Effect of the Million Hearts Cardiovascular Disease Risk Reduction Model on Initiating and Intensifying Medications (2024)

Key Points

Question Does the Million Hearts Cardiovascular Disease Risk Reduction Model increase the initiation or intensification of statins or antihypertensive medications among high-risk Medicare patients?

Findings In this prespecified secondary analysis of a cluster randomized, pragmatic trial that included 125 436 Medicare patients, the rate of initiation or intensification was 37% for patients enrolled by organizations paid to assess and reduce cardiovascular risk vs 32% for patients enrolled by organizations that were not, indicating a statistically significant difference.

Meaning The pay-for-performance model in this study modestly improved the use of medications for patients with suboptimal cholesterol or blood pressure levels, although baseline use of medications was high.

Importance The Million Hearts Cardiovascular Disease (CVD) Risk Reduction Model pays provider organizations for measuring and reducing Medicare patients’ cardiovascular risk.

Objective To assess whether the model increases the initiation or intensification of antihypertensive medications or statins among patients with blood pressure or low-density lipoprotein (LDL) cholesterol levels above guideline thresholds for treatment intensification.

Design, Setting, and Participants This prespecified secondary analysis of a cluster-randomized, pragmatic trial included primary care and cardiology practices, health care centers, and hospital-based outpatient departments across the US. Participants included Medicare patients who were enrolled into the model in 2017 by participating organizations and who were at high risk and at medium risk of a myocardial infarction or stroke in 10 years. Patient outcomes were analyzed for 1 year postenrollment (through December 2018) using an intent-to-treat design. Analysis began November 2019.

Interventions US Centers for Medicare & Medicaid Services paid organizations for risk stratifying Medicare patients and reducing CVD risk among high-risk patients through discussing risk scores, developing individualized risk reduction plans, and following up with patients twice yearly.

Main Outcomes and Measures Initiating or intensifying statin or antihypertensive therapy within 1 year of enrollment, measured in Medicare Part D claims, and LDL cholesterol and systolic blood pressure levels approximately 1 year after enrollment, measured in usual care and reported to Centers for Medicare & Medicaid Services via a data registry (data complete for 51% of high-risk enrollees). The study’s primary outcome (incidence of first-time myocardial infarction and stroke) is not reported because the trial is ongoing.

Results A total of 330 primary care and cardiology practices, health care centers, and hospital-based outpatient departments and 125 436 Medicare patients were included in this analysis. High-risk patients in the intervention group had a mean (SD) age of 74 (4.1), 15 213 (63%) were male, 21 657 (90%) were receiving antihypertensive medication at baseline, and 16 558 (69%) were receiving statins. Almost all (21 791 [91%]) high-risk intervention group patients had above-threshold systolic blood pressure level (>130 mm Hg), LDL cholesterol level (>70 mg/dL), or both. Patients in the intervention group with these risk factors were more likely than control patients (8127 [37.3%] vs 4753 [32.4%]; adjusted difference in percentage points, 4.8; 95% CI, 2.9-6.7; P < .001) to initiate or intensify statins or antihypertensive medication. Centers for Medicare & Medicaid Services did not pay for CVD risk reduction for medium-risk enrollees, but initiation or intensification rates for these enrollees were also higher in the intervention vs control groups (12 668 [27.9%] vs 7544 [24.8%]; adjusted difference in percentage points, 3.1; 95% CI, 1.9-4.3; P < .001). Among high-risk enrollees with clinical data approximately 1 year after enrollment, LDL cholesterol level was slightly lower in the intervention vs control groups (mean [SD], 89 [31.8] vs 91 [32.1] mg/dL; adjusted difference in percentage points, −1.8; 95% CI, −2.9 to −0.6; P = .002), as was systolic blood pressure (mean [SD], 133 [15.7] vs 135 [16.4] mm Hg; adjusted difference in percentage points, −1.7; 95% CI, −2.8 to −0.6; P = .003).

Conclusions and Relevance In this study, a pay-for-performance model led to modest increases in the use of CVD medications in a range of organizations, despite high medication use at baseline.

Although many risk factors for cardiovascular disease (CVD) have declined throughout the past 50 years, CVD remains a leading cause of death and disability in the United States.^1,2 Improvements in diet and exercise, smoking cessation, and appropriate use of preventive medications could substantially reduce the burden of CVD.^3,4 In 2017, the US Centers for Medicare & Medicaid Services (CMS) launched the Million Hearts CVD Risk Reduction Model to reduce the incidence of first-time myocardial infarctions and strokes among Medicare patients.⁵ In this model, CMS pays provider organizations (1) to assess Medicare patients’ risk of having a myocardial infarction or stroke over 10 years⁶ and (2) for reducing that risk among high-risk patients (those with a 30% or higher risk). CMS is testing this model in a pragmatic, cluster-randomized trial including primary care and cardiology practices, health care centers, and hospital outpatient departments throughout the US.

Although the Million Hearts Model does not prescribe how provider organizations should reduce CVD risk, one possible strategy is to increase use of statins or antihypertensive medications. Guidelines recommend patients consider statins if they have low-density lipoprotein (LDL) cholesterol level of 70 mg/dL or higher (to convert to millimoles per liter, multiply by 0.0259) and a 10-year CVD risk greater than 7.5%. Similarly, guidelines recommend antihypertensive medication for patients with systolic blood pressure level of 130 mm Hg or higher and a 10-year risk greater than 10%.^7-9 Each of these medications can reduce CVD events by 15% to 25%.³

This study examines whether the Million Hearts Model affected a prespecified, intermediate outcome¹⁰: initiation or intensification of statins or antihypertensive medications within 1 year of enrollment among high-risk patients indicated for these medications. We also tested whether the model increased CVD medications for patients with medium risk (15% to 30% likelihood of a CVD event over 10 years). CMS does not pay for risk reduction for this medium-risk group, but clinicians may still recommend medications if, through risk stratification, they become more aware of elevated risk. Finally, for high-risk patients, we tested whether the model was associated with reduced LDL cholesterol or systolic blood pressure levels. We did not test the study’s primary outcome (first-time incidence of myocardial infarction or stroke over 5 years) because the trial is ongoing.

The trial protocol can be found in Supplement 1. From May 2015 to April 2016, CMS solicited applications from organizations throughout the country to participate in the model. Organizations were eligible if they had 1 or more practitioner (physician, physician assistant, or nurse practitioner) billing for Medicare Part B services and with a certified electronic health record. CMS enrolled all eligible organizations that agreed to model provisions. CMS used a minimization procedure, equivalent to random assignment,¹¹ to allocate half the organizations to the intervention group and half to the usual care control group. This procedure ensured the 2 groups were similar in size, location, and expected number of Medicare patients (eMethods 1 in Supplement 2). The trial was approved by the RAND institutional review board and is registered with ClinicalTrials.gov (NCT04047147).

The model will run for 5 years, starting in 2017, and was powered to detect a 7% effect on incidence of first-time myocardial infarction and stroke among high-risk patients. Our study estimates effects on secondary outcomes (medications, systolic blood pressure, and LDL cholesterol) that we hypothesized might improve within the 2.5 years covered in this analysis (January 2017 to October 2019).

Starting in January 2017, organizations enrolled patients into the model during routine office visits. The model is ongoing, but the current analysis includes patients enrolled on or before December 31, 2017. Using a registry, the organizations submitted to CMS the demographic and clinical data needed to calculate each patient’s 10-year CVD risk, officially enrolling the patient in the model. Patients were eligible for the model if they had Medicare Parts A and B, were aged 40 to 79 years, had not had a myocardial infarction or stroke, were not in hospice, and did not have end-stage kidney disease. Because the purpose of the model is to evaluate a public benefit, the Medicare program, patient consent was not required¹²; however, patients could opt out.

The intervention organizations agreed to (1) risk stratify all eligible Medicare patients using a standard 10-year CVD risk assessment tool⁶ and (2) provide cardiovascular care management services to high-risk patients. Those services included discussing risk scores with patients; developing individualized risk reduction plans incorporating patients’ goals and preferences; conducting annual in-person risk assessments using a longitudinal risk calculator designed for this model¹³; and following up with patients at least twice more each year. Organizations chose how to meet these model provisions, including how to engage individual clinicians.

To incentivize changes, CMS paid intervention organizations $10 for each patient risk stratified and, in 2017, $10 per high-risk patient per month for care management services. Starting in 2018, CMS paid $0 to $10 per high-risk patient per month depending on the organization’s success in reducing the mean risk score among high-risk patients ($0 for a mean decline of <2 percentage points, $5 for a 2- to 10-point decline, and $10 for >10 point decline).

CMS sent intervention organizations semiannual reports describing performance enrolling patients and reducing CVD risk. CMS also offered quarterly peer-to-peer learning sessions.

CMS paid control organizations that submitted clinical data ($20 per patient per submission) that CMS used to calculate Medicare patients’ CVD risk at enrollment and annually through 2019. CMS did not report CVD risk scores to control organizations or ask organizations to calculate risk scores themselves. To limit CMS outlays, CMS allowed up to 20 clinicians per control organization to enroll patients. CMS did not apply a similar cap to the intervention group because of concerns a cap might limit intervention-group participation.

This study includes 3 patient populations. The first was all Medicare patients enrolled in the Million Hearts Model in 2017 who had medium or high CVD risk at enrollment and who had Part D coverage. For this population, we assessed CVD risk and medication use at baseline and medication use in the year after enrollment. The second population was the subset of the first who were also candidates for initiation or intensification of statins or antihypertensive medications because, at enrollment, they had systolic blood pressure level of 130 mm Hg or higher, LDL cholesterol level of 70 mg/dL or higher, or both (levels above guideline thresholds for treatment intensification among people with high estimated CVD risk). For this second population, we estimated model effects on medication initiation or intensification. The third population was the subset of the second population for whom we had follow-up clinical data approximately 1 year after enrollment (eMethods 2 in Supplement 2). For this population, we assessed follow-up systolic blood pressure and LDL cholesterol level.

This article’s main outcome is whether patients with clinical risk factors (population 2) initiated or intensified statin or antihypertensive therapy within a year of enrolling in the model, as measured in Part D claims. Patients met this composite outcome if (1) they had LDL cholesterol levels of 70 mg/dL or higher at baseline and either initiated statin therapy (filled a statin prescription in the year after enrollment but not in the 4 months before) or intensified statin therapy (filled a prescription for a statin at a higher intensity or dose in the year after enrollment than was filled in the 4 months before) or (2) had systolic blood pressure levels of 130 mm Hg or higher at baseline and either initiated antihypertensive medications or intensified antihypertensive medications (added a new antihypertensive medication or increased the dosage or strength of an existing one). We also conducted sensitivity tests (1) with a blood pressure level cutoff to 140 mm Hg or higher and (2) trimming the intervention group to mimic the 20-clinician cap applied to the control group. Results from these tests (eTable 1 in Supplement 2) were very similar to the main results. We also measured the proportion of enrollees (population 1) who, within a year of enrollment, took statins (any, and by intensity level) or antihypertensive medications.

For high-risk enrollees, we measured mean follow-up lipid values (total, high-density lipoprotein, and LDL cholesterol levels in mg/dL) and systolic blood pressure (in mm Hg). Organizations collected these clinical data in the usual care setting a mean of 13 months after enrollment and submitted them to CMS via registry (eMethods 2 in Supplement 2).

We used logistic regressions to estimate model effects on initiation or intensification of CVD medications and linear regressions to measure associations with blood pressure and cholesterol (see eTable 2 in Supplement 2 for the covariates). We adjusted standard errors for clustering within organizations and used a P < .05 threshold for significance. We did not adjust for multiple comparisons.

CMS enrolled 516 organizations and randomized 260 to the intervention group and 256 to the control group (Figure 1). In both arms, slightly more than one-third of organizations did not participate in the model either because they withdrew or did not enroll any patients. The 330 participating organizations included primary care practices, cardiology and multispecialty practices, health care centers, and hospital outpatient departments and ranged in size from solo-practitioner practices to large organizations with multiple sites and more than 20 clinicians. Despite attrition, the intervention and control organizations that participated were similar on key characteristics (Table 1).

Participating intervention organizations were more likely than nonparticipants to be primary care practices (89 [52%] vs 33 [37%]) or to have been participating in other CMS initiatives when they applied for the model (88 [51%] vs 35 [39%]) (eTable 3 in Supplement 2).

More medium- and high-risk patients enrolled in the intervention group (n = 74 904) than the control group (n = 50 532) because of the 20-clinician cap that applied only to the control group (Figure 1).

The intervention and control groups were similar in demographics, CVD risk factors, recent office visits, location, and type of organization that enrolled them both for medium- and high-risk patients (Table 2).

Among high-risk patients in the intervention group, the mean (SD) age at baseline was 74 (4.1) years, 1762 (7%) were Black, 15 213 (63%) were male, and 2889 (12%) were enrolled in Medicaid (Table 2). The mean (SD) CVD risk score was 40.1 (8.8). A total of 17 328 intervention group enrollees (73%) had above-threshold LDL cholesterol level (≥70 mg/dL), and 17 624 (74%) had elevated systolic blood pressure level (≥130 mm Hg). In addition, 16 558 (69%) took statins and 21 657 (90%) took antihypertensive medication at baseline. Patients had a mean (SD) of 10 (7.6) office visits in the year before enrollment.

Compared with the high-risk group, the medium-risk intervention group enrollees had lower mean (SD) CVD risk scores (21.5 [4.2]) and a smaller proportion with elevated systolic blood pressure level (26 840 [53%]). However, the medium-risk group had a larger proportion of enrollees with an above-threshold LDL cholesterol level (40 685 [80%]), possibly owing to the smaller proportion receiving statins (31 031 [61%]). A total of 40 578 medium-risk intervention group enrollees (80%) took antihypertensive medication at baseline.

In both the intervention and control groups, 89% to 90% of all medium- and high-risk enrollees were also candidates to initiate or intensify statins or antihypertensive medication because of their systolic blood pressure or LDL cholesterol levels. As a result, the baseline characteristics for this population look very similar to those for all medium- and high-risk enrollees, except mean blood pressure and LDL cholesterol levels are higher (eTable 4 in Supplement 2).

The intervention and control organizations submitted follow-up data for only 52% (9592 of 18 307) and 45% (5149 of 11 565) of eligible high-risk patients, respectively (eMethods 2 in Supplement 2). Among this subset, the intervention and control groups were similar in baseline demographics, CVD risk factors, recent office visits, and type of organization that enrolled them (eTable 5 in Supplement 2).

The unadjusted likelihood of initiating or intensifying statins or antihypertensive medication increased steeply in the first year after enrollment (the prespecified time period for the effect analysis) and more gradually afterwards (Figure 2). The intervention group rate increased more quickly than the control group’s in the first year, with the differences between the groups staying roughly steady in later months.

Among high-risk enrollees, the regression-adjusted rate of initiation or intensification of statins or antihypertensive medication was 4.8 percentage points higher in the intervention group than the control group after 1 year of enrollment (Table 3; 95% CI, 2.9-6.7). Rates of initiation and intensification for statins and antihypertensive medication, individually, were also higher in the intervention vs control groups, with a larger adjusted difference for statins (5.2 percentage points; 95% CI, 3.6-6.9) than for antihypertensive medication (2.5 percentage points; 95% CI, 0.9-4.2). For both statins and antihypertensive medication, regression-adjusted differences were larger for initiation than intensification (Table 3). Overall, rates of statin use were 2.3 percentage points higher in the intervention group than the control group (95% CI, 1.3-3.2), with differences concentrated in high-intensity statins.

Among medium-risk enrollees, rates of initiation or intensification of statins or antihypertensive medication were also higher in the intervention group within 1 year of enrollment (Table 3; 27.9% vs 24.8%; adjusted difference, 3.1 percentage points; 95% CI, 1.9-4.3).

Among high-risk enrollees with follow-up clinical data (a mean of 13 months after enrollment), mean systolic blood pressure was 1.2% lower in the intervention vs control groups (mean [SD], 133 [15.7] vs 135 [16.4] mm Hg; adjusted difference of −1.7 mm Hg; 95% CI, −2.8 to −0.6; P = .003). Total cholesterol level was 1.2% lower (mean [SD], 165 [37.7] vs 167 [38.0] mg/dL; adjusted difference of −1.9 mg/dL; 95% CI, −3.2 to −0.7; P = .002), LDL cholesterol level was 2.0% lower (mean [SD], 89 [31.8] vs 91 [32.1] mg/dL; adjusted difference of −1.8 mg/dL; 95% CI, −2.9 to −0.6 ; P = .003), and high-density lipoprotein cholesterol level was not different (mean [SD], 48 [14.7] vs 48 [15.0] mg/dL; adjusted difference of 0.1 mg/dL; 95% CI, −0.3 to 0.5; P = .61) (Table 3).

In this secondary analysis of a large pragmatic trial where the primary outcome (first-time myocardial infarction or stroke) is not yet reported, 90% of high-risk Medicare patients already took antihypertensive medications, statins, or both at baseline. This high degree of baseline treatment is consistent with enrollees’ frequent office visits (averaging 10 per person) in the year before enrollment, providing opportunities to identify and treat CVD risk. Despite high rates of baseline treatment, 74% of patients did not have a blood pressure level at target, and 73% did not have an LDL cholesterol level below 70 mg/dL. This pattern, consistent with previous observations about clinical inertia, suggests there is room for reducing CVD risk through medications, either through increasing patients’ adherence or through intensifying medications.^14-16 Indeed, both the intervention and control group patients increased use of CVD medications in the year after enrollment, suggesting further CVD risk management was important even among the usual care control group. However, among high-risk patients with suboptimal risk factors, the Million Hearts Model increased the likelihood of initiating or intensifying statins or antihypertensive medications by 5 percentage points (or 15%). The model effects were larger (in percentage terms) for initiation than for intensification. However, because a much larger share of the population was eligible for intensification, increases in it drove the populationwide effect. Among high-risk enrollees with follow-up data a mean of 13 months after enrollment, the model was also associated with roughly 1% reductions in systolic blood pressure level and 2% reductions in LDL cholesterol level. The model also had positive spillover to the much larger medium-risk population, increasing CVD medications by 3 percentage points (or 13%) for this group.

One likely mechanism for these effects is that the model increased clinicians’ awareness of modifiable CVD risk in their patient panel. According to a survey we fielded to randomly selected clinicians (eMethods 3 in Supplement 2), the model substantially increased the extent to which clinicians reported that they risk stratified at least half of Medicare patients (69% vs 41%; eTable 6 in Supplement 2). Further, most (73%) intervention-group clinicians reported that greater use of risk stratification helped them better identify CVD risk among patients (eTable 7 in Supplement 2). This mechanism would explain increases in medications in the medium-risk group. Although CMS did not pay for risk reduction in this group, clinicians did risk stratify these patients. Another complementary explanation is that patients, more aware of their CVD risk, would be more willing to take medications.

These results largely align with smaller-scale trials introducing CVD risk stratification into clinical care, either as a standalone intervention or as part of a broader quality improvement initiative. A 2017 meta-analysis found providing CVD risk scores to patients, clinicians, or both can increase initiation or intensification of antihypertensive medications and statins by about 5 percentage points and reduce systolic blood pressure level, total cholesterol level, and LDL cholesterol level by 3 mm Hg, 4 mg/dL, and 1 mg/dL, respectively, similar to the effects we observed.¹⁷

Our study contributes to the pay-for-performance literature by showing modest incentives can improve use of CVD medications in varied clinical settings. Pay-for-performance incentives to clinicians or provider organizations have had mixed effects, especially when assessed using experimental designs.^18-21 The incentives in the Million Hearts Model have some features that might contribute to their effectiveness: they rely on an outcome (risk of myocardial infarction or stroke) that is important for patients and clinicians; they focus solely on CVD and so are not diluted by other quality measures; they align with many clinicians’ views of good clinical care (complementing intrinsic motivation); and they reward improvement rather than attaining prespecified targets.

Because this trial is implemented in hundreds of organizations throughout the US, results are likely broadly generalizable.²² However, participants volunteered to participate and so might differ from other organizations in systematic ways (such as motivation or capacity) that could influence effects. Further, the substantial attrition, often due to challenges submitting clinical data to the registry, limits the generalizability and scalability of the results.²³ Organizations that remained in the model were disproportionately likely to have participated in other CMS initiatives, suggesting these organizations have capacity (eg, health IT) to readily report data to CMS or that other CMS initiatives provide resources (such as funds to hire care managers) that aid model implementation.

This study provides support for guidelines recommending routine CVD risk score calculation.⁹ The evidence base for that so far has been limited, supported mainly by expert opinion, small clinical trials, and observations that benefits of CVD medications scale with absolute CVD risk.²⁴ This study shows that routine risk assessment likely prompts initiation or intensification of medications to help overcome clinical inertia. However, it is too early to tell whether these differences are large enough to meaningfully reduce CVD events.

First, we did not measure effects on patients’ adherence to medications, one way the model could improve outcomes.^25,26 Second, the 35% attrition of organizations, the 20-clinician cap applied selectively to control group clinicians, and the fact that organizations enrolled patients via data upload (introducing opportunities for selection), all raise the possibility that intervention and control patients would differ in ways that might drive apparent model effects on medication use. We found strong balance on many baseline characteristics, including CVD risk factors and medication use, although it remains possible the groups differed in unobserved ways that influence effect estimates. Third, we defined potential candidates for antihypertensive medication initiation or intensification based on 1 systolic blood pressure reading at enrollment, which might not be sufficient to identify persistently elevated blood pressure.⁸ Finally, although we observed medication use for all enrollees who retained Part D coverage, about half of high-risk patients were missing follow-up clinical data, increasing the risk of bias for estimates for blood pressure and cholesterol.

The Million Hearts Model’s targeted incentives and supports led to modest increases in use of statins and antihypertensive medication for high-risk patients and were associated with slight decreases in systolic blood pressure and LDL cholesterol levels. The model also had positive spillover to the medium-risk group, increasing medication use by similar amounts, even though provider organizations were not separately paid for risk reduction in this group. This pay-for-performance model is promising for improving appropriate use of preventive medications in a wide range of organizations, even among patients with high rates of treatment at baseline.

Corresponding Author: G. Greg Peterson, PhD, MPA, Mathematica, 1100 First St, NE, Washington, DC 20002 (gpeterson@mathematica-mpr.com).

Accepted for Publication: April 5, 2021.

Published Online: June 2, 2021. doi:10.1001/jamacardio.2021.1565

Author Contributions: Dr Peterson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Peterson, Pu, Kranker, Rose, Blue, Markovitz, McCall, Markovich.

Acquisition, analysis, or interpretation of data: Peterson, Pu, Magid, Barterian, Kranker, Barna, Conwell, Rose, Blue, Markovitz, Markovich.

Drafting of the manuscript: Peterson, Pu, Barterian, Barna, Markovitz.

Critical revision of the manuscript for important intellectual content: Peterson, Pu, Magid, Kranker, Conwell, Rose, Blue, McCall, Markovich.

Statistical analysis: Peterson, Pu, Barterian, Kranker, Barna, Blue, Markovitz.

Obtained funding: Peterson, Kranker.

Administrative, technical, or material support: Peterson, Pu, Barterian, Kranker, Conwell, Markovich.

Supervision: Peterson, McCall, Markovich.

Conflict of Interest Disclosures: None reported.

Funding/Support: The work was funded under a contract with the US Centers for Medicare & Medicaid Services to conduct an independent evaluation of the Million Hearts Cardiovascular Disease Risk Reduction Model (contract HHSM-500-2014-00034I).

Role of the Funder/Sponsor: US Centers for Medicare & Medicaid Services reviewed the design for the evaluation and early drafts of the manuscript and supported Mathematica’s decision to submit the manuscript for publication.

Disclaimer: The statements contained in this article are solely those of the authors and do not necessarily reflect the views or policies of the US Centers for Medicare & Medicaid Services.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank other Mathematica staff who helped collect, process, or analyze the study data; US Centers for Medicare & Medicaid Services staff who helped to describe the model design and implementation; and clinicians from the participating organizations who shared their experiences with us.