Abstract
Background
Atherosclerotic cardiovascular disease (ASCVD) causes most deaths in the United States and accounts for the highest healthcare spending. The association between the modifiable risk factors (MRFs) of ASCVD and pharmaceutical expenditures are largely unknown.
Methods and Results
We examined the association between MRFs and pharmaceutical expenditures among adults with ASCVD using the 2012 and 2013 Medical Expenditure Panel Survey. A 2‐part model was used while accounting for the survey's complex design to obtain nationally representative results. All costs were adjusted to 2013 US dollars using the gross domestic product deflator. The annual total pharmaceutical expenditure among those with ASCVD was $71.6 billion, 33% of which was for medications for cardiovascular disease and 14% medications for diabetes mellitus. The adjusted relationship between MRFs and pharmaceutical expenditures showed significant marginal increase in average annual pharmaceutical expenditure associated with inadequate physical activity ($519 [95% confidence interval (CI), $12–918; P=0.011]), dyslipidemia ($631 [95% CI, $168–1094; P=0.008]), hypertension: ($1078 [95% CI, $697–1460; P<0.001)], and diabetes mellitus ($2006 [95% CI, $1470–2542]). Compared with those with optimal MRFs (0–1), those with average MRFs (2–3) spent an average of $1184 (95% CI, $805–1564; P<0.001) more on medications, and those with poor MRFs (≥4) spent $2823 (95% CI, $2338–3307; P<0.001) more.
Conclusions
Worsening MRFs were proportionally associated with higher annual pharmaceutical expenditures among patients with established ASCVD regardless of non‐ASCVD comorbidity. In‐depth studies of the roles played by other factors in this association can help reduce medication‐related expenditures among ASCVD patients.
Keywords: coronary heart disease, cost, modifiable risk factors, pharmaceutical expenditure
Subject Categories: Quality and Outcomes, Health Services, Risk Factors
Introduction
Cardiovascular disease (CVD) remains the leading cause of morbidity and mortality globally,1 and atherosclerotic cardiovascular disease (ASCVD) is the most common type, accounting for >370 000 deaths annually.2 In addition to the impact on mortality, ASCVD causes significant loss of quality of life and is responsible for the highest healthcare expenditure for any single class of disease.1 In 2011, the estimated annual direct cost for CVD and stroke in the United States was about $196 billion.1 Incidence, associated morbidity, disability, mortality, and healthcare costs have been shown to depend largely on modifiable risk factors (MRFs) for CVD.3, 4, 5, 6
Expenditures on prescription medication in the US general population formed 9.8% of total healthcare spending in 2014,7 and in 2012, 26.1% of total healthcare expenditures among adults with CVD were for medications.8 Although few studies have reported incremental healthcare costs associated with worsening cardiovascular risk factor profiles among patients with diagnosed CVD,9, 10 the association between MRFs and pharmaceutical expenditures in the United States has not been examined.
Considering the economic impact of ASCVD and the significant contribution of pharmaceutical expenditure to overall healthcare expenditure, we aimed to examine the association between MRFs and pharmaceutical expenditures (both overall and medication‐specific) in a nationally representative population with established ASCVD.
Methods
Study Design and Population
We conducted a retrospective study of US adults aged ≥40 years with established ASCVD using data from the 2012 and 2013 Medical Expenditure Panel Survey (MEPS) database. MEPS, sponsored by the Agency for Healthcare Research and Quality (AHRQ), is a national survey of individuals, families, their medical providers (for medical conditions), and employers, to obtain patients' healthcare resource utilization and expenditure. Each year, the MEPS Household Components (MEPS‐HC) sample is drawn from respondents of the previous year's National Health Interview Survey. It has an overlapping panel design, with each panel composed of randomly sampled noninstitutionalized US civilians. Participants are interviewed every 6 months over a period of 30 months, and their responses are reported annually to provide nationally representative estimates of sociodemographic characteristics, medical conditions, and healthcare utilization and costs.11, 12 Interviews were conducted over the telephone with participants, and their physicians, hospitals, and pharmacies were contacted to obtain additional healthcare use and cost data. The AHRQ researchers assigned person weights and variance estimation strata to participants after data collection to reflect survey nonresponse and population totals of the participants surveyed.13
We merged the MEPS‐HC full‐year consolidated, medical conditions, and prescribed medicines files for 2012 and 2013 for this study.14, 15 Pooling this 2‐year data afforded us a larger and analyzable population of adults with ASCVD. Race and ethnicity were determined using the MEPS‐defined categories that allow respondents to report multiple Hispanic ethnicities.
Because MEPS consists of publicly available, de‐identified data files, this study was exempt from institutional review board, in accordance with US Department of Health and Human Services guidelines. Written consent was obtained from participants to contact them for interviews and to contact their healthcare providers (clinicians and pharmacies).
We classified participants as having ASCVD using the International Statistical Classification of Diseases, Ninth Revision, Clinical Modification (ICD‐9 CM) diagnosis of the condition (Table S1)16, 17 and/or self‐reported history of diagnosis of coronary heart disease, angina, myocardial infarction, and/or stroke. Our study population was limited to noninstitutionalized US adults with established ASCVD who were ≥40 years at the time of the survey (ASCVD is uncommon among younger adults), had a body mass index (calculated as weight in kilograms divided by height in meters squared) of ≥18.5 kg/m2 (underweight individuals generally represent a sicker population),18 and who had a final person‐weight >0 to be representative of the national population at the time of the survey (Figure S1).
Pharmaceutical Expenditures
During the household interview, respondents supplied the name of any prescribed medicine that they or their family members purchased or otherwise obtained during the reference period. They were also asked for permission to obtain payment data and other information from pharmacies. With the written permission from participants, pharmacies were contacted to obtain information on the medication name, national drug code, strength, quantity, date filled, and amount paid. The multisourced, person‐level medication information was then included in MEPS prescribed medicine files and linked to the Multum Lexicon databases by AHRQ researchers to assign the drugs into classes.19 The codes we used to group CVD, diabetes, and other classes of medications are shown in Table S2. More details on the collection and management of pharmaceutical data in MEPS are provided elsewhere.20 For each drug prescribed, the exact dollar amount paid was reported. Using these cost data, we calculated expenditures specific to the different classes of drugs. All expenditures were adjusted to constant 2013 US dollars using the gross domestic product deflator.
Modifiable Risk Factors and Comorbidity Burden
The MRFs examined in this study included inadequate physical activity, obesity, smoking, dyslipidemia, hypertension, and diabetes mellitus. We used responses from the self‐administered questionnaire to determine the MRF status of participants, and we classified each as a binary variable (favorable [0] versus unfavorable [1]). Any participant who did not engage in moderate vigorous physical activity 5 times a week; had a BMI ≥30 kg/m2; was a smoker at the time of interview; or reported a diagnosis of a cholesterol disorder, hypertension, or diabetes mellitus was classified as having unfavorable MRFs. Based on the presence of these individual risk factors, survey participants were categorized as poor (≥4 cardiovascular risk factors), average (2–3 cardiovascular risk factors), or optimal (0–1 cardiovascular risk factor). We also determined the contribution of comorbidity to the association between MRFs and pharmaceutical expenditures and compared the marginal pharmaceutical expenditures associated with comorbidity and those associated with MRF profile. Participants' comorbidity burden was assessed using the grouped Charlson Comorbidity Index (GCCI), which has been described elsewhere.21, 22 For our analyses, however, we modified the GCCI by excluding acute myocardial infarction, congestive heart failure and peripheral vascular disease from our estimation of GCCI score, since these were included in our definition of ASCVD; and diabetes mellitus, since it was considered a cardiovascular risk factor. We had 3 categories for GCCI: no comorbidity (0), 1 long‐term condition (1), and ≥2 long‐term conditions present other than CVD and/or diabetes mellitus (2; Table S2).
Covariates
We considered age, sex, race/ethnicity, and family income as factors that were also associated with pharmaceutical expenditures; therefore, we used these as covariates in the determination of the association between MRFs and pharmaceutical expenditures. Participant age as of the last day of the survey was classified into 3 categories: 40 to 64, 65 to 74, and ≥75 years. We had 5 categories of race and ethnicity: non‐Hispanic white, non‐Hispanic black, Hispanic, Asian, and other (American Indian, Alaska Native, and those who reported multiple races/ethnicities). There were 4 categories of family income level expressed as a proportion of the federal poverty level: <100%, 100% to 200%, 200% to 400%, and ≥400%.
Statistical Analysis
All analyses were conducted using STATA 14 (StataCorp). We accounted for the complex sampling design of MEPS in all our analyses using the final person‐weight and variance estimations (person sampling and strata). We determined the prevalence of each MRF and of the 3 MRF profiles and used χ2 statistics to test for their variations across different sociodemographic characteristics. Because the distribution of cost data is often right‐skewed secondary to the high proportion of people with no expenditure, we used a 2‐part model to estimate the marginal pharmaceutical expenditure associated with each MRF (unfavorable versus favorable), the 3 MRF profiles, and the modified GCCI. The 2‐part model consists of (1) a binary choice model fit for the probability of observing a positive‐versus‐zero annual expenditure on medications using the probit command and (2), contingent on having >$0 annual pharmaceutical expenditure, a generalized linear model (with γ distribution and a log link) was fitted for the >$0 expenditure to estimate the effect of MRFs on pharmaceutical expenditures.23, 24 To determine the appropriate distribution of the generalized linear model in the 2‐part model, we used the modified Park test.25 We used the margins postestimation command to determine the marginal and absolute pharmaceutical expenditures associated with the predictor variables in the 2‐part model. The use of the 2‐part model and the margins in STATA allowed us to estimate robust standard errors, 95% confidence intervals (CIs) and P values associated with each estimate of marginal expenditure (α level of significance was 0.05). This also helped avoid the difficulties associated with retransformation when ordinary least squares with the log scale was used in analyzing highly skewed expenditure data such as MEPS. In addition, because the svy: twopm command can be used in STATA to incorporate the complex design of MEPS in our analyses, our results can be generalizable to the US population. However, the 2‐part model did not cause the data to become normal; the skewness and kurtosis tests for normality among those with positive cost value (ie, cost >0$) showed the distribution is nonnormal (joint P<0.001).
Results
Sample Characteristics
From 2012 to 2013, there were 75 914 respondents in MEPS, representing an average annual national estimate of 314.6 million individuals—similar to the average of the projected US civilian population of 2012 and 2013. Among this surveyed population, 4248 adults aged ≥40 years with a BMI ≥18 had ASCVD (an annual equivalent of 21.9 million US adults). The mean age was 67.7 years (SD 21.6), and 44.8% were female.
The distribution of demographic characteristics for the study population is shown in Table S3. The prevalence and the variation of the prevalence of each MRF are shown in Table 1. Hypertension was the most common MRF among ASCVD adults, with a prevalence of 81.5%. Smoking and obesity were significantly more prevalent among adults aged 40 to 65 years, whereas the prevalence of dyslipidemia was higher among those aged 65 to 79 years; men were more likely to have hyperlipidemia than women. Most of the study participants (49.9%) reported an average MRF profile, and with the exception of sex, there was significant variation in the prevalence of MRF profiles across different sociodemographic characteristics (Table S4).
Table 1.
Prevalence of Individual Modifiable Risk Factors Across Sociodemographic Characteristics of Adults Aged ≥40 Years Living With ASCVD, MEPS 2012–2013
| No. of Survey Participants | Prevalence, % (95% CI) | ||||||
|---|---|---|---|---|---|---|---|
| Inadequate Physical Activity | Smoking | Obesity | Hyperlipidemia | Hypertension | Diabetes Mellitus | ||
| Overall | 4248 | 65.8 (63.6–68.0) | 16.6 (14.8–18.5) | 39.1 (36.8–41.5) | 77.1 (75.1–79.0) | 81.5 (79.8–83.1) | 33.0 (30.9–35.1) |
| Age group, y | |||||||
| 40–64 | 1647 | 63.7(60.4–66.9) | 28.5 (25.3–32.0) | 48.5 (45.1–52.0) | 71.2 (68.4–73.8) | 75.0 (72.0–77.8) | 32.5 (29.5–35.6) |
| 65–79 | 1203 | 59.7 (55.6–63.7) | 14.3 (11.6–17.5) | 41.5 (37.3–45.9) | 85.9 (82.4–88.8) | 85.7 (81.8–88.9) | 37.0 (33.0–41.2) |
| ≥80 | 1398 | 73.6 (69.9–77.0) | 4.4 (3.2–6.0) | 26.0 (22.3–30.0) | 76.4 (72.3–80.2) | 85.5 (82.8–87.9) | 30.1 (26.9–33.6) |
| P valuea | <0.001b | <0.001b | <0.001b | <0.001b | <0.001b | 0.024b | |
| Sex | |||||||
| Men | 2345 | 61.7 (58.8–64.5) | 18.2 (15.9–20.8) | 39.1 (35.8–42.4) | 80.4 (77.8–82.7) | 82.4 (79.9–84.6) | 33.3 (30.4–36.4) |
| Women | 1903 | 71.0 (68.0–73.8) | 14.5 (12.3–17.0) | 9.2 (36.1–42.4) | 73.0 (69.8–76.0) | 80.4 (77.9–82.7) | 32.5 (29.5–35.7) |
| P value | <0.001b | 0.014b | 0.945 | <0.001b | 0.273 | 0.708 | |
| Race/ethnicity | |||||||
| Non‐Hispanic white | 3191 | 64.7 (61.9–67.3) | 17.0 (15.0–19.3) | 37.7 (35.0–40.5) | 78.8 (76.3–81.1) | 80.5 (78.5–82.4) | 29.1 (26.5–31.8) |
| Non‐Hispanic black | 485 | 70.5 (66.7–74.1) | 20.0 (16.7–23.7) | 48.6 (43.6–53.6) | 69.4 (65.5–73.1) | 88.9 (85.7–91.5) | 40.7 (36.7–44.7) |
| Asian | 107 | 58.8 (47.8–69.0 | 5.3 (2.4–11.4) | 8.6 (4.6–15.4) | 81.7 (73.0–88.0) | 78.7 (68.4–86.4) | 46.1 (37.0–55.5) |
| Hispanic | 375 | 69.4 (64.3–74.0) | 10.9 (7.7–15.3) | 47.7 (42.8–52.6) | 70.4 (66.0–74.5) | 80.8 (76.9–84.2) | 45.1 (40.6–49.8) |
| Other | 90 | 76.5 (66.7–84.1) | 17.8 (9.9–29.9) | 38.3 (26.2–52.0) | 79.3 (68.5–87.1) | 82.2 (69.4–90.4) | 63.2 (47.8–76.3) |
| P value | 0.008b | 0.001b | <0.001b | <0.001b | 0.004b | <0.001b | |
| Family income levelc | |||||||
| <100% | 646 | 72.2 (67.5–76.5) | 28.9 (24.7–33.6) | 40.5 (36.6–44.6) | 77.9 (73.2–81.9) | 82.6 (78.6–86.0) | 34.9 (30.5–39.4) |
| 100–200% | 1027 | 69.1 (65.0–72.9) | 17.4 (14.7–20.4) | 37.8 (33.9–41.8) | 76.8 (73.3–80.0) | 86.3 (83.3–88.7) | 35.0 (31.4–38.9) |
| 200–400% | 1185 | 67.6 (64.1–70.9) | 13.6 (10.8–17.1) | 42.2 (37.9–46.6) | 76.5 (72.2–80.3) | 82.0 (79.2–84.5) | 33.5 (29.9–37.3) |
| >400% | 1390 | 59.0 (55.0–62.9) | 12.7 (10.0–15.9) | 36.8 (33.0–40.8) | 77.4 (74.1–80.5) | 77.1 (73.1–80.6) | 30.1 (26.2–34.3) |
| P value | <0.001b | <0.001b | 0.146 | 0.960 | <0.001b | 0.212 | |
| GCCI | |||||||
| 0 | 2866 | 63.3 (60.7–65.9) | 15.4 (13.4–17.8) | 36.2 (33.6–38.8) | 75.3 (72.9–77.6) | 79.2 (77.1–81.2) | 30.5 (28.1–32.9) |
| 1 | 867 | 69.0 (64.8–72.9) | 20.9 (17.4–24.9) | 45.8 (40.9–50.7) | 78.8 (74.2–82.8) | 86.4 (82.6–89.5) | 36.8 (32.5–41.4) |
| 2 | 515 | 74.5 (68.6–79.6) | 15.6 (11.6–20.8) | 44.3 (38.5–50.3) | 84.0 (79.5–87.6) | 86.0 (80.0–90.4) | 40.3 (33.4–47.6) |
| P value | <0.001b | 0.026b | <0.001b | 0.007b | 0.003 | 0.004b | |
ASCVD indicates atherosclerotic cardiovascular disease; CI, confidence interval; GCCI, grouped Charlson Comorbidity Index; MEPS, Medical Expenditure Panel Survey.
χ2 Statistic used to test difference in proportions between respondents.
Statistically significant.
Family income expressed as a proportion of the federal poverty level.
Annual Pharmaceutical Expenditures Among Those With ASCVD
Of the 4248 adults with ASCVD in 2012–2013, 95.4% used prescription medications; 86.8% used ≥1 CVD medication. The annual per capita pharmaceutical expenditures among adults with ASCVD was $3432. On average, 34% of total pharmaceutical expenditure was spent on CVD medications ($1139; 95% CI, $1063–1215), and 14% ($482; 95% CI, $407–556) was spent on antidiabetic medications. The residual 52% ($1786) was for medications other than those for CVD and diabetes (Figure 1). When projected using MEPS's complex designs, the 2012–2013 annual total pharmaceutical spending among those with ASCVD was an estimated $71.6 billion; $23.8 billion was spent on CVD medications, $10 billion was spent on diabetes medications, and approximately $37.8 billion was spent on other medications.
Figure 1.
Annual per capita pharmaceutical expenditures for different medication classes among adults with ASCVD, MEPS 2012–2013. All costs are in 2013 US dollars. ASCVD indicates atherosclerotic cardiovascular disease; CNS, central nervous system; CVD, central nervous system; GI, gastrointestinal; MEPS, Medical Expenditure Panel Survey; RS, respiratory system.
Effects of MRFs on Annul Pharmaceutical Expenditures Among Those With ASCVD
The unadjusted and adjusted marginal pharmaceutical expenditures associated with the presence versus absence of individual MRFs is shown in Table 2. Inadequate physical activity, dyslipidemia, hypertension, and diabetes mellitus were statistically significantly associated with total pharmaceutical expenditure after adjusting for sociodemographic characteristics and burden of comorbid conditions. The marginal pharmaceutical expenditure associated with diabetes mellitus was the highest of all MRFs at $2006 (95% CI, $1470–$2542; P<0.001). When different categories of medication expenditures were examined, dyslipidemia, hypertension, and diabetes mellitus were significantly associated with increased CVD medication expenditures; obesity and diabetes mellitus were associated with increased expenditures for diabetes medication; and inadequate physical activity and hypertension were significantly associated with an increase in non‐CVD, nondiabetic medication expenditures.
Table 2.
Marginal Pharmaceutical Expenditures Associated With Each MRF: Results From the 2‐Part Econometric Model, MEPS 2012–2013
| MRF | Univariate (Unadjusted) | Model 1a | Model 2b | |||
|---|---|---|---|---|---|---|
| Marginal Expenditures (95% CI) | P Value | Marginal Expenditures (95% CI) | P Value | Marginal Expenditures (95% CI) | P Value | |
| Overall pharmaceutical expenditures | ||||||
| Inadequate vs optimal physical activity | 864 (336–1391) | 0.001c | 893 (430–1355) | <0.001c | 520 (121–918) | 0.011c |
| Obesity vs normal BMI | 1271 (726–1817) | <0.001c | 1146 (675–1616) | <0.001c | 349 (−106 to 8030 | 0.132 |
| Currently smoking vs nonsmoker | 719 (29–1409) | 0.007c | 411 (−267 to 1090) | 0.233 | 400 (−237 to 1037) | 0.217 |
| Dyslipidemia vs no dyslipidemia | 1031 (396–1666) | 0.002c | 1143 (627–1659) | <0.001c | 631 (168–1094) | 0.008c |
| Hypertension vs no hypertension | 1679 (1261–2096) | <0.001c | 1679 (1302–2056) | <0.001c | 1079 (697–1460) | <0.001c |
| Diabetes mellitus vs no diabetes mellitus | 2557 (1997–3117) | <0.001c | 2564 (2020–3108) | <0.001c | 2006 (1470–2542) | <0.001c |
| CVD medication expenditures | ||||||
| Inadequate vs optimal physical activity | 118 (−32 to 268) | 0.123 | 142 (−8 to 291) | 0.063 | 73 (−76 to 222) | 0.333 |
| Obesity vs normal BMI | 223 (74–3730 | 0.004c | 272 (123–422) | <0.001 | 141 (−24 to 305) | 0.093 |
| Currently smoking vs nonsmoker | 65 (−148 to 278) | 0.551 | 145 (−73 to 363) | 0.190 | 81 (−111 to 273) | 0.408 |
| Dyslipidemia vs no dyslipidemia | 649 (502–795) | <0.001c | 631 (489–774) | <0.001c | 506 (350–663) | <0.001c |
| Hypertension vs no hypertension | 574 (426–723) | <0.001c | 567 (420–715) | <0.001c | 394 (230–558) | <0.001c |
| Diabetes mellitus vs no diabetes mellitus | 357 (202–512) | <0.001c | 406 (254–557) | <0.001c | 235 (77–393) | 0.004c |
| Diabetes medication expenditures | ||||||
| Inadequate vs optimal physical activity | 253 (120–386) | <0.001c | 255 (121–388) | <0.001c | 10 1 (−23 to 226) | 0.110 |
| Obesity vs normal BMI | 502 (315–688) | <0.001c | 476 (297–656) | <0.001c | 227 (94–360) | 0.001c |
| Currently smoking vs nonsmoker | 53 (−171 to 2765) | 0.643 | −15 (−213 to 183) | 0.882 | 116 (−80 to 312) | 0.245 |
| Dyslipidemia vs no dyslipidemia | 168 (10–325) | 0.037c | 178 (23–333) | 0.024c | −40 (−194 to 114) | 0.608 |
| Hypertension vs no hypertension | 311 (153–469) | <0.001c | 328 (198–459) | <0.001c | 77 (−75 to 228) | 0.320 |
| Diabetes mellitus vs no diabetes mellitus | 1399 (1193–1606) | <0.001c | 1540 (1145–1934) | <0.001c | 1296 (1108–1484) | <0.001c |
| Other medication expenditures | ||||||
| Inadequate vs optimal physical activity | 443 (−25 to 911) | 0.063 | 480 (135–825) | 0.007c | 319 (16–622) | 0.039c |
| Obesity vs normal BMI | 520 (61–979) | 0.027c | 340 (−25 to 704) | 0.068 | 77 (−315 to 469) | 0.698 |
| Currently smoking vs nonsmoker | 776 (187–1366) | 0.01c | 324 (−181 to 828) | 0.207 | 240 (−248 to 728) | 0.333 |
| Dyslipidemia vs no dyslipidemia | 22 (−614 to 658) | 0.946 | 160 (−296 to 616) | 0.490 | −144 (−568 to 280) | 0.504 |
| Hypertension vs no hypertension | 626 (314–938) | <0.001c | 604 (329–878) | <0.001c | 492 (219–765) | <0.001c |
| Diabetes mellitus vs no diabetes mellitus | 697 (275–1120) | 0.001c | 611 (242–979) | 0.001c | 386 (13–759) | 0.042c |
All cost are in 2013 US dollars. BMI indicates body mass index; CI, confidence interval; CVD, cardiovascular disease; MEPS, Medical Expenditure Panel Survey; MRF, modifiable risk factor.
Model 1: Each MRF was used as a predictor and adjusted for age, sex, race/ethnicity, and income level (all variable entered as categorical variables).
Model 2: All MRFs were entered simultaneously and adjusted for age, sex, race/ethnicity, income level, and Charlson Comorbidity Index (all variables entered as categorical variables).
Statistically significant.
The average annual pharmaceutical expenditure among patients with ASCVD who had an optimal MRF profile was $1400 (95% CI, $1073–1728) compared with $2672 (95% CI, $2332–3013) among those with an average MRF profile and $4516 (95% CI, $4067–4965) among those with a poor MRF profile. The unadjusted and adjusted marginal pharmaceutical expenditures associated with MRF profiles are shown in Table 3. After accounting for demographics, income status, and comorbid conditions, the annual pharmaceutical expenditure was $1184 (95% CI, $805–1564) higher among those with average MRF profiles and $2823 (95% CI, $2338–3307) higher among ASCVD patients with poor MRF profiles compared with those with optimal MRF profiles (P<0.001). Similar trends were noted when mean expenditures for specific medication classes (CVD, diabetes, and other) were considered.
Table 3.
Per Capita Marginal Pharmaceutical Expenditures Associated With Grouped MRFs and Charlson Comorbidity Index Among Adults With ASCVD, MEPS 2012–2013
| MRF Profile | Unadjusted | Adjusteda | ||
|---|---|---|---|---|
| Overall pharmaceutical expendituresb (95% CI) | ||||
| Optimal MRF | Reference | P value | Reference | P‐value |
| Average MRF | 1272 (844–1700) | <0.001c | 1184 (804–1564) | <0.001c |
| Poor MRF | 3115 (2645–3586) | <0.001c | 2823 (2338–3308) | <0.001c |
| CVD‐medication expenditures (95% CI) | ||||
| Optimal MRF | Reference | P value | Reference | P value |
| Average MRF | 396 (200–591) | <0.001c | 406 (231–582) | <0.001c |
| Poor MRF | 791 (589–993) | <0.001c | 848 (653–1043) | <0.001c |
| Diabetes medication expenditures (95% CI) | ||||
| Optimal MRF | Reference | P value | Reference | P value |
| Average MRF | 136 (84–188) | <0.001c | 131 (77–185) | <0.001c |
| Poor MRF | 969 (797–1141) | <0.001c | 943 (760–1126) | <0.001c |
| Other medication expenditures (95% CI) | ||||
| Optimal MRF | Reference | P value | Reference | P value |
| Average MRF | 622 (199–1045) | 0.004c | 526 (134–918) | 0.009c |
| Poor MRF | 1167 (773–1560) | <0.001c | 807 (434–1180) | <0.001c |
| Modified Grouped CCI | Unadjusted | Adjustedd | ||
|---|---|---|---|---|
| Overall pharmaceutical expendituresb (95% CI) | ||||
| 0 | Reference | P value | Reference | P value |
| 1 | 2603 (1723–3483) | <0.001c | 2272 (1519–3025) | <0.001c |
| 2 | 2518 (1665–3371) | <0.001c | 2068 (1349–2787) | <0.001c |
| CVD‐medication expenditures (95% CI) | ||||
| 0 | Reference | P value | Reference | P value |
| 1 | 53 (−130 to 236) | 0.567 | 21 (−153 to 195) | 0.811 |
| 2 | 132 (−70 to 334) | 0.2c | 30 (−169 to 229) | 0.77 |
| Diabetes medication expenditures (95% CI) | ||||
| 0 | Reference | P value | Reference | P value |
| 1 | 159 (21–297) | <0.024c | 29 (−101 to 160) | <0.001c |
| 2 | 375 (94–656) | 0.009c | 206 (1–411) | 0.048 |
| Other medication expenditures (95% CI) | ||||
| 0 | Reference | P value | Reference | P value |
| 1 | 2312 (1514–3111) | <0.001c | 2015 (1412–2617) | <0.001c |
| 2 | 1903 (1274–2532) | <0.001c | 1773 (1189–2357) | <0.001c |
All costs are in 2013 US dollars. ASCVD indicates atherosclerotic cardiovascular disease; CCI, Charlson Comorbidity Index; CI, confidence interval; CVD, cardiovascular disease; MRF, modifiable risk factors; MEPS, Medical Expenditure Panel Survey.
Adjusted for age, sex, race/ethnicity, income level, CCI.
All Costs are in 2013 USD.
Statistically significant.
Adjusted for age, sex, race/ethnicity, income level, and grouped MRFs category.
Figure 2 describes the association among both MRF profiles and increasing burden of comorbid conditions with annual pharmaceutical expenditures (overall and medication‐specific). The lowest annual drug costs were observed among those with optimal MRF profiles and no major comorbid conditions ($1386). In contrast, ASCVD patients with poor MRF profiles and a GCCI of 2 had the highest annual pharmaceutical expenditures ($6948). Across worsening comorbidity, individuals with poor MRF profiles incurred the highest pharmaceutical expenditures (Figure 2). In a subanalysis of specific CVD medications, annual costs of antihyperlipidemics, antihypertensives, and coagulation modifiers significantly increased with worsening MRF profile across all categories of comorbidity burden (Figure 3).
Figure 2.
Mean pharmaceutical expenditures associated with grouped MRFs across different levels of grouped CCI among those with ASCVD, MEPS 2012–2013. Mean expenditures were estimated using the person weight and variance estimation stratum and person sampling unit of MEPS 2012–2013. All costs are in 2013 US dollars. ASCVD indicates atherosclerotic cardiovascular disease; CCI, Charlson Comorbidity Index; CVD, cardiovascular disease; MEPS, Medical Expenditure Panel Survey; MRF, modifiable risk factor.
Figure 3.
Mean CVD medication expenditures and their association with MRF profiles across different levels of grouped CCI among those with ASCVD, MEPS 2012–2013. Mean expenditures were estimated using the person weight, variance estimation stratum, and person sampling unit of MEPS 2013. All costs are in 2013 US dollars. ASCVD indicates atherosclerotic cardiovascular diseases; CCI, Charlson Comorbidity Index; CVD, cardiovascular disease; MEPS, Medical Expenditure Panel Survey; MRF, modifiable risk factor.
Discussion
In a nationally representative population, our study demonstrated that in 2012–2013, adults with established ASCVD spent $284 billion on health care per annum. Of this, $71.6 billion was spent on medications. The contributions of medications for CVD and diabetes mellitus to the overall pharmaceutical expenditures were $23.8 billion (34%) and $10 billion (14%), respectively. More than half (52%) of the overall pharmaceutical expenditure was spent on non‐CVD, nondiabetes medications. The details of the 2012–2013 per capita pharmaceutical expenditures examined in the context of the mean healthcare expenditures in our study population are shown in Figure S2. The overall pharmaceutical expenditure was significantly associated with individual MRFs; inadequate physical activity, dyslipidemia, hypertension, and diabetes mellitus as well as worsening MRF profile were all associated with higher pharmaceutical expenditure. Similar patterns were observed when expenditures for specific medication classes (CVD, diabetes, and others) were examined. These associations persisted even after accounting for underlying comorbid conditions.
Many studies have found that current MRFs are important drivers of future morbidity and mortality among individuals with established CVD in a dose‐response fashion.26, 27, 28 Although studies within and outside the United States have attempted to estimate the incremental healthcare expenditures associated with individual cardiovascular MRFs among those without established CVD,29, 30 no study has detailed the potential economic impact of CVD MRFs on pharmaceutical costs, which remains one of the largest contributors to overall healthcare expenditures. Sullivan et al showed that among persons with cardiometabolic risk factor clusters (BMI ≥25 and any 2 of hypertension, hyperlipidemia, and diabetes mellitus), 34% of total healthcare expenditure was for prescription medications in 2006,10 and depending on the source of payment, the proportion could be as high as 49%.8 In Australia, Ademi et al showed that obesity, hypertension, and diabetes mellitus are predictors of higher pharmaceutical expenditures among persons with or at risk of CVD.29
The findings of Ademi et al are similar to ours. Our study of the US adults with ASCVD in 2012–2013 showed that individual cardiovascular MRFs such as inadequate physical activity, dyslipidemia, hypertension, and diabetes mellitus were significantly associated with higher pharmaceutical expenditures of $519, $631, $1078, and $2006, respectively, compared with adults without the respective risk factors, after accounting for sociodemographic factors and comorbidity. In addition, we found significantly higher pharmaceutical expenditures (all medications, CVD‐specific medications, and non‐CVD pharmaceutical expenditures) associated with poor cardiovascular MRF profiles among persons with established ASCVD in a nationally representative cohort, after accounting for underlying comorbid conditions. The adjusted average annual pharmaceutical expenditure was highest among those with a poor MRF profile ($4516) and lowest for those with an optimal MRF profile ($1400).
Another significant finding of our study is that although the study population consisted of those with established ASCVD, non‐CVD medications contributed the most toward total pharmaceutical cost. This is comparable to the findings of the Cooper Center Longitudinal Study by Willis et al, in which they reported that average annual non‐CVD healthcare cost was higher than CVD‐related healthcare cost and that overall cost increased with worsening MRF profile.30 These findings are not unexpected considering the complex interplay among CVD, MRFs, and associated comorbidities31 in increasing healthcare resource utilization and cost.
Some studies have demonstrated an association between comorbidity and healthcare costs.32, 33, 34 In our study, comorbidity was found to be associated with higher pharmaceutical expenditure among people with established ASCVD. Beyond that, our study also demonstrated that the burden of cardiovascular MRF may have a higher impact on pharmaceutical expenditure than the burden of other comorbid conditions among those with ASCVD. The marginal pharmaceutical expenditures associated with worsening MRF profiles were larger than those associated with a higher burden of comorbid conditions (see Table 3). This underscores the importance of addressing MRF prevention in the general population, especially because the prevention of MRFs reduces the risk of ASCVD. This approach can potentially lead to pharmaceutical cost saving and, ultimately, reduced healthcare spending.
As projected, 40.5% of the US population will have a type of CVD by 2030 and will spend $818 billion in annual direct medical cost—a significant leap from the $217 billion spent in 2010.1 It is imperative that all possible means targeted at stalling or halting this projected economic impact of CVD be explored today. An integrated approach to the management of ASCVD, its associated MRFs, and other comorbidities could be explored to reduce spending. The concept of “bundle payment” in the “one cycle of care”35 may be adapted to prescription medications and explored among ASCVD patients, perhaps taking the approach of “one price for a year's supply” to prescriptions used in managing chronic conditions. The use of the payment‐for‐outcome approach in prescription medication can also encourage pharmacists and clinicians to improve on quality, rather than quantity, of services. Careful monitoring of healthcare costs and outcomes must be implemented for early identification of negative trends and prompt institution of mitigation measures.
This study has several strengths. First, MEPS's careful design and execution involved multilevel verification of information collected from participants.36 Second, the oversampling of minority race/ethnicity such as Hispanic and black make results generalizable to all races in the United States. Third, the large sample size allowed us to adequately characterize persons with ASCVD by MRFs and yet have enough participants to estimate marginal expenditures associated with MRFs. The results of our study, however, must be interpreted in the context of the following limitations. First, we were limited to the 3‐digit ICD‐9‐CM code used to map medical conditions, which means our observed prevalence of ASCVD may be underestimated. Second, because cardiovascular MRFs were self‐reported, the true national prevalence is likely underestimated.37 Third, the use of the Multum Lexicon drug classification system, although apt for drug classification, may overestimate expenditures for CVD medications because some CVD medications may have non‐CVD uses. Fourth, MEPS does not account for over‐the‐counter prescription expenditures, and this may also underestimate the average pharmaceutical expenditures. Fifth, MEPS was conducted among noninstitutionalized US civilians, and thus our results are nonrepresentative of the entire US population. Finally, although we comprehensively controlled for variables chosen based on existing knowledge of factors associated with higher pharmaceutical expenditures and scientific selection of statistical models using Akaike's criteria, there may be unobserved characteristics that affect the outcomes studied, causing residual confounding; factors such as health insurance, clinicians, pharmacists, healthcare organization, drug companies, and even patient behaviors are important determinants of pharmaceutical expenditures that could not be assessed in this study.
Conclusion
Some individual MRFs and worsening MRF profiles among persons with established ASCVD are associated with a higher annual pharmaceutical expenditure. Future studies are needed to demonstrate whether patient‐centered pragmatic approaches to manage and prevent MRFs will curtail the rising pharmaceutical expenditures (on all medications and CVD‐specific medications) among those with ASCVD.
Disclosures
Dr Nasir is on the Advisory Board for Quest Diagnostic, and he is a consultant for Regeneron. There is no other potential conflict of interest relevant to this study.
Supporting information
Table S1. ICD‐9 CM codes of diseases classified as ASCVD, and Multum Lexicon Drug Codes for CVD, Diabetic, and other Major Medication Classes
Table S2. Charlson Comorbidity Index Scoring System
Table S3. Sociodemographic Characteristics of Adults Aged ≥40 Years Living With Atherosclerotic Cardiovascular Disease, Medical Expenditure Panel Survey 2012–2013
Table S4. Prevalence of Modifiable Risk Factor Profile Across Sociodemographic Characteristics of Adults Aged ≥40 Years Living With Atherosclerotic Cardiovascular Disease, Medical Expenditure Panel Survey 2012–2013
Figure S1. Flow chart showing the study population selection, Medical Expenditure Panel Survey 2012–2013.
Figure S2. Mean healthcare expenditure across modifiable risk factor profile and Charlson Comorbidity Index.
(J Am Heart Assoc. 2017;6:e004996 DOI: 10.1161/JAHA.116.004996.)28600400
References
- 1. Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman M, de Ferranti S, Després JP, Fullerton HJ, Howard VJ, Huffman MD; American Heart Association Statistics Committee and Stroke Statistics Subcommittee . Heart disease and stroke statistics—2015 update: a report from the American Heart Association. Circulation. 2015;131:e29–e322. [DOI] [PubMed] [Google Scholar]
- 2. CDC, NCHS . Underlying cause of death 1999–2013 on CDC WONDER online database, released 2015. Data are from the Multiple Cause of Death Files, 1999‐2013, as compiled from data provided by the 57 vital statistics jurisdictions through the Vital Statistics Cooperative Program. Available at: https://wonder.cdc.gov/ucd-icd10.html. Accessed February 3, 2015.
- 3. Qiu M, Shen W, Song X, Ju L, Tong W, Wang H, Zheng S, Jin Y, Wu Y, Wang W, Tian J. Effects of prediabetes mellitus alone or plus hypertension on subsequent occurrence of cardiovascular disease and diabetes mellitus: longitudinal study. Hypertension. 2015;65:525–530. [DOI] [PubMed] [Google Scholar]
- 4. Cornier MA, Despres JP, Davis N, Grossniklaus DA, Klein S, Lamarche B, Lopez‐Jimenez F, Rao G, St‐Onge MP, Towfighi A, Poirier P. Assessing adiposity: a scientific statement from the American Heart Association. Circulation. 2011;124:1996–2019. [DOI] [PubMed] [Google Scholar]
- 5. Williams L. Third report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel III) final report. Circulation. 2002;106:3143. [PubMed] [Google Scholar]
- 6. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, Jones DW, Materson BJ, Oparil S, Wright JT Jr, Roccella EJ. Seventh report of the joint national committee on prevention, detection, evaluation, and treatment of high blood pressure. Hypertension. 2003;42:1206–1252. [DOI] [PubMed] [Google Scholar]
- 7. Center for Disease Control . Health expenditures. 2016. Available at: http://www.cdc.gov/nchs/fastats/health-expenditures.htm. Accessed November 28, 2016.
- 8. Salami JA, Valero‐Elizondo J, Osondu CU, Ogunmoroti O, Post JM, Younus A, Latif MA, Ahmad R, Arrieta A, Spatz ES, Rana JS. Per capita proportion of total health care expenditures on pharmaceuticals among US adults with cardiovascular disease: 2012 Medical Expenditure Panel Survey. Circ Cardiovasc Qual Outcomes. 2016;9(Suppl 2):A230. [Google Scholar]
- 9. Willis BL, DeFina LF, Bachmann JM, Franzini L, Shay CM, Gao A, Leonard D, Berry JD. Association of ideal cardiovascular health and long‐term healthcare costs. Am J Prev Med. 2015;49:678–685. [DOI] [PubMed] [Google Scholar]
- 10. Sullivan PW, Ghushchyan V, Wyatt HR, Hill JO. The medical cost of cardiometabolic risk factor clusters in the United States. Obesity (Silver Spring). 2007;15:3150–3158. [DOI] [PubMed] [Google Scholar]
- 11. Cohen JW, Monheit AC, Beauregard KM, Cohen SB, Lefkowitz DC, Potter D, Arnett RH III. The Medical Expenditure Panel Survey: a national health information resource. Inquiry. 1996;33:373–389. [PubMed] [Google Scholar]
- 12. Medical Expenditure Panel Survey . Agency for Healthcare Research and Quality Web site. Available at: http://www.meps.ahrq.gov/mepsweb. Accessed November 30, 2015.
- 13. Medical Expenditure Panel Survey: Survey Background. Available at: http://meps.ahrq.gov/mepsweb/about_meps/survey_back.jsp. Accessed December 15, 2015.
- 14. Medical Expenditure Panel Survey: 2013 full‐year consolidated data file. Available at: http://meps.ahrq.gov/mepsweb/data_stats/download_data_files_detail.jsp?cboPufNumber=HC-163. Accessed November 30, 2015.
- 15. Medical Expenditure Panel Survey: 2013 prescribed medicine file. Available at: http://meps.ahrq.gov/mepsweb/data_stats/download_data_files_detail.jsp?cboPufNumber=HC-160A. Accessed November 30, 2015.
- 16. International Classification of Disease: Disease of the Circulatory System. Available at: http://www.icd9data.com/2013/Volume1/390-459/default.htm. Accessed November 30, 2015.
- 17. American Heart Association: Coronary Artery Disease. Available at: http://www.heart.org/HEARTORG/Conditions/More/MyHeartandStrokeNews/Coronary-Artery-Disease-Coronary-Heart-Disease_UCM_436416_Article.jsp#. Accessed November 30, 2015.
- 18. Florez H, Castillo‐Florez S. Beyond the obesity paradox in diabetes: fitness, fatness, and mortality. JAMA. 2012;308:619–620. [DOI] [PubMed] [Google Scholar]
- 19. Cerner . Multum lexicon drug classification system. Available at: http://www.cerner.com/cerner_multum/. Accessed November 20, 2016.
- 20. Salami JA, Warraich H, Valero‐Elizondo J, Spatz ES, Desai NR, Rana JS, Virani SS, Blankstein R, Khera A, Blaha MJ, Blumenthal RS. National trends in statin use and expenditures in the US adult population from 2002 to 2013: insights from the Medical Expenditure Panel Survey. JAMA Cardiol. 2017;2:56–65. [DOI] [PubMed] [Google Scholar]
- 21. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373–383. [DOI] [PubMed] [Google Scholar]
- 22. de Groot V, Beckerman H, Lankhorst GJ, Bouter LM. How to measure comorbidity: a critical review of available methods. J Clin Epidemiol. 2003;56:221–229. [DOI] [PubMed] [Google Scholar]
- 23. Belotti F, Deb P, Manning WG, Norton EC. twopm: two‐part models. Stata J. 2015;15:3–20. [Google Scholar]
- 24. Hardin J, HIilbe J. Generalized Linear Models and Extensions. 3rd ed College Station: StataCorp LP: Stata Press; 2012. [Google Scholar]
- 25. Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001;20:461–494. [DOI] [PubMed] [Google Scholar]
- 26. Espeland MA, Glick HA, Bertoni A, Brancati FL, Bray GA, Clark JM, Curtis JM, Egan C, Evans M, Foreyt JP, Ghazarian S. Impact of an intensive lifestyle intervention on use and cost of medical services among overweight and obese adults with type 2 diabetes: the action for health in diabetes. Diabetes Care. 2014;37:2548–2556. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Kahn R, Robertson RM, Smith R, Eddy D. The impact of prevention on reducing the burden of cardiovascular disease. Circulation. 2008;118:576–585. [DOI] [PubMed] [Google Scholar]
- 28. Murakami Y, Okamura T, Nakamura K, Miura K, Ueshima H. The clustering of cardiovascular disease risk factors and their impacts on annual medical expenditure in Japan: community‐based cost analysis using Gamma regression models. BMJ Open. 2013;3:e002234. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Ademi Z, Liew D, Hollingsworth B, Steg G, Bhatt DL, Reid CM. Predictors of annual pharmaceutical costs in Australia for community‐based individuals with, or at risk of, cardiovascular disease. Am J Cardiovasc Drugs. 2010;10:85–94. [DOI] [PubMed] [Google Scholar]
- 30. Willis BL, DeFina LF, Bachmann JM, Franzini L, Shay CM, Gao A, Leonard D, Berry JD. Association of ideal cardiovascular health and long‐term healthcare costs. AM J Prev Med. 2015;49:678–85. [DOI] [PubMed] [Google Scholar]
- 31. Leckie EV, Withers RF. Obesity and depression. J Psychosom Res. 1967;11:107–115. [DOI] [PubMed] [Google Scholar]
- 32. Simpson SH, Corabian P, Jacobs P, Johnson JA. The cost of major comorbidity in people with diabetes mellitus. Can Med Assoc J. 2003;168:1661–1667. [PMC free article] [PubMed] [Google Scholar]
- 33. Dalal AA, Shah M, Lunacsek O, Hanania NA. Clinical and economic burden of patients diagnosed with COPD with comorbid cardiovascular disease. Respir Med. 2011;105:1516–1522. [DOI] [PubMed] [Google Scholar]
- 34. Hodgson TA, Cai L. Medical care expenditures for hypertension, its complications, and its comorbidities. Med Care. 2001;39:599–615. [DOI] [PubMed] [Google Scholar]
- 35. Froimson MI, Rana A, White RE, Marshall A, Schutzer SF, Healy WL, Naas P, Daubert G, Iorio R, Parsley B. Bundled payments for care improvement initiative: the next evolution of payment formulations: AAHKS Bundled Payment Task Force. J Arthroplasty. 2013;28:157–165. [DOI] [PubMed] [Google Scholar]
- 36. Cohen SB. Design strategies and innovations in the Medical Expenditure Panel Survey. Med Care. 2003;41:III‐5–III‐12. [DOI] [PubMed] [Google Scholar]
- 37. National Center for Health Statistics (US) . Evaluation of National Health Interview Survey Diagnostic Reporting. Hyattsville, United States: US Department of Health & Human Services; 1994. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1. ICD‐9 CM codes of diseases classified as ASCVD, and Multum Lexicon Drug Codes for CVD, Diabetic, and other Major Medication Classes
Table S2. Charlson Comorbidity Index Scoring System
Table S3. Sociodemographic Characteristics of Adults Aged ≥40 Years Living With Atherosclerotic Cardiovascular Disease, Medical Expenditure Panel Survey 2012–2013
Table S4. Prevalence of Modifiable Risk Factor Profile Across Sociodemographic Characteristics of Adults Aged ≥40 Years Living With Atherosclerotic Cardiovascular Disease, Medical Expenditure Panel Survey 2012–2013
Figure S1. Flow chart showing the study population selection, Medical Expenditure Panel Survey 2012–2013.
Figure S2. Mean healthcare expenditure across modifiable risk factor profile and Charlson Comorbidity Index.