Prevalence and predictors of cost-related medication nonadherence in individuals with cardiovascular disease: Results from the Behavioral Risk Factor Surveillance System (BRFSS) survey

Abstract

Medication nonadherence is highly prevalent among patients with chronic cardiovascular disease. Poor adherence has been associated with increased morbidity and mortality. Medication cost is a major driver for medication nonadherence. Utilizing data from the 2016 to 2018 Behavioral Risk Factor Surveillance System (BRFSS) survey, we estimated the prevalence of cost-related medication nonadherence (CRMNA) among the overall population and among individuals who reported a history of diabetes, atherosclerotic cardiovascular disease (ASCVD), or hypertension. We then performed multivariable logistic regression to analyze sociodemographic factors associated with CRMNA. Our study population consisted of 142,577 individuals of whom 24% were older than 65 years, 47% were men, 66% were White, 17% Black, 35% had hypertension, 13% had diabetes mellitus, and 10% had ASCVD. CRMNA was reported in 10% of the overall population, 12% among those with hypertension, 17% among those with diabetes, and 17% among those with ASCVD. Age below 65 years, female gender, unemployment, lower income, lower educational attainment, having at least 1 comorbidity, and living in a state that did not expand Medicaid were independently associated with CRMNA. The prevalence of CRMNA increased with greater number of these high-risk sociodemographic factors. We conclude that the prevalence of CRMNA is 10% among U.S. adults overall and is higher among those with common chronic diseases. Risk factors associated with CRMNA should be addressed in order to improve adherence rates and health outcomes among high-risk individuals.

1. Introduction

Optimal management of chronic cardiovascular disease including ischemic heart disease, hypertension, diabetes, and stroke is contingent upon the receipt of long-term medications proven to reduce morbidity and mortality.(Arnett et al., 2019; Grundy et al., 2019; Whelton et al., 2018) However, poor medication adherence among those with cardiovascular conditions has been estimated to approach up to 43%.(Naderi et al., 2012) Medication nonadherence has been associated with increased all-cause and cardiovascular mortality in patients with history of myocardial infarction, hypertension, and hyperlipidemia.(Bitton et al., 2013; Balkrishnan, 2005; Ho et al., 2006; De Vera et al., 2014; Ho et al., 2008) Additionally, medication nonadherence has been linked with greater healthcare costs attributed to complications from poor disease control, need for medical and surgical therapeutic escalation, and increased hospitalization.(Esposito et al., 2009; Sun et al., 2008; Pittman et al., 2011)

Cost-related medication nonadherence (CRMNA) is one potential avenue for improvement. It can be defined as delays in filling prescription medications, skipping doses, or taking less than the prescribed dosage due to cost related issues.(Pierre-Jacques et al., 2008) It is the most common reason for medication nonadherence, followed by side effect concerns.(McHorney and Spain, 2011) determinants of CRMNA include factors such as household income and out of pocket costs - which are in turn impacted by prescription drug prices, insurance coverage, healthcare system formularies, and available alternatives.(Iuga and McGuire, 2014) Additionally, patient-related factors including health-literacy, patient-physician communication, and lifestyle preferences may indirectly influence rates of CRMNA.(Iuga and McGuire, 2014) CRMNA may be particularly important in the United States, given higher prescription drug costs compared to other Organization of Economic Cooperation and Development member countries.(Sarnak et al., 2017) Due to an aging population, a rising cardiovascular disease prevalence, and up-trending medication costs, the burden of CRMNA is expected to grow.(Keehan et al., 2020)

Similarly to medication nonadherence in general, CRMNA has been linked with higher mortality in patients with chronic diseases, including cardiovascular disease.(Van Alsten and Harris, 2020) A relatively limited number of population studies have examined the prevalence of CRMNA among individuals with cardiovascular disease.(Lee et al., 2018; Khera et al., 2019a; Marcum et al., 2013) Moreover, most of the available studies were conducted prior to the implementation of major provisions of the affordable care act. In this study, we use data from a large, contemporary, nationally representative sample for the purpose of describing the prevalence of CRMNA in the general population and among those with atherosclerotic cardiovascular disease. We also describe the prevalence of CRMNA among individuals with cardiovascular risk factors including diabetes and hypertension. Additionally, we investigate sociodemographic factors associated with CRMNA, to identify factors that can be used to screen for CRMNA and to inform potential system-wide interventions.

2. Methods

The Behavioral Risk Factor Surveillance System (BRFSS) survey is a nationwide telephone-based questionnaire survey established by the Centers for Disease Control and Prevention to evaluate chronic health conditions, health-related risk behaviors, and the use of preventive services among U.S. adults. The BRFSS is administered to a random sample that is representative of U.S. adult residents over the age of 18 years. It is conducted in all 50 U.S. states, the District of Columbia, and the US territories of Puerto Rico, Guam, and the Virgin Islands, making it the largest telephone-based survey in the world. We utilized cross-sectional data from the 2016, 2017, and 2018 BRFSS surveys. Variables were recorded according to self-reported responses to survey items. This approach has been previously validated.(Li et al., 2012; Pierannunzi et al., 2013) Institutional Review Board approval was waived as BRFSS data is deidentified and publicly available.

Appropriate CDC-provided survey weights were used for all analyses in order to account for the complex survey design of BRFSS, and to correctly extrapolate our results into nation-level estimates. The overall study population included all individuals with data available for CRMNA. CRMNA was defined by answering “Yes” to the question “Not including over the counter (OTC) medications, was there a time in the past 12 months when you did not take your medications as prescribed because of cost?”. Our population was stratified into two groups according to whether individuals reported CRMNA. Baseline characteristics were summarized using weighted percentages. The prevalence of CRMNA was first calculated for the overall population. CRMNA prevalence was then calculated for those with history of atherosclerotic cardiovascular disease (ASCVD), for those with history of diabetes, and for those with history of hypertension. Participants were considered to have ASCVD if they reported ever having coronary heart disease, myocardial infarction, or stroke. Diabetes mellitus was identified if participants were ever told they had diabetes. Hypertension was defined if participants reported having been told they have high blood pressure by a doctor, nurse, or other health professional.

We then examined the association between each of hypertension, diabetes mellitus and ASCVD and CRMNA using univariable and multivariable logistic regression models, adjusting for age, gender, race/ethnicity, education, employment status, income, living in a state that expanded Medicaid, healthcare coverage, presence of at least one comorbidity, and spoken language. States were classified as having expanded Medicaid if they had done so by the year 2020 according to data from the Kaiser Family Foundation.(Status of State Action on the Medicaid Expansion Decision, 2020) At the time of study analysis, states that had not yet expanded Medicaid included Alabama, Florida, Georgia, Kansas, Mississippi, South Dakota, Tennessee, Texas, Wisconsin, and Wyoming.

We then analyzed factors associated with CRMNA in the overall population using multivariable logistic regression models with the following pre-specified variables: Age under 65 years, gender, race/ethnicity, highest level of education attainment, employment status, household income, living in a state with Medicaid expansion, having health insurance, presence and type of healthcare coverage, language spoken, and having at least 1 comorbidity (including hypertension, diabetes mellitus, COPD, hyperlipidemia, cancer, chronic kidney disease, and ASCVD). Variables independently associated with CRMNA were labeled as high-risk features. Individuals were then categorized into 4 quartiles based on their number of high-risk features. The proportion of individuals who reported CRMNA within each quartile was calculated and represented graphically using a bar chart for both the general US population and for those with history of ASCVD.

As sensitivity analyses, given the modulatory effect of Medicare among the elderly and the anticipated higher prevalence of CRMNA among the uninsured, we evaluated the factors associated with CRMNA among individuals older than 65 years and those with healthcare coverage.

All analyses were conducted using Stata version 13.1 (StataCorp, College Station, TX). A p-value <0.05 was considered statistically significant and was defined a priori.

3. Results

Our study population included 142,577 adults with complete information on the presence of CRMNA. Baseline characteristics are summarized in Table 1. Overall, 24% were above the age of 65, 47% were men, 66% were White and 17% Black, 35% reported hypertension, 13% reported diabetes mellitus, and 10% reported ASCVD. CRMNA was more prevalent among individuals with lower household income (20% with income < $10,000 vs. 10% with income > $75,000), the unemployed (18% vs. 9% employed, P < 0.00001), those with < high school education (16% vs. 6% with at least college education, P < 0.00001), those without healthcare coverage (23% vs. 9% with healthcare coverage, P < 0.00001), and current smokers (17% vs. 8% never smokers, P < 0.00001). Hispanic and black individuals reported higher rates of CRMNA compared to White individuals (13 and 12% vs. 9%, P < 0.00001), as did women compared to men (12% vs. 8%, P < 0.00001). Individuals living in Medicaid non-expansion states reported higher prevalence of CRMNA (11% vs. 9%, P < 0.00001). CRMNA was also more prevalent among those living in rural areas compared to suburban and urban locations (10% vs. 8%, P = 0.0009) (Table 1).

Table 1.

Baseline characteristics of the study population by cost related medication nonadherence.

	Cost related medication nonadherence		P value
	No (n = 129,717)	Yes (n = 12,860)
Age (years)			< 0.00001
18–34	19,274 (90%)	2161 (10%)
35–44	14,301 (88%)	1930 (12%)
45–54	20,036 (88%)	2574 (11.5%)
55–64	28,230 (89%)	3208 (11%)
≥65	47,876 (93%)	2987 (7%)
Gender			< 0.00001
Men	56,305 (92%)	4421 (8%)
Women	73,335 (88%)	8427 (12%)
Race/ethnicity			< 0.00001
White	94,800 (91%)	8055 (9%)
Black	16,445 (88%)	2162 (12%)
Hispanic	9834 (87%)	1524 (13%)
Other	6203 (89%)	829 (11%)
Education			< 0.00001
< high school	10,268 (84%)	1936 (16%)
High school – Some college	71,624 (90%)	7962 (10%)
≥ college	47,358 (94%)	2903 (6%)
Income			< 0.00001
<10,000	5066 (80%)	1196 (20%)
$10,000–15,000	5470 (82%)	1182 (18%)
$15,000–20,000	7988 (82%)	1658 (18%)
$20,000–25,000	10,209 (85%)	1746 (15%)
$25,000–35,000	11,420 (87%)	1420 (13%)
$35,000–50,000	15,267 (90%)	1443 (10%)
$50,000–75,000	17,052 (93%)	1033 (7%)
>$75,000	35,548 (90%)	1250 (10%)
Employment status			< 0.00001
Employed	62,695 (91%)	5399 (9%)
Unemployed	21,440 (82%)	4661 (18%)
Student	3204 (94%)	258 (6%)
Retired	41,159 (94%)	2430 (6%)
Presence of healthcare coverage			< 0.00001
Yes	121,261 (91%)	10,215 (9%)
No	8072 (77%)	2601 (23%)
Type of healthcare coverage			< 0.00001
Plan purchased through employer or union	50,419 (93%)	3192 (7%)
Plan that you or family member purchases	12,813 (90%)	1191 (10%)
Medicare	38,341 (90%)	3629 (10%)
Medicaid	9052 (86%)	1388 (14%)
Tricare	4402 (96%)	147 (4%)
Alaska Native, Indian Health	529 (90%)	43 (10%)
Service, Tribal Health Services
Other	3496 (89%)	391 (11%)
None	255 (89%)	47 (11%)
Presence of primary care physician
Yes	108,732 (90%)	10,309 (10%)
No	20,545 (89%)	2508 (11%)
Living in a state with Medicaid expansion			< 0.00001
Yes	87,144 (91%)	7656 (9%)
No	41,571 (89%)	5087 (11%)
Living in a rural area			0.0009
Yes	17,139 (90%)	1421 (10%)
No	37,795 (92%)	2750 (8%)
Smoking status			< 0.00001
Never smoked	71,217 (92%)	5725 (8%)
Former smoker	35,695 (90%)	3283 (10%)
Current smoker	17,779 (83%)	3374 (17%)
Hypertension			< 0.00001
Yes	16,330 (88%)	2091 (12%)
No	21,903 (92%)	1926 (8%)
Diabetes mellitus			< 0.00001
Yes	18,560 (83%)	3044 (17%)
No	107,083 (91%)	9207 (9%)
Hyperlipidemia			< 0.00001
Yes	22,066 (88%)	1914 (12%)
No	14,136 (92%)	1843 (8%)
Chronic obstructive pulmonary disease			< 0.00001
Yes	10,837 (78%)	2703 (22%)
No	118,309 (91%)	10,052 (9%)
Atherosclerotic cardiovascular disease			< 0.00001
Yes	15,412 (83%)	2569 (17%)
No	112,905 (91%)	10,003 (9%)

Altogether, 10% of our study population reported CRMNA which translates to approximately 26 million individuals assuming a U.S. census of 255,200,373 adults over the age of 18 years. CRMNA was more prevalent among individuals with established cardiovascular risk factors or ASCVD. CRMNA was reported by 17% of individuals with ASCVD history, corresponding to 4 million U.S. adults. Additionally, 17% of individuals with diabetes and 12% of individuals with hypertension reported CRMNA, corresponding to 11 and 6 million U.S. adults respectively. Diabetes, hypertension, and ASCVD were each significantly associated with CRMNA, even after adjusting for known confounders. (Table 2).

Table 2.

Multivariable-adjusted^a odds ratio (95% confidence interval) for the association of hypertension, diabetes, or atherosclerotic cardiovascular disease and cost-related medication nonadherence.

Variable	Unadjusted	Adjusted
Hypertension	1.56 (1.35, 1.79)	1.65 (1.33, 2.03)
Diabetes mellitus	2.06 (1.90, 2.23)	2.11 (1.90, 2.35)
Atherosclerotic cardiovascular disease	2.08 (1.89, 2.27)	1.91 (1.69, 2.17)

^aAdjusted for patient’s age < 65 years, gender, race/ethnicity, highest level of education attainment, employment status, household income, living in a state with Medicaid expansion, having health insurance, presence and type of healthcare coverage, spoken language, and presence of at least one comorbidity.

In multivariable models, the following sociodemographic factors were independently associated with CRMNA (Table 3): Age < 65 years (OR 1.85 [95% CI 1.59–2.15]), female gender (OR 1.27 [95% CI 1.16–1.40]), unemployment (OR 1.49 [95% CI 1.32–1.67]), lower income (3.89 [95% CI 3.11–4.85] for household income < $10,000 vs. > $75,000), and having at least 1 medical comorbidity (OR 1.50 [95% CI 1.28–1.74]). Factors associated with reduced odds of CRMNA included at least college education (OR 0.80 [95% CI 0.67–0.96]) and living in a Medicaid expansion state (OR 0.91 [95% CI 0.83–0.99]). Race was not associated with CRMNA and neither was the type of health coverage plan.

Table 3.

Factors associated with cost related medication nonadherence.

Variable	Odds ratio (95% confidence interval)
Age < 65 years	1.85 (1.59, 2.15)
Gender
Men	1 (ref)
Women	1.27 (1.16,1.40)
Race/ethnicity
White	1 (ref)
Black	0.93 (0.82, 1.05)
Hispanic	1.15 (0.94, 1.39)
Other	1.12 (0.91, 1.37)
Education
Less than high school	1 (ref)
High school – Some college	0.90 (0.78, 1.05)
Greater than college	0.80 (0.67, 0.96)
Income
<$10,000	3.89 (3.11, 4.85)
$10,000–$15,000	3.96 (3.23, 4.84)
$15,000–$20,000	4.01 (3.31, 4.79)
$20,000–$25,000	3.70 (3.09, 4.43)
$25,000–$35,000	3.46 (2.85, 4.20)
$35,000–$50,000	2.54 (2.16, 2.99)
$50,000–$75,000	1.69 (1.44, 1.98)
>$75,000	1 (ref)
Employment status
Employed	1 (ref)
Unemployed	1.49 (1.32, 1.67)
Student	0.62 (0.46, 0.82)
Retired	0.89 (0.76, 1.03)
Living in a state with Medicaid expansion Type of healthcare coverage	0.91 (0.83, 0.99)
Plan purchased through employer or union	1.27 (0.64, 2.50)
Plan that you or family member purchases	1.54 (0.78, 3.06)
Medicare	1.47 (0.75, 2.90)
Medicaid	1.09 (0.55, 2.15)
Tricare	0.62 (0.29, 1.32)
Alaska Native, Indian Health Service, Tribal	1.00 (0.41, 2.48)
Health Services
Other	1.16 (0.58, 2.34)
None	1 (ref)
First language
Spanish	0.64 (0.38, 1.08)
Presence of comorbidity a
Yes	1.50 (1.28, 1.74)

^aComorbidities included were hypertension, diabetes, atherosclerotic cardiovascular disease, hyperlipidemia, and chronic obstructive pulmonary disease.

The number of high-risk features (age < 65, female gender, < high school education, annual household income < $75,000, unemployment, having at least one comorbidity, and living in a state without Medicaid expansion) was summed for all individuals. The population was then split into four groups according to the number of high-risk features: 0–2, 3, 4, and ≥ 5. The median number of high-risk features in the overall cohort was 4 (IQR: 3–4). 17% had 0 to 2 high-risk features, 33% had 3 high-risk features, 31% had 4 high-risk features and 19% had ≥5 high-risk features. In the overall population, the prevalence of CRMNA increased from 4% among those with 0 to 2 high-risk features to 17% among those with at least 5 high-risk features (p < 0.001) (Fig. 1). Among those with established ASCVD, the prevalence of CRMNA was 6% for those with 0 to 2 high-risk features, 12% for those with 3 high-risk features, 17% for those with 4 high-risk features, and 29% for those with at least 5 high-risk features (Fig. 1).

Fig. 1.

Prevalence of cost-related medication nonadherence with increasing number of high-risk features.

The prevalence of CRMNA was 7% for those older than 65. Among those older than 65, Hispanic race (OR 2.00 [95% CI 1.01–3.94]), unemployment (OR1.66 [95% CI 1.20–2.29]), and low income (3.57 [95% CI 2.11–6.02] for household income < $10,000 vs. > $75,000) were associated with CRMNA in multivariable regression analyses. Among those with healthcare coverage, 9% reported CRMNA. Within this subpopulation, female sex (OR 1.27 [95% CI 1.16–1.40]), unemployment (OR 1.49 [95% CI 1.32–1.67]), lower income (OR 3.89 [95% CI 3.11–4.85] for household income less $10,000 vs. > $75,000), and having at least one comorbidity (OR 1.50 [95% CI 1.28–1.74]) were associated with higher odds of CRMNA, while at least college education (OR 0.80 [95% CI 0.67–0.96]) and living in a state with Medicaid expansion (OR 0.91 [95% CI 0.83–0.99]) were associated with lower odds of CRMNA in the multivariable regression analyses.

4. Discussion

In a nationally representative U.S. dataset, we found that the prevalence of CRMNA was 10%, and this prevalence was greater among those with ASCVD, diabetes mellitus, or hypertension. Sociodemographic factors including younger age, female gender, low education, low annual household income, unemployment, having at least 1 comorbidity, and living in a state without Medicaid expansion are significantly associated with higher CRMNA. The prevalence of CRMNA incrementally increased with greater numbers of these high-risk features. Those with 0–2 risk factors had a CRMNA prevalence of 4% and 6% in the overall and ASCVD population respectively, compared to 17% and 29% for those with at least 5 risk factors.

Medication cost presents an important barrier to medication adherence and may lead to morbidity and mortality, preventable hospitalizations, and increased healthcare costs.(Cutler et al., 2018; Dunbar-Jacob et al., 2003; Osterberg and Blaschke, 2005) Our findings demonstrate that there is a substantial prevalence of CRMNA in the U.S., reinforcing findings from previous studies. In a cross-sectional study of the National Health Interview Survey, Khera et al. reported a CRMNA prevalence of 12.6% among individuals with ASCVD.(Khera et al., 2019b) In a smaller sample from the Health, Aging, and Body Composition study the prevalence of CRMNA was 7.7%.(Marcum et al., 2013) Of note, our study findings point to a comparatively higher rate of CRMNA (10% in the overall population and 17% among individuals with ASCVD). In addition to further establishing the prevalence of CRMNA among a contemporary, nationally representative sample, our study adds to the literature by identifying individual risk factors that were associated with CRMNA - which could be used for screening patients at risk and informing policy-level interventions aimed at curbing CRMNA.

Individual-level factors that correlated with higher risk of CRMNA in our study were mostly socioeconomic variables such as unemployed status, lower household income, and lower level of educational attainment. Perhaps surprising was the lack of association of CRMNA with race. Rather, our results implicate socioeconomic factors as the main drivers of CRMNA. Employment status takes on special importance not only as a potential source of income, but also because the vast majority of Americans receive health insurance through their employers.(Med. Benefits, 2019) The correlation between lower educational attainment and higher CRMNA may be related to a number of factors including earning potential, health literacy, and net wealth (as opposed to income). While these findings are not surprising, the high prevalence is again notable. Among individuals with household incomes < $10,000, approximately one in five responders reported CRMNA. Among individuals with household incomes > $75,000, 10% reported CRMNA, indicating a significant prevalence across a wide spectrum of household incomes. The burden of CRMNA will likely increase as a result of the economic impacts and rise of unemployment during the COVID-19 pandemic.(Coibion et al., 2020)

Consistent with our findings, prior studies have shown that women have a higher risk of medication non-adherence compared to their male counterparts.(Lee and Khan, 2015; Zhang et al., 2017) These factors can be related to socioeconomic factors, access to care, and also attitudes and beliefs towards their treatment.(Lee and Khan, 2015; Zhang et al., 2017) Having at least 1 comorbidity was also associated with CRMNA likely because such individuals are on a complex regimen of multiple costly medications. Of note, the relative burden of CRMNA is likely to vary based on the specific disease being treated with diseases requiring more expensive prescription medications likely being associated with higher CRMNA compared to those that can be managed with generics such as statin therapy. It is noteworthy that diabetes was associated with a higher adjusted OR – which could be related to higher medication cost and regimen complexity compared to hypertension and ASCVD.(Herkert et al., 2018) Age younger than 65 years being associated with higher CRMNA supports findings from the Khera et al.(Khera et al., 2019b) This is of great importance as older adults are theoretically at greater risk of CRMNA due to higher number of comorbidities and a substantial percentage living on a fixed income. That Americans >65 years old were less likely to report CRMNA is most likely related to the protective benefits of Medicare.(Gellad et al., 2011; Sommers et al., 2017; Assessing Medicare Part D After 10 Years, 2016; Brown and Bussell, 2011)

Our findings may be related to high prescription drug costs in the U. S. compared to other developed nations – the drivers of which include pharmaceutical pricing, a lack of purchasing power, over-utilization of non-generic medication, and delays in the introduction of generic medications.(Sarnak et al., 2017) Moreover, the higher prevalence of CRMNA among individuals with low socioeconomic status potentially points to a patchy social safety net, high rates of uninsured and underinsured Americans, and substantial insurance deductibles and co-pays. (Sommers et al., 2017)

Of note, the affordable care act was designed to ameliorate many of these problems and is currently being challenged in the United States Supreme Court.(Musumeci, 2020) One of the provisions of the affordable care act was the expansion of Medicaid eligibility to include individuals with incomes up to 138% of the federal poverty line (FPL), which has provided affordable healthcare access to an estimated 18 million additional Americans in the process.(Garfield and Orgera, 2019) At the time of writing this article, only 39 states have chosen to expand Medicaid.(Garfield and Orgera, 2019) In states that have not expanded Medicaid, there is a gap in government provided aid affecting those who have household incomes too high to qualify for Medicaid, but too low to deem them eligible for federal subsidies to health insurance market-places.(Garfield and Damico, 2017; Solomon, 2021) Nearly 2.5 million US adults fall within this coverage gap.(Garfield and Damico, 2017; Solomon, 2021) Previous studies show Medicaid expansion to be associated with improved measures of healthcare access and affordability, but few have examined the association with CRMNA specifically.(Guth et al., 2020) We found that living in a Medicaid expansion state to be associated with a lower likelihood of CRMNA. This in line with findings by Levine et al., who found that the burden of CRMNA in stroke survivors decreased appreciably after ACA implementation.(Levine et al., 2018) Additionally, Sommers et al. showed that Medicaid expansion was associated with fewer skipped medications as well as lower difficulty with medical bills among individuals aged 19 to 64 with incomes below 138% of the FPL.(Sommers et al., 2016)

Our results emphasize the relevance of socioeconomic factors in the management of chronic cardiovascular disease. Clinicians should actively look for CRMNA and work to address it, whether by providing cheaper generic medications or deploying social work resources to aid patients.(Kleinsinger, 2018) Moreover, rather than conceptualizing medical care in simple physiological terms and compartmentalizing it to individual factors, a holistic approach accounting for the social environment should be undertaken. CMRNA may also influenced by patient perceptions of the medication importance as well as life stressors. Strong patient-physician communication can help identify those at risk for CRMNA, address patient concerns, and produce solutions that work for the particular individual. Furthermore, policy level interventions such as Medicare and the Medicaid expansion portion of the affordable care act seem to have had a favorable impact on CRMNA and should be supported. Additional efforts should be made to create safety nets for those at highest risk of CRMNA.

4.1. Study limitations and strengths

The primary strength of our study is its large sample size and diverse, nationally representative study population. One of the significant limitations includes the survey nature of our data set and the lack of granular data related to medication adherence. All variables included were based on self-reported responses to survey items. We could not account for variables not part of the BRFSS survey such as medication cost, regimen complexity, and insurance structure. Our definition of ASCVD was also limited and did not include patients with peripheral arterial disease, stable coronary artery disease, or transient ischemic attacks. Therefore, the true prevalence for ASCVD may be underestimated. Furthermore, we could not assess disease severity or medical symptoms. Lastly, we cannot rule out residual confounding factors.

5. Conclusion

CRMNA is especially prevalent among patients with established ASCVD and risk factors such as diabetes or hypertension. Sociodemographic factors associated with CRMNA include younger age, female gender, low education, low annual household income, unemployment, having at least 1 comorbidity, and living in a state without Medicaid expansion. Identification of these risk factors is important for screening individuals at risk for CRMNA and for developing policy solutions.