Background Prior studies have demonstrated that patients with high-risk acute myocardial infarction (AMI) are less likely to receive guideline-directed medications during hospitalisation. It is unknown if this paradox persists following discharge. We aimed to assess if persistence with guideline-directed medications post discharge varies by patients’ risk following AMI.
Methods Data were analysed from two prospective, multicentre US AMI registries. The primary outcome was persistence with all prescribed guideline-directed medications (aspirin, β-blockers, statins, angiotensin-antagonists) at 1, 6 and 12 months post discharge. The association between risk and medication persistence post discharge was assessed using multivariable mixed-effect models.
Results Among 6434 patients with AMI discharged home, 2824 were considered low-risk, 2014 intermediate-risk and 1596 high-risk for death based upon their Global Registry of Acute Coronary Event (GRACE) 6-month risk score. High-risk was associated with a lower likelihood of receiving all appropriate therapies at discharge compared with low-risk patients (relative risk (RR) 0.90; 95% CI 0.87 to 0.94). At 12 months, the rate of persistence with all prescribed therapies was 61.5%, 57.9% and 45.9% among low-risk, intermediate-risk and high-risk patients, respectively. After multivariable adjustment, high-risk was associated with lower persistence with all prescribed medications (RR 0.87; 95% CI 0.82 to 0.92) over follow-up. Similar associations were seen for individual medications. Over the 5 years of the study, persistence with prescribed therapies post discharge improved modestly among high-risk patients (RR 1.05; 95% CI 1.03 to 1.08 per year).
Conclusions High-risk patients with AMI have a lower likelihood of persistently taking prescribed medications post discharge as compared with low-risk patients. Continued efforts are needed to improve the use of guideline-directed medications in high-risk patients.
Statistics from Altmetric.com
Among patients with acute myocardial infarction (AMI), guidelines and performance measures aim to improve quality of care delivered by encouraging provision of evidence-based medications in all eligible patients.1 Prior studies have demonstrated that high-risk patients with AMI often do not receive guideline-directed medications during hospitalisation, a phenomenon that has been referred to as the ‘risk-treatment paradox’.2 ,3 However, little is known about whether such paradox exists for use of prescribed medications following hospital discharge.
Both provision of appropriate medications and continued use of these medications are necessary to realise their potential to reduce the risk of mortality and recurrent AMI. While it is known that physicians are less likely to optimally manage high-risk patients with AMI at the time of discharge,2 ,3 it is unknown whether long-term use of these guideline-directed medications differs by patients’ risk after discharge or whether use of guideline-directed medical therapy post discharge has improved over time. Identifying such treatment gaps can enable targeted interventions to improve the use and persistence with cardiac medications.
The objective of this study was to assess persistence with guideline-directed therapies during longitudinal follow-up in two large, prospective, multicentred registries of patients with AMI. We assessed prescription of aspirin, statins, β-blockers and ACE inhibitors/angiotensin receptor blockers (ACEI/ARBs) to patients with AMI at low, intermediate and high risk for all-cause mortality based on the Global Registry of Acute Coronary Event (GRACE) risk score at hospital discharge. We then sought to describe persistence with these medications in the year following hospital discharge, as well as assess temporal trends in persistence with these cardiac medications across risk strata over the course of this study.
The analytical cohort for this study was derived from the Prospective Registry Evaluating Myocardial Infarction: Events and Recovery (PREMIER) and Translating Research Investigating Underlying Disparities in AMI Patients Health Status (TRIUMPH) registries. Both are prospective, multicentre, observational registries of patients with AMI. PREMIER enrolled 2498 patients from 19 US medical centres between 1 January 2003 and 28 June 2004 and TRIUMPH enrolled 4340 patients from 24 US medical centres between 11 April 2005 and 31 December 2008 (31 sites in total; 12 sites participated in both registries). Both registries had identical inclusion and exclusion criteria and employed the same standards in data collection and follow-up. Their study designs have been previously described and further details have been provided in the online supplementary appendix.4 ,5
Study design and cohort
This study was performed as a retrospective analysis of prospectively collected data. We included all patients enrolled in PREMIER and TRIUMPH who were discharged home and had data available on discharge medications (figure 1). We excluded patients with documented contraindications including active bleeding on arrival or concomitant warfarin use at discharge for aspirin; heart rate <50 bpm, 2nd/3rd degree heart block, systolic blood pressure<100 mm Hg for β-blockers; moderate/severe aortic valve stenosis or systolic blood pressure <90 mm Hg for ACEI/ARBs, or documented allergies or patient refusal. Contraindications were prospectively abstracted from medical records. For ACEI/ARBs, we included only patients with LV systolic dysfunction (EF <40%) at the time of AMI.6
Risk was assessed among included patients using GRACE discharge risk-score. This model predicts all-cause mortality in both ST-segment elevation myocardial infarction (STEMI) and non-ST segment elevation myocardial infarction (NSTEMI) patients at various time points post discharge, ranging from 6 months to 4 years (c-statistic >0.75).7 ,8 Components of GRACE score include age, presenting heart rate, presenting systolic blood pressure, initial serum creatinine, elevated cardiac biomarkers, ST segment depression, in-hospital percutaneous coronary intervention (PCI), in-hospital coronary artery bypass grafting (CABG), history of congestive heart failure and history of myocardial infarction. We calculated GRACE scores separately based on presenting diagnosis (STEMI/NSTEMI) and stratified the cohort into three risk categories (low, intermediate, high) based on established and validated criteria (see online supplementary table S1).9
At hospital discharge, we assessed whether eligible patients were appropriately prescribed aspirin, statin, β-blockers and ACEI/ARBs. A composite measure was calculated as prescription of all guideline-directed medications for which the patient was eligible. Hence, a patient eligible for two medications prescribed both medications was classified as receiving all guideline-directed care, but a patient eligible for four medications prescribed only three or less was classified as not receiving all guideline-directed care.
Next, we ascertained patients’ persistence with prescribed guideline-directed medications post discharge (ie, among those patients discharged on the corresponding medication). Medication usage at follow-up was based on patients’ self-report assessed via in-person or telephone interviews at 1, 6 and 12 months postindex AMI. At each of these interviews, patients were asked to collect all their current medications and read each medicine to the interviewer. Since we evaluated only self-reported use of medications (yes/no), we used the term ‘persistence’ as opposed to adherence, which quantifies intensity of medication use.10–13 Persistence was assessed for the composite endpoint (ie, all prescribed guideline-directed medications as defined above) as well as for use of aspirin, statins, β-blockers and ACEI/ARBs individually.
Previous studies examining validity of self-reported medication use have found it to be comparable with other direct measures of persistence such as blood drug levels.13–16 Additionally, in real-world clinical practice, medication use among patients is predominantly assessed by self-report.
Other covariates included in our model are detailed in online supplementary appendix.
Baseline characteristics were compared between low-risk, intermediate-risk and high-risk patients using χ2 or Fisher's exact tests for categorical variables and one-way analysis of variance for continuous variables. Relative risk (RR) regression was used to assess association between risk-category and prescription of all eligible guideline-directed therapies at discharge, adjusting for enrolling site and other covariates (provided in supplementary material). Since prescribing medications was not a ‘rare event,’ we used a modified Poisson regression model to estimate RRs directly, as opposed to ORs obtained by logistic regression, which overestimates RRs in such situations. Because Poisson model misspecifies the distribution of outcome, robust variance estimation was used to provide correct standard errors.17
Next, association between risk and post-discharge persistence with prescribed medications was assessed, both for composite of all prescribed medications and for each prescribed medication individually. Effect of risk on persistence was assessed with generalised linear mixed models, using the modified Poisson framework described above but including random effects to accommodate within-patient correlations among repeated assessments at 1, 6 and 12 months. The model included fixed effects for risk-category, enrolling site, all covariates, follow-up month and patient-level random intercepts for each of the three follow-up time points. An unstructured covariance matrix for the random intercepts was used to accommodate within-patient correlation between follow-up assessments. Interaction between risk and time was also assessed; however, no interaction achieved the 0.05 significance level for any outcome (p values ranged from 0.09 to 0.55), so only main effects are presented here. Sensitivity analyses were performed using logistic regression instead of Poisson models to evaluate robustness of our findings; results were consistent and so are not reported.
Finally, we assessed temporal trends in use of guideline-directed therapies at and post discharge across risk-strata. Observed rates of medication use were calculated within 6-month intervals across the study duration (2003–2008), stratified by patient-risk. Temporal trends, adjusting for patient factors, were statistically assessed by augmenting the above regression models with a linear term for year and a year-by-risk group interaction.
Details regarding missing data analyses are provided in online supplementary appendix. All analyses were conducted using SAS V.9.4 (SAS Institute, Cary, North Carolina, USA) and R V.184.108.40.206 Our models were fitted with SAS PROC GLIMMIX. All statistical tests were evaluated at a two-sided significance level of 0.05.
Between 1 January 2003 and 31 December 2008, a total of 6838 patients with AMI were enrolled in PREMIER and TRIUMPH registries. We excluded patients who died prior to discharge (n=41), patients discharged to hospice or nursing homes (n=146) and patients on whom data regarding discharge medications were lacking (eg, those transferred to another acute or non-acute care facility, patients who signed out against medical advice or were lacking data on final disposition, n=217; figure 1). The final analytical cohort comprised of 6434 patients of which 2824 (43.9%) patients were low-risk, 2014 (31.3%) intermediate-risk and 1596 (24.8%) high-risk by GRACE risk-score. Overall, follow-up data were available on 5684 (88.3%) patients.
Table 1 summarises baseline characteristics of the study cohort stratified by risk. Compared with patients in low-risk and intermediate-risk strata, high-risk patients were older, more likely to be women, less likely to be Caucasian and had a greater burden of all comorbidities. High-risk patients were more likely to be insured and less likely to report avoiding medications due to cost, compared with low-risk and moderate-risk patients. Furthermore, high-risk patients were less likely to undergo diagnostic coronary angiography and coronary revascularisation (PCI/CABG) during index AMI.
Therapy use at discharge
Table 2 shows rates of medication use at discharge among eligible patients in the study population stratified by risk. Overall, 1136 (71.2%) high-risk patients received all guideline-directed therapies at discharge as compared with 1583 (78.6%) intermediate-risk patients and 2385 (84.5%) low-risk patients (p<0.001). Following multivariable adjustment, high-risk was associated with a lower rate of receipt of all eligible therapies at discharge compared with low-risk patients (RR 0.90; 95% CI 0.87 to 0.94). Similarly, intermediate-risk was also associated with lower rate of receipt of all eligible therapies at discharge (RR 0.95; 95% CI 0.92 to 0.98) as compared with low-risk patients.
Persistence with therapy after discharge
Among patients discharged on guideline-directed therapies, increasing risk was associated with lower persistence with all prescribed medications at 1 month (low-risk vs intermediate-risk vs high-risk patients: 70.4% vs 63.8% vs 51.0%), 6 months (63.9% vs 59.3% vs 45.8%) and 12 months after index AMI (61.5% vs 57.9% vs 45.9%; figure 2). After multivariable adjustment, high-risk was significantly associated with lower persistence with all guideline-directed therapies across follow-up compared with low-risk patients (RR 0.87; 95% CI 0.82 to 0.92), although intermediate-risk was not (RR 0.98; 95% CI 0.94 to 1.01; table 3). Other variables associated with lower persistence with all guideline-directed therapies post discharge are shown in online supplementary figure S1. This included socioeconomic factors such as non-white race, lack of high school education and insurance, other chronic medical conditions (eg, diabetes, congestive heart failure, chronic kidney disease), depression and higher anginal burden assessed by the Seattle Angina Questionnaire.
Individual observed persistence rates for aspirin, statins, β-blockers and ACEI/ARBs post discharge were also lowest among high-risk patients, but tended to be comparable between low-risk and intermediate-risk patients (figure 3). Following multivariable adjustment, high-risk was associated with lower persistence, but in general, intermediate-risk was not (table 3).
Temporal trends in medication use at and after discharge
Figure 4 shows medication use at and post discharge over the study duration from 2003 to 2008 (no data were available between July 2004 and July 2005, the period between end of PREMIER enrolment and start of TRIUMPH enrolment). From 2003 to 2008, we observed a modest increase in receipt of all guideline-directed therapies at discharge for all risk-groups. Although the increase was greatest among high-risk patients, interaction by risk-group did not quite achieve statistical significance (p=0.06). Similarly, persistence with all prescribed therapies post discharge increased modestly and significantly over time for all risk-groups; and again, although the increase was greatest among high-risk patients, the interaction by risk-group was not significant (p=0.06; table 4).
Our objective was to determine the receipt of cardiac medications at discharge and persistence with their use during longitudinal follow-up in a large, prospective cohort of patients with AMI stratified by their mortality risk. We found an association between high-risk and a lower receipt of indicated cardiac medications at hospital discharge. Furthermore, among patients discharged on appropriate therapies, high-risk patients had lower persistence with medications over the year following AMI. Largest drop in medication persistence occurred in the first month post discharge. Less than half of high-risk patients persisted with all guideline-directed therapies at 12 months post discharge. Additionally, examination of temporal trends in discharge prescription rates and persistence with these therapies in high-risk patients revealed a trivial increase over time. To our knowledge, this is one of the first studies addressing the concept of ‘risk-persistence paradox’ among guideline-directed medical therapies in a prospective cohort of patients with AMI post discharge.
Annually over 715 000 Americans have an AMI. With dramatic improvements in acute care, majority of patients with AMI survive index hospitalisation.19 Thus, long-term use of optimal secondary prevention therapies is a major focus of care. To date, efforts to improve use of evidence-based medications among patients with AMI have focused primarily on administration of appropriate medications during hospitalisation and at discharge. Prior studies describe lower rates of medication use among high-risk patients with AMI during hospitalisation and at discharge.2 ,3 Our study extends findings of prior reports and evaluates persistence in the year following AMI to all guideline-directed therapies prescribed at discharge among patients stratified by risk. While we observed high rates of prescription of guideline-directed therapies at discharge, even among high-risk patients, persistence with these therapies post discharge was suboptimal and lowest among high-risk patients who would potentially benefit most. For example, after adjusting for a wide range of clinical characteristics that influence medication persistence, among 10 high-risk patients two were not taking aspirin, two were not taking β-blocker, three were not taking statin and three were not taking ACEI/ARB by 1 month post discharge. These rates continued to decline over the observation time period of 1 year. While persistence rates were also dismal in low-risk and intermediate-risk groups, these groups showed better persistence with medications compared with high-risk population. Our findings highlight the need to emphasise on persistence with therapy post discharge among all patients with AMI and particularly for high-risk patients with AMI.
Non-persistence is known to be a marker of adverse cardiovascular outcomes irrespective of baseline-risk and other comorbidities.13 Hence, efforts to bolster persistence with therapy are important potential opportunities to improve patient outcomes, and lack of natural improvements in persistence is demonstrated by the marginal increase in persistence rates we observed over the entire study duration in all risk-categories. Our findings suggest that educational and persistence intervention programmes should target all patients, but particularly those at higher risk. Interestingly, for all medications, the largest drop in persistence was observed over the first month after discharge. This likely represents primary medication non-adherence; that is, failure in filling prescriptions issued at discharge. An alternative explanation may include early onset medication side-effects leading to their discontinuation. However, the high rates of discontinuation observed are unlikely to be explained entirely by side-effects.
Primary non-adherence to cardiovascular medications has been previously described in literature as well. In particular, this holds true for medications for chronic conditions such as coronary artery disease, hypertension and diabetes.20 ,21 Furthermore, patients not filling medications have a higher 1-year mortality.20 These findings are comparable with our results as we also found similar rates of medication non-persistence at 1 month post discharge, and higher-risk patients at greater risk of mortality were least likely to be taking appropriate medications at 1 month post discharge. Hence, interventions addressing persistence should target patients early post discharge. Addition of new multiple medications, as is commonly seen after hospitalisations,22 may be accompanied by inadequate patient education,23 follow-up and continuity of care. Accordingly, improved efforts to facilitate transition in care from hospital into the community are needed. Discharge medication counselling, involvement of pharmacists and ancillary staff with education and postdischarge follow-up may provide reliable solution to this issue.
Another factor that may contribute to low persistence rates could be lack of continued emphasis on importance of medications by physicians. Physician support for use of a drug is an important factor to patients in their long-term medication use.24 Prior literature has suggested an underestimation of patient-risk by treating physicians.25 Other studies have suggested underuse due to concerns of higher risk of side-effects in high-risk patients compounded by risks of polypharmacy.26 While such concerns may be well founded, benefits of therapy are more likely to outweigh risks in such patients who have the greatest potential to benefit from therapy.27 ,28 Incorporation of validated risk-score models, such as the GRACE score, in clinical practice to identify high-risk patients can assist physicians in decision-making.
Our findings should be interpreted in the light of several potential limitations. First, our finding of lower persistence with therapies among high-risk patients could be secondary to the development of side-effects or contraindications to specific therapies (eg, a new gastrointestinal bleed, hypotension). However, we limited our analysis to eligible patients at discharge and also adjusted for a number of possible contraindications to therapy such as kidney disease (ACEI/ARB) and lung disease (β-blockers). Second, persistence with medication was determined through patient self-report and could introduce recall bias. However, patients were asked to read the labels of all their medications in an attempt to minimise recall bias in ascertaining medication persistence. Nonetheless, self-report is used routinely in clinical practice to assess medication persistence and has been validated in other studies against direct persistence measures such as blood drug levels and pill counts.15 ,16 Additionally, self-report would bias our findings towards null suggesting that real-world persistence is likely to be lower than we report. Third, we did not have follow-up data on roughly 10% of patients. While our use of imputation and likelihood-based estimation ameliorates any missing-data biases associated with observed factors, it is possible that non-ignorable biases may remain (eg, medication persistence may be lower among patients missing medication use data).29 However, as noted in the methods, we did evaluate the potential impact of this bias by imputing a value of ‘not persistent’ for patients with incomplete data, and the results did not change appreciably, so we believe the impact of non-ignorable missing data is small. Fourth, residual confounding in determining the association between risk and treatment persistence cannot be eliminated. Finally, compared with a retrospective description of patients with AMI enrolled in a large, national cohort, patient characteristics were comparable suggesting that our results are generalisable to a wide range of patients seen in clinical characteristics.30
The results of our study show that high-risk patients with AMI have lower persistence rates to guideline-directed therapy in the year following discharge. While trends in the use of therapy at discharge are improving, there have been marginal changes in use after discharge. Further research is necessary to understand reasons for low persistence with therapy after discharge. Additionally, targeted efforts towards improving the use of outpatient therapy are necessary.
What is already known on this subject?
Patients with acute myocardial infarction (AMI) at high risk for mortality are less likely to be discharged on optimal medical therapy compared with intermediate-risk and low-risk patients.
What might this study add?
Among patients discharged on optimal therapy, overall persistence with medications is low at 1 year post discharge.
Persistence rates are lowest among high-risk patients, with only 45% of high-risk patients taking all prescribed therapies at 1 year post discharge.
Improvement in medication persistence over a 5-year observation period was minimal.
How might this impact on clinical practice?
Efforts to improve medication persistence after AMI hospitalisation are needed.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Funding TRIUMPH was supported by a grant from the National Institutes of Health (National Heart, Lung, Blood Institute): Washington University School of Medicine (SCCOR Grant #P50HL077113 -01). PREMIER was sponsored by CV Therapeutics, Palo Alto, California, USA.
Competing interests DLB discloses the following relationships—Advisory Board: Elsevier Practice Update Cardiology, Medscape Cardiology, Regado Biosciences; Board of Directors: Boston VA Research Institute, Society of Cardiovascular Patient Care; Chair: American Heart Association Get With The Guidelines Steering Committee; Honoraria: American College of Cardiology (Editor, Clinical Trials, Cardiosource), Belvoir Publications (Editor in Chief, Harvard Heart Letter), Duke Clinical Research Institute (clinical trial steering committees), Harvard Clinical Research Institute (clinical trial steering committee), HMP Communications (Editor in Chief, Journal of Invasive Cardiology); Population Health Research Institute (clinical trial steering committee), Slack Publications (Chief Medical Editor, Cardiology Today's Intervention), WebMD (CME steering committees); Data Monitoring Committees: Duke Clinical Research Institute; Harvard Clinical Research Institute; Mayo Clinic; Population Health Research Institute; Research Grants: Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, Roche, Sanofi Aventis, The Medicines Company; Unfunded Research: FlowCo, PLx Pharma, Takeda. JS has received research grants from NIH, AHA, Genentech, Gilead, Amorcyte, Evaheart and Lilly. He serves as a consultant on the United Healthcare Cardiac Scientific Advisory Board and to Amgen, Novartis, Jannsen, and Gilead. He owns the copyright to the SAQ, KCCQ, and PAQ and has an equity interest in Health Outcomes Sciences.
Ethics approval IRB at all centres participating in TRIUMPH and PREMIER registries.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.