Changes in Cardiovascular Care Provision After the Affordable Care Act
Joseph A. Ladapo, MD, PhD; and Dave A. Chokshi, MD, MSc
Coronary heart disease (CHD) and stroke are leading causes of death in the United States,1 and physicians play a central role in treating modifiable risk factors among patients seeking clinical care. The Affordable Care Act (ACA) is designed to reduce barriers to preventive care, including preventive cardiovascular care, by eliminating marginal cost sharing on high-value screening tests identified by the US Preventive Services Task Force (USPSTF).2 The design of this particular provision of the ACA is unique, however, because it is unlike other policies and programs for cardiovascular disease (CVD) reduction that directly target the behavior of patients or healthcare providers—such as cigarette taxes, calorie postings on menus, or Medicare’s recently launched physician incentive program for preventive cardiovascular care.3,4
The ACA’s economic incentives for USPSTF services indirectly target physicians’ provision of preventive care by reducing patients’ out-of-pocket costs.Because physicians are gatekeepers to these services, directly assessing their patterns of preventive care provision in response to the ACA is critical to informing the design of future policies for clinical prevention. For example, policy approaches to increasing the population prevalence of a preventive cardiovascular test or therapy may differ depending on whether the service is provided during 5%, 15%, or 50% of clinical visits in which a physician sees a potentially eligible patient. Data on physicians’ provision of preventive services may also be more persuasive to physicians than the population prevalence of preventive service penetration, since the former more directly reflects physicians’ decision making.
To help inform future policies for preventive care in CVD, we used nationally representative data from the National Ambulatory Medical Care Survey (NAMCS) and National Hospital Ambulatory Medical Care Survey (NHAMCS) and examined physicians’ early response to the ACA. We primarily focused on physician visits for patients enrolled in private insurance plans because these plans insure the majority of the US population.
We used a quasi-experimental, difference-in-differences (DID) approach to examine the impact of the ACA on physicians’ provision of preventive cardiovascular care. The DID approach measures changes in an outcome associated with a policy change, after accounting for secular trends in that outcome, as reflected in a control group that is not exposed to the policy change.5 In this study, physician visits with insured patients in the target population were defined according to criteria proposed by the USPSTF, and the control group was composed of physician visits with patients who were: a) insured but ineligible for the preventive service by USPSTF guidelines (eg, patients close to but not meeting an age threshold for a screening test) or b) self-pay/uninsured and eligible for the preventive service. We selected these control groups because we hypothesized that physicians’ provision of preventive cardiovascular care in these encounters would likely capture temporal trends unrelated to the ACA,6,7 and we formally tested this hypothesis by comparing trends among the target and control populations in preventive cardiovascular services prior to the ACA’s provisions (evidence of a difference in slopes would suggest a difference in trends prior to the ACA). These tests showed no significant differences among our target and control groups for any of the preventive cardiovascular services we evaluated, which supports our selection of control groups.
We analyzed data collected in the NAMCS and NHAMCS from 2006 to 2013. The National Center for Health Statistics (NCHS) and the CDC conduct the NAMCS and NHAMCS annually on a nationally representative sample of visits to office-based physicians, hospital-based outpatient clinics, and emergency departments in the United States.8 Data are collected on patients’ symptoms, comorbidities, and demographic characteristics; physicians’ diagnoses; medications ordered or provided; and medical services provided. The NAMCS and NHAMCS intake materials allow physicians and staff to record up to 3 reasons for each visit and 3 diagnoses related to the visit, in addition to capturing several other major comorbid diagnoses (coded by NCHS staff using the International Classification of Diseases, Ninth Revision, Clinical Modification).
Data on outpatient hospital departments and community health centers from the NHAMCS were unavailable from 2012 to 2013, but the majority of ambulatory care is performed in office-based visits and captured by the NAMCS (93% of visits during 2006-2011 occurred in the office rather than hospital outpatient departments, and 99% of office visits occurred outside of community health centers). However, we adjusted for the absence of these 2 care sites in our regression analyses and used the 2006 to 2011 ratio of total visits to nonhospital outpatient/non–community health center visits to adjust the 2012 and 2013 estimates of care provision.9 Our analyses included only preventive care visits (including screening and general exams), routine visits for chronic problems, or visits to primary care physicians (including physicians in family practice, internal medicine, and obstetrics/gynecology) or to a patient’s primary care clinic, because we viewed these visit types as reasonable opportunities for the provision of preventive cardiovascular care.
From 2006 to 2013, the physician and hospital/outpatient clinic response rates in the NAMCS and NHAMCS ranged from 54% to 65% and 84% to 90%, respectively, and item nonresponse rates were generally 5% or less in both surveys.
Study Population and Primary Measures
We evaluated 11 preventive cardiovascular services for adult patients (≥18 years) that received A or B ratings from the USPSTF prior to the ACA’s implementation. Our target populations were primarily defined as patients who were privately insured unless otherwise stated, although we did not exclude patients concurrently enrolled in public plans as the ACA expanded preventive cardiovascular benefits in Medicare and Medicaid. Including these public plans therefore provides a more accurate assessment of the ACA’s overall effects. In addition, we evaluated tobacco screening and cessation treatment among pregnant women enrolled in Medicaid because of the importance of this particular policy and its unique treatment in the ACA as a provision applicable to all Medicaid plans rather than only Medicaid expansion plans. We excluded physician visits by patients who had previously been diagnosed with the condition that was subject to screening (eg, visits by patients with diabetes were excluded from our analysis of glycated hemoglobin [A1C] use for diabetes screening). In our analysis of tobacco use screening, all adults were included, irrespective of their prior tobacco use.
The preventive cardiovascular services we evaluated were: a) use of A1C to screen patients with hypertension for type 2 diabetes10; b) obesity management with counseling about diet/nutrition, exercise, or weight reduction11; c) measuring blood pressure in adults without diagnosed hypertension12; d) lipid testing among men 35 years or older or in high-risk men aged 20 to 34 years (high-risk defined by USPSTF as history of diabetes, previous CHD or atherosclerosis, family history of CVD, tobacco use, hypertension, or obesity by body mass index ≥30 kg/m2)13; e) lipid testing among women age 45 or older or high-risk women aged 20 to 44 years (high-risk defined similarly in women and men)13; f) aspirin therapy to prevent myocardial infarction in men aged 45 to 79 years14; g) aspirin therapy to prevent stroke in women aged 55 to 79 years14; h) tobacco use screening among pregnant women enrolled in Medicaid15,16; i) tobacco use screening among adults enrolled in private plans15; j) smoking cessation advice/counseling in pregnant smokers enrolled in Medicaid15,16; and k) smoking cessation advice/counseling in smokers enrolled in private plans.15
In our control groups, we examined physicians’ provision of these same preventive cardiovascular services in visits with: a) asymptomatic adults with normal blood pressure, b) self-pay/uninsured obese adults, c) self-pay/uninsured adults without diagnosed hypertension, d) low-risk men aged 20 to 34 years, e) low-risk women aged 20 to 44 years, f) men aged 30 to 44 years, g) women aged 45 to 54 years, h) nonpregnant women enrolled in Medicaid, i) self-pay/uninsured adults, j) nonpregnant female smokers enrolled in Medicaid, and k) self-pay/uninsured smokers.
We hypothesized that these were plausible control groups because they likely captured temporal trends in physicians’ preventive cardiovascular care but would not be directly affected by the ACA during our study timeframe because patients either did not meet USPSTF guideline criteria or did not have an eligible insurance plan. To identify our control groups, we also considered their clinical similarity to patients in the target population. When possible, we used age to distinguish between the target and control populations (eg, men 45 years or older were eligible for aspirin therapy whereas men younger than this age were generally not) because selection based on age was used in prior DID evaluations of the ACA.17 When it was not appropriate to use an age cutoff (eg, A1C screening is based on presence or absence of hypertension among adults and is not age-specific; Medicaid tobacco use screening targeted pregnant women rather than nonpregnant women), we distinguished our target and control patients by the presence or absence of a relevant comorbidity or risk factor. When neither of these options were available or appropriate, we identified clinically similar patients and distinguished the target and control cohorts by whether they were enrolled in private plans (and therefore eligible for the provision) or were considered self-pay/uninsured (and therefore ineligible for the provision). Examples of this included measuring blood pressure in adults without diagnosed hypertension (applying age eligibility criteria was not appropriate, nor was using a control group composed of patients with diagnosed hypertension, since these patients have an established diagnosis and would no longer be eligible for screening). The validity of our control population selections was assessed in comparisons of time trends between target and control populations during the pre-ACA period, and there were no significant differences, a finding that supports our hypothesis.
All ACA provisions took effect on September 23, 2010, for privately insured patients, but we used a start date of October 1 due to data availability. For pregnant women in Medicaid, ACA tobacco coverage was required as of October 1, 2010.
To further account for patient and clinical characteristics that may be associated with physicians’ use of preventive cardiovascular services, we extracted information on patient age, sex, race/ethnicity, US Census region (Northeast, Midwest, South, and West), urban or rural setting, and important comorbidities (identified using visit diagnoses and reasons for visit) known to increase the risk of adverse cardiovascular events (hypertension, coronary artery disease, diabetes, chronic kidney disease, and chronic obstructive pulmonary disease).18 To account for a possible increase in the complexity of self-pay/uninsured patients after the ACA took effect, we adjusted for a measure of continuity of care, as defined by whether a patient was an established patient or a new patient in the physician’s practice.19,20
All analyses accounted for the complex sampling design of the NAMCS and NHAMCS.21 We estimated simple and multivariate DID logistic regressions and estimated the predicted probability of a preventive cardiovascular service (as a dummy variable) to examine the impact of the ACA on the physicians’ care patterns (see eAppendix [eAppendix available at ajmc.com). Our models were generally implemented as: Preventive cardiovascular service = β0 + β1Target population visit + β2Post ACA + β3Target population visit × Post ACA
where Preventive cardiovascular service, Target population visit, and Post ACA are indicator variables and Post ACA specifically captures the period from October 1, 2010, to December 31, 2013.
The coefficient on the interaction term between Target population visit and Post ACA estimates the impact of interest. Specifically, this coefficient captures the differences among preventive cardiovascular service rates between target population visits and control population visits in the time period before the ACA’s implementation and compares them with the differences after the policy change; the coefficient therefore represents the independent relationship between the ACA and physicians’ preventive cardiovascular service rates. We tested for differences in pre-2010 testing trends among target population visits and control population visits by estimating multivariate logistic regressions limited to the period between January 1, 2006, and September 30, 2010, and including an interaction variable between time and our indicator for target population visits. The coefficient for this variable was not significant for any of the preventive cardiovascular services we examined. Multivariate logistic regression models also adjusted for patients’ clinical risk factors and demographic characteristics, insurance, geographic region, and urban/rural setting. Analyses were performed using Stata version 14 (StataCorp LLC; College Station, Texas).
Because of the concern that self-pay/uninsured patients may be sicker (or healthier overall; see Decker et al,22 for example) in the post-ACA period than in the period before the law took effect23 and that such differences could bias our results, we assessed changes in the overall cardiovascular risk of patients with private insurance versus patients who are self-pay/uninsured. Specifically, we computed the atherosclerotic cardiovascular disease (ASCVD) scores for privately insured patients and self-pay/uninsured patients from 2006 to 2013 and constructed linear regression models to test for an interaction between the post-ACA period and self-pay/uninsured status.24 Lipid levels and blood pressure were not uniformly available in our data, so we imputed these values with age- and sex-adjusted population values in the United States (see eAppendix Table 1).25,26 These models included covariates similar to those of our primary models, but patients with CHD were omitted (3.9% of overall population) because the ASCVD is only applicable to patients without CHD. We did not find evidence of an interaction (P = .45 for interaction), indicating that the overall cardiovascular risk level of the self-pay/uninsured cohort was not significantly different over time.
Annual Number and Prevalence of Preventive Cardiovascular Services
We present overall trends in physicians’ use of preventive cardiovascular services during clinic visits for diabetes screening, obesity therapy, hypertension screening, cholesterol screening, aspirin therapy, tobacco use, and smoking cessation treatment in Figure 1. Characteristics of the population are reported in eAppendix Table 2. The rates of most preventive cardiovascular services tended to be flat or to trend modestly upward for physician visits with target and control patients, although rates of obesity treatment and smoking cessation advice/counseling tended to trend downward.
Changes in Preventive Cardiovascular Services Post ACA
Using our DID regression models, we found that in the period after the ACA’s provisions took effect for preventive cardiovascular services, there was no significant change in the use of obesity treatment (47.2% in 2006-2010 [third quarter] to 40.3% in 2010 [fourth quarter]-2013; DID, –1.3 per 100 visits; 95% CI, –21.9 to 19.4), cholesterol screening in men (10.3% in 2006-2010 to 8.9% in 2010-2013; DID, +0.4 per 100 visits; 95% CI, –2.2 to 3.0), cholesterol screening in women (8.4% in 2006-2010 to 7.9% in 2010-2013; DID, +0.4 per 100 visits; 95% CI, –1.3 to 2.0), aspirin therapy in women (8.8% in 2006-2010 to 10.0% in 2010-2013; DID, +0.4 per 100 visits; 95% CI, –1.3 to 2.0), tobacco use screening in pregnant women (67.7% in 2006-2010 to 70.3% in 2010-2013; DID, +2.5 per 100 visits; 95% CI, –5.4 to 10.5), smoking cessation advice in pregnant smokers (17.6% in 2006-2010 to 13.3% in 2010-2013; DID, –7.2 per 100 visits; 95% CI, –17.7 to 3.2), or smoking cessation advice in adult smokers (23.0% in 2006-2010 to 17.8% in 2010-2013; DID, +0.8 per 100 visits; 95% CI, –6.2 to 7.7) (Table).
The ACA’s 2010 provisions were associated with an increase in the use of diabetes screening (3.9% in 2006-2010 to 7.6% in 2010-2013; DID, +3.5 per 100 visits; 95% CI, 1.1-5.9), tobacco use screening in adults (64.4% in 2006-2010 to 74.5% in 2010-2013; DID, +11.6 per 100 visits; 95% CI, 4.8-18.3), aspirin therapy in men (11.1% in 2006-2010 to 13.5% in 2010-2013; DID, +2.9 per 100 visits; 95% CI, 1.1-4.6), and hypertension screening (73.2% in 2006-2010 to 76.4% in 2010-2013; DID, +9.9 per 100 visits; 95% CI, 2.8-16.9).
We performed sensitivity analyses for cholesterol screening and aspirin therapy that used a self-pay/uninsured cohort of similar age and sex to address any confounding potentially related to age differences in our primary analyses. This analysis was not possible for aspirin therapy in men because pre-ACA trends differed between these 2 groups. These sensitivity analyses yielded similar results to our main analyses, showing no significant change in cholesterol screening in men (DID, +2.2 per 100 visits; 95% CI, –1.9 to 6.2), cholesterol screening in women (DID, +0.4 per 100 visits; 95% CI, –4.0 to 4.7), or aspirin therapy in women (DID, +1.7 per 100 visits; 95% CI, –0.4 to 3.8).
Using the Benjamini-Hochberg procedure to adjust for multiple testing, where P values are sorted in ascending order and critical limits are estimated for each P value based on its rank, total number of statistical tests, and false discovery rate (0.05), we found that changes in tobacco use, hypertension, and diabetes screening remained significant, whereas the change in aspirin therapy among men was no longer significant.
Our results indicate that physicians’ provision of cardiovascular preventive care increased for USPSTF-recommended services following enactment of the ACA, based primarily on findings in patients with employer-sponsored health plans. Out of 11 services examined, 4 (diabetes screening, tobacco use screening, aspirin therapy among men, and hypertension screening) were found to be used more frequently by physicians. Our findings also indicate that overall, physician visits for preventive cardiovascular care were increasing prior to the ACA—and this trend continued following its passage. The sex disparity for aspirin use underscores wider concerns about disparities in cardiovascular care among women versus men.27
To our knowledge, this study is the first to analyze changes in use of cardiovascular preventive care after the ACA using physician visit–level data. The physician focus in our analysis allows a more direct assessment of clinical decision making in response to policy changes encoded in the ACA, such as elimination of marginal cost sharing for USPSTF-recommended preventive services.
Comparing our results with those in prior studies using patient-level data (eg, from the Medical Expenditure Panel Survey), we found similarly mixed, although generally modest, effects on uptake of USPSTF-recommended preventive services after the ACA.28-30 This may reflect the limited consequences that eliminating patient cost sharing has on physician decision making. In March 2014, only about 43% of the population reported awareness that the ACA eliminated out-of-pocket expenses for preventive services.31 The fact that overall physician visits for preventive cardiovascular services increased during the timeframe studied (2006-2013) may mean that other factors were more important in influencing physician uptake. For example, a growing emphasis on cardiovascular prevention through insurance payer policies and professional society guidelines may have been more significant than ACA provisions.
With respect to the increase in certain preventive services (diabetes screening, tobacco use screening, aspirin therapy among men, and hypertension screening), it is difficult to pinpoint the mechanism of effect, particularly because they are services that could be delivered within the context of a routine primary care visit. Patients may not have experienced cost sharing for these services even before the ACA—other than their co-payment for the office visit itself, which would not be affected by the ACA preventive services provisions. The causes of these temporal trends require further investigation with more granular data on cost sharing at the visit level.
Our study supports the argument that the ACA’s cost-sharing provisions are an effective way to increase uptake of clinical preventive services, although overall levels of service provision were still lower than those recommended by the USPSTF and these gaps increase the population risk of CVD. Our findings are in contrast to some earlier evaluations of the ACA that found minimal or no effects on preventive care. However, the absolute effects of the ACA’s preventive cardiovascular care provisions were often modest. Physician decision making may be more sensitive to more proximal factors such as educational interventions, enhanced reimbursement for preventive services, or ease of operational processes, such as referrals for smoking cessation advice or point-of-care A1Ctesting. For services delivered during a preponderance of clinic visits, such as hypertension screening or tobacco use screening, strategies may differ and revolve around implementation of practice-level processes that ensure near-universal screening. For those offered during a lower proportion of visits, such as aspirin therapy or diabetes screening, clinical decision support (eg, electronic health record defaults) may be more effective.
Our study has several limitations. We were unable to account for the presence of grandfathered plans exempt from some ACA provisions, patient or physician awareness of ACA provisions, or the effects of insurer medical loss ratio regulations that may have increased overall preventive service provision. Our findings may therefore underestimate (or overestimate, particularly in the cases of hypertension and tobacco use screening, where control populations were self-pay/uninsured) the effect of the policy change on physicians’ provision of preventive cardiovascular services. In addition, diffusion of high-deductible insurance plans may have exerted indiscriminate downward pressure on appropriate and inappropriate preventive care, a finding that was demonstrated in the RAND health insurance experiment.32 Related to this, if private plans or state Medicaid programs that were otherwise exempt from the ACA’s provisions chose to reduce cost sharing on preventive cardiovascular services in response to a changing climate of health reform, these shifts would cause us to understate the ACA’s effects. We also performed multiple statistical tests, and our findings should be interpreted in this context. Further, we did not have data on patients’ incomes, and some research suggests that patients from lower income groups—and the physicians who care for them—may be more sensitive to the elimination of marginal cost sharing than patients from higher income groups.
Physicians’ provision of cardiovascular preventive care increased for some USPSTF-recommended services following enactment of the ACA. The results of our direct assessment of physicians’ clinical decision making in response to policy changes encoded in the ACA support the notion that cost-sharing provisions are an effective way to increase uptake of evidence-based clinical preventive services, although substantial gaps in preventive care persist. The sex disparity in aspirin use also underscores wider concerns about poorer-quality cardiovascular care in women versus men. Other interventions, including those with an educational, reimbursement-based, or practice-level focus, may be complementary approaches to influencing physician decision making and reducing the population burden of CVD.
Dr Joseph Ladapo 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.