Author + information
- Received February 11, 2016
- Revision received May 2, 2016
- Accepted June 9, 2016
- Published online August 23, 2016.
- Sameer Bansilal, MD, MSa,
- Jose Maria Castellano, MD, PhDa,b,c,
- Ester Garrido, MPHa,d,
- Henry G. Wei, MDe,
- Allison Freeman, MSe,
- Claire Spettell, PhDe,
- Fernando Garcia-Alonso, MD, PhDd,
- Irene Lizano, PhDd,
- Renee J.G. Arnold, PharmDa,
- Jay Rajda, MD, MBAe,
- Gregory Steinberg, MBChBe and
- Valentin Fuster, MD, PhDa,b,∗ ()
- aIcahn School of Medicine at Mount Sinai School, New York, New York
- bNational Centre for Cardiovascular Research, Madrid, Spain
- cHospital Universitario Monteprincipe, Grupo HM, Madrid, Spain
- dFerrer, Barcelona, Spain
- eAetna Inc., Hartford, Connecticut
- ↵∗Reprint requests and correspondence:
Dr. Valentin Fuster, Cardiovascular Institute, Mount Sinai Heart, One Gustave L. Levy Place, Box 1030, New York, New York 10029.
Background Although guideline-recommended therapies reduce major adverse cardiovascular events (MACE) in patients after myocardial infarction (MI) or those with atherosclerotic disease (ATH), adherence is poor.
Objectives The goal of this study was to determine the association between medication adherence levels and long-term MACE in these patients.
Methods We queried the claims database of a large health insurer for patients hospitalized for MI or with ATH. The primary outcome measure was a composite of all-cause death, MI, stroke, or coronary revascularization. Using proportion of days covered for statins and angiotensin-converting enzyme inhibitors, patients were stratified as fully adherent (≥80%), partially adherent (≥40% to ≤79%), or nonadherent (<40%). Per-patient annual direct medical (ADM) costs were estimated by using unit costs from 2 national files.
Results Data were analyzed for 4,015 post-MI patients and 12,976 patients with ATH. In the post-MI cohort, the fully adherent group had a significantly lower rate of MACE than the nonadherent (18.9% vs. 26.3%; hazard ratio [HR]: 0.73; p = 0.0004) and partially adherent (18.9% vs. 24.7%; HR: 0.81; p = 0.02) groups at 2 years. The fully adherent group had reduced per-patient ADM costs for MI hospitalizations of $369 and $440 compared with the partially adherent and nonadherent groups, respectively. In the ATH cohort, the fully adherent group had a significantly lower rate of MACE than the nonadherent (8.42% vs. 17.17%; HR: 0.56; p < 0.0001) and the partially adherent (8.42% vs. 12.18%; HR: 0.76; p < 0.0001) groups at 2 years. The fully adherent group had reduced per-patient ADM costs for MI hospitalizations of $371 and $907 compared with the partially adherent and nonadherent groups.
Conclusions Full adherence to guideline-recommended therapies was associated with a lower rate of MACE and cost savings, with a threshold effect at >80% adherence in the post-MI population; at least a 40% level of long-term adherence needs to be maintained to continue to accrue benefit. Novel approaches to improve adherence may significantly reduce cardiovascular events.
It is estimated that there are 83.6 million patients in the United States with established atherosclerotic (coronary, cerebrovascular, and peripheral artery) disease (1). Furthermore, approximately 735,000 Americans experience a myocardial infarction (MI) every year, and 210,000 have a recurrent event (2). The use of evidence-based and guideline-recommended medications for the secondary prevention of cardiovascular (CV) disease was estimated to be responsible for one-half of the overall 50% reduction in mortality from CV disease observed over the past 2 decades (3). Amazingly, this substantial reduction in mortality has been achieved despite patients not receiving proven medical therapies to the fullest extent, with nearly one-half of the patients being nonadherent with their prescribed regimen 2 years after experiencing a CV event (4). In the longer term, for patients with documented atherosclerosis, medication adherence is <50% (5). The potential for further improvement in CV outcomes through improved medication adherence is therefore a tantalizing prospect. Improved outcomes will also have a direct impact on the immense financial burden associated with CV disease.
Although nonadherence with evidence-based secondary prevention medications is common in patients with established atherosclerotic disease, studies on long-term outcomes are limited. The goal of the present paper was to study the association between levels of medication adherence and long-term major adverse cardiovascular events (MACE), resource utilization, and cost differences in an acute post-MI cohort and in a complementary chronic atherosclerosis cohort using a large U.S.-based health insurance database.
Materials and Methods
This non-concurrent cohort study was conducted by using 2010 to 2013 medical and pharmaceutical claims obtained from Aetna Commercial and Medicare Advantage population databases. These databases consist of enrollment records from a large, geographically diverse, insured population. These records were linked, allowing for comprehensive tracking of patients’ use of health care resources and clinical outcomes over time and across providers. Enrollment files contain individualized demographic and health insurance plan characteristics, such as age, sex, geographic region, type of health insurance plan, and enrollment status. Medical claims included detailed information about inpatient and outpatient care, including date and place of service diagnosis codes according to the International Classification of Diseases (ICD)-Ninth Revision-Clinical Modification, and procedure codes, such as those from Current Procedural Terminology, Fourth Edition. Pharmacy claims files included information on National Drug Code, dispense date, quantity dispensed and supplied, and copayment amounts. In addition, Symmetry episode risk group scores (Ingenix, Inc., Eden Prairie, Minnesota; 2008 proprietary risk prediction algorithm), publicly available data from the U.S. Census 2010 file, and self-reported multisource race/ethnicity data were included in the analysis. For the cost data, 2 separate, federally funded and nationally representative cost databases were analyzed: the Healthcare Cost and Utilization Project and the Medicare files (Medicare Physician Fee Schedule and Medicare Outpatient Prospective Payment System). We also performed a sensitivity analysis of the results by using simple average unit costs for stroke, MI, atherosclerosis or angina, revascularization procedures, and CV test categories.
The post-MI cohort included adults who initiated both statin and angiotensin-converting enzyme (ACE) inhibitor medications after a hospitalization discharge for MI according to ICD codes 410.x (excluding codes when the fifth digit was 2) and 411.1 with a length of stay of >2 days, between January 1, 2010, and February 28, 2013. Patients were included in the cohort if they had continuous eligibility for both medical and prescription drug benefits from Aetna during 6 months before and after the MI. The discharge date of the MI hospitalization was identified as the index date.
The atherosclerosis cohort included adults who initiated both statin and ACE inhibitor medications and also had 2 coronary, cerebrovascular, or peripheral artery disease ICD codes (claims) within 1 category or a revascularization code between January 1, 2010, and December 31, 2010. Patients were included in the cohort if they had continuous eligibility for both medical and prescription drug benefits from Aetna from January 1, 2010, to December 31, 2011. The first statin and ACE inhibitor refill during 2010 was identified as the index date. For both cohorts, an “out-of-hospital” first medication fill during the period of interest was considered as an inclusion requirement.
Members were excluded from both cohorts if they met any of the following criteria: pregnancy; patients with diagnosis codes indicating psychoses, dementia, bipolar disorder, major depressive disorder (severe with psychotic behaviors), or alcohol/substance abuse; and patients living in a nursing home, hospice, or respite care. Patients who had a refill for angiotensin-receptor blocker (ARB) medication in 2010 were also excluded (to avoid a potential inclusion bias with patients who could have potentially switched to an ARB due to ACE inhibitor intolerance).
Follow-up was through December 31, 2013, and was truncated at the following: 1) disenrollment from the “health care benefits plan” (equivalent to lost to follow-up); 2) death; or 3) end of the follow-up period.
We used clinical judgment to identify the initial set of candidate variables. Independent variables that met the p < 0.05 threshold of significance used for adjustment in the study were as follows: sociodemographic characteristics; pharmacy copayment and medication use; and comorbidities and comorbidity scores. The use of preventive services is reportedly associated with increased medication adherence and has been introduced in different studies as a proxy for the “healthy adherer” effect. As an example, a primary care physician visit was used for an influenza vaccination as a healthy adherer proxy variable. The specific variables related to each category are shown in Online Table 1 (6).
In the present study, medication adherence was estimated by calculating the proportion of days covered (PDC) for both statins and ACE inhibitors (class of drugs) after the index date. The PDC calculation is made on the basis of the fill dates and days supplied for each fill of a prescription and have been extensively validated as a measure of medication adherence in CV medicine. The denominator refers to the number of days between the first fill and the end of the follow-up period being measured; the numerator is the number of days covered by the prescription fills during the denominator period. For refills identified near the end of the observation period, only the days supplied between that refill date and the end of the follow-up period were counted. On the basis of previous evidence regarding standardized methods for measuring adherence to multiple medications, patients were considered to be adherent if they were refilling both ACE inhibitor and statin prescriptions concurrently (7). The medication adherence assessment period was 6 months for the post-MI cohort and 12 months for the atherosclerosis cohort. Patients were categorized into 1 of 3 groups on the basis of their PDCs by using standard thresholds: ≥80% (fully adherent); 40% to 79% (partially adherent); and <40% (nonadherent) (8,9).
The primary outcome measure was defined as MACE, which included all-cause mortality or hospitalization for nonfatal MI; stroke; or coronary revascularization.
Secondary outcomes included hospitalizations for other CV atherosclerosis-related or angina diagnoses; emergency department (ED) visits; ED visits for cardiac atherosclerosis-related causes; the total number of outpatient visits to a cardiologist and those with CV testing, such as stress tests, echocardiograms, and diagnostic coronary angiograms; and cost differences between adherence groups for each outcome.
All outcomes were assessed by using ICD or Current Procedural Terminology, Fourth Edition, codes. Details regarding codes used are shown in Online Appendix.
Descriptive analyses were conducted to compare sociodemographic characteristics, comorbidities, concomitant medication use, and copayments between adherence exposure groups. Analysis of variance was used to identify statistically significant differences in continuous variables, and Pearson’s chi-square test was used for categorical variables.
Cumulative MACE incidence functions for the 3 PDC categories were compared by using Cox proportional hazards regression. Incidence function was censored if the subject’s benefit eligibility terminated or if he or she reached the end of the study period. A set of independent variables, including demographic characteristics, comorbid conditions, concomitant medication use and copayment, and preventive service use, were assessed in the model and were included when they remained significant at a 0.05 significance level. To test the robustness of the findings, a sensitivity analysis was also performed excluding hospitalizations for MACE that occurred during the adherence assessment period in the atherosclerosis cohort. The Akaike information criterion was used to assess the model fit. Hazard ratios (HRs) and 95% confidence intervals (CIs) are reported.
The number of total hospitalizations for MACE, MIs, strokes, hospitalizations for other CV atherosclerosis or angina diagnoses, coronary revascularization procedures, all-cause ED visits, cardiac ED visits, outpatient visits to a cardiologist, and those with CV testing were compared by using negative binomial regression and logistic regression models. These analyses were similarly adjusted as mentioned previously. An offset variable indicating each subject’s time in the study was included to normalize for varying eligibility time frames. Pearson chi-square statistics and the Akaike information criterion were used to assess the goodness of fit of the models. Event rates per 100 person-years, adjusted ratios, and 95% CIs are reported.
For the cost data, a third-party payer economic perspective was adopted. The cost estimates, on the basis of 2012 data, were converted to 2015 U.S. dollars on the basis of the Consumer Price Index for hospital inpatient services. Per-person cost differences between adherence groups for each outcome were estimated by multiplying each outcome event rate by each unit cost–weighted average or simple average for ED visits and outpatient visits. Distribution of medical claims codes from the Aetna Commercial and Medicare Advantage population databases provided information to weight unit costs from national databases. Weighted unit costs for these categories were derived considering the different distribution of diagnostic and procedure codes in each cohort, thereby culminating in different costs for the same outcome. The Healthcare Cost and Utilization Project provided information on costs of inpatient services (stroke, MI, CV atherosclerosis, or angina) and procedure codes (inpatient revascularization procedures). The Medicare Outpatient Prospective Payment System database provided 2015 outpatient services geometric mean costs according to Current Procedural Terminology, Fourth Edition, codes (outpatient revascularization procedures and CV tests). Costs for outpatient and inpatient physician services were estimated on the basis of the Medicare Physician Fee Schedule.
Statistical analyses were performed by using SAS version 9.2 (SAS Institute, Inc., Cary, North Carolina). The Sterling Institutional Review Board (registration number IRB00001790) approved this study.
In the post-MI cohort, 1,031 (26%) patients were classified as nonadherent, 1,263 (31%) as partially adherent, and 1,721 (43%) as fully adherent. For the atherosclerosis cohort, 3,571 (28%) patients were classified as nonadherent, 4,941 (38%) as partially adherent, and 4,464 (34%) as fully adherent.
The baseline characteristics of both study cohorts are shown in Table 1. Interestingly, the partially adherent post-MI cohort had the highest rates of diabetes, hypertension, hyperlipidemia, peripheral disease, and chronic renal failure among the 3 groups, translating into a higher Charlson comorbidity score. The average PDC for both medications together according to adherence group was approximately 93%, 62%, and 21%, respectively, in the post-MI cohort.
In the atherosclerosis cohort, there was an increase in the prevalence of depression, diabetes, chronic renal failure, and Charlson comorbidity score with decreasing levels of adherence. In the atherosclerosis cohort, average PDC according to adherence group was 90%, 62%, and 19%, respectively, for both classes of drugs.
Major clinical outcomes
In the post-MI cohort, the fully adherent group had a significantly lower risk of experiencing MACE than the nonadherent group (18.9% vs. 26.3%; HR, 0.73; 95% CI: 0.61 to 0.87; p = 0.0004) and the partially adherent group (18.9% vs. 24.7%; HR: 0.81; 95% CI: 0.69 to 0.96; p = 0.02). No statistical difference was observed between the nonadherent and partially adherent groups (p = 0.22) (Figure 2, Table 2).
In the atherosclerosis cohort, the fully adherent group had a significantly lower rate of MACE than the nonadherent (8.42% vs. 17.17%; HR: 0.56; p < 0.0001) and the partially adherent (8.42% vs. 12.18%; HR: 0.76; p < 0.0001) groups at 2 years, and the partially adherent group also had a significantly lower risk than the nonadherent group (HR: 0.73; 95% CI: 0.67 to 0.80; p < 0.0001). The sensitivity analysis of the atherosclerosis cohort showed that the difference between the fully adherent (HR: 0.73; 95% CI: 0.63 to 0.84; p < 0.0001) and partially adherent (HR: 0.80; 95% CI: 0.70 to 0.92; p = 0.0021) groups compared with the nonadherent group persisted; however, the difference between the fully adherent and partially adherent cohorts was attenuated to a statistically insignificant level (HR: 0.91; 95% CI: 0.79 to 1.04; p = 0.18) (Figure 3, Table 2).
In the post-MI cohort, the fully adherent cohort exhibited a significant reduction in the rate of hospitalizations of the composite outcome compared with the nonadherent group (adjusted ratio: 0.72; 95% CI: 0.62 to 0.85; p = 0.0001) and the partially adherent group (adjusted ratio: 0.81; 95% CI: 0.69 to 0.94; p = 0.007). Although the partially adherent group reported a reduction in hospitalizations compared with the nonadherent group, this finding was not statistically significant (adjusted ratio: 0.90; 95% CI: 0.76 to 1.05; p = 0.18).
In the atherosclerosis cohort, the rate of hospitalizations for this outcome exhibited a different pattern compared with the post-MI cohort in terms of significantly fewer hospitalizations in the fully adherent group compared with the nonadherent group, and in the partially adherent group compared with the nonadherent group. Table 3 provides further details regarding the impact of levels of adherence on various combinations of the primary outcome measure in both cohorts.
No significant differences were found in the rate of hospitalizations due to CV atherosclerosis or angina in the post-MI cohort (Table 4). In the atherosclerosis cohort, the fully adherent cohort exhibited a significant reduction in the rate of hospitalizations of the composite outcome compared with the nonadherent group (adjusted ratio: 0.52; 95% CI: 0.46 to 0.58; p < 0.0001) and the partially adherent group (adjusted ratio: 0.71; 95% CI: 0.63 to 0.79; p < 0.0001). In addition, the partially adherent group had a significantly lower reduction than the nonadherent group (adjusted ratio: 0.73; 95% CI: 0.66 to 0.81; p < 0.0001).
A significant reduction was noted in all-cause ED visits with increasing levels of adherence in the post-MI cohort and the atherosclerosis cohort. There was no significant difference in the rates of outpatient cardiac visits or CV testing between any pairwise comparisons among the 3 groups in any of the study cohorts (Table 4).
The post-MI estimated weighted unit costs were $14,874 for stroke hospitalizations, $17,593 for MI hospitalizations, $18,853 for angina and CV atherosclerosis, $23,439 for revascularization procedures (considering both outpatient and inpatient), and $1,120 for CV tests. Simple average unit costs were $14,739 for stroke hospitalizations, $22,122 for MI hospitalizations, $19,490 for angina and CV atherosclerosis, $24,433 for revascularization procedures, and $2,145 for CV tests.
For the atherosclerosis cohort, the estimated weighted unit costs were $12,971 for stroke hospitalizations, $16,501 for MI hospitalizations, $20,612 for angina and CV atherosclerosis, $22,186 for revascularization procedures, and $1,142 for CV tests. Simple average unit costs were $17,811 for stroke hospitalizations, $21,266 for MI hospitalizations, $18,474 for angina and CV atherosclerosis, $24,991 for revascularization procedures, and $2,108 for CV tests. The unit costs for ED visits (CV and all-cause) and for outpatient visits were $492 and $215, respectively, for both cohorts.
In the post-MI cohort, full adherence to statins and ACE inhibitors was associated with reduced per-patient annual direct medical costs associated with hospitalizations for MI of $369 and $440 and for revascularization procedures of $539 and $844 compared with partial and nonadherence, respectively. The sensitivity analysis found per-patient annual direct medical cost differences of $465 and $553 for MI, and $562 and $880 for revascularization procedures, compared with partial and nonadherence.
In the atherosclerosis cohort, full adherence was related to a reduced per-patient annual direct medical cost associated with hospitalization for MI of $116 and $215 and for revascularizations of $288 and $799 compared with partial and nonadherence, respectively. The reduction in hospitalizations due to CV atherosclerosis or angina in this cohort was estimated at $371 and $907 compared with partial and nonadherence. The sensitivity analysis found per-patient annual direct medical cost differences of $149 and $276 for MI, $325 and $900 for revascularization procedures, and $333 and $813 for angina or atherosclerosis compared with partial and nonadherence. Table 5 provides further details regarding the cost outcomes.
We evaluated the impact of adherence levels in post-MI and atherosclerosis patients enrolled in a large U.S. insurer plan who had continuous health insurance, including prescription drug coverage. Four major conclusions were made from these results. First, only 43% of patients post-MI were fully adherent to guideline-indicated therapy (Central Illustration). In the longer term, only 34% were fully adherent. Second, being fully adherent reduced the risk of MACE by >25% compared with nonadherence and by at least 20% compared with partial adherence. Third, to accrue benefit required a very high level of adherence (>80%) in the acute phase in post-MI patients and ≥40% long-term adherence in the more chronic atherosclerosis model. Fourth, better medication adherence led to important cost savings.
Previous evidence has shown that guideline-indicated secondary preventive medication can mitigate risk for future events by >50% (3). In the real world, widespread use of these medications in patients after MI has contributed substantially to reductions in CV mortality. However, the benefits of these therapies are limited by patient adherence to treatment. Multiple large surveys (10), prospective registries (5), retrospective observational studies (11), meta-analyses, and data from randomized trials (12) have conclusively established that at least 50% of patients are not adequately adherent to their prescribed medications within the first 1 to 2 years after their index event. Few previous studies have analyzed medication adherence in a secondary prevention population, specifically including patients with cerebrovascular disease and peripheral artery disease. The international REACH (Reduction of Atherothrombosis for Continued Health) registry in secondary prevention patients showed that adherence to medication at both baseline and 1 year was significantly associated with the lowest incidence of adverse outcomes (p < 0.01) (5,13). Newby et al. (14) analyzed the use of evidence-based therapies during the period from 1995 to 2002 for patients with documented coronary artery disease in the Duke Databank for Cardiovascular Disease. They showed that consistent use of CV medication in patients with coronary artery disease was associated with statistically significant lower adjusted mortality. Our results confirm that increasing adherence to CV medications is associated with a significantly decreased risk of MACE hospitalizations in a secondary prevention cohort.
The post-MI cohort showed that compared with patients who were nonadherent, those who were fully adherent to their prescribed secondary prevention medications had significantly better event-free survival, with a 27% risk reduction of MACE. However, patients who were partially adherent did not enjoy the same protective benefit compared with nonadherent subjects, highlighting the need for very high levels of adherence (>80%) in the acute phase.
The atherosclerosis cohort displayed a different pattern, with fully adherent patients exhibiting a 44% and 24% risk reduction of MACE compared with nonadherent and partially adherent patients, respectively. In addition, the partially adherent group presented a 27% risk reduction over nonadherent patients. The sensitivity analysis added to the validity of the study and confirmed the pattern between the fully adherent and partially adherent groups compared with the nonadherent group. The difference between the fully adherent and partially adherent groups in the atherosclerosis cohort was attenuated, not reaching statistical significance because the rate of hospitalizations decreased in the second year in all groups and became too low to detect statistical significance. It seems, therefore, that one needs to maintain at least ≥40% adherence in the long term to continue to accrue benefit. It remains true that there is a linear relationship between higher adherence (>80%) levels and risk reduction for MACE. These findings have profound and lasting policy implications for accountable care organizations and quality measure targets.
The total costs of CV disease in 2010 were estimated to be $444 billion, representing approximately 17% of overall national health expenditures, with direct health care costs comprising about 61% in the United States (15). In addition, nearly one-third of all health care is attributed to inpatient hospital services (16). In the present study, we estimated per-person cost differences between the adherence groups in both cohorts. Full adherence to statins and ACE inhibitors was associated with reduced per-patient annual direct medical costs. This difference was mainly driven by a lower rate of CV events and revascularization procedures with higher adherence. Drivers and directionality of the impact of use and cost of health services were similar to results reported in previous research by Roebuck et al. (17) and Jha et al. (18).
The World Health Organization found that patient-, provider-, and system-level factors each may play a role in suboptimal adherence (19). Most interventions designed at improving quality of care and monitoring patients after an MI have focused on efforts to improve prescribing and patient knowledge at the time of hospital discharge (20). In contrast, value-based insurance designs that lower copayments of highly cost-effective medications or provide disease management programs aim to increase long-term medication use. However, data are lacking from randomized controlled studies evaluating the effectiveness of these combined strategies (21). In addition, many of the post-MI interventions with the most successful adherence are extremely labor-intensive, involving strategies such as individual counseling, medication education, pharmacy post-discharge programs, or visiting nurse or nurse practitioner–based services. In the era of dramatic changes to the health care economic landscape in the United States, such as the Balanced Budget Act of 1997 and, more recently, the Patient Protection and Affordable Care Act of 2010, efforts to improve health care quality must be coupled with efforts to contain costs. Ito et al. (22) evaluated the efficiency of various interventions (e.g., mailed education, disease management, a polypill) and combinations of these interventions. Results showed that a polypill combined with mailed education could be cost-effective and potentially cost-saving. Other modeling exercises have replicated this potential cost-saving with the use of a polypill (23).
Our study should be interpreted in the context in which it was performed. It was conducted by using a large, geographically diverse sample of patients with health care benefits from a U.S. national health insurer, including Medicare Advantage and commercial insurance, spanning varied employer-sponsored benefit plans. Our findings were attained on the basis of insurance and pharmacy claims data, and they benefit from the strengths of such systematic data collection regarding medications, procedures, and hospitalizations but also suffer from the limitations inherent in the use of insurance claims data. In fact, possession of a current filled prescription is a necessary, although not always sufficient, condition for therapy adherence. Underestimation may be related to patients paying out-of-pocket, and overestimation may occur because claims data do not actually confirm if the patient is taking the medication as prescribed. However, the requirement that patients filled at least 1 prescription after hospital discharge and the extended follow-up through 2 years help to mitigate these concerns. Grymonpre et al. (24) studied the validity of a similar measure for ACE inhibitor use among elderly patients and reported 95% concordance with pill counts. High correlation between claims-based measures and pill counts suggests that the rate of medications refill is usually consistent with the rate at which patients consume them.
Furthermore, although aspirin was of particular interest in the context of studying the impact of multiple medications in patients at CV risk, it is primarily obtained over-the-counter and thus not reliably measured in an insurance claims dataset. Medical claims data also lack the richness of clinical data in measuring clinical outcomes, including cause of death. The lack of benefit for many of the secondary outcomes is likely related to inadequate power and reduced specificity regarding these events through claims-based data. Furthermore, our study's primary outcome was measured from the time immediately after the refill or post-discharge, thus leading to some overlap within the adherence assessment period. However, as previously noted, we performed a sensitivity analysis excluding hospitalizations that occurred during the adherence assessment period in the atherosclerosis population; this approach helped to mitigate concerns of overlapping adherence and event evaluation periods in the largest cohort. It could also be argued that the fully adherent cohort seemed to have a lower risk profile at baseline, and the partially adherent cohort seemed to have a higher risk profile, which may have contributed to our findings; but, as previously stated, our outcomes were adjusted for multiple factors, including health-seeking behaviors, in an attempt to overcome the issues of confounding inherent in an observational cohort study. This set of variables was individually assessed in the model, and included if and when they reached statistical significance.
Thus, associations in both studies between medication nonadherence and adverse hospitalizations were independent of important potential confounding variables, such as age, pharmacy copayment, and CV risk factors. In addition, the consistent association between nonadherence and MACE hospitalizations may be the result of a healthy adherer effect, whereby adherent patients are probably less likely to engage in risky behaviors and more likely to follow medical recommendations. We specifically included variables related to health-seeking behaviors, such as flu vaccination in the adherence period data, to adjust for this confounder. We believe that these approaches added to the richness and validity of our findings. Although both medications are guideline-recommended, the requirement of both a statin and an ACE inhibitor refill limits practical generalizability to some extent.
Being fully adherent to guideline-recommended therapies was associated with a lower rate of MACE compared with partial or nonadherence; it was also associated with lower costs. There seemed to be a threshold effect for this benefit at >80% adherence in the post-MI population; ≥40% levels of adherence need to be maintained in the long term to continue to accrue benefit. Novel approaches to improve adherence may significantly reduce CV events post-MI.
COMPETENCY IN SYSTEMS-BASED PRACTICE: Better medication adherence may be associated with a reduction in major CV events and cost savings.
TRANSLATIONAL OUTLOOK: Strategies that improve medication adherence should be evaluated in randomized trials to assess their true impact on CV outcomes.
For a supplemental table and details regarding codes, please see the online version of this article.
Ms. Garrido, Dr. Alonso, and Dr. Lizano are employees of Ferrer. Dr. Rajda, Ms. Freeman, and Dr. Spettell are employees of Aetna Inc. Drs. Wei and Steinberg were employees of Aetna Inc. at the time this research was performed.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- confidence interval
- hazard ratio
- International Classification of Diseases
- major adverse cardiovascular events
- myocardial infarction
- proportion of days covered
- Received February 11, 2016.
- Revision received May 2, 2016.
- Accepted June 9, 2016.
- American College of Cardiology Foundation
- Go A.S.,
- Mozaffarian D.,
- Roger V.L.,
- et al.,
- American Heart Association Statistics Committee and Stroke Statistics Subcommittee
- Mozaffarian D.,
- Benjamin E.J.,
- Go A.S.,
- et al.,
- American Heart Association Statistics Committee and Stroke Statistics Subcommittee
- Task Force Members,
- Montalescot G.,
- Sechtem U.,
- et al.
- Ho P.M.,
- Bryson C.L.,
- Rumsfeld J.S.
- Farkouh M.E.,
- Boden W.E.,
- Bittner V.,
- et al.
- Newby L.K.,
- LaPointe N.M.,
- Chen A.Y.,
- et al.
- Heidenreich P.A.,
- Trogdon J.G.,
- Khavjou O.A.,
- et al.
- ↵Gonzalez JM. National health care expenses in the US civilian noninstitutionalized population, 2011. Statistical Brief #425. November 2013. Rockville, MD: Agency for Healthcare Research and Quality. Available at: https://meps.ahrq.gov/data_files/publications/st425/stat425.pdf. Accessed June 14, 2016.
- Roebuck M.C.,
- Liberman J.N.,
- Gemmill-Toyama M.,
- et al.
- Jha A.K.,
- Aubert R.E.,
- Yao J.,
- et al.
- Sabaté E.
- Rogers A.M.,
- Ramanath V.S.,
- Grzybowski M.,
- et al.,
- American College of Cardiology Foundation
- Choudhry N.K.,
- Rosenthal M.B.,
- Milstein A.
- Becerra V.,
- Gracia A.,
- Desai K.,
- et al.
- Grymonpre R.E.,
- Didur C.D.,
- Montgomery P.R.,
- et al.