Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2013 Dec 16.
Published in final edited form as: Value Health. 2012 Apr 10;15(3):10.1016/j.jval.2012.01.008. doi: 10.1016/j.jval.2012.01.008

Simulated Value Based Insurance Design Applied to Statin Use by Medicare Beneficiaries with Diabetes

Amy Davidoff 1, Ruth Lopert 2, Bruce Stuart 1, Thomas Shaffer 1,3, Jennifer Lloyd 3, J Samantha Shoemaker 1
PMCID: PMC3864093  NIHMSID: NIHMS515005  PMID: 22583449

Abstract

Objective

To examine cost responsiveness and total costs associated with a simulated “value based” insurance design (VBID) for statin therapy in a Medicare population with diabetes.

Methods

Four-year panels constructed from the 1997–2005 Medicare Current Beneficiary Survey selected by self-report or claims-based diagnoses of diabetes in Year 1, and use of statins in Year 2 (N= 899). We computed number of 30-day statin prescription fills, out-of-pocket (OOP) and third party drug costs, and Medicare Part A and B spending.

Methods

Multivariate ordinary least squares regression models predicted statins fills as a function of OOP costs, and a generalized linear model with log link predicted Medicare spending as a function of number of fills, controlling for baseline characteristics. Estimated coefficients were used to simulate changes in fills associated with copayment caps from $25 to $1, and to compute changes in 3rd-party payments and Medicare cost offsets associated with incremental fills. Analyses were stratified by patient cardiovascular event risk.

Results

A simulated OOP price of $25[$1] increased plan drug spending by $340[$794], and generated Medicare Part A/B savings of $262[$531]. Medicare Part A/B savings were greater for higher risk patients, generating a net savings for the plans.

Conclusions

Reducing statin copayments for Medicare beneficiaries with diabetes resulted in modestly increased use, and reduced medical spending. The VBID simulation strategy met financial feasibility criteria, but only for higher risk patients.

Keywords: diabetes, Medicare, medication adherence, cost offsets

Introduction

Value based insurance designs (VBID) have been proposed as a means of slowing the dramatic rise in healthcare spending associated with the treatment of chronic health conditions1,2. VBID incorporates financial incentives into prescription drug or other insurance benefits in order to encourage the initiation of, and adherence to, prevention and treatment regimens for which higher utilization levels are expected to yield clinical benefits which will in turn generate downstream savings in the medical care budget. The types of financial incentives that have been suggested include the reduction or complete elimination of cost-sharing, as well as incentives or rewards for high levels of adherence3.

Evaluating the success of a VBID strategy depends on the perspective taken in the analysis. The societal perspective focuses on the net change in total spending associated with changes in both medication use and spending on other medical care. It may also incorporate the value of any incremental survival or improvement in quality of life associated with increased medication use. However, given that most health insurance is provided through either private markets, or public programs with budget constraints, it may be more useful to consider the perspective of the third party payer. This perspective limits the focus to direct spending, and in particular, spending by the payer.

Figure 1 illustrates the hypothetical process through which VBID functions. Starting in the upper right, application of VBID reduces patient cost sharing for the targeted medication, shifting more of the payment to the third party payer. Assuming price responsiveness, this reduced OOP price will increase medication use at the margin (upper left). The combination of increased medication use and increased share of payment for all medications (not just the marginal medication use) results in a net increase in 3rd party spending on medications (lower left). With appropriate targeting, the increased use of medications is presumed to improve health status, and reduce the need for medical care services and spending. If the savings in medical spending to third parties are larger than the increase in spending on medications, then VBID is a desirable strategy.

Figure 1.

Figure 1

Open in a new tab

Value Based Insurance Design at the margin, 3rd party.

The success of a VBID strategy depends on the magnitude of two key relationships – the behavioral responsiveness of medication use to out of pocket price, and the relationship between medication use and health outcomes that affect medical care spending. VBID may be applied universally – in other words, to all patients for whom a drug might be indicated, or selectively to targeted subgroups of patients most likely to benefit clinically or to respond more readily to benefit design incentives. As the ability to target the benefit structure increases, the larger are the cost savings likely to accrue to the insurer.

Treatments for patients with diabetes have become a focus for VBID inquiries because of the evidence of cost savings associated with appropriate medication use. The use of statins to prevent or delay the secondary complications of diabetes has been shown to improve clinical outcomes for older adults4–7. There is also increasing evidence that improved adherence to this class of medications reduces medical costs8–15. In a recent observational study that included a variety of controls for confounding, Stuart et al. (2011)16 identified a savings of $832 associated with 10% higher adherence to statins, when measured over a three year period.

If improved medication adherence is cost-saving, the question is whether patients are sufficiently responsive to out-of-pocket (OOP) price changes such that reduced co-pays can motivate a sufficiently large increase in medication use. The price elasticity of demand associated with cost sharing and prescription drug expenditures for management of chronic conditions generally ranges between −0.2 to −0.617,18, depending on the study design, drug, and population. This indicates that a 10% decrease in out of pocket spending is associated with a 2–6% increase in spending on prescription drugs. Several studies have focused on antidiabetic medications, finding changes in copayments to be associated with changes in medication use.19–21. Goldman et al. (2004) found a 25% reduction in days supplied of antidiabetic medications when copayments were doubled22 Some studies have examined the response to increases in out of pocket prices. For example, a longitudinal study of statin users found that a 100% co-payment increase resulted in reduced statin adherence rates of 2.6% and 1.1% for new and continuing users, respectively23. In contrast, a 30% reduction in copayments from a large employer resulted in a 3.4% increased utilization of statins24. In general, Medicare beneficiaries purchasing maintenance medications used for the long-term treatment of chronic conditions have been found to have a low level of response to cost sharing incentives18. Chernew et al. (2008)24 reported price elasticities for diabetic drugs (−0.136) and statins (−0.182), estimates that were at the lower end of the range previously reported for medications in patients with other chronic conditions.

There have been several types of studies focused on demonstrating the potential for VBID interventions. An early study by Rosen et al25 used a Markov model to simulate how the elimination of cost sharing would increase drug use in diabetics, reduce disease progression, and ultimately reduce medical care utilization and spending. However, the estimates used to populate various nodes in the model were taken from the literature, and may not have been specific to the study sample or environment. Other studies have drawn on VBID interventions, principally observational studies or small convenience samples where single employers, such as Pitney Bowes, implemented a VBID program change26. Some more recent studies have included comparison groups27–29, with results pending for two prospective controlled trials30,31. To date, there is evidence that VBID-related reductions in cost sharing result in small but statistically significant increases in medication use. However, no studies have also demonstrated medical cost offsets to the increased cost borne by the insurer. Furthermore, these studies examine the effects of VBID in employee populations, but have not yet addressed the potential for VBID in the Medicare population.

In this study we explored the potential cost savings associated with a simulated VBID strategy related to statin adherence for an elderly population with diabetes. Extending prior research examining medical cost savings associated with statin adherence,16 we used a sample of Medicare beneficiaries with diabetes selected from the nationally representative Medicare Current Beneficiary Survey, following them for up to three years after baseline assessment. We estimated multivariate models of demand for statins as a function of OOP price, and then estimated the relationship between statin fills and Medicare spending. We used microsimulation to assess the total and Medicare cost saving potential associated with reductions in cost sharing for statins. We also stratified our estimates using a modified application of the Framingham Risk Calculator as a way to examine the role of targeting.

METHODS

Data and Study Sample

Data from the 1997–2005 Medicare Current Beneficiary Survey (MCBS) was used for this study. The MCBS enrolls approximately 4,000 Medicare beneficiaries every year and follows these individuals for up to three additional years. Data collected from this survey include self-reports of demographic and socioeconomic characteristics, health and functional status, as well as utilization and cost of all health services including prescription drugs32. Information about prescription drug use is provided by the respondent who is asked to keep a log of pharmacy receipts, insurance claims, and used medication containers. Each prescription fill or refill is recorded as a separate event and includes information about the drug name, strength, and quantity of doses. Although the date of the prescription fill and the number of days treatment supplied are not recorded. The MCBS survey is linked to Medicare administrative data including Part A and B claims.

Six pooled panels were constructed of beneficiaries inducted into the MCBS between 1997–2002 who had a diagnosis of diabetes in their baseline year. Identification of diabetes cases was based on claims and self-report. An algorithm developed by CMS for the Chronic Condition Warehouse (CCW) was used to identify diabetes cases from claims33. The CCW criteria for diabetes include beneficiaries with an ICD-9 code for 250.xx (diabetes), 357.2 (polyneuropathy in diabetes), 362.01 (background diabetic retinopathy), 362.02 (proliferative diabetic retinopathy NOS), 366.41 (diabetic cataract) on one or more inpatient hospital, skilled nursing facility (SNF), or home health claims or two outpatient hospital or physician claims in any position33. Patients who reported being given a diagnosis of diabetes by a physician where also included as cases. Self-reported diabetes is considered the gold standard for identifying individuals when clinical indicators are not available34,35.

Individuals excluded from the sample include 1) beneficiaries living in long term care (LTC) facilities during their baseline year, although beneficiaries who were admitted to LTC in subsequent years were retained, 2) beneficiaries enrolled in capitated Medicare health plans at any time during participation in MCBS, as well as dual eligibles who have both Medicaid and Medicare. Individuals in LTC were excluded due to differences in the way the MCBS is administered in these settings. Individuals enrolled in capitated Medicare health plans do not have Part A and B claims, which were necessary in determining Medicare costs. Lastly, dual eligibles were excluded as they were found to be disproportionately represented among the higher-risk group, and they faced minimal OOP prices.

There were 2,949 beneficiaries who had diabetes during their baseline year. Among those, the final sample comprised a cohort of individuals who reported filling prescriptions for statins during their first full year in the MCBS following their fall induction survey (N=899). This drug user cohort was then followed until they completed their MCBS participant or were lost to follow up, admitted to a LTC facility for a long-term stay, or died; the final observation year was 2005.

Measures

We had two dependent variables in the multivariate models – the number of 30-day statin fills and Medicare expenditures for Part A and Part B services. Statin prescription fills were determined by drug names and therapeutic class indicators. Although days’ supply of prescription fills is not available in the MCBS, pills counts are available and were used to measure drug use for each subject. About 16% of medication fills were missing pill counts; these values were imputed based on information in a beneficiaries’ medication regime using a two-stage imputation procedure. We divided the number of pills in each fill by 30 to calculate the number of standardized 30-pill fills. As most statins are dosed once daily, this measure should accurately distinguish the number of monthly fills, with 36 expected to be the maximum over a three-year period. Medicare expenditures for Part A and B services were measured over the same timeframe as the statin fills measure. To account for inflation, all dollar values were converted to constant 2005 dollars using the Consumer Price Index-urban price index36.

The key independent variable of interest in the statin fill demand model was OOP price per fill. To measure this we summed the OOP payments by beneficiaries associated with statin fills in the base year and divided by the number of 30 pill fills in that year. Ideally we would have an exogenous price measure, not based on observed prices. However, since our question relates to VBID among users of statins, the observed OOP price should reflect the price faced at the margin for these users. In the cost-offset model the key independent variable of interest was the number of statin fills. We did not include the OOP price in the models, but included indicators for supplemental medical and drug benefits by source of coverage, as these would be expected to have independent effects on medical spending.

There were a number of covariates used in our regression models to control for confounding. These included demographic characteristics such as age, sex, race, marital status, education, income and census region, which are believed to influence medication use and Medicare spending. A number of health status and disease severity variables were used to avoid indication bias associated with prescribing. Self-reported measures used included overall health, height and weight which were used to compute body-mass index [BMI], number of limitations in activities of daily living [ADLs], and diabetes type. Medicare administrative data provided variables to identify current and former recipient of Social Security Disability Insurance (SSDI). Lastly, Medicare claims were used to create measures of diabetes complications [ICD-9 codes 250.1x to 250.9x] and various common comorbidities, including chronic renal failure, hypertension, ischemic heart disease, cardiac failure, hyperlipidemia, chronic obstructive pulmonary disease, and osteoarthritis, plus a count of hierarchical coexisting conditions or HCCs. Finally, we included the number of days followed in the community and dummy variables to capture reason for any censoring (loss to follow up, LTC admission, death), as well as year of induction into the MCBS (a proxy for temporal trends in diabetes and availability of treatment options), and a dummy variable that indicated if the diabetes case was identified by self-report only. Except for the spending, drug utilization and censoring variables, all covariates were measured during the baseline year.

The analysis was conducted for the full cohort of diabetes patients, and then stratified according to 10-year risk of a cardiovascular event (stroke, acute myocardial infarction). The risk stratification used a variant of the Framingham Risk Calculator [based on the Simple Model with Office-Based non-laboratory Predictors], which assigns points based on the following risk factors: age, body mass index, systolic blood pressure (untreated and treated), smoking status, and diabetes status37. Variables from the MCBS used to create high and lower-risk groups included self-reported age, gender, body mass index (based on height and weight), current smoking status, and diabetes status as well as claims-based measures of diabetes and hypertension. Because the MCBS does not record systolic blood pressure (SBP) measures, it was assumed that individuals with hypertension (70% of the sample) had an untreated SBP of 140. Individuals without hypertension were assumed to have a treated SBP of 120. Points for each item were summed to create a risk score for each beneficiary. According to the Framingham website, all participants in this sample were above a 30% 10-year risk of a cardiovascular event. The distribution of risk factor scores was examined by gender and a cut point for the higher-risk group was determined (risk score>=21 for men and risk score>=24 for women) leaving about 40% of the sample in the higher-risk group (N=350) and 60% in the lower-risk group (N=549). This cut point was used to stratify the analysis to determine if higher-risk individuals were more price responsive to out of pocket prices for statins, and to see if higher-risk individuals had larger cost-offsets associated with statin use than lower risk individuals.

Statistical Analysis and Simulation

We used bivariate analyses and multivariate regressions to examine the relationship between the OOP price and the cumulative number of statin fills in the demand model, and in the cost offset model, we estimated the effect of the cumulative number of statins fills on cumulative Medicare Part A and B spending. The statin demand model was estimated using ordinary least squares regression, and the cost offset model was a generalized linear model with a gamma distribution and log link to approximate the right skewed distribution of Medicare costs. Because virtually all subjects had some Medicare spending over the three-year observation period, we estimated one-part models, replacing the small number of zero observations with a value of $1.

To simulate the effects of a VBID, we first generated a deterministic prediction for our dependent variables, based on observed OOP price and all other variables. Then we capped the OOP price at various levels, and generated a predicted value for the number of statin fills. We computed total and plan prescription drug spending, assuming that the total spending per standard fill did not change, but that for some users, the number of fills would increase and the spending per fill would shift from the user to the drug plan. The predicted number of fills was then substituted into the cost offset model and a new Medicare Part A and B spending value generated. We computed the change in predicted statin fills from baseline using the simulated OOP prices. Similarly we computed the change in plan drug spending and Medicare Part A and B spending associated with the simulated OOP prices. The predictions used the observed values for each individual, with the exception of the OOP price. This approach differs from the more common practice of generating predictions using sample means, and permits us to examine heterogeneity in magnitude of response. Analyses used SAS (9.2) or Stata (version 10). The study protocol was approved by the University of Maryland Baltimore Institutional Review Board.

RESULTS

Descriptive Findings

The sample included 899 Medicare beneficiaries with diabetes who used statins during their baseline year, with 350 identified as higher-risk. Table 1A reports sample means and proportions for the dependent and independent measures, overall and stratified by risk status. Overall, patients filled an average of 24.4 out of a maximum 36 fills, with a slightly higher value among higher-risk patients (25.5 compared to 23.7 fills among the lower-risk). Mean cumulative medical spending over the three-year observation period was $36,690, with higher spending among the higher-risk ($40,041) compared to the lower-risk group ($34,553). Table 2A reports results from the ordinary least squares regressions of OOP price on the number of statin fills, overall and stratified by risk status. For each dollar increase in the OOP price, there is a reduction of 0.086 statin fills (p<.0001); the estimated elasticity is −0.101. The estimated effects are somewhat larger for the higher-risk compared to the lower-risk, with elasticity estimates of −0.111 and −0.089 respectively. The full set of regression coefficients is provided in Appendix Table 1. Although the models overall are highly significant, relatively few individual variables are significant. Increasing education levels are associated with increased fills, while the presence of diabetes complications is associated with lower adherence, suggesting that lower adherence in the past may be associated with the development of diabetes complications.

Table 1.

Table1A. Mean Values for Independent and Dependent Variables for Full Sample and by CVD Risk†
By CVD Risk
Full cohort
(N= 899)		 High Risk
(N= 350)		 Low Risk
(N=549)
Mean s.e. Mean s.e. Mean s.e.
Medicare Part A & B spending, cumulative 3 year (2005 $) 36,690 1653 40,041 2840 34,553 2010
Statin fills, cumulative 3 year, 30 day 24.4 0.4 25.5 0.7 23.7 0.5
1st yr OOP Spending per statin fill ($) 28.8 1.0 26.8 1.6 30.0 1.4
Table1B. Characteristics of Sample Cohort: Medicare Beneficiaries with Diabetes and Baseline Statin Use
By CVD Risk
Full cohort
(N= 899)		 High Risk
(N= 350)		 Low Risk
(N=549)
Mean or % s.e. Mean or % s.e. Mean or % s.e.
Count of other drugs used (Mean) 8.6 0.2 9.1 0.3 8.2 0.2
Supplemental medical insurance only (%) 19.8 1.3 18.0 2.1 20.9 1.7
Supplemental Medical and Rx insurance (%) (not mutually exclusive)
 Employer plan 51.6 1.7 51.4 2.7 51.7 2.1
 Other Private 2.0 0.5 2.3 0.8 1.8 0.6
 Other Public 8.6 0.9 10.0 1.6 7.7 1.1
 Self-purchased 15.7 1.2 15.1 1.9 16.0 1.6
No Supplemental Medical and Rx insurance (%) 7.9 0.9 9.1 1.5 7.1 1.1
Age (%)
 Age < 65 (SSDI) 9.0 1.0 12.9 1.8 6.6 1.1
 65–69 27.8 1.5 31.1 2.5 25.7 1.9
 70–74 22.1 1.4 26.6 2.4 19.3 1.7
 75–79 20.8 1.4 18.0 2.1 22.6 1.8
 80+ 20.2 1.3 11.4 1.7 25.9 1.9
 ≥65 former SSDI 18.8 1.3 25.4 2.3 14.6 1.5
Female (%) 47.8 1.7 46.9 2.7 48.5 2.1
Race
 Non-Hispanic white 84.1 1.2 80.3 2.1 86.5 1.5
 Non-Hispanic black 9.0 1.0 12.9 1.8 6.6 1.1
 Hispanic 4.1 0.7 2.9 0.9 4.9 0.9
 Other 2.8 0.5 4.0 1.0 2.0 0.6
Marital status (%)
 Married 64.3 1.6 63.1 2.6 65.0 2.0
 Widowed 9.6 1.0 12.0 1.7 8.0 1.2
 Single 26.1 1.5 24.9 2.3 27.0 1.9
Education (%)
 No high school 12.9 1.1 13.7 1.8 12.4 1.4
 Some high school 16.5 1.2 15.4 1.9 17.1 1.6
 High school grad 32.9 1.6 34.3 2.5 32.1 2.0
 Some higher education 37.5 1.6 36.3 2.6 38.3 2.1
Annual income (%)
 ≤ 100% FPL 7.5 0.9 8.0 1.5 7.1 1.1
 101% to 199% FPL 32.6 1.6 34.9 2.6 31.1 2.0
 200% to 299% FPL 26.4 1.5 26.0 2.3 26.6 1.9
 ≥300% FPL 33.4 1.6 30.9 2.5 35.0 2.0
Census region (%)
 East 20.9 1.4 20.3 2.2 21.3 1.7
 Midwest 26.7 1.5 29.1 2.4 25.1 1.9
 South 39.5 1.6 40.3 2.6 39.0 2.1
 West 12.9 1.1 10.3 1.6 14.6 1.5
Diabetes Type (%)
 Type 1 10.5 1.0 12.6 1.8 9.1 1.2
 Type 2 83.0 1.4 81.5 2.3 83.9 1.8
Self-report health status (%)
 Excellent/very good 25.8 1.5 18.0 2.1 30.8 2.0
 Good 34.3 1.6 37.1 2.6 32.4 2.0
 Fair 27.8 1.5 31.1 2.5 25.7 1.9
 Poor 12.0 1.1 13.7 1.8 10.9 1.3
Body mass index (%)
 <18.5 (underweight) 0.2 0.2 0.3 0.3 0.2 0.2
 18.5 – 24.9 (normal) 20.1 1.3 5.1 1.2 29.7 2.0
 25.0 – 29.9 (overweight) 40.3 1.6 10.3 1.6 59.4 2.1
 30.0 – 34.9 (obese 1) 24.4 1.4 56.0 2.7 4.2 0.9
 35.0 – 39.9 (obese 2) 9.6 1.0 20.9 2.2 2.4 0.6
 40.0+ (obese 3) 2.8 0.5 6.6 1.3 0.4 0.3
Comorbidities (%)
 Diabetes complications 23.6 1.4 26.0 2.3 22.0 1.8
 Chronic kidney disease 5.2 0.7 6.9 1.4 4.2 0.9
 Hypertension 70.3 1.5 80.0 2.1 64.1 2.0
 Ischemic heart disease 53.2 1.7 50.3 2.7 55.0 2.1
 Congestive heart failure 24.1 1.4 30.0 2.5 20.4 1.7
 Hyperlipidemia 75.1 1.4 71.7 2.4 77.2 1.8
 COPD 10.8 1.0 14.3 1.9 8.6 1.2
 Osteoarthritis 14.0 1.2 16.3 2.0 12.6 1.4
  Cognitive impairment 10.7 1.0 12.6 1.8 9.5 1.3
  Psychiatric conditions 7.8 0.9 10.0 1.6 6.4 1.0
ADL limitations (Mean) 0.8 0.0 1.0 0.1 0.7 0.1
Current Smoker (%) 9.7 1.0 24.9 2.3 0.0 0.0
MCBS induction year (%)
 1997 10.2 1.0 8.0 1.5 11.7 1.4
 1998 12.1 1.1 10.3 1.6 13.3 1.5
 1999 15.8 1.2 14.6 1.9 16.6 1.6
 2000 17.4 1.3 19.1 2.1 16.2 1.6
 2001 19.9 1.3 22.3 2.2 18.4 1.7
 2002 24.6 1.4 25.7 2.3 23.9 1.8
Community months (Mean) 32.3 0.3 32.4 0.5 32.2 0.4
Died during follow-up 10.2 1.0 13.1 1.8 8.4 1.2
Entered LTC facility 3.7 0.6 4.9 1.2 2.9 0.7
Lost to Follow-up 10.8 1.0 6.6 1.3 13.5 1.5
Self reported diabetes only 8.5 0.9 7.4 1.4 9.1 1.2
Open in a new tab

Source: Medicare Current Beneficiary Survey 1997–2005

†

OOP: Out of Pocket; CVD: Cardiovascular Disease; High risk: High risk for cardiovascular event; Low risk: Low risk for cardiovascular event

Table 2.

Table 2A. Estimated effects of Out of pocket price per fill on cumulative 3 year statin fills†
Cohort N Mean OOP price ($) Mean fills (N) Coef Std. Err. P>|z| Price elasticity Model R-squared
Overall 899 28.77 24.41 −0.086 0.015 0.000 −0.101 0.317
High risk 350 26.81 25.50 −0.106 0.023 0.000 −0.111 0.455
Low risk 549 30.02 23.72 −0.070 0.019 0.000 −0.089 0.298
Table 2B. Estimated effects of statin fills on 3 year cumulative Medicare spending
Mean fills (N) Medicare Part A & B Spending ($) Marginal Effect incremental fill Std. Err. P>|z|
Overall 899 24.41 36,690 −159.57 73.24 0.03
High risk 350 25.50 40,041 −279.54 111.66 0.01
Low risk 549 23.72 34,553 −97.47 91.56 0.29
Open in a new tab

Source: Medicare Current Beneficiary Survey 1997–2005

†

OOP: Out of Pocket; High risk: High risk for cardiovascular event; Low risk: Low risk for cardiovascular event

Appendix Table 1.

Estimated Effects of Out of Pocket Price on 3 Year Cumulative Statin Fills†

Overall
High Risk
Low Risk
Variables Coefficient s.e. p value Coefficient s.e. p value Coefficient s.e. p value
OOP spend per fill, year 2 −0.086 0.015 0.000 −0.106 0.023 0.000 −0.070 0.019 0.000
Count other prescription drugs −0.090 0.090 0.315 −0.225 0.175 0.199 −0.087 0.108 0.418
Age (ref = 65–69)
 under65 2.733 2.285 0.232 5.279 3.105 0.090 0.769 3.275 0.815
 age70_74 −1.320 1.175 0.261 −2.071 1.736 0.234 −1.729 1.661 0.298
 age75_79 −1.108 1.164 0.341 −0.117 1.865 0.950 −1.683 1.560 0.281
 age80up −1.899 1.143 0.097 −2.305 2.186 0.293 −1.818 1.424 0.202
Formerly disabled −2.507 1.456 0.085 −4.458 1.909 0.020 −1.451 2.260 0.521
Female −1.339 0.917 0.145 −2.035 1.528 0.184 −0.552 1.279 0.666
Race/ethnicity (ref = White, non-Hispanic)
 Black −0.765 1.621 0.637 0.446 2.008 0.824 −1.998 2.521 0.428
 Hispanic −1.058 1.920 0.582 −5.520 3.354 0.101 0.762 2.398 0.751
 Other −0.056 2.152 0.979 2.518 3.302 0.446 −2.799 2.408 0.246
Marital status (ref = Married)
 Widowed −1.987 1.523 0.192 −3.585 2.243 0.111 −0.528 2.277 0.817
 Single 0.618 1.005 0.539 −0.817 1.693 0.630 1.115 1.283 0.385
Education (ref = No high school)
 Some high school 1.495 1.504 0.321 2.529 2.668 0.344 0.174 1.688 0.918
 High school graduate 3.272 1.273 0.010 2.317 1.981 0.243 3.681 1.650 0.026
 College graduate 3.206 1.289 0.013 2.535 2.051 0.217 3.816 1.649 0.021
Income (ref=poor)
 Near poor (100–200% FPL) 1.240 1.679 0.460 5.195 3.009 0.085 −2.248 1.979 0.257
 Middle income (200–400% FPL) 0.426 1.709 0.803 2.690 3.180 0.398 −2.489 2.093 0.235
 Upper income (300% FPL +) 1.356 1.723 0.432 4.069 3.238 0.210 −1.601 2.091 0.444
Region (ref = Northeast)
 Midwest −1.578 1.057 0.136 −0.659 1.827 0.719 −2.625 1.414 0.064
 South −0.763 1.007 0.449 −0.950 1.762 0.590 −1.244 1.324 0.348
 West −1.060 1.215 0.383 −2.224 2.149 0.301 −1.436 1.553 0.356
Type 1 Diabetes 0.373 1.294 0.773 2.626 2.173 0.228 0.058 1.684 0.973
Self report health status (ref = Excellent,
 Good 0.307 0.982 0.755 0.719 1.666 0.666 0.282 1.241 0.821
 Fair 0.996 1.142 0.384 2.021 2.079 0.332 0.296 1.389 0.831
 Poor 1.790 1.610 0.266 1.934 2.551 0.449 1.983 2.217 0.371
BMI (ref = normal)
 <18.5 (underweight) −1.924 8.865 0.828 −18.630 4.259 0.000 10.190 4.442 0.022
 25.0 – 29.9 (overweight) −0.584 1.036 0.573 −2.232 3.217 0.488 −0.270 1.143 0.813
 30.0 – 34.9 (obese 1) 1.123 1.181 0.342 −4.926 3.148 0.119 −0.753 3.190 0.814
 35.0 – 39.9 (obese 2) −1.225 1.606 0.446 −7.402 3.440 0.032 0.615 3.249 0.850
 40.0+ (obese 3) −2.956 2.167 0.173 −8.478 3.949 0.033 −10.650 5.861 0.070
Comorbidities & complications
 Diabetes complications −0.620 1.021 0.544 0.890 1.774 0.616 −1.280 1.340 0.340
 Diabetic kidney disease −2.920 1.636 0.075 −7.233 2.304 0.002 −0.917 2.317 0.693
 Hypertension 0.546 0.921 0.554 −2.278 1.779 0.201 1.244 1.247 0.319
 Ischemic heart disease 0.738 0.851 0.386 0.362 1.465 0.805 0.743 1.103 0.501
 Congestive heart failure 1.471 0.980 0.134 3.075 1.593 0.055 0.965 1.331 0.469
 Hyperlipidemia −0.738 0.944 0.434 1.905 1.492 0.203 −2.366 1.304 0.070
 COPD −1.177 1.207 0.330 0.049 2.002 0.980 −0.787 1.577 0.618
 Osteoarthritis −2.238 1.049 0.033 −3.140 1.899 0.099 −1.659 1.341 0.216
 Cognitive impairment 1.254 1.499 0.403 0.378 2.257 0.867 1.309 2.038 0.521
 Psychiatric conditions −2.030 1.742 0.244 −2.038 2.744 0.458 −3.422 2.447 0.163
ADL Limitations (Count) 0.479 0.376 0.202 1.148 0.648 0.078 0.222 0.489 0.650
Smoker, current 1.570 1.362 0.249 −1.944 2.226 0.383 0.000 0.000
Cohort year (ref = 1997)
 Panel98 0.963 1.622 0.553 −2.255 2.672 0.399 2.462 2.126 0.248
 Panel99 −0.382 1.587 0.810 0.499 2.646 0.851 −0.550 2.052 0.789
 Panel00 0.981 1.560 0.530 1.361 2.659 0.609 −0.236 2.067 0.909
 Panel01 0.628 1.525 0.681 −2.046 2.539 0.421 2.312 2.069 0.264
 Panel02 2.513 1.502 0.095 2.825 2.564 0.271 3.228 1.980 0.104
Survivorship
 Months in community 0.837 0.080 0.000 0.987 0.113 0.000 0.808 0.148 0.000
 Died 2.637 1.831 0.150 3.994 2.511 0.113 2.838 3.050 0.353
 LTC facility admission −3.022 1.849 0.103 −3.430 2.882 0.235 −0.034 2.878 0.991
 Lost to follow-up 2.333 2.178 0.284 5.143 3.048 0.093 2.416 3.597 0.502
Self reported diagnosis only 0.500 1.596 0.754 0.038 3.168 0.990 −0.777 1.860 0.676
Constant −2.485 4.131 0.548 −1.842 6.187 0.766 1.420 6.436 0.825
Observations 899 350 549
R-squared 0.317 0.455 0.298
Open in a new tab

Source: Medicare Current Beneficiary Survey 1997–2005

†

High risk: High risk for cardiovascular event; Low risk: Low risk for cardiovascular event

Table 2B reports results from the multivariate regressions of statin fills on Medicare spending. Overall, the estimated effect of an additional 30-pill fill is a reduction of $159.57 (p=.03) in Medicare spending. For the higher-risk patients the estimated effect is almost double, at $279.54 (p>.01) while the estimated effect for the lower-risk is smaller ($97.47, p=.29) and not significantly different than zero. The full set of regression coefficients is reported in Appendix Table 2.

Appendix Table 2.

Estimated Effects of Statin Fills on 3 Year Cumulative Medicare Spending†

Overall
High Risk
Low Risk
Coefficient s.e. p value Coefficient s.e. p value Coefficient s.e. p value
Statin fills, 3 year cumulative −0.006 0.003 0.027 −0.010 0.004 0.010 −0.004 0.004 0.284
Count other prescription drugs 0.035 0.007 0.000 0.049 0.014 0.000 0.030 0.009 0.001
Age (ref = 65–69)
 Under65 −0.302 0.157 0.054 −0.115 0.226 0.609 −0.218 0.226 0.335
 Age70_74 0.127 0.101 0.207 0.067 0.145 0.643 0.228 0.148 0.123
 Age75_79 0.123 0.102 0.228 0.076 0.159 0.631 0.215 0.133 0.106
 Age80up 0.037 0.103 0.718 0.067 0.180 0.709 0.098 0.126 0.438
Formerly disabled 0.104 0.124 0.405 0.028 0.176 0.875 0.062 0.177 0.728
Female 0.107 0.076 0.159 0.003 0.117 0.982 −0.007 0.108 0.948
Race/ethnicity (ref = White, non-Hispanic)
 Black −0.102 0.125 0.414 −0.108 0.165 0.512 0.025 0.180 0.890
 Hispanic −0.217 0.174 0.210 −0.359 0.359 0.317 −0.193 0.197 0.327
 Other 0.444 0.234 0.058 0.395 0.313 0.208 0.669 0.368 0.069
Marital status (ref = Married)
 Widowed 0.001 0.119 0.991 −0.334 0.184 0.070 0.209 0.168 0.213
 Single −0.046 0.084 0.586 −0.171 0.141 0.225 −0.019 0.109 0.865
Education (ref = No high school)
 Some high school −0.017 0.121 0.890 −0.421 0.197 0.033 0.280 0.148 0.059
 High school graduate 0.093 0.115 0.418 −0.132 0.169 0.433 0.369 0.147 0.012
 College graduate 0.182 0.115 0.115 −0.007 0.178 0.969 0.452 0.147 0.002
Income (ref= Poor)
 Near poor (100–200% FPL) 0.119 0.136 0.380 0.163 0.205 0.426 0.306 0.191 0.108
 Middle income (200–400% FPL) 0.128 0.150 0.393 −0.002 0.223 0.992 0.324 0.207 0.117
 Upper income (300% FPL +) 0.209 0.154 0.175 0.217 0.241 0.368 0.303 0.211 0.151
Region (ref = Northeast)
 Midwest −0.233 0.105 0.027 −0.351 0.166 0.034 −0.147 0.140 0.291
 South −0.227 0.103 0.028 −0.326 0.161 0.043 −0.169 0.128 0.185
 West −0.152 0.129 0.236 −0.391 0.204 0.056 −0.040 0.160 0.802
Supplemental insurance (ref = None)
 Medical only 0.203 0.129 0.117 0.361 0.215 0.093 0.191 0.161 0.236
 Employment related w/rx 0.109 0.109 0.316 0.217 0.189 0.251 0.170 0.133 0.204
 Private, self pay w/rx 0.113 0.116 0.331 0.119 0.194 0.540 0.101 0.141 0.476
 Public, other 0.324 0.160 0.043 0.424 0.235 0.072 0.243 0.210 0.246
Type 1 Diabetes 0.244 0.122 0.045 0.535 0.188 0.004 −0.055 0.146 0.705
Self report health status (ref = Excellent,
 Good 0.025 0.089 0.783 −0.140 0.141 0.321 0.055 0.106 0.603
 Fair 0.171 0.105 0.102 0.106 0.157 0.500 0.182 0.131 0.164
 Poor 0.155 0.135 0.250 −0.061 0.222 0.782 0.386 0.179 0.031
BMI (ref = Normal)
 <18.5 (underweight) −0.625 0.225 0.006 −0.009 0.431 0.984 −0.921 0.358 0.010
 25.0 – 29.9 (overweight) −0.153 0.098 0.118 0.205 0.262 0.434 −0.184 0.102 0.072
 30.0 – 34.9 (obese 1) −0.246 0.115 0.032 −0.268 0.292 0.360 0.281 0.253 0.267
 35.0 – 39.9 (obese 2) −0.192 0.141 0.173 −0.158 0.316 0.617 0.435 0.281 0.122
 40.0+ (obese 3) −0.540 0.181 0.003 −0.427 0.336 0.203 −1.203 0.553 0.029
Comorbidities & complications
 Diabetes complications 0.225 0.079 0.004 0.317 0.133 0.017 0.119 0.106 0.260
 Diabetic kidney disease 0.526 0.131 0.000 0.436 0.207 0.035 0.540 0.199 0.007
 Hypertension 0.176 0.083 0.035 0.313 0.163 0.056 0.276 0.107 0.010
 Ischemic heart disease 0.428 0.073 0.000 0.402 0.121 0.001 0.479 0.090 0.000
 Congestive heart failure 0.426 0.083 0.000 0.365 0.133 0.006 0.446 0.117 0.000
 Hyperlipidemia 0.006 0.085 0.946 0.089 0.131 0.496 0.049 0.107 0.650
 COPD 0.118 0.092 0.202 0.035 0.158 0.823 0.216 0.144 0.133
 Osteoarthritis 0.070 0.093 0.448 0.195 0.157 0.214 0.066 0.129 0.609
 Cognitive impairment −0.080 0.119 0.500 −0.214 0.192 0.266 −0.035 0.152 0.820
 Psychiatric conditions 0.069 0.133 0.602 −0.190 0.195 0.329 0.221 0.175 0.206
ADL Limitations (Count) 0.092 0.031 0.003 0.140 0.045 0.002 0.096 0.041 0.020
Smoker, current 0.074 0.115 0.524 −0.014 0.195 0.945 0.000 0.000
Cohort year (ref = 1997)
 Panel98 −0.051 0.135 0.708 −0.046 0.248 0.853 −0.022 0.166 0.894
 Panel99 −0.055 0.128 0.669 0.178 0.241 0.459 −0.132 0.154 0.392
 Panel00 −0.052 0.127 0.682 −0.093 0.236 0.694 −0.049 0.148 0.742
 Panel01 0.108 0.120 0.366 −0.019 0.232 0.936 0.127 0.141 0.367
 Panel02 0.069 0.118 0.555 −0.169 0.231 0.464 0.171 0.140 0.223
Survivorship
 Months in community 0.011 0.009 0.227 0.011 0.012 0.356 0.012 0.013 0.343
 Died 0.612 0.179 0.001 0.500 0.226 0.027 0.609 0.234 0.009
 LTC facility admission 0.720 0.195 0.000 0.815 0.202 0.000 0.686 0.366 0.061
 Lost to follow-up −0.440 0.193 0.023 −0.656 0.271 0.016 −0.330 0.256 0.197
Self reported diagnosis only −0.443 0.156 0.005 −0.584 0.257 0.023 −0.355 0.193 0.065
Constant 8.813 0.455 0.000 9.060 0.700 0.000 8.070 0.582 0.000
Observations 899 350 549
Open in a new tab

Source: Medicare Current Beneficiary Survey 1997–2005

†

OOP: Out of Pocket; High risk: High risk for cardiovascular event; Low risk: Low risk for cardiovascular event

The simulated effects of VBID are illustrated in Figure 2. Figure 2A shows the simulated effects of capping the OOP price at various levels on changes in 3rd party payer spending on statins. For example, relative to the baseline prediction of $1280, third party spending on statins increased by $340 when the OOP price was capped at $25, and by $794 when OOP price was capped at $1. Estimates vary slightly by risk status. Figure 2B illustrates the effects of VBID on changes in Medicare A & B spending. For the overall sample, implementation of VBID pricing reduced Medicare spending by $262 (price cap at $25) to $531 (price cap at $1). The effect for the higher-risk subgroup was substantially higher, ranging from a reduction of $558 to $1193. The net effect of changes in 3rd party spending in these two categories is illustrated in Figure 2C. Overall, the reduction in OOP price associated with VBID generated a net increase in 3rd party spending, beginning with the $25 cap and increasing as the OOP price cap diminished. The net effect was a savings to 3rd party payers for the higher-risk group, ranging from $249 to $415 across the various price caps. For the lower-risk groups the results were inverted, with net increases in 3rd party spending as price caps were lowered.

Figure 2.

Figure 2

Open in a new tab

A–C: Simulated Effects of VBID OOP Price of Standardized Statins Fill Caps on Medicare Beneficiaries†

†VBID: Value Based Benefit Design; OOP: Out of Pocket; High risk: High risk for cardiovascular event; Low risk: Low risk for cardiovascular event

DISCUSSION

The results of this study provide the first empirical evidence that a VBID strategy may be effective within the Medicare population, and in particular, demonstrate that its selective application could deliver net savings to the Medicare program38. Previous studies have reported on changes in drug use associated with cost sharing, while others have reported on reduced medical costs associated with medication adherence. In this study we integrate the two components, and simulate the net effects from the perspective of Medicare as a third party payer. Our simulated cost offsets were substantial for the higher-risk subgroup, with a net savings of $415 per patient associated with a $1 copayment, but did not yield net offsets for the group of diabetes patients overall. These findings suggest that a VBID strategy may be very attractive within the Medicare program if it is feasible to target the benefit. However, further research, including intervention studies, to confirm the response and net costs of a VBID strategy of this type, is warranted.

Within the current structure of the Medicare Part D system, managed care prescription drug plans (MA-PDs) would have the capacity to benefit from any net savings in medical expenditures, as the medical and prescription drug benefits are integrated. The introduction of VBID for MA-PDs would clearly present a competitive advantage for these types of plans over their standalone Part D plan counterparts. That said, reducing cost sharing for the sickest enrollees may also lead to adverse risk selection, with the sickest beneficiaries favoring MA-PDs offering it. In addition, some MA-PDs may balk at introducing such mechanisms where it is not possible to ensure that targeted beneficiaries will remain in the MA-PD for sufficient time for the plan to benefit from the downstream offsets.

While the logistics of implementing VBID for MA-PD enrollees may be relatively straightforward, applying VBID mechanisms to other types of drug coverage may be more complex. Downstream savings in medical care costs reaped by the Medicare FFS program clearly present no incentive for ex-ante increases in coverage by standalone PDPs under Part D. Other mechanisms would need to be devised to provide the necessary incentives to reduce copayments, for example, increased levels of federal subsidy to Part D plans to cover the additional plan costs for targeted beneficiaries. Employer sponsored plans that receive the retiree drug subsidy could be further subsidized for implementing a VBID cost-sharing structure, and the Federal Employee Health Benefits Program could easily adopt these changes as well.

Targeting VBID incentives to higher-risk beneficiaries presents a number of important logistical issues that would need to be addressed, such as how patients are designated as eligible, and how the alternative copayment structure is implemented at the time prescriptions are filled39. In addition, targeting VBID copayment reductions may generate perceptions of inequity as individual patients are charged different copayments for the same medications. Concerns about inequity may be dispelled if insurers emphasize the role of clinical need in targeting the benefit. Of potentially greater concern is that targeting by clinical status could potentially create perverse incentives. It could, for example, discourage some individuals from attempting to lose weight or quit smoking if they thought it could lead to the loss of their preferred VBID status.

While our example demonstrates the feasibility of a VBID strategy, the study is subject to limitations. The study period for our data ended in 2005. While behavioral responses to price changes are unlikely to vary, and the clinical impacts and medical cost savings may not have changed dramatically, the costs of statin therapies for Medicare beneficiaries have changed as a result of patent expirations, and perhaps more importantly, following the introduction of Medicare Part D in 2006. Hence it is important to consider how changes in market price – both total and OOP – affect the VBID calculus. In our analysis we used observed OOP payments, and applied successively generous caps. The observed payments reflected the range of prices over the study period, and should have captured reductions in the price of statins associated with generic competition starting in 2002. As generic drugs tend to be subject to even lower copayments in more recent years – either through tiered formularies or through the pricing policies of big-box stores such as Walmart – a VBID strategy would have to reduce OOP prices even further to elicit increased use. On the other hand, the third party payer would then only be at risk for a smaller increase in its share of the total drug costs, while the downstream savings in medical spending should remain unaffected. Hence, our results may overstate the increased cost to insurers to participate in a VBID strategy, and understate the net savings in medical care costs.

The study is observational, and the estimates of price response and medical care offsets are based on cross sectional and temporal variation. Despite extensive controls for confounding in the multivariate models, there is the potential for bias associated with unobserved health or behavioral factors. In addition, we note that our measures of drug use and spending are not based on insurance claims, and hence may be subject to reporting error. The estimated elasticity is a key element of the simulation. Our elasticity estimates, in the range of −0.1, are even lower than those reported in the literature. To the extent that they are biased downward, then our simulation provides a conservative measure of the cost net savings associated with a VBID strategy. Finally, we did not attempt to model any costs associated with the implementation, monitoring, and evaluation of the VBID strategy in this simulation; identifying low-cost methods to meet these program requirements is important to the feasibility of a VBID program.

The prevalence of diabetes in the Medicare population was approximately 8.2 million in 2009, while forecasts suggest an increase to almost 15 million by 2034, and an almost four-fold increase in annual diabetes-related expenditure40. Nevertheless, while statin therapy is recommended for most diabetics, additional examples likely would be needed to justify the expense of implementing a VBID approach. In a related study16 examining medical care spending offsets associated with the use of oral anti-diabetics and angiotensin converting enzyme inhibitor (ACE-I) or angiotensin receptor blocker (ARB) in the same sample of diabetic patients, we failed to find parallel savings41–43.

Of course, VBID only focuses on the net costs of increased medication use – whether it is from the third party or societal perspective. There is no attempt to value and incorporate any survival or quality of life benefit from the patient perspective. In addition, there may be benefits to employers related to reduced illness and absenteeism which our estimates cannot capture because they are focused on the insurer44. If these were to be included in the calculus, many additional treatments might be identified as appropriate for VBID.

Finally, VBID can be used to increase use of selected therapies that are demonstrated to be both clinically effective and likely to lead to reduced medical care spending. However, it is important to recognize that VBID applies a higher standard for the designated services than is currently applied in the Medicare program. Medicare currently covers some therapies for which there is only limited empirical evidence, or evidence of relatively modest clinical benefit, and Congress has been reluctant to mandate that new evidence generated through comparative effectiveness research be used to make coverage decisions45. Instead of setting a uniquely high bar for VBID, a better strategy might be to reduce cost sharing for effective services, even without evidence of a medical cost offset. The increased cost associated with higher use of those services could be offset by limiting coverage or introducing a higher level of cost-sharing for services that have evidence of limited benefit or value.46,47 Services with a limited evidence base could be covered only through an evidence development strategy, and ultimately this would help to identify additional candidates for VBID. Adoption of VBID may be the needed catalyst to rationalize use of evidence in coverage decisions.

Acknowledgments

We thank Newell McElwee and Linda Simoni-Wastila for thoughtful suggestions on an earlier version of this paper.

Sources of financial support for manuscript: The research for this study was supported by a grant from the Robert W. Johnson Foundation Health Care Financing and Organization. Additional support was provided by Pfizer Inc. Ms. Lloyd’s effort was supported by a grant provided by the National Institute on Aging grant no. T32 4G000262.

References

  • 1.Fendrick AM, Smith DG, Chernew ME, Shah SN. A benefit-based copay for prescription drugs: Patient contribution based on total benefits, not drug acquisition cost. Am J Manag Care. 2001;7:861–867. [PubMed] [Google Scholar]
  • 2.Chernew ME, Rosen AB, Fendrick AM. Value-based insurance design. Health Aff (Millwood) 2007;26:w195–203. doi: 10.1377/hlthaff.26.2.w195. [DOI] [PubMed] [Google Scholar]
  • 3.Giuffrida A, Gravelle H. Paying patients to comply: An economic analysis. Health Econ. 1998;7:569–579. doi: 10.1002/(sici)1099-1050(1998110)7:7<569::aid-hec376>3.0.co;2-s. [DOI] [PubMed] [Google Scholar]
  • 4.Brown AF, Mangione CM, Saliba D, Sarkisian CA, California Healthcare Foundation/American Geriatrics Society Panel on Improving Care for Elders with Diabetes Guidelines for improving the care of the older person with diabetes mellitus. J Am Geriatr Soc. 2003;51:S265–80. doi: 10.1046/j.1532-5415.51.5s.1.x. [DOI] [PubMed] [Google Scholar]
  • 5.Odegard PS, Capoccia K. Medication taking and diabetes: A systematic review of the literature. Diabetes Educ. 2007;33:1014–29. doi: 10.1177/0145721707308407. discussion 1030–1. [DOI] [PubMed] [Google Scholar]
  • 6.American Diabetes Association. Standards of medical care in diabetes--2010. Diabetes Care. 2010;33(Suppl 1):S11–61. doi: 10.2337/dc10-S011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Kamari Y, Bitzur R, Cohen H, Shaish A, Harats D. Should all diabetic patients be treated with a statin? Diabetes Care. 2009;32(Suppl 2):S378–83. doi: 10.2337/dc09-S344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Golan L, Birkmeyer JD, Welch HG. The cost-effectiveness of treating all patients with type 2 diabetes with angiotensin-converting enzyme inhibitors. Ann Intern Med. 1999;131:660–667. doi: 10.7326/0003-4819-131-9-199911020-00005. [DOI] [PubMed] [Google Scholar]
  • 9.Balkrishnan R, Rajagopalan R, Camacho FT, Huston SA, Murray FT, Anderson RT. Predictors of medication adherence and associated health care costs in an older population with type 2 diabetes mellitus: A longitudinal cohort study. Clin Ther. 2003;25:2958–2971. doi: 10.1016/s0149-2918(03)80347-8. [DOI] [PubMed] [Google Scholar]
  • 10.Brandle M, Davidson MB, Schriger DL, Lorber B, Herman WH. Cost effectiveness of statin therapy for the primary prevention of major coronary events in individuals with type 2 diabetes. Diabetes Care. 2003;26:1796–1801. doi: 10.2337/diacare.26.6.1796. [DOI] [PubMed] [Google Scholar]
  • 11.Hepke KL, Martus MT, Share DA. Costs and utilization associated with pharmaceutical adherence in a diabetic population. Am J Manag Care. 2004;10:144–151. [PubMed] [Google Scholar]
  • 12.Gilmer TP, O’Connor PJ, Rush WA, et al. Predictors of health care costs in adults with diabetes. Diabetes Care. 2005;28:59–64. doi: 10.2337/diacare.28.1.59. [DOI] [PubMed] [Google Scholar]
  • 13.Rosen AB, Hamel MB, Weinstein MC, Cutler DM, Fendrick AM, Vijan S. Cost-effectiveness of full medicare coverage of angiotensin-converting enzyme inhibitors for beneficiaries with diabetes. Ann Intern Med. 2005;143:89–99. doi: 10.7326/0003-4819-143-2-200507190-00007. [DOI] [PubMed] [Google Scholar]
  • 14.Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43:521–530. doi: 10.1097/01.mlr.0000163641.86870.af. [DOI] [PubMed] [Google Scholar]
  • 15.Stuart BC, Simoni-Wastila L, Zhao L, Lloyd JT, Doshi JA. Increased persistency in medication use by U.S. medicare beneficiaries with diabetes is associated with lower hospitalization rates and cost savings. Diabetes Care. 2009;32:647–649. doi: 10.2337/dc08-1311. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Stuart B, Davidoff A, Lopert R, Shaffer T, Samantha Shoemaker J, Lloyd J. Does medication adherence lower medicare spending among beneficiaries with diabetes? Health Serv Res. 2011 doi: 10.1111/j.1475-6773.2011.01250.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Goldman DP, Joyce GF, Zheng Y. Prescription drug cost sharing: Associations with medication and medical utilization and spending and health. JAMA. 2007;298:61–69. doi: 10.1001/jama.298.1.61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Gilman BH, Kautter J. Impact of multitiered copayments on the use and cost of prescription drugs among medicare beneficiaries. Health Serv Res. 2008;43:478–495. doi: 10.1111/j.1475-6773.2007.00774.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Huskamp HA, Deverka PA, Epstein AM, Epstein RS, McGuigan KA, Frank RG. The effect of incentive-based formularies on prescription-drug utilization and spending. N Engl J Med. 2003;349:2224–2232. doi: 10.1056/NEJMsa030954. [DOI] [PubMed] [Google Scholar]
  • 20.Goldman DP, Joyce GF, Escarce JJ, et al. Pharmacy benefits and the use of drugs by the chronically ill. JAMA. 2004;291:2344–2350. doi: 10.1001/jama.291.19.2344. [DOI] [PubMed] [Google Scholar]
  • 21.Gibson TB, Mark TL, McGuigan KA, Axelsen K, Wang S. The effects of prescription drug copayments on statin adherence. Am J Manag Care. 2006;12:509–517. [PubMed] [Google Scholar]
  • 22.Goldman DP, Joyce GF, Escarce JJ, et al. Pharmacy benefits and the use of drugs by the chronically ill. JAMA. 2004;291:2344–2350. doi: 10.1001/jama.291.19.2344. [DOI] [PubMed] [Google Scholar]
  • 23.Gibson TB, Mark TL, McGuigan KA, Axelsen K, Wang S. The effects of prescription drug copayments on statin adherence. Am J Manag Care. 2006;12:509–517. [PubMed] [Google Scholar]
  • 24.Chernew M, Gibson TB, Yu-Isenberg K, Sokol MC, Rosen AB, Fendrick AM. Effects of increased patient cost sharing on socioeconomic disparities in health care. J Gen Intern Med. 2008;23:1131–1136. doi: 10.1007/s11606-008-0614-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Rosen AB, Hamel MB, Weinstein MC, Cutler DM, Fendrick AM, Vijan S. Cost-effectiveness of full medicare coverage of angiotensin-converting enzyme inhibitors for beneficiaries with diabetes. Ann Intern Med. 2005;143:89–99. doi: 10.7326/0003-4819-143-2-200507190-00007. [DOI] [PubMed] [Google Scholar]
  • 26.Mahoney JJ. Lowering patient drug acquisition costs to improve health outcomes and costs. Manag Care Interface. 2006;19:13. [PubMed] [Google Scholar]
  • 27.Zeng F, An JJ, Scully R, Barrington C, Patel BV, Nichol MB. The impact of value-based benefit design on adherence to diabetes medications: A propensity score-weighted difference in difference evaluation. Value Health. 2010;13:846–852. doi: 10.1111/j.1524-4733.2010.00730.x. [DOI] [PubMed] [Google Scholar]
  • 28.Choudhry NK, Fischer MA, Avorn J, et al. At pitney bowes, value-based insurance design cut copayments and increased drug adherence. Health Aff (Millwood) 2010;29:1995–2001. doi: 10.1377/hlthaff.2010.0336. [DOI] [PubMed] [Google Scholar]
  • 29.Maciejewski ML, Farley JF, Parker J, Wansink D. Copayment reductions generate greater medication adherence in targeted patients. Health Aff (Millwood) 2010;29:2002–2008. doi: 10.1377/hlthaff.2010.0571. [DOI] [PubMed] [Google Scholar]
  • 30.Spaulding A, Fendrick AM, Herman WH, et al. A controlled trial of value-based insurance design – the MHealthy: Focus on diabetes (FOD) trial. Implement Sci. 2009;4:19. doi: 10.1186/1748-5908-4-19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Choudhry NK, Brennan T, Toscano M, et al. Rationale and design of the post-MI FREEE trial: A randomized evaluation of first-dollar drug coverage for post-myocardial infarction secondary preventive therapies. Am Heart J. 2008;156:31–36. doi: 10.1016/j.ahj.2008.03.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Centers for Medicare and Medicaid Services. Medicare Current Beneficiary Survey. Available at: http://www.cms.gov/mcbs/. Accessed December 14, 2009.
  • 33.Buccaneer Computer Systems and Services, Inc. Chronic Condition Warehouse User Manual. Available at: http://ccwdata.org/downloads/CCW%20User%20Manual.pdf. Accessed June 15, 2009.
  • 34.Hebert PL, Geiss LS, Tierney EF, Engelgau MM, Yawn BP, McBean AM. Identifying persons with diabetes using medicare claims data. Am J Med Qual. 1999;14:270–277. doi: 10.1177/106286069901400607. [DOI] [PubMed] [Google Scholar]
  • 35.Saydah SH, Geiss LS, Tierney E, Benjamin SM, Engelgau M, Brancati F. Review of the performance of methods to identify diabetes cases among vital statistics, administrative, and survey data. Ann Epidemiol. 2004;14:507–516. doi: 10.1016/j.annepidem.2003.09.016. [DOI] [PubMed] [Google Scholar]
  • 36.U.S. Department of Labor, Bureau of Labor Statistics. Consumer Price Index: All Urban Consumers. Available at: www.bls.gov/cpi. Accessed June 15, 2009.
  • 37.D’Agostino RBS, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: The framingham heart study. Circulation. 2008;117:743–753. doi: 10.1161/CIRCULATIONAHA.107.699579. [DOI] [PubMed] [Google Scholar]
  • 38.Encinosa WE. Value-based insurance design in medicare. Appl Health Econ Health Policy. 2009;7:149–154. doi: 10.1007/BF03256148. [DOI] [PubMed] [Google Scholar]
  • 39.Murphy L, Carloss J, Brown RE, Heaton E, Carino T, Fendrick AM, Chernew M, Rosen AB. Value-Based Insurance Design in the Medicare Prescription Drug Benefit/An Analysis of Policy Options. Available at: http://www.avalerehealth.net/research/docs/Value-Based_Insurance_Design_in_the_Medicare_Prescription_Drug_Benefit.pdf. Accessed March 8, 2011.
  • 40.Huang ES, Basu A, O’Grady M, Capretta JC. Projecting the future diabetes population size and related costs for the US. Diabetes Care. 2009;32:2225–2229. doi: 10.2337/dc09-0459. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Aubert RE, Yao J, Xia F, Garavaglia SB. Is there a relationship between early statin compliance and a reduction in healthcare utilization? Am J Manag Care. 2010;16:459–466. [PubMed] [Google Scholar]
  • 42.McFarlane SI, Muniyappa R, Francisco R, Sowers JR. Clinical review 145: Pleiotropic effects of statins: Lipid reduction and beyond. J Clin Endocrinol Metab. 2002;87:1451–1458. doi: 10.1210/jcem.87.4.8412. [DOI] [PubMed] [Google Scholar]
  • 43.Davignon J. Beneficial cardiovascular pleiotropic effects of statins. Circulation. 2004;109:III39–43. doi: 10.1161/01.CIR.0000131517.20177.5a. [DOI] [PubMed] [Google Scholar]
  • 44.Nicholson S. The effect of cost sharing on employees with diabetes. Am J Manag Care. 2006;12(Spec no):SP20–6. [PubMed] [Google Scholar]
  • 45.Pearson SD, Bach PB. How medicare could use comparative effectiveness research in deciding on new coverage and reimbursement. Health Aff (Millwood) 2010;29:1796–1804. doi: 10.1377/hlthaff.2010.0623. [DOI] [PubMed] [Google Scholar]
  • 46.Fendrick AM, Smith DG, Chernew ME. Applying value-based insurance design to low-value health services. Health Aff (Millwood) 2010;29:2017–2021. doi: 10.1377/hlthaff.2010.0878. [DOI] [PubMed] [Google Scholar]
  • 47.Neumann PJ, Auerbach HR, Cohen JT, Greenberg D. Low-value services in value-based insurance design. Am J Manag Care. 2010;16:280–286. [PubMed] [Google Scholar]

RESOURCES