Eliminating Medication Copayments for Low-Income Older Adults at High Cardiovascular Risk: A Randomized Controlled Trial

Background:

One in eight people with heart disease has poor medication adherence that, in part, is related to copayment costs. This study tested whether eliminating copayments for high-value medications among low-income older adults at high cardiovascular risk would improve clinical outcomes.

Methods:

This randomized 2×2 factorial trial studied 2 distinct interventions in Alberta, Canada: eliminating copayments for high-value preventive medications and a self-management education and support program (reported separately). The findings for the first intervention, which waived the usual 30% copayment on 15 medication classes commonly used to reduce cardiovascular events, compared with usual copayment, is reported here. The primary outcome was the composite of death, myocardial infarction, stroke, coronary revascularization, and cardiovascular-related hospitalizations over a 3-year follow-up. Rates of the primary outcome and its components were compared using negative binomial regression. Secondary outcomes included quality of life (Euroqol 5-dimension index score), medication adherence, and overall health care costs.

Results:

A total of 4761 individuals were randomized and followed for a median of 36 months. There was no evidence of statistical interaction (P=0.99) or of a synergistic effect between the 2 interventions in the factorial trial with respect to the primary outcome, which allowed us to evaluate the effect of each intervention separately. The rate of the primary outcome was not reduced by copayment elimination, (521 versus 533 events, incidence rate ratio 0.84 [95% CI, 0.66–1.07], P=0.162). The incidence rate ratio for nonfatal myocardial infarction, nonfatal stroke, and cardiovascular death (0.97 [95% CI, 0.67–1.39]), death (0.94 [95% CI, 0.80 to 1.11]), and cardiovascular-related hospitalizations (0.78 [95% CI, 0.57 to 1.06]) did not differ between groups. No significant between-group changes in quality of life over time were observed (mean difference, 0.012 [95% CI, –0.006 to 0.030], P=0.19). The proportion of participants who were adherent to statins was 0.72 versus 0.69 for the copayment elimination versus usual copayment groups, respectively (mean difference, 0.03 [95% CI, 0.006–0.06], P=0.016). Overall adjusted health care costs did not differ ($3575 [95% CI, –605 to 7168], P=0.098).

Conclusions:

In low-income adults at high cardiovascular risk, eliminating copayments (average, $35/mo) did not improve clinical outcomes or reduce health care costs, despite a modest improvement in adherence to medications.

Registration:

URL: https://www.clinicaltrials.gov; Unique identifier: NCT02579655.

Clinical Perspective.

What Is New?

• This trial examined the effect of eliminating medication copayments for people with a broad set of chronic diseases relevant to cardiovascular outcomes, and for a comprehensive list of high-value medications.

• It tests the hypothesis that a small improvement in adherence to effective medications by eliminating copayments, with the potential to improve blood pressure, glycemic control, and cardiovascular risk, could lead to better clinical outcomes.

What Are the Clinical Implications?

• In low-income older adults at high cardiovascular risk, eliminating medication copayments resulted in a small improvement in adherence, but did not improve clinical outcomes or quality of life, or reduce health care costs.

• Policy makers will need to compare the magnitude of out-of-pocket payments for usual level of copayment associated with government insurance in Alberta, Canada, with those associated with their own insurance plans.

Editorial, see p 1515

Although prescription medications are a key element of managing and preventing chronic diseases, poor adherence to these medications is an issue.^1–3 One major reason, affecting 1 in 8 Americans with cardiovascular disease,⁴ is cost-related nonadherence, which is associated with worse health outcomes.⁵ Although cost-related nonadherence occurs in individuals without medication insurance, it can also occur in those with insurance who are subject to cost-sharing.⁵

Most individuals with medication insurance are subject to some form of cost-sharing (ie, copayments or deductibles), requiring that patients pay a portion of the medication cost. Copayments are typically 20% to 30% of the total cost of the medication. Other plans use deductibles, which may be as high as 5% to 20% of household income. In a large survey of Canadian adults with chronic diseases, up to 20% identified financial barriers to obtaining their medications, and patients with these barriers were 30% to 55% less likely to use statins than those without.⁶

The effect of cost-sharing on medication adherence was identified as uncertain in a 2007 Cochrane systematic review.¹ In 2 large cluster randomized trials that tested the effect of eliminating coinsurance after myocardial infarction (MI) for a restricted set of medications, both showed a small increase in adherence (from 3.3% to 6%), and 1 study showed an 11% reduction in rates of total major vascular events or revascularization.^7,8 However, the effect of eliminating copayments for patients with a broader set of chronic diseases or for a comprehensive list of high-value medications is unknown.

This trial aimed to determine the effect of eliminating copayment for medications for cardiovascular risk reduction on health outcomes and overall health care costs in low-income older adults. The goal was to increase initiation and adherence to a broad range of medications used to reduce the risk of cardiovascular events including those used to manage hypertension, hypercholesterolemia, diabetes, coronary artery disease, and heart failure.

Methods

The ACCESS study (Assessing Outcomes of Enhanced Chronic Disease Care Through Patient Education and a Value-Based Formulary Study) tested the effect of a value-based formulary that eliminates copayment for medications for chronic diseases on health outcomes and health care costs for low-income older adults at high cardiovascular risk.⁹ This pragmatic, parallel group, open-label, factorial randomized controlled trial design examined the effect on cardiovascular outcomes, mortality, and hospitalizations for cardiovascular-related ambulatory care–sensitive conditions over a 3-year follow-up period. The trial is reported according to CONSORT (Consolidated Standards of Reporting Trials) guidelines (Supplemental Methods S1). Minimal changes were made to the study protocol after commencement.¹⁰ This study was approved by the University of Calgary’s Conjoint Health Research Ethics Board, and all participants provided informed consent. The data used for outcome ascertainment in this study belong to Alberta Health and are protected by the Alberta Health Information Act and are therefore not permitted to be released to third parties outside the study team. The other trial data (surveys) could be made available upon reasonable request.

Setting

The ACCESS trial recruited community-based participants living in Alberta, Canada. All participants had public health insurance, which provides physician and hospital services free of charge to all residents. Participants were recruited by signs in physician offices and pharmacies across urban and rural Alberta, with letters sent to patients who had recent health system interactions (hospitalization or from a cardiac catheterization registry), with advertising in television, radio, and social media, or by word of mouth.¹¹

Interested individuals called the study recruitment line, which was staffed by professional customer service representatives, who answered basic questions and administered eligibility questions. For those who were interested and eligible, their information was forwarded to the study coordinating center, who then contacted participants to complete intake questions, obtain informed consent, and arrange for completion of baseline study questionnaires (by mail or electronic means, as per participant preference). Once these were received by the study coordinating center, participants were randomized (see Sequence Generation, Allocation Concealment, and Blinding).

Eligibility Criteria

Inclusion criteria included age ≥65 years, coverage by provincially sponsored seniors’ medication insurance, high cardiovascular risk on the basis of 1 or more of coronary artery disease, stroke, chronic kidney disease, or heart failure, or 2 or more of current smoking, diabetes, hypertension, or high cholesterol, and household income <$50 000 Canadian dollars per year. Exclusion criteria included additional insurance coverage that reduced cost sharing, receiving medications administered by a nurse or facility, or the inability to participate in self-management modules because of cognitive impairment or a language barrier (unless they had a family member proxy).

Intervention

This 2×2 factorial trial tested the effect of 2 interventions, eliminating copayments for high-value preventive medications and a tailored self-management education and support program, which is reported in a separate article. All participants in the copayment elimination arm were exempted from copayments on high-value preventive medications at the dispensing pharmacy through changes made to their existing government-sponsored medication insurance plan. Specifically, copayments were eliminated on 15 classes of medications that have been shown to improve clinically relevant outcomes related to MI, stroke, other vascular disease, or chronic kidney disease including statins, other cholesterol-lowering drugs, β-blockers, ACEi (angiotensin-converting enzyme inhibitors), ARB (angiotensin receptor blockers), calcium channel blockers, diuretics, other antihypertensive medications, antiarrhythmic agents, nitrates, anticoagulants, antihyperglycemic medications, and antiplatelet agents (Supplemental Methods S2). During the study period, if new agents from eligible classes of medication were added to the government-sponsored formulary, copayments for these medications were also waived.

Control Arm (Usual Copayment)

The control arm was the usual universal public pharmaceutical insurance plan for seniors, which is provided without premium to people ages ≥65 years and includes access to formulary medications with a 30% copayment to a maximum of $25 Canadian dollars per medication dispensation.

Outcomes

The primary outcome was the composite rate of all-cause mortality, MI, stroke, coronary revascularization, and hospitalizations for cardiovascular-related ambulatory care-sensitive conditions (ie, heart failure, coronary artery disease, diabetes, hypertension, and chronic kidney disease, including recurrent events), each defined using validated algorithms applied to administrative health data (Supplemental Methods S3). Secondary outcomes included individual components of the primary end point (including recurrent events), major adverse cardiovascular events (nonfatal MI, nonfatal stroke, and cardiovascular death [a post hoc outcome]), medication adherence, overall quality of life (measured using the EQ-5D [EuroQol 5 dimensions] index score), and health care costs.

Sequence Generation, Allocation Concealment, and Blinding

Participants were randomized in a 1:1:1:1 fashion (using variable block sizes) by a computer-generated randomization algorithm, ensuring concealed randomization. Four treatment groups were generated (copayment elimination only, individualized self-management only, both, and control). Randomization was stratified by age (<70 years, ≥70 years), annual income (<$30 000, $30 000–$50 000); and sex (male, female).

Blinding was not possible for this intervention. To reduce risk of ascertainment bias, administrative data codes defining the primary and secondary outcomes were specified in advance (Supplemental Methods S3), and analysis was conducted in a blinded fashion until all outcomes were finalized.

Data Collection

The components of the primary outcome, medication adherence, and health care cost outcomes, were assessed using provincial administrative health data (including vital statistics, hospitalizations, emergency department visits, pharmacy and physician claims, and all health care costs).¹² This dataset has been used for many observational studies^13–16 and for assessing outcomes in a randomized controlled trial of over 20 000 individuals.¹⁷ Each component of the primary outcome was assessed using validated algorithms applied to administrative health data, all with positive predictive values and sensitivity >70%^18–20 (Supplemental Methods S3).

Medication adherence was measured by the number of days dispensed/number of days between prescription renewals.^21,22 Quality of life was assessed by survey, completed at baseline and at the end of the study using the EQ-5D-5L index score. The index score was estimated using the Canadian algorithm,²³ which is anchored by convention at 1.0 for full health and 0.0 for death²²; as such, participants who died during follow-up were assigned a score of 0.0.

Total health care costs included medication costs (including any copayment), all-cause hospitalizations, emergency department visits, physician visits, outpatient procedures, diagnostic imaging, and laboratory testing. Case mix grouper methods and ambulatory case costing methods were used to estimate hospital and outpatient costs, respectively.¹²

Statistical Analysis

The use of a Poisson model was explored to test the main effects of the intervention on the primary outcome. However, there was evidence of overdispersion (greater variance than mean) threatening the assumptions of the Poisson model, reflected by a P value <0.0001 from the likelihood-ratio test of the dispersion parameter α being equal to 0. Therefore, as prespecified, adjusted rates using negative binomial regression were used to account for the overdispersion,²⁴ and a negative binomial model was used to test the main effects of the effect of the intervention on the rate of the primary outcome.

All analyses controlled for the variables upon which randomization was stratified (age, sex, income).²⁴ Given that this was a factorial trial with 2 distinct interventions, analyses also controlled for receipt of the other study intervention (self-management and education intervention) by including this as a covariate in the models.

Prespecified secondary analyses included time-to-event analysis with Kaplan-Meier plots, and use of the cumulative incidence function to account for the competing risk of death.

Participants with medication supplies to cover ≥80% of observed treatment days were considered adherent (proportion of days covered; PDC80).²⁵ Adherence was assessed over the full study period for 2 commonly prescribed classes of covered medications that guidelines recommend for use in nearly all patients eligible for this study (statins, and ACEi/ARB).²⁶ Between-group differences in the binary medication adherence variable were assessed using log binomial regression (generalized linear models with a log link). Mixed models were used to compare end of study EQ-5D index scores. In secondary analyses, multiple imputation using chained equations was used to impute for missing EQ-5D data. Mixed effects models were fitted using random intercepts with an unstructured within-patient covariance structure to analyze the change in EQ-5D index scores from baseline to study end. Generalized linear models were used to compare total costs across groups, using a gamma distribution and log-link function.

The target sample size was 4714 participants, which was estimated to provide 80% power to identify a 12% absolute reduction in the primary outcome, assuming a primary outcome event rate of 14 per 100 participant years, α=5%, average follow-up of 3 years, and no important interaction between the 2 interventions.⁹ Because recurrent events were included within the primary outcome, the sample size was estimated for Poisson regression analysis.²⁷ All analyses used the intention-to-treat principle. Subgroup analyses were done using prespecified groups, including demographic characteristics, baseline chronic diseases, and income status.

Results

Participants

A total of 8870 potentially eligible individuals were assessed, excluding 2793 who were not eligible, and 1316 who did not complete enrollment, leaving 4761 participants who were randomized (Figure 1). Most participants (37%) were recruited from health care providers, followed by those who received study information after a health system interaction (33%), television, radio, and social media advertising (12%), and word of mouth (18%).¹¹ Participants were followed for a median of 36 months, and primary outcome data were available for 99.6% of participants; the remainder emigrated from Alberta during the study. Recruitment occurred between November 2015 and June 2018, with follow-up ending on March 31, 2021—the end of the fiscal year for administrative data.

Figure 1.

Participant flow.

Baseline characteristics were well-balanced between arms, with small differences in self-reported coronary artery disease and heart failure (Table 1). Mean age was 74.4 years (SD, 6.4 years), 57.6% of participants had annual income <$30 000, and the mean EQ-5D index score was 0.69 (SD, 0.22). More than half of participants had diabetes (55.5%) and coronary disease (50.3%), and 10.2% of participants were current smokers. At baseline, 68.4% of participants self-reported full adherence to statin therapy.

Table 1.

Baseline Characteristics for Participants, Usual Copayment Versus Copayment Elimination

Clinical Outcomes

Using a negative binomial model, there was no evidence of statistical interaction (P=0.99) or of a synergistic effect between the 2 interventions in the factorial trial (self-management intervention and copayment elimination) with respect to the primary outcome, which allowed us to evaluate the effect of each intervention separately. Of participants in the copayment elimination and the usual copayment arms, 85.9% and 85.8% of participants had no primary outcome event, 9.6% and 8.8% had 1 primary outcome event, and 4.5% and 5.3% had 2 or more primary events, respectively (P=0.43) (Table S1). Of the primary outcome events, 40% were related to cardiovascular-related ambulatory care-sensitive hospitalizations, and of those, 68.8% were due to heart failure.

The number or rate of primary outcome events did not differ among those who received copayment elimination (521 events total) compared with usual copayment participants (533 events total) (incidence rate ratio [IRR], 0.84 [95% CI, 0.66–1.07], P=0.16) (Table 2). A comparison of incidence rate ratios derived by Poisson and negative binomial regression is available in Table S2.

Table 2.

Primary and Secondary Outcomes, Copayment Elimination Compared With Usual Copayment, Using Negative Binomial Regression

Results were similar when a time to first event approach was used (hazard ratio, 0.99 [95% CI, 0.85–1.16], P=0.94) (Figure S1), and considering the risk of competing events (Figure S2). There were no differences between arms in the incidence rate ratio for major adverse cardiovascular events (nonfatal MI, nonfatal stroke, cardiovascular death) (0.97 [95% CI, 0.67–1.39]) (a post hoc analysis), death (0.94 [95% CI, 0.80–1.11]), or hospitalizations for cardiovascular-related ambulatory care–sensitive conditions (0.78 [95% CI, 0.57–1.06]) (Table 2).

Subgroup Analyses

Prespecified subgroup analyses are reported in Figure 2. There was no evidence of statistically significant effect modification by any of the characteristics of interest.

Figure 2.

Prespecified subgroup analyses, copayment elimination versus usual copayments, using negative binomial regression. ACEi indicates angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; and IRR, incidence rate ratio.

Quality of Life

All participants completed an EQ-5D survey at baseline. At study end, 3559 (74.8%) patients had completed an EQ-5D survey or had died (where a 0 is assigned to the EQ-5D score). A total of 1202 (25.2%) patients had true missing EQ5D scores, including 318 (13.4%) in the copayment elimination group and 884 (37.2%) in the usual copayment group. EQ-5D index scores declined significantly over time in both arms, driven by the conventional practice of assigning a utility of 0 to those who died (–0.092 [95% CI, –0.104 to –0.080] and –0.104 [95% CI, –0.118 to –0.090]) for the copayment elimination and usual copayment groups, respectively. There was no difference between arms (0.012 [95% CI, –0.006 to 0.030], P=0.19), regardless of the method used to take account of missing data (Table S3).

Adherence to Statins and ACEi/ARB

The proportion of participants who received at least 1 statin prescription during the 3 years was 0.85 and 0.84 for the copayment elimination and usual copayment arms, respectively (mean difference, 0.01 [95% CI, –0.01 to 0.03], P=0.32) (Table S4). The proportion of all participants who were adherent (PDC80) to statins was 0.72 and 0.68 for the copayment elimination and usual copayment groups, respectively (mean difference, 0.03 [95% CI, 0.006–0.06], P=0.017). The proportion of all participants who were adherent (PDC80) to ACEi/ARB was 0.66 and 0.63 for the copayment elimination and usual copayment groups, respectively (mean difference, 0.03 [95% CI, 0.007–0.061], P=0.01) (Table S4).

Health Care Costs

During the entire study period, the government insurer paid an average of CAD $4002/person (95% CI, $3840–$4154) and $2847/person (95% CI, $2732–$2963) for covered medications for the copayment elimination and usual copayment groups, respectively (Table 3). This difference (mean, $1155/person [95% CI, $964–$1345], P<0.001; median, $901 [interquartile range, $441– $1512])—or a mean difference of $35/mo for patients (over an average of 33.5 months) was due to taking on patient copayments (a difference of $24/mo) but also because of an increase in medication use by participants in the copayment elimination arm (a change of $11/mo) (Table 3). There was no difference in the cost for medications that were used by patients outside the 15 covered medication classes (mean difference, –$152 [95% CI, –$731 to $427], P=0.61). There was no difference in overall health care adjusted costs for the copayment elimination and usual copayment arms (mean difference, $3575 [95% CI, –$605 to $7168], P=0.098).

Table 3.

Mean Health Care Unadjusted Costs in Canadian Dollars Over Full Study Period, Copayment Elimination Compared With Usual Copayment

Discussion

Among low-income older adults at high cardiovascular risk, elimination of copayments for high-value preventive medications reduced mean participant-borne medication costs by an average of $35/mo, but did not reduce the rate of the primary outcome, composed of death, adverse cardiovascular events, and hospitalization for potentially avoidable cardiovascular-related conditions. Elimination of copayments also did not affect overall quality of life or total health care costs, although it did modestly improve adherence to statins and ACEi/ARB. No subgroups were identified that benefited from copayment elimination, even those with the lowest baseline incomes.

The results should be considered in the context of other published studies. Two cluster randomized controlled trials have compared copayment elimination for participants after MI, with varying results. In the MI-FREEE study (Post-Myocardial Infarction Free Rx Event and Economic Evaluation),⁷ copayment elimination for statins, β-blockers, ACEi, or ARB after MI did not significantly reduce rates of the trial’s primary outcome (first major vascular event or revascularization), but improved medication adherence by 4% to 6% across medication classes and lowered rates of total major vascular events or revascularization by 11% (hazard ratio, 0.89 [95% CI, 0.80–0.99], P=0.03). In the ARTEMIS trial (Affordability and Real-World Antiplatelet Treatment Effectiveness After Myocardial Infarction Study),⁸ provision of a voucher to offset medication copayments led to a 3.3% absolute increase in participant-reported persistence with P2Y₁₂ inhibitors, but no significant reduction in 1-year adverse cardiovascular outcomes compared with no voucher (adjusted hazard ratio, 1.07 [95% CI, 0.93–1.25]).

The RAND study (Research and Development) compared comprehensive health care insurance (including for medications) with 3 different cost-sharing strategies,²⁸ and found that full insurance increased the use of antihypertensive medications among people with hypertension (20% absolute increase) and significantly decreased diastolic blood pressure (−1.9 mm Hg [95% CI, −3.5 to −0.3 mm Hg]), compared with each of the 3 cost-sharing strategies.²⁹ The CLEAN-MEDS trial examined the effect of providing a list of 128 “essential medicines” without charge to outpatients who reported cost barriers.³⁰ Adherence to treatment was 11% higher in those randomized to receive free distribution (95% CI, 4.9%–18.4%), but the effect on disease-specific surrogate outcomes (blood pressure, hemoglobin A1C/glycated hemoglobin concentration, cholesterol) was variable. Last, in a 1-year randomized controlled trial involving 479 individuals with uncontrolled hypertension, eliminating copayments for antihypertensive medications did not improve blood pressure control (P=0.36).³¹

This study has several strengths that should be considered. It evaluated an important policy option (eliminating copayments for high-value medications) in a large group of community-based older adults at high cardiovascular risk. The randomized design and complete follow-up (>99.6% of participants) have minimized the risk of confounding and selection bias.

This study has 3 main limitations. First, although it assessed a widely implemented policy (copayments for medications), it is possible that the copayment avoided (an average difference of $35/mo in out-of-pocket payments) was not costly enough to consistently reduce medication use, perhaps because the government-sponsored plan for the usual copayment group limited the maximum copayment to $25 per prescription dispensation. Second, adherence was relatively high even in the usual copayment group, which may have reduced the opportunity for copayment elimination to increase adherence sufficiently to improve clinical outcomes. Third, the sample size was based on a projected annual composite primary outcome event rate of 14 per 100 participant years, whereas the observed rate was 8.4 per 100 participant years, suggesting that this study may have been underpowered due to enrolling a healthier cohort than expected. These findings and the associated 95% CIs do argue against a large benefit of copayment elimination for a population-based medication insurance plan that has a copayment similar in magnitude to that used within the usual copayment group in this study, although a smaller benefit cannot be excluded.

Eliminating copayments may be effective for a subset of participants with severe financial barriers, or for people with very low baseline adherence, but it was not possible to determine this in our study. Although this trial did not demonstrate large changes in adherence no improvement in clinical outcomes, given the modest copayments associated with government medication insurance in Alberta, Canada, the findings may not be generalizable to countries like the United States, where copayments and other forms of cost-sharing can be significantly higher. Policy makers should compare the magnitude of out-of-pocket payments with usual copayments associated with government insurance in Alberta, Canada, with those in their own insurance plans when interpreting how these results may apply to their setting.

In low-income adults at high cardiovascular risk in Alberta, Canada, eliminating copayments of approximately $35 per month did not improve clinical outcomes or reduce health care costs, despite a modest improvement in adherence to medications. This information will assist health policy makers in creating formulary rules for insurance plans.

Article Information

Sources of Funding

This study was funded by a Canadian Institutes of Health Research Foundation Grant (No. 201509-FDN-353640 to B.J.M.), an Alberta Innovates Collaborative Research & Innovation Opportunity Team Grant to the Interdisciplinary Chronic Disease Collaboration, and a Major Research Grant from the Clinical Research Fund from the Cumming School of Medicine, University of Calgary. This study is based in part on data provided by Alberta Health. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta. Neither the Government of Alberta nor Alberta Health expresses any opinion in relation to this study.

Disclosures

M.T. has received honoraria for lectures not related to the topic of the current article from AstraZeneca and received honoraria donated to charity. C.M. is an employee of Alberta Health, presently in the capacity of Assistant Deputy Minister of the Pharmaceutical Branch. R.T.T. is a consultant for Shoppers Drug Mart and Emergent Biosolutions. R.T.T. has received investigator-initiated arms-length grants from Merck, Sanofi, AstraZeneca, and Pfizer. The other authors report no conflicts.

Supplemental Material

Expanded Methods S1–S3

Figures S1 and S2

Tables S1–S4

References 32–34

Supplementary Material

cir-147-1505-s001.pdf ^{(1.3MB, pdf)}