data available for the first year. A multivariable logistic regression model was used with generalized estimating equations to account for clustering of patients within hospitals and to adjust for the listed variables. The composite “all medication” adherence outcome required at least 80% adherence to all of whichever of the 4 medication classes that the patient was prescribed at discharge. To allow for differences in the likelihood of adhering to the different medications, as well as the possibility that there may be an interaction in the likelihood of adhering to some medications (ie, some combinations may be more or less difficult to adhere to), an additional term was included in the model to describe the patient’s set of medications.
P < .05 was considered statistically significant for all tests. Analyses were performed using SAS software, version 9.3 (SAS Institute Inc).
Of 17 419 Medicare patients with Part A, B, and D coverage discharged alive after MI, 4674 (27%) were dual eligible, and 12 745 (73%) had Medicare coverage only. All dual-eligible patients received the LIS, but of the Medicare-only patients, only 1370 (11%) received the LIS. Dual-eligible patients were more likely to be female (64% vs 49%) and nonwhite (29% vs 6%) and to have a greater prevalence of comorbid conditions such as hypertension, diabetes, and kidney disease. Indicators of SES, as determined by patient zip code, were lower for dual-eligible patients. A detailed comparison of patient characteristics is provided in Table 1.
In-hospital use of evidenced-based therapies was lower among dual-eligible patients, including less frequent revascularization for NSTEMI (58% vs 65%; P < .001) and primary PCI for STEMI (77% vs 81%; P < .001) in the absence of documented contraindication. Drug-eluting stents were also used less frequently for dual-eligible patients (53% vs 59%; P < .001). At discharge, dual-eligible patients were slightly less likely to be prescribed statin (86% vs 88%; P < .001) or P2Y12 inhibitor therapy (71% vs 74%; P < .001) in the absence of clinical contraindications to their use, although there was no difference in P2Y12 inhibitor use among patients receiving PCI (96% vs 96%).
Observed rates of readmission, death, and MACE were higher among dual-eligible patients compared with Medicare-only patients (Table 2 and Figure 1). Nearly two-thirds of dual-eligible patients had a readmission event within 1 year, and nearly one-third had a MACE event. After adjustment for differences in patient characteristics and in-hospital treatment, dual-eligible status remained associated with a higher risk for all outcomes. To exclude patients residing in long-term care facilities, we performed a sensitivity analysis examining only patients discharged to home after the index hospitalization (n = 13 947, 23.6% of whom were dual eligible). Rates of each of the outcomes were lower than in the full population, but differences between dual-eligible and Medicare-only patients persisted both before and after adjustment, as detailed in eTable 1 and eTable 2 in the Supplement.
Differences between dual-eligible and Medicare-only patients in 1-year mortality were present across subgroups of age (≤75 vs >75 years), sex, race (nonwhite vs white), and MI type (NSTEMI vs STEMI). The higher risk of mortality in dual-eligible patients was more pronounced in patients who were younger, white, or who presented with STEMI.
Among patients surviving 1 year after discharge and enrolled in the Part D prescription coverage plan for the full year (n = 11 481, 22.5% of whom were dual eligible), only 50% to 60% were adherent to each of the 4 classes of guideline-recommended medications prescribed at discharge (Table 3 and Figure 2). Dual-eligible patients were more likely than Medicare-only patients to be adherent to each medication, before and after adjustment for clinical and treatment variables. In addition, adherence to all medications (≥80% adherence to all of whichever of the 4 medication classes that the patient was prescribed) was greater among dual-eligible patients (hazard ratio [HR], 1.55; 95% CI, 1.39-1.74).
To assess the potential contribution of the LIS in determining postdischarge adherence, we then subdivided the Medicare-only population into those who received the LIS and those who did not and compared these groups to dual-eligible patients. The LIS group was associated with higher rates of adherence compared with the Medicare-only non-LIS population, whether the LIS was obtained via Medicaid coverage (dual eligible) or via low income but not meeting criteria for Medicaid (Medicare-only plus LIS).
Compared with Medicare-only patients, patients with dual Medicare and Medicaid eligibility had worse short- and long-term outcomes after MI despite the additional financial support provided by Medicaid. While prior studies have shown a similar association between low SES and worse outcomes,10-12 the present study is novel in demonstrating higher rates of postdischarge medication adherence among patients with dual eligibility, presumably owing to the lower copayment burden in this population. In addition, we observed disparities in in-hospital treatment that may be modifiable, such as lower rates of reperfusion for STEMI, revascularization for NSTEMI, drug-eluting stent use, and prescription of evidence-based medications at discharge.
Our study indicated that dual-eligible patients hospitalized for MI are disproportionately female and nonwhite and have higher rates of comorbid conditions. This is consistent with the overall dual-eligible population.2,22 Dual-eligible patients are also, by definition, of lower SES than Medicare-only older adults. Since sex, race, and SES have all previously been linked to lower utilization of cardiovascular procedures,9,23 it is not surprising that dual-eligible patients in our study had lower rates of reperfusion for STEMI, revascularization for NSTEMI, and drug-eluting stent use even in the absence of contraindications. Prior studies demonstrating lower procedural rates among patients with lower SES have postulated a variety of causes, including lack of access to care, treatment at lower-volume hospitals, and discrimination on the basis of income, insurance status, race, ethnicity, or sex.8,10,24 Our study is unique in examining a patient population that not only has inpatient and outpatient insurance coverage, but also part D prescription medication coverage. The existence of treatment disparities despite this insurance coverage may indicate that the disparities are not based on the ability to pay for medical care, and are instead related to race, sex, or other unmeasured factors. Alternatively, clinicians may not be aware of the support provided to dual-eligible patients or might believe that other barriers to adherence exist. If clinicians anticipate poor adherence to prescribed medications such as dual-antiplatelet therapy, fewer evidence-based treatments may be provided.
Though the high comorbidity burden of dual-eligible patients has been well described in the literature,22 to our knowledge there are no prior studies examining outcomes after an acute condition such as MI. We found that dual-eligible patients had markedly increased risk of death and MACE at 1 year. Differences in demographics and comorbidities may partially explain these worse outcomes, but disparities persisted even after extensive statistical adjustment. Multiple prior studies have demonstrated disparities in MI outcomes related to SES,11,25 even in countries with universal health care26 and in the US Medicare population.12 Dual-eligibility status is a marker of low income and possibly other medical and social factors that may affect health outcomes. Our study indicates that the reduced patient financial burden associated with additional Medicaid coverage does not eliminate these disparities.
Dual-eligible patients have higher rates of physical disability and need for long-term care than Medicare-only patients, for which we could not fully adjust. To focus only on a home-dwelling population, we performed a sensitivity analysis excluding all patients not discharged to home from the index hospitalization. Adjusted risks of death and MACE remained significantly higher for this group.
Studies using administrative data have previously identified that dual-eligible patients are at high risk for admission overall and for 30-day readmission after hospitalization,4,27 but readmission rates after MI have not been reported. After adjustment, dual-eligible patients had a 16% increased risk of readmission at 30 days and 19% at 1 year compared with Medicare-only patients.
Hospitals wishing to reduce MI readmissions should target dual-eligible patients for quality improvement interventions. Use of drug-eluting stents and PCI are associated with reduced readmission rates,28,29 and these therapies were underused among dual-eligible patients in this study. Patients with lower quality of life and depression also are more likely to be readmitted.29 Dual-eligible patients have higher rates of physical disability and mental illness,22 so readmission reduction efforts may require longitudinal and multifactorial interventions.
Adherence to medications after MI affects survival.30-32 Lower-income and older patients are at greater risk of cost-related medication nonadherence,13,33 so poor medication adherence is a potential cause of outcome disparities for dual-eligible patients. Instead, our analysis indicated that dual-eligible patients were more likely to adhere to secondary prevention medications than Medicare-only patients, though overall adherence to all prescribed cardiovascular medications was low for both groups. The LIS, a benefit for some Medicare Part D enrollees that reduces or eliminates drug copayments, may explain this improved adherence among dual-eligible patients. All dual-eligible patients receive the LIS, as do Medicare beneficiaries with income between 135% and 150% of the federal poverty line. Studies of the effect of the LIS generally show lower out-of-pocket costs and improved long-term adherence.14,34,35 In our study, adherence was similar among dual-eligible patients and Medicare-only LIS recipients, but greater for both groups compared with Medicare-only patients. This indicates that the improved adherence among dual-eligible patients is likely driven by the LIS.
Despite these data, there may be a perception among clinicians that dual-eligible patients are less likely to adhere to medications owing to cost. This may contribute to lower usage rates of revascularization and drug-eluting stents, owing to concerns about discontinuation of dual-antiplatelet therapy. Our analysis indicates that these concerns should not be limited to the dual-eligible population. Nonadherence is common for all patients, and interventions to improve adherence should be applied uniformly.
Our study has several potential limitations. It is retrospective and observational and thus cannot determine if dual eligibility, the LIS, or low SES are causally related to the observed outcomes. Unmeasured confounders cannot be excluded. In particular, we are unable to assess postdischarge behaviors, with the exception of medication adherence. All patients were treated at hospitals participating in a national quality-improvement registry. This may lead to more consistent and homogeneous application of evidence-based therapies, which may therefore underestimate treatment and outcomes disparities at nonparticipating hospitals. Adherence was assessed using prescription fill data. Though we did not have access to more direct measures of adherence such as pill counts, prescription fill methods have good correlation to other measures of adherence and have been associated with clinical outcomes in other studies.30,36 Our analysis was limited to the Medicare population and may not be applicable to other low-income populations. Though poverty is a primary requirement for Medicaid assistance, specific eligibility criteria vary by state and were not assessed in this study. Similarly, benefits provided to dual-eligible patients vary by state and income level, but our analysis was unable to characterize the specific financial support received by each patient. Finally, the effect of Medicaid expansion due to the ACA was not assessed by this study, and more inclusive eligibility criteria may substantively alter the dual-eligible population.
Older adults with dual Medicare-Medicaid eligibility were less likely to receive primary PCI for STEMI or revascularization for NSTEMI during the index MI hospitalization and had worse short- and long-term outcomes than patients with Medicare coverage alone. These disparities were not driven by medication nonadherence; in fact, rates of adherence at 1 year were greater among dual-eligible patients than they were among Medicare-only patients. Despite the additional support provided by Medicaid coverage, patients of low SES are a vulnerable population that may benefit from interventions to optimize post-MI outcomes, including efforts to improve use of evidence-based therapies.
Corresponding Author: Jacob A. Doll, MD, Duke Clinical Research Institute, Duke University School of Medicine, 2400 Pratt St, Durham, NC 27705 (email@example.com).
Accepted for Publication: June 27, 2016.
Published Online: August 17, 2016. doi:10.1001/jamacardio.2016.2724.
Author Contributions: Drs Doll and Wang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Doll, Wang.
Acquisition, analysis, or interpretation of data: Doll, Hellkamp, Goyal, Sutton, Peterson, Wang.
Drafting of the manuscript: Doll, Sutton.
Critical revision of the manuscript for important intellectual content: Doll, Hellkamp, Goyal, Sutton, Peterson, Wang.
Statistical analysis: Doll, Hellkamp.
Obtained funding: Peterson, Wang.
Administrative, technical, or material support: Peterson.
Study supervision: Doll, Wang.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Peterson reports grants and personal fees from Janssen and personal fees from Boehringer Ingelheim, Merck, Sanofi, Bayer, and Astra Zeneca outside the scope of the submitted work. Dr Wang reports research grant support from Eli Lilly, Daiichi Sankyo, Astra Zeneca, Bristol Myers Squibb, Boston Scientific, Gilead, Glaxo Smith Kline, and Regeneron; consulting services from Eli Lilly, Astra Zeneca, and Premier. No other disclosures are reported.
Funding/Support: This project was supported by grant number U19HS021092 from the Agency for Healthcare Research and Quality (AHRQ) (Dr Doll).
Role of the Funder/Sponsor: The AHRQ had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Create a personal account or sign in to: