Impact of a Copayment Reduction Intervention on Medication Persistence and Cardiovascular Events in Hospitals With and Without Prior Medication Financial Assistance Programs.

Background Hospitals commonly provide a short-term supply of free P2Y12 inhibitors at discharge after myocardial infarction, but it is unclear if these programs improve medication persistence and outcomes. The ARTEMIS (Affordability and Real-World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study) trial randomized hospitals to usual care versus waived P2Y12 inhibitor copayment costs for 1-year post-myocardial infarction. Whether the impact of this intervention differed between hospitals with and without pre-existing medication assistance programs is unknown. Methods and Results In this post hoc analysis of the ARTEMIS trial, we examined the associations of pre-study free medication programs and the randomized copayment voucher intervention with P2Y12 inhibitor persistence (measured by pharmacy fills and patient report) and major adverse cardiovascular events using logistic regression models including a propensity score. Among 262 hospitals, 129 (49%) offered pre-study free medication assistance. One-year P2Y12 inhibitor persistence and major adverse cardiovascular events risks were similar between patients treated at hospitals with and without free medication programs (adjusted odds ratio 0.93, 95% CI, 0.82-1.05 and hazard ratio 0.92, 95% CI, 0.80-1.07, respectively). The randomized copayment voucher intervention improved persistence, assessed by pharmacy fills, in both hospitals with (53.6% versus 44.0%, adjusted odds ratio 1.45, 95% CI, 1.20-1.75) and without (59.0% versus 48.3%, adjusted odds ratio 1.46, 95% CI, 1.25-1.70) free medication programs (Pinteraction=0.71). Differences in patient-reported persistence were not significant after adjustment. Conclusions While hospitals commonly report the ability to provide free short-term P2Y12 inhibitors, we did not find association of this with medication persistence or major adverse cardiovascular events among patients with insurance coverage for prescription medication enrolled in the ARTEMIS trial. An intervention that provided copayment assistance vouchers for 1 year was successful in improving medication persistence in hospitals with and without pre-existing short-term medication programs. Registration URL: https://www.clini​caltr​ials.gov/. Unique identifier: NCT02406677.

RCT Entities:   Population Interventions Outcomes

The Affordability and Real‐World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study

major adverse cardiovascular events

The Post‐Myocardial Infarction Free RX Event and Economic Evaluation Trial

Clinical Perspective

What Is New?

Hospitals frequently use medication assistance programs to reduce the cost of medications after myocardial infarction, but these programs were not associated with an improvement in P2Y12 inhibitor persistence, whereas the randomized copayment intervention tested in the ARTEMIS (Affordability and Real‐World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study) was associated with improved persistence.

What Are the Clinical Implications?

Hospitals and payors may consider strategies that provide long‐term reduction of copayments for critical secondary prevention medications, rather than short‐term support at the time of discharge.

Introduction

Medication non‐adherence after myocardial infarction (MI) is common and is associated with adverse cardiovascular outcomes.1, 2, 3, 4 Guidelines recommend P2Y12 inhibitors for at least 1 year after an MI,5, 6 but premature discontinuation of P2Y12 inhibitors can lead to stent thrombosis and recurrent MI with often fatal outcomes.7, 8 Among hospitals in the United States, there is significant variability in medication adherence after MI, and patients discharged from hospitals with low adherence rates have higher incidence of major adverse cardiovascular events (MACE).9 In response to these concerns, hospitals have enacted varying strategies to promote optimal secondary prevention medication use. Many of these strategies have not been formally evaluated and, in those that have been, their effectiveness is mixed.10, 11, 12 Medication cost is a common and significant barrier to adherence.13 For patients with financial barriers to medication adherence, some hospitals offer a free 30‐day supply of medications at discharge to promote initial adherence and prevent rehospitalizations. The prevalence of use and potential impact of such free short‐term medication programs on longitudinal medication persistence and patient outcomes are unknown.

The ARTEMIS (Affordability and Real‐World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study) trial randomized hospitals to usual care versus the ability to waive copayment out‐of‐pocket costs for P2Y12 inhibitor fills for 1 year.14 The intervention enhanced guideline‐adherent P2Y12 inhibitor selection and longitudinal medication persistence, but did not significantly reduce MACE.15 Participating hospitals were surveyed on strategies used to promote optimal medication use before and after ARTEMIS study participation, including whether hospitals had the ability to provide a free 30‐day supply of a P2Y12 inhibitor to patients at MI discharge. We hypothesized that patients treated at hospitals with pre‐existing free medication programs would already have better medication persistence and outcomes compared with those treated at hospitals without pre‐existing free medication programs. Thus, these hospitals would less likely benefit from the randomized ARTEMIS copayment intervention than hospitals without pre‐existing free mediation programs.

Methods

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subjects confidentiality protocols may be sent to the Duke Clinical Research Institute at tracy.wang@duke.edu. The ARTEMIS trial was a cluster‐randomized trial assessing the impact of a copayment reduction intervention on 1‐year P2Y12 inhibitor persistence and MACE in patients hospitalized for acute MI. The design of ARTEMIS has been previously reported in detail.14, 15 In brief, hospitals were eligible for randomization if they treated at least 50 MI patients annually and had both clopidogrel and ticagrelor available for clinical use. In both randomized hospital groups, enrolled patients were aged >18 years, hospitalized for MI, treated with a P2Y12 inhibitor during the hospitalization, and had any US‐based health insurance with a prescription drug plan. Overall, 26 006 patients were screened for the ARTEMIS study, and enrollment rates were higher at intervention hospitals (6436 of 18 803 screened patients, 34%) than control hospitals (4565 of 20 436 screened patients, 22.3%). Of 11 001 patients enrolled from 287 hospitals between June 6, 2015 and September 20, 2016, we excluded patients who died during the index admission or withdrew from the study before discharge (n=25), were not discharged on clopidogrel or ticagrelor (n=874), or were treated at sites that did not complete the pre‐enrollment hospital survey questions (n=512 patients at 25 sites). Our final study population included 9590 patients from 262 hospitals (Figure 1). The study was approved by the institutional review boards at Duke University and all participating sites. All patients provided written informed consent for participation in the ARTEMIS study.

Figure 1

Study population.

ARTEMIS indicates Affordability and Real‐World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study trial.

Sites randomized to the copayment intervention provided subjects with a voucher that allowed them to fill clopidogrel or ticagrelor at any pharmacy without out‐of‐pocket copayment costs for 1‐year post‐MI. Decisions on P2Y12 inhibitor drug choice and treatment duration were left to the treating clinician. The protocol did not prohibit the use of any other strategies to optimize medication use; sites in either arm could continue pre‐existing programs, or initiate new interventions, during participation in ARTEMIS. We therefore performed a post hoc, non‐prespecified analysis of the ARTEMIS study to assess associations of pre‐existing assistance programs with medication persistence and outcomes.

Before randomization and patient enrollment, each site completed a survey of their hospital's practices related to transition of care and medication persistence promotion; the survey was completed by a clinician involved in the discharge of patients with MI at that hospital who was familiar with institutional discharge practices. The site survey included a question: “Does your hospital provide medication assistance to patients who cannot afford the prescribed P2Y12 inhibitor therapy?” Hospitals that answered “we provide drug for a short period (30 days) sponsored by the hospital,” or “we provide drug for a short period sponsored by someone else (eg, pharmaceutical company samples)” were defined as hospitals offering free short‐term medication assistance before ARTEMIS study participation. All other responses (“provide discount vouchers,” “apply on patient's behalf for prescription assistance programs,” “refer patient to pharmacies/clinics that have prescription assistance, discount generic programs, or samples,” “provide other medication assistance,” or “no assistance provided”) did not provide patients with a free short‐term supply of medication, therefore these hospitals were not defined as providing free medication assistance. Additional questions focused on other medication use optimization strategies, discharge practices, and hospital quality improvement systems. One year after the site completed study enrollment, 210 sites (80%) completed a follow‐up survey that included the same question.

P2Y12 inhibitor persistence, measured by patient report, and MACE were co‐primary end points of the ARTEMIS study. As patient report overestimates medication persistence,16 persistence was also defined as no gap in P2Y12 inhibitor supply >30 days using pharmacy fill data from Symphony Health Solutions, which captures pharmacy claims data from ≈90% of retail, 60% of mail‐order, and 70% of specialty pharmacies in the United States.17 Pharmacy fill data were available for 7942 patients (82.8%). Persistence was measured at 90‐days and 1‐year post‐MI. For patients who died or had missing persistence data before 1 year, the last observation carried forward method was used for persistence up to 1 year. MACE was defined as the composite of all‐cause death, MI, or stroke within 1‐year post‐MI. Events were independently adjudicated by physicians at the coordinating center as previously described.14

All patients completed a survey at enrollment which included items related to medication‐taking behavior and financial burden. Missingness rates were low (<4%) for all of these items.

We compared characteristics of hospitals with and without pre‐study free short‐term medication programs, and well as the demographic, clinical, and treatment characteristics of patients treated at these hospitals. Categorical variables were presented as frequencies and differences between groups were assessed using Chi‐square test or an exact test as appropriate. Continuous variables were presented as median (interquartile range) and compared using the Wilcoxon rank‐sum test.

We compared outcomes of patients treated at hospitals with and without pre‐study free short‐term medication programs. P2Y12 inhibitor persistence comparisons were adjusted for patient and hospital characteristics using logistic regression models with parameters estimated using generalized estimating equations to account for within‐hospital clustering. MACE outcomes were compared using Cox regression models with adjustment for patient and hospital characteristics and robust standard errors to account for within‐hospital clustering. Covariates included age, sex, race, insurance payor, region, pre‐enrollment site MI volume and proportion of ticagrelor use, an indicator for randomization scheme (2:1 versus 1:1), an indicator variable for randomization status (intervention versus control), as well as a propensity score that estimated likelihood of treatment at a hospital with pre‐study free medication programs using a logistic regression model containing 51 covariates selected a priori based on clinical relevance (Table S1, Figure S1). All covariates were included in the model regardless of whether there were statistically significant differences between groups. In a sensitivity analysis, we assessed the association of free medication programs with outcomes among hospitals randomized to the copayment voucher intervention or usual care by developing separate logistic regression models for each group.

We then assessed the impact of the randomized intervention on persistence and MACE among hospitals with and without pre‐study free medication programs using similar methodology, though a separate propensity score was developed for the likelihood of assignment to the intervention group (versus control) (Table S1, Figure S2). We then tested the interaction of pre‐study free medication programs and randomization groups with regard to persistence and MACE. Among hospitals randomized to the intervention, we assessed for any imbalance in use of the copayment waiving voucher between patients treated at hospitals with and without pre‐study free medication programs.

One year after sites completed study enrollment, sites completed a follow‐up survey that again asked: “Does your hospital provide medication assistance to patients who cannot afford the prescribed P2Y12 inhibitor therapy?” We calculated the proportion of sites now defined as providing free short‐term medication assistance. Among sites that did not offer pre‐study free medications, we calculated the proportion of sites that converted to providing free short‐term medication assistance.

Missingness was <1% for all clinical variables, <5% for all patient‐reported variables, and <10% for all hospital survey questions. We imputed missing medical history, home medications, admission features, and in‐hospital events to the mode. Socioeconomic variables, laboratory values, and weight were imputed to age‐, sex‐, and race‐specific modes for categorical variables and medians for continuous variables. A P≤0.05 was considered statistically significant for all tests. Analyses were performed using SAS software, version 9.4 (SAS Institute Inc).

Results

Among 262 hospitals participating in the ARTEMIS trial, 49% (n=129 centers enrolling 5051 patients) reported pre‐study free short‐term medication assistance (ie, the ability to provide patients with a free 30‐day supply of P2Y12 inhibitors at discharge). These included 29 hospitals (11.1%) that covered the cost of the drug directly, 81 (30.9%) that provided drug that was sponsored by a pharmaceutical company, and 19 (7.3%) that had both options available.

Hospitals with and without free medication programs had no significant differences in size, teaching status, geographic distribution, and profit status, however hospitals with free short‐term medication programs more often included a case manager or care coordinator in the discharge process and filled outpatient P2Y12 inhibitor prescriptions before the patient left the hospital; alternative medication cost reduction strategies, such as the provision of commercial discount vouchers, were also used more frequently at these hospitals (Table 1). Several non‐financial strategies were commonly used by hospitals with and without pre‐study free medication programs, including: a discharge medication list providing rationale for each medication, phone contact within 1 week of discharge, and a 24‐hour call center for patient concerns (Table 1).

Table 1

Site Characteristics and Adherence Promotion Strategies at Hospitals With and Without Pre‐Study Free Medication Programs

Variable	Free Medication Program	No Free Medication Program	P Value
n	129	133
Site characteristics
Total hospital beds, median (IQR)	381 (276, 601)	393 (270, 555)	0.66
Teaching hospital	29%	24%	0.40
Profit status			0.27
Not‐for‐profit	81%	73%
For‐profit	9%	14%
Government	11%	13%
Region			0.92
Northeast	18%	20%
Midwest	33%	29%
South	38%	40%
West	12%	12%
Medication use optimization strategies
Routine participation in patient dischargea
Pharmacist	38%	32%	0.28
Care coordinator/case manager	84%	57%	<0.01
Social worker	45%	37%	0.18
Dedicated transition of care nurse	32%	36%	0.51
Routinely screens patients fora
Medication non‐adherence	48%	42%	0.35
Ability to afford medications	75%	65%	0.08
Call pharmacy to check P2Y12 inhibitor cost	17%	23%	0.22
Provides commercial discount vouchers for P2Y12 inhibitorb	66%	53%	0.03
Apply for P2Y12 inhibitor prescription assistance on patient's behalfc	40%	30%	0.08
On‐site pharmacy fills P2Y12 inhibitor before patient leaves the hospitala	33%	17%	<0.01
Discharge medication list describing rationale for each medicationa	84%	72%	0.05
Phone contact within 1 wk of dischargea	65%	51%	0.02
24‐h call center for patient concernsa	57%	52%	0.44

IQR indicates interquartile range.

Hospital performs this for >50% of all myocardial infarction patients before hospital discharge.

Discount vouchers include coupons distributed by pharmaceutical manufacturers, prescription benefit managers, pharmacies, or marketing companies.

Prescription assistance programs are generally funded by pharmaceutical companies to provide lower cost medications to applicants that demonstrate financial need.

Patients enrolled at free medication hospitals were more likely to be of white race and present with non–ST‐segment–elevation myocardial infarction, and to be treated with ticagrelor during hospital admission and at time of discharge (Table 2). Overall, half of the patients reported financial hardship related to filling prescription medications, 17% had not filled a prescription because of cost within the prior 3 months, and cost was frequently ranked as an “extremely important” factor in medication decision‐making (Figure 2). Compared with patients treated at hospitals without free medication programs, patients treated at hospitals with free medication programs more often considered cost “extremely important” (49% versus 45%, P<0.01) and reported financial hardship (51% versus 49%, P<0.05).

Table 2

Characteristics of Patients Admitted With Myocardial Infarction to Sites With and Without Pre‐Existing Free Medication Programs

Variable	Free Medication Program	No Free Medication Program	P Value
n	5051	4539
Patient demographics
Age, y, median (IQR)	62 (54–70)	62 (54–70)	0.55
Men	69%	67%	0.26
Race
White	89%	87%	<0.01
Black	9%	11%	<0.01
Other	3%	3%	0.86
Insurance payor			0.29
Private	63%	64%	0.41
Medicare	43%	43%	0.67
Medicaid	9%	9%	0.43
Other	10%	8%	0.04
Medical history
Hypertension	69%	69%	0.57
Diabetes mellitus	33%	32%	0.59
Dyslipidemia	58%	59%	0.64
Dialysis	2%	2%	0.65
Prior MI	21%	20%	0.19
Prior PCI	26%	25%	0.16
Prior CABG	11%	11%	0.23
Prior TIA/stroke	7%	7%	0.40
Prior heart failure	7%	8%	0.15
Current/recent smoker	35%	32%	0.02
Presentation and treatment
STEMI	45%	47%	0.02
Cardiogenic shock	2%	3%	0.27
Cardiac arrest	3%	3%	0.72
Diagnostic angiography	98%	98%	0.07
PCI	88%	90%	0.08
CABG	1%	2%	0.06
P2Y12 inhibitor usea
Home P2Y12 inhibitor use	14%	15%	0.39
In‐hospital
Clopidogrel	48%	57%	<0.01
Ticagrelor	62%	55%	<0.01
At discharge
Clopidogrel	43%	52%	<0.01
Ticagrelor	57%	48%	<0.01
Patient survey responses
Medication cost is extremely important	49%	45%	<0.01
Financial hardship related to medications	51%	49%	0.05
Not filled prescription because of cost in past 90 d	17%	17%	0.49

CABG indicates coronary artery bypass graft; IQR, interquartile range; MI, myocardial infarction; PCI, percutaneous coronary intervention; STEMI, ST‐segement–elevation myocardial infarction; and TIA, transient ischemic attack.

All patients were treated during hospitalization with clopidogrel and/or ticagrelor (switching during the hospitalization was permitted), and clopidogrel or ticagrelor at time of discharge.

Figure 2

Responses to the baseline patient survey on medication cost and cost‐related non‐adherence, administered to all subjects at time of enrollment in ARTEMIS (The Affordability and Real‐World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study) (n=9590).

Overall, persistence to P2Y12 inhibitors was 96% at 90 days and 86% at 1 year when assessed by patient report. Persistence assessed by pharmacy fills was 72% at 90 days and 52% at 1 year. Pre‐study free short‐term medication programs were not associated with differences in short‐ or long‐term medication persistence rates or MACE in either unadjusted or adjusted analyses (Table 3, Figure S3). Outcomes of patients treated at hospitals with pre‐study free medication programs were not significantly different from those treated at hospitals without free medication programs, when examined separately by randomized arm (Tables S2 and S3).

Table 3

The Association of Hospital Use of Pre‐Study Free Medication Programs (Vs No Free Medication Programs) With P2Y12 Inhibitor Persistence and MACE

Outcome	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
90‐d persistence
Patient report	1.18 (0.92–1.51)	1.11 (0.89–1.40)
Pharmacy fill	1.01 (0.85–1.19)	0.98 (0.83–1.15)
1‐y persistence
Patient report	1.03 (0.87–1.22)	1.01 (0.86–1.18)
Pharmacy fill	0.95 (0.83–1.10)	0.93 (0.82–1.05)

HR indicates hazard ratio; MACE, major adverse cardiovascular events; and OR, odds ratio.

However, the randomized copayment intervention led to increased 1‐year medication persistence, in both hospitals with and without pre‐existing medication assistance programs (Table 4). This effect persisted after multivariable adjustment for pharmacy‐based persistence (Table 4). The intervention did not lead to a significant change in MACE in either hospital group. Among hospitals randomized to the copayment voucher intervention, all patients received a copayment waiving voucher at discharge, and voucher usage rates over the next year were similar at hospitals with and without pre‐study free medication programs (73.2% versus 71.9%, P=0.29).

Table 4

The Association of the Randomized Copayment Reduction Intervention With Outcomes Among Patients Treated at Hospitals With and Without Pre‐Existing Free Medication Programs

Outcome at 1 Y	Pre‐Study Hospital Ability to Provide Free Medicationa	Intervention	Usual Care	Adjusted OR/HR (95% CI)	P interaction
P2Y12 Inhibitor persistence (patient‐report)	Yes	87.2%	83.0%	1.25 (0.98–1.59)	0.85
	No	87.4%	84.2%	1.18 (0.96–1.44)
P2Y12 inhibitor persistence (pharmacy)	Yes	53.6%	44.0%	1.45 (1.20–1.75)	0.71
	No	59.0%	48.3%	1.46 (1.25–1.70)
MACE	Yes	10.2%	10.3%	1.24 (0.98–1.57)	0.21
	No	10.7%	10.8%	1.04 (0.86–1.27)

HR indicates hazard ratio; MACE, major adverse cardiovascular events; and OR, odds ratio.

Hospital reports ability to provide free P2Y12 inhibitor for a short period (30 days) to patients who cannot afford the prescribed P2Y12 inhibitor therapy, sponsored by the hospital or an external organization.

At 1 year after completing study enrollment, 210/262 (80.2%) hospitals completed a follow‐up survey. More hospitals (59.5%, n=125) now reported ability to provide short‐term free P2Y12 inhibitor at discharge. Among hospitals completing the follow‐up survey without free medication programs at baseline, 53/107 (49.5%) converted to providing free medication programs; this was not significantly different between sites randomized to the copayment intervention versus usual care (43.5% versus 54.1%, P=0.28).

Discussion

Before participation in the ARTEMIS trial, approximately half of hospitals already offered a free short‐term supply of P2Y12 inhibitors to patients who cannot afford treatment. However, patients treated at these hospitals did not have better 1‐year medication persistence or MACE rates compared with patients treated at hospitals that did not offer this type of short‐term medication assistance. In contrast, we found that the ARTEMIS intervention—waived P2Y12 inhibitors copayments for the guideline‐recommeneded 1‐year course—was associated with improved persistence, and this effect was similar in both hospitals with and without pre‐existing free short‐term medication programs. This suggests that longer‐term financial assistance strategies may be more effective than short‐term support in improving patient persistence to medications.

Hospitals used varying strategies to promote medication persistence after discharge. In our survey, many hospitals invest resources in screening for potential non‐adherence, providing medication reminder tools and/or pharmacist‐led education. In particular, interventions addressing financial barriers to medication use are common; these include dedicated resources to assess drug affordability, troubleshoot financial barriers, and facilitate medication filling. Many hospitals in our study have taken on the cost of providing free short‐term medication supplies; surprisingly, this did not vary by hospital size or profit status, both of which may correlate with resource richness. Hospitals provide this cost assistance for many reasons. Early discontinuation of P2Y12 inhibitors is common and can be catastrophic,8, 18 but also often predictable and preventable. Hospitals that improve post‐discharge persistence can avoid adverse outcomes leading to costly readmissions. Assistance programs may also respond to patient need, since hospitals with assistance programs treated a higher percentage of patients with self‐reported financial hardship related to medications. Finally, hospitals with pre‐existing medication assistance programs were more likely to treat MI patients with higher potency P2Y12 inhibitors. This observed association may be bi‐directional: higher copayment costs relative to clopidogrel may have motivated hospitals to help defray these costs, while the existence of assistance programs may encourage clinicians to more frequently prescribe higher‐cost agents. Hospitals with pre‐existing free medication programs also more frequently used other inteventions to promote medication adherence, yet were not associated with higher rates of medication persistence. Potential explanations include the following: (1) These programs, while available, may have been infrequently used or suboptimally deployed and thus did not reach patients with the most potential benefit; (2) The short duration of medication support may not be sufficient to influence long‐term medication persistence and outcomes; (3) ARTEMIS enrolled insured patients with prescription drug coverage, therefore the impact of pre‐existing free medication programs may be underestimated. Patients without insurance may be the primary target of hospital‐based free medication programs, but were not captured in our data. However, even in this insured study population, most patients in our study reported financial hardship related to medications and 17% reported recently failing to fill a medication because of cost concerns. Finally, hospitals randomized to the ARTEMIS intervention may not have used alternative free medication programs because the study intervention provided full copayment support. It is therefore important to note that pre‐existing free medications programs were not associated with improved persistence even when examining only control hospitals, where patients did not receive copayment support from ARTEMIS.

We anticipated a lesser impact of the randomized copayment intervention among hospitals already providing free short‐term medication assistance. However, the uptake of the randomized intervention was similar in hospitals with and without pre‐existing programs, and the intervention was associated with a similar magnitude of improved persistence for patients treated in hospitals with and without pre‐study free medication programs. The MI FREEE (Post‐Myocardial Infarction Free RX Event and Economic Evaluation) trial randomized patients to full coverage of secondary prevention medications, resulting in 4% to 6% higher adherence rates and lower vascular event rates.19 Combined, these studies argue for systematic efforts to reduce long‐term out‐of‐pocket medication costs for patients post‐MI regardless of the existence of alternative hospital‐based strategies. In contrast, the HeartStrong study did not improve medication adherence with a behavioral economic approach that included financial incentives for adherence.11 In our study, neither pre‐study free medication programs nor the randomized intervention were associated with improvement in MACE rates. This is consistent with some prior randomized trials in which interventions improved adherence or persistence rates without impact on clinical outcomes.10, 15 Despite this uncertain impact on clinical outcomes, a greater proportion of hospitals provided free medication programs when surveyed at the end of the study, with half of the hospitals previously without free medication programs converted to now offering free short‐term medication assistance.

Medication cost remains an important consideration following MI. Patients in ARTEMIS commonly reported concerns about medication cost and financial hardship even before discharge from the MI admission. It is possible that cost‐reduction programs, either through pre‐existing medication assistance programs or the randomized voucher intervention, resulted in financial and quality‐of‐life benefits for patients that were unrelated to medication persistence or recurrent cardiovascular events. From a resource usage perspective, hospitals using these programs need to identify and target individuals at the greatest risk of cost‐related non‐adherence, and triage patients with non‐financial adherence barriers to other supportive measures that might be more effective. Further study is needed to identify an optimal cost‐reduction strategy, alone or in combination with non‐financial interventions, for improving medication‐taking habits and clinical outcomes.

This study has several important limitations. ARTEMIS was limited to sites in the United States and patients with US‐based health insurance. The impact of cost‐reduction strategies in other contexts cannot be assessed. Short‐term copayment reduction could be more effective among patients without insurance coverage for prescription medication. In addition, free medication programs were identified by hospital self‐report. We could not assess the quality or scope of these programs, such as patient eligibility and how drug was delivered to the patient. It is possible that these programs could be effective if delivered systematically to appropriate patients. This post hoc analysis used rigorous propensity models that encompassed a wide spectrum of patient‐ and hospital‐level covariates, but may still be limited by unmeasured confounders. The randomized copayment voucher intervention resulted in numerically higher persistence when measured by both patient report (which was the prespecified co‐primary end point of the ARTEMIS trial) and pharmacy fills, but only pharmacy fill persistence reached statistical significance after adjustment. This may be related to recall bias and overestimation of persistence by self‐report. Medication persistence was measured by pharmacy fills using a linked pharmacy claims data source that encompasses the majority of US pharmacies, however, persistence may be underestimated if filled at a pharmacy that is not submitting claims data, and pharmacy fills may also misclassify persistence, though this misclassification should be non‐differential and not impact the study findings.20, 21 Our analysis could be biased toward the null if lower income patients who are more likely to benefit from a cost‐reduction strategy are less likely to use pharmacies that participate in the Symphony Health data aggregator system. When considering both assessment methods, our results support an overall significant impact of the randomized intervention on persistence, similar to the primary ARTEMIS analysis. We were unable to specifically assess the impact of free medication programs on initiation of P2Y12 inhibitors, since our first assessment of persistence occurred at 90 days. Finally, some sites reported changing their ability to provide free medication programs in the follow up survey at 1 year after completion of ARTEMIS enrollment. Some of these changes could have occurred during the study period, thereby biasing our analysis towards the null. For this, and the other reasons noted above, we cannot exclude the possibility that some hospital free medication programs may be effective.

Conclusions

Before participation in the ARTEMIS trial, approximately half of hospitals already offered a free short‐term supply of P2Y12 inhibitors to patients with myocardial infarction. However, patients enrolled in the ARTEMIS trial—with insurance coverage for prescription medications—at these hospitals did not have better 1‐year medication persistence or MACE rates compared with patients treated at hospitals that did not offer this type of short‐term medication assistance. In contrast, eliminating patient copayments for 1 year, was effective in improving persistence to P2Y12 inhibitors in hospitals with and without pre‐existing free short‐term medication programs. Health systems may consider systematic and long‐term strategies to reduce cost burden for their patients rather than short‐term interventions.

Sources of Funding

This study was supported by a research grant from AstraZeneca to the Duke Clinical Research Institute. The trial protocol and data collection forms were designed and written by the academic investigators. Duke Clinical Research Institute, Durham, North Carolina, served as the coordinating center and was responsible for all study data collection and analyses. The sponsor reviewed this article but had no role in the data analysis.

Disclosures

Anstrom reports consulting honoraria from AstraZeneca. Cannon reports research grant support from Amgen, Arisaph, Boehringer Ingelheim, Bristol‐Meyers Squibb, Daiichi Sankyo, Janssen, Merck, and Takeda and consulting honoraria for Alnylam, Amarin, Amgen, Arisaph, AstraZeneca, Boehringer Ingelheim, Bristol‐Myers Squibb, Eisai, GlaxoSmithKline, Janssen, Kowa, Lipimedix, Merck, Pfiser, Regeneron, Sanofi, and Takeda. Henry reported receiving a steering committee honorarium for ARTEMIS from AstraZeneca. Fonarow reports consulting honoraria from Abbott, Amgen, AstraZeneca, Bayer, Janssen, Novartis. Choudhry reported receiving research grant support to Brigham and Women's Hospital from Merck, Sanofi, AstraZeneca, CVS Health, and Medisafe Inc. Fonseca was an employee and shareholder of AstraZeneca during the conduct of this study. Bhalla and Eudicone are employees of AstraZeneca. Peterson reported receiving grants and/or personal fees from Bayer Pharmaceuticals, Janssen Pharmaceuticals, AstraZeneca, Genentech, and the American Heart Association GWTG–Stroke Analytic and has served as a consultant/advisory board member for Janssen, Boehringer Ingelheim, Sanofi, Bayer, Merck, AstraZeneca, Signal Path, and Venable. Wang reported receiving research grant support to the Duke Clinical Research Institute from AstraZeneca, Bristol‐Myers Squibb, Cryolife, Pfizer, Portola, and Regeneron and receiving consulting honoraria from AstraZeneca and Sanofi. The remaining authors have no disclosures to report.

Table S1–S3

Figure S1–S3

Click here for additional data file.