Sean D. Sullivan, PhD

 Value-based insurance design has been suggested as an effective approach to ensure access to highvalue medications in health insurance markets. Premera Blue Cross, a large regional health plan, implemented a value-based formulary (VBF) for pharmaceuticals in 2010 that explicitly used cost-effectiveness analysis to inform medication co-payments. This study assesses the impact of a VBF on adherence and patient and health plan expenditures on 3 chronic disease states: diabetes, hypertension, and hyperlipidemia.

 Interrupted time series design of employer-sponsored plans from 2006 to 2013. Beneficiaries exposed to the VBF formed the intervention group, and beneficiaries in similar plans without any changes in pharmacy benefits formed the control group.

We measured medication expenditures from member, health plan, and member-plus-health plan (overall) perspectives and medication adherence as proportion of days covered. We conducted an exploratory analysis of medication utilization classifying medications according to whether co-payments moved up or down in the year following VBF implementation.

For the diabetes cohort, there was a statistically significant reduction in member and overall expenditures of $5 per member per month (PMPM) and $9 PMPM, respectively. For the hypertension cohort, there was a statistically significant reduction in member expenditures of $4 PMPM and an increase in health plan expenditures of $3 PMPM. There were no statistically significant effects on hyperlipidemia cohort expenditures or on medication adherence in any of the 3 disease cohorts. Exploratory analyses suggest that patients in the diabetes and hyperlipidemia cohorts were switching to higher-value medications.

 A VBF can ensure access to high-value medications while maintaining affordability.

Am J Manag Care. 2017;23(3 Suppl):S46-S53

A value-based formulary (VBF) estimates the value of individual medications and aligns medication co-payments with value. We evaluated the impact of a formulary on adherence and patient and health plan expenditures in 3 chronic disease states: diabetes, hypertension, and hyperlipidemia. 

For the diabetes cohort, member and overall expenditures decreased.

For the hypertension cohort, there was a cost shift from member to plan. 

There were no statistically significant effects on hyperlipidemia cohort expenditures or on medication adherence in any of the 3 disease cohorts. 

A VBF can ensure access to high-value medications while maintaining affordability.

Rising prescription drug expenditures in recent years have prompted healthcare payers to look for ways to achieve greater value in pharmaceutical care. Some proposed solutions have included the adoption of frameworks for assessing value,

 varying payments to manufacturers based on the effectiveness of a drug according to each clinical indication (indication-specific pricing),

 and aligning medication utilization with value through cost-sharing redesign.

 This last policy, known as value-based insurance design (VBID), advocates for the redesign of cost sharing such that patient out-of-pocket expense aligns with the value—both health benefit and cost&mdash;of an intervention rather than solely on the cost of the intervention.

Recently, there has been federal interest in the exploration of the effects of VBID for patients with particular chronic conditions.

 Beginning in January 2017, the Center for Medicare & Medicaid Innovation (CMMI) began granting Medicare Advantage plans in 7 states the ability to pilot VBID for enrollees with at least 1 of 7 chronic conditions, including diabetes, hypertension, and certain cardiovascular diseases. One of the approaches permitted by CMMI for these pilot programs is the reduction of cost sharing for high-value services, including pharmaceuticals. In 2018, this pilot program will expand to 3 additional states and 2 additional chronic conditions

; there also is current Congressional discussion to expand the pilot to all states.

According to CMMI, the goals of these pilot programs are to improve health outcomes and attain healthcare cost savings or cost neutrality. Specifically, applicants are required to actuarially demonstrate “net savings to Medicare and no net cost increases to enrollees over the course of the model test.”

 However, findings from previous trials of VBID implemented among patients with chronic diseases suggest that co-payment reductions alone may not necessarily reduce plan costs. The Post-Myocardial Infarction Free Rx Event and Economic Evaluation trial was a randomized trial to solely test the effect of targeted medication co-payment reduction on expenditures.

This study randomized health plans to offer co-payment waivers for certain cardiovascular medications for patients who had been discharged from a hospital with a diagnosis of a myocardial infarction. The study found that although adherence improved moderately (4%-6%) and there were improvements in certain secondary cardiovascular health outcomes, plan medication expenditure increased significantly while plan nonmedication health expenditures were not significantly reduced. The results of other observational studies also show that co-payment reduction alone improves adherence (by 1.5%-9.4%) but increases health plan medication expenditures and has no significant effect on health plan nonmedication expenditures.

A value-based formulary (VBF) is a fuller realization of VBID that incorporates both increases in co-payments for some drugs and decreases in co-payments for other drugs according to value estimates generated by cost-effectiveness analyses. Drugs that have higher estimated value are placed in lower co-payment tiers to incentivize use, and drugs that have lower estimated value are placed in higher co-payment tiers to disincentivize use. In 2010, Premera Blue Cross, the largest private health plan in Washington state, implemented a VBF for their employees and dependents. Details regarding the design and implementation of the VBF have been reported previously.

 A 3-year analysis of the impact of the Premera VBF has shown that the VBF is able to shift plan members toward higher-value drugs and decrease health plan and total (health plan plus member) medication expenditures while not having any measurable adverse effects on hospitalizations, emergency department visits, or nonmedication expenditures.

 Given the interest in value-based approaches among patients with chronic diseases, we evaluated the 3-year effects of the VBF on medication adherence and expenditures among individuals with diabetes, hypertension, or hyperlipidemia.

The study population comprised employees and dependents aged 0 to 64 years who were covered under employer-sponsored preferred provider organization plans administrated by Premera Blue Cross. The intervention group consisted of employees and dependents of Premera in a plan that implemented the VBF on July 1, 2010. The control group included employees and dependents of 5 employersponsored plans administered by Premera that did not have any changes in pharmacy benefits over the entire study period. These plans were chosen based on similarity to the intervention group prior to VBF implementation in industry classification, member geography of residence, and medication co-payment tiers.

The analysis was performed at the individual-member level. For each member in our sample, we obtained monthly measures on demographics (age, sex, zip code of residence, relationship to employee), prescriptions fills (National Drug Code, generic code number, number of days’ supply, date dispensed, place of purchase [retail or mail order pharmacy]), expenditures (amount paid by member, amount paid by health plan), and plan characteristics (benefit renewal month and medical benefit relativity value). The medical benefit relativity value is an index of medical benefit generosity commonly used in health insurance actuarial analyses that takes into account a large number of plan cost-sharing and utilization characteristics (eg, deductibles, co-payments, co-insurance, out-of-pocket maximums, prior authorization, quantity limits).

 The values range between 0 and 1; a value of 0.75 indicates that a health plan pays 75% of medical expenses and the member pays the remaining 25% for a typical market basket of healthcare interventions.

Because the VBF benefit change was implemented across the entire population of members, we aimed to assess the effect of the VBF on this population by constructing prevalent disease cohorts. For these cohorts, individuals were not required to have been given a new diagnosis or to be newly initiating a drug of interest to be included. We identified disease cohorts using 

International Classification of Diseases, Ninth Revision, Clinical Modification

 diagnosis codes (diabetes mellitus: 250.xx; hypertensive disease: 401. xx-404.xx; disorders of lipoid metabolism: 272.xx).

Patients were included in the analysis if they had a diagnosis of interest in the 4-year period prior to VBF implementation and if they had filled a retail or mail order drug of interest based on the American Hospital Formulary Service system of therapeutic classification—the most widely used drug classification system in the United States and Canada.

 The drugs of interest for the diabetes cohort were oral antidiabetic agents: alpha-glucosidase inhibitors (68:20.02), biguanides (68:20.04), dipeptidyl peptidase IV inhibitors (68:20.05), meglitinides (68:20.16), sodium-glucose cotransporter 2 inhibitors (68:20.18), sulfonylureas (68:20.20),thiazolidinediones (68:20.28). Nonoral antidiabetic agents, such as insulins, were not included in the analysis because there is no valid approach to measuring adherence of the nonoral agents using claims data. The drugs of interest for the hypertension cohort were oral antihypertensive agents: central alpha-agonists (24:08.16), beta-adrenergic blocking agents (24:24), calcium-channel blocking agents (24:28), renin-angiotensin-aldosterone system inhibitors (24:32), loop diuretics (40:28.08), potassium-sparing diuretics (40:28.16), thiazide diuretics (40:28.20), and thiazide-like diuretics (40:28.24). The drugs of interest for the hyperlipidemia cohort were antilipemic agents (24:06:xx). The first claim for a drug of interest was considered a patient’s index date, and both adherence and expenditure were measured beginning from this date. Patients were followed until they lost insurance eligibility or the study period ended.

We measured medication adherence using the proportion of days covered (PDC) metric for each of the 3 disease cohorts.

 The PDC is defined as the proportion of days that a given cohort member in a given month had at least 1 disease cohort—specific drug of interest based on the initial prescription fill date(s) and the days’ supply.

 Specifically, the denominator was calculated as the sum of days from the index date or first day of the month (whichever occurs last) to the last day of the month or date of disenrollment from the plan (whichever occurs first). The PDC numerator was calculated as the sum of the days that a member had at least 1 disease cohort—specific drug of interest during the period defined in the denominator. This results in a continuous PDC measure between 0 and 1.

Consistent with standard PDC calculations, in the case of early refills, we assumed that the prior supply was taken fully before the new supply was initiated.

 Hence, the new supply is added to the prior supply in the calculation of the numerator, allowing for a “carryover” in the calculation. Drug supply extending beyond the end of 1 month was added to the drug supply for the following month. Drug supply extending beyond the end of the study period was truncated at the end of the study period such that the maximum PDC is 1. For expenditures, we measured member, health plan, and overall medication expenditures for the disease cohort—specific drugs of interest per member per month (PMPM).

Because we aimed to estimate the average effect of the VBF among those subject to the VBF policy, we sought to compare the observed outcomes among VBF members with the expected outcomes for the same group of VBF members had the VBF not been implemented. We obtained the expected estimate by using the contemporaneous observed outcomes in the 5 control plans that were not exposed to the VBF, after adjusting for the covariate distributions in both the groups. We confirmed the similarity of the control group to the intervention group in pre-VBF outcomes trends by examining both the statistical significance and magnitude of the regression coefficients that represented the differential trends in the groups prior to VBF implementation.

We generated expected outcomes by using generalized linear models. To model PDC, we used a binomial distribution with logit link with iterated, reweighted least-squares optimization of the deviance. To model medication expenditures, we used 2-part models. We first used binomial distribution and logit link to model the probability of incurring expenditures and gamma distribution with log link to model expenditures among those who have incurred expenditures. We assessed overall model fit by using the following goodness-of-fit tests: Pearson’s correlation test, Pregibon link test, and a modified Hosmer-Lemeshow test.

 We modeled correlations between monthly observations within members using robust variance estimators.

 We generated standard errors and confidence intervals for all estimates using 1000 bootstrap replications.

We adjusted for individual-level characteristics (sex, age), zip code—level characteristics (bachelor’s degree, household income, urban residence, African American race, Washington state residence), plan-level characteristics (medical benefits relativity value, benefit renewal month), fixed effects for calendar months (January to December), and study period (early pre-VBF period, late pre-VBF period, and post-VBF period).

 Because observation begins from the index date and adherence likely decreases from the time since that date, we also included an indicator of time since index date in our PDC models to control for potential differences in the start of therapy between the intervention and control groups.

All analyses were done with Stata version 13.1 (StataCorp, College Station, Texas). Estimates with 

 values less than .05 were considered statistically significant. This study was approved by the institutional review board at the University of Washington and the Premera data privacy board.

We utilized an interrupted time series design, with the interruption coinciding with the policy implementation date occurring on July 1, 2010. This design utilizes characteristics and outcomes of the intervention group in the pre-policy periods and the characteristics and outcomes of the control group in the pre-policy and postpolicy periods to control for confounding.

 We divided our analysis into 3 periods: 3 years before to 1 year before VBF implementation (early pre-VBF period: July 2006 to June 2009), 1 year before VBF implementation (late pre-VBF period: July 2009 to June 2010), and immediately after VBF implementation to 3 years after (post-VBF period: July 2010 to June 2013). This equals 84 months of observation. We conducted additional sensitivity analyses that excluded 1, 3, and 6 months immediately prior to and after VBF implementation to assess potential anticipatory or delayed filling of medications.

In the first month of the late pre-VBF period (July 2009), the intervention group had an increase of $5 in the pharmacy co-payment for 2 co-payment tiers. The pharmacy benefits in the intervention group are described in detail in eAppendix Tables A and B [

]. In contrast, there were no changes in the pharmacy benefits for the control group over the entire period. We controlled for all benefit changes other than the implementation of the VBF policy in our statistical analyses.

Exploratory Analyses: Association of the VBF and Medication Utilization by Change in Co-Payment Tier 

Because the VBF is expected to increase utilization of drugs moved into lower co-payment tiers and decrease utilization of drugs moved into higher co-payment tiers, we separately describe the association of the VBF with the days’ supply (PMPM) of medications moved into lower co-payment tiers, higher co-payment tiers, or no change in tier in the VBF. We limited these exploratory analyses to observations 1 year before and after VBF implementation to reduce potential tier misclassification. Occasionally, drugs may be reassigned to co-payment tiers on a quarterly basis after VBF implementation to reflect updated cost-effectiveness estimates. Over a longer duration, this may result in misclassification if medications are initially labeled as “no change” in co-payment tier when the drug was later moved into a lower co-payment tier.

We calculated the mean days’ supply PMPM in the intervention and control groups in the pre- and postpolicy periods by taking the sum of the days’ supply and dividing by the number of personmonths of observation in the respective groups and periods. The association between the VBF and these medication utilization measures was calculated as the difference between the mean days’ supply PMPM comparing the pre- and postpolicy periods for the intervention and control groups separately and then the difference between these pre-/post differences comparing the intervention and control groups.

The intervention and control groups included 230 (8634 membermonths) and 660 members (23,330 member-months), respectively, for the diabetes cohort; 844 (36,315 member-months) and 2592 members (104,139 member-months), respectively, for the hypertension cohort; and 606 (19,485 member-months) and 2269 members (71,015 member-months), respectively, for the hyperlipidemia cohort. Individuals in the intervention cohorts were, on average, younger and resided in zip codes with higher median household incomes (

). As specified a priori, we adjusted for these and other demographic and socioeconomic characteristics. Rates of enrollment or attrition (results not shown) did not differ between VBF and control groups in the pre- and postpolicy periods.

Changes in Mean Co-Payments Due to the VBF

The VBF reduced mean co-payments for antidiabetic and antihypertensive medications by $5 (standard deviation [SD] = $8) and $6 (SD = $8), respectively, and increased mean co-payments for antihyperlipidemic medications by $6 (SD = $21), comparing the year before and after VBF implementation (eAppendix Tables C-E). Antidiabetic medications were either moved into lower co-payment tiers or had no changes in co-payment tier. Antihypertensive and antihyperlipidemic medications were moved up or down or had no change in co-payment tier.

Changes in Medication Adherence and Expenditures After Adjusting for Secular Trends 

The VBF policy had no statistically significant impact on adherence in any of the 3 disease cohorts (

). Comparing the observed expenditures in the intervention group with the expected expenditures in this group had the VBF policy not been implemented, we found that for the diabetes cohort, member expenditures decreased significantly by $5 PMPM (63%; 

 <.001), health plan expenditures decreased nonsignificantly by $4 PMPM (13%; 

 = .282), and total medication expenditures decreased significantly by $9 PMPM (24%; 

). For the hypertension cohort, member expenditures decreased significantly by $4 PMPM (103%; 

 <.001), health plan expenditures increased significantly by $3 PMPM (29%; 

 <.001), and total medication expenditures decreased nonsignificantly by $1 PMPM (7%; 

 = .118). For the hyperlipidemia cohort, member expenditures did not change (

 = .471), health plan expenditures decreased nonsignificantly by $3 PMPM (8%; 

 <.085), and total medication expenditures decreased nonsignificantly by $3 PMPM (7%; 

For the descriptive comparisons of mean days’ supply changes in the year before and after VBF implementation for the intervention and control groups in the diabetic cohort, we found that the intervention group had an increase in days’ supply PMPM for medications moved into lower co-payment tiers and a decrease in days’ supply PMPM for medications with no change in co-payment tiers (

). In the hypertension cohort, we found that the intervention group had an increase in days’ supply for medications moved into lower co-payment tiers and no change in days’ supply for medications with no change in co-payment tiers or that moved into higher tiers. In the hyperlipidemia cohort, we found that the intervention group had an increase in days’ supply for medications moved into lower co-payment tiers and a decrease in days’ supply for medications with no change in co-payment tiers or that moved into higher tiers.

We found that the VBF decreased overall co-payments for antidiabetic and antihypertensive medications but increased overall co-payments for antihyperlipidemic medications. These co-payment changes did not have a statistically significant effect on medication adherence as measured by pharmacy claims. However, the percentage changes in adherence for the hypertension and hyperlipidemia cohorts (3% and 5%, respectively) were in line with the improved adherence findings from other VBID studies.

One possible reason as to why this same magnitude of adherence change was not statistically significant was that, as an intervention that increased and decreased co-payments, adherence may have increased for patients on highervalue drugs and decreased for patients on lower-value drugs. This would tend to increase the variance of the estimates, potentially contributing to the nonstatistically significant adherence findings.

The VBF did have measureable effects on medication expenditures. For the diabetes cohort, we found statistically significant reductions in expenditures due to the VBF from the member and total medication perspectives. In contrast, for the hypertension cohort, member medication expenditures decreased while health plan expenditures increased significantly.

The different mean overall co-payment changes for each of the 3 disease cohorts, along with no significant effects on adherence, yet different effects on expenditures, prompted the additional exploratory analyses. The VBF is expected to increase utilization of medications moved into lower co-payment tiers and decrease utilization of medications moved into higher co-payment tiers; therefore, overall utilization may not change. Further, adherence primarily measures changes in the intensive margin of use, the level of utilization among users. In contrast, co-payment changes due to the VBF can also affect the extensive margin of use (an increase or decrease in the total number of users). Hence, for our exploratory analyses, we classified medications by changes in co-payment tiers and we utilized days’ supply PMPM as a more inclusive measure of changes in utilization.

In these exploratory analyses, we found that for the diabetes and hyperlipidemia cohorts, there was an increase in days’ supply for medications subject to lower co-payments (moved into lower tier) and decreased days’ supply for medications subject to higher co-payments (moved into higher tier or no change in tier). For the hypertension cohort, the days’ supply increased for medications subject to lower co-payments and did not change for medications subject to higher copayments. These exploratory results suggest that individuals in diabetes and hyperlipidemia cohorts may have decreased utilization of lower-value drugs and increased utilization of higher-value drugs.

Our results suggest that the expected impact of a VBF depends on 4 factors: 1) the magnitude and direction of co-payment changes, 2) the sensitivity of individuals to co-payment changes (ie, elasticity of demand), 3) the incremental effect of the change in medication utilization on costs and outcomes (ie, whether individuals are shifted toward medications with lower costs and better health outcomes or both), and 4) the relative number of individuals on high- versus low-value drugs prior to VBF implementation. The second, third, and fourth factors translate the direct effect of the VBF on co-payment changes into downstream effects on medication utilization, individual patient costs, and outcomes and population costs and outcomes, respectively.

In this study, among the diabetes and hypertension cohorts, mean co-payments were both reduced by the same amount. However, the diabetic members appeared to be more sensitive to the price changes and switched from low-value to high-value drugs. If, for the diabetes cohort, the higher-value drugs were less costly than the lower-value drugs, then a shift to higher-value drugs would explain the member and total savings. In contrast, despite the same mean reduction in co-payment, members with hypertension did not seem to be as price-sensitive: they continued to use the same amount of low-value drugs and slightly increased their use of high-value drugs. This could result in the observed cost shift from member to health plan if the increased health plan expenditure due to the increased utilization of high-value drugs more than overcomes the increased cost sharing for low-value drugs. This is particularly possible given that mean co-payments decreased overall for the antihypertensive drugs.

Because value in a VBF setting is measured as a ratio of costs relative to health benefits, the expected net effect of a VBF may be to reduce costs or improve health outcomes, or both. In this study, we measured medication adherence and expenditures; however, nonmedication expenditures and health outcomes were not measured. Yet, these are important outcomes that help to understand the overall impact of the VBF policy. Our previous work had analyzed the effect of the VBF on nonmedication health expenditures and emergency department visits, hospitalization, and office visits, and had found no clinically and no statistically significant effects. For this study among these smaller chronic disease cohorts, our preliminary power calculations revealed that we would be insufficiently powered to detect any differences in these nonmedication outcomes.

A second limitation is that the exploratory analyses that revealed associations between days’ supply changes and co-payment tier changes were descriptive in nature and were intended for hypothesis generation and to guide future analyses. Future analyses with larger sample sizes should consider inferential statistics in order to test the hypotheses generated from these exploratory analyses.

A third limitation is the use of pharmacy claims as a proxy measure of medication adherence. Medication adherence is a complex health behavior, making measurement challenging. However, current healthcare quality indicators, such as the CMS Star Ratings, use claims-based measures of medication adherence.

A fourth limitation is that our sample was drawn from a working- age population and their dependents; thus, our results may not necessarily be generalizable to other populations, such as the poor, elderly, or chronically ill. This is an important limitation in light of the CMMI VBID pilot studies that will be conducted in the Medicare population. The effect of a VBF on such a population will need to be validated.

Articles

A value-based formulary was implemented that used cost-effectiveness analysis to inform medication co-payments. Diabetes cohort expenditures decreased by $9 per member per month.

Impact of a Value-Based Formulary in Three Chronic Disease Cohorts

Risk-sharing agreements (RSAs) between drug manufacturers and payers link coverage and reimbursement to real-world performance or utilization of medical products. These arrangements have garnered considerable attention in recent years. However, greater use outside the United States raises questions as to why their use has been limited in the US private sector, and whether their use might increase in the evolving US healthcare system. 

 To understand current trends, success factors, and challenges in the use of RSAs, we conducted a review of RSAs, interviews, and a survey to understand key stakeholders’ experiences and expectations for RSAs in the US private sector.

Trends in the numbers of RSAs were assessed using a database of RSAs. We also conducted in-depth interviews with stakeholders from pharmaceutical companies, payer organizations, and industry experts in the United States and European Union. In addition, we administered an online survey with a broader audience to identify perceptions of the future of RSAs in the United States. 

Most manufacturers and payers expressed interest in RSAs and see potential value in their use. Due to numerous barriers associated with outcomes-based agreements, stakeholders were more optimistic about financial-based RSAs. In the US private sector, however, there remains considerable interest—improved data systems and shifting incentives (via health reform and accountable care organizations) may generate more action.

 In the US commercial payer markets, there is continued interest among some manufacturers and payers in outcomes-based RSAs. Despite continued discussion and activity, the number of new agreements is still small.

	This research assessed whether risk-sharing agreements between US commercial payers and pharmaceutical companies may become of greater interest with improved data systems and shifting incentives (via health reform and accountable care organizations). Key findings: 

		While such agreements offer numerous advantages to payers, pharmaceutical companies, patients, and society, their growth remains stagnant in the United States, due in large part to the additional effort required to implement and adjudicate risk-sharing agreements as compared to traditional rebates. 

		In the US private sector, however, there remains considerable interest: improved data systems and shifting incentives may generate more action.

Risk-sharing agreements (RSAs) between drug manufacturers and payers—also called performance-based risk-sharing arrangements (PBRSAs), managed entry agreements, patient access schemes, and coverage with evidence development (CED), among other terms&mdash;link coverage and reimbursement levels to real-world performance or utilization of medical products.

 These arrangements have garnered considerable attention in recent years: a recent analysis of the University of Washington Performance-Based Risk-Sharing (PBRS) Database,

 including 148 arrangements over the last 2 decades, concluded that although the overall pace of adoption seems to be slowing, several new countries have begun to implement RSAs.

 In the United States, the largest number of arrangements has been in the Medicare program; the uptake of private sector RSAs seems stagnant despite growing interest in the general principle of “paying for performance.” The aim of this study was to assess the state of and prospects for private sector RSAs in the United States, considering by comparison experiences from other countries, most commonly involving public payers.

RSAs offer a number of potential advantages,

including: 1) reducing the risk to payers of making a suboptimal purchase; 2) providing earlier access to medications for patients; 3) creating international pricing efficiency, especially in a world with external reference pricing and parallel trade; and 4) generating evidence on what works in the real world. These potential advantages, coupled with the higher volume of use outside the United States, raise questions as to why they have seen limited use in the United States,

 as well as what arrangements might gain traction with the new emphasis on accountable care organizations (ACOs), which operate under stronger incentives to manage patient care efficiently.

Previous work in this area has characterized the types of RSAs, trends in their adoption, economic incentives and rationale, and the drivers and barriers for adoption, but has not focused on current thinking in the US private sector. A broad distinction has been made between “outcomes-based” RSAs and “financial-based” RSAs. For the former, the manufacturer provides or agrees to rebates, refunds, or price adjustments if their product fails to meet agreed-upon clinical outcome targets. For the latter, reimbursement is tied to financial measures (eg, total sales) or to utilization. Here, we focus on the US private sector marketplace and particularly on the level of interest in outcomes-based agreements; specifically, we sought to understand the barriers to and drivers of developing private sector RSAs in the United States. We also explored whether the current situation reflects a broader demand by regulators and payers for greater real-world data and comparative evidence.

To assess the validity of the general perception of the lack of uptake of RSAs in the United States, we conducted a review of the University of Washington’s PBRS Database to assess the number of US agreements and trends in the number of agreements over time. For comparison, we also examined the number of agreements for the rest of the world. At the time of the study, the database included information on RSAs initiated from January 1993 to December 2013, with information on the parties participating in the arrangement, the type of arrangement, and details of the arrangement and/or outcomes, as available.

To understand current trends, success factors, and challenges in the use of RSAs, we conducted key informant interviews and an online survey to understand key stakeholders’ experiences with RSAs both inside and outside the United States as well as lessons for the US private sector situation. We used purposive sampling to identify 16 key stakeholders from pharmaceutical companies, payer organizations, and industry experts in the United States and European Union. We developed the interview guide and survey based on previous work in the area

 and through iterative review and revision by a group of content experts. The key informant interviews consisted of in-depth, 1-hour telephone interviews conducted between October 2013 and May 2014. Interview topics included: experience with RSAs, motivating factors, administrative considerations, effectiveness of agreements, and potential for use in the future (

). To identify themes from the interviews, we assessed the number of interviewees who stated or generally agreed with key statements across all interviews.

The survey was administered via the Web in May 2014, and consisted of 8 questions that asked respondents about their expectations for the future of RSAs in the United States and to identify and rank key barriers to their use (

). We identified themes by assessing the number of respondents who provided similar answers to the surveys. We invited a convenience sample of 37 individuals to complete the survey. Given the small sample size, we summarized responses using descriptive statistics.

Our review of the PBRS Database identified 148 RSAs worldwide from the late 1990s and 2013. Only 18 of the 148 total arrangements in the University of Washington Database, or 12.2%, represent US RSAs: 11 of these were public sector coverage with evidence development schemes, while only 7 were private sector RSAs. A review of the number of agreements annually suggests a little growth in the number of agreements in the United States. To date, RSAs have been more frequently employed in single-payer systems across Europe, Canada, and Australia.

Of the 16 individuals targeted for our key stakeholder interviews, we were able to schedule and conduct interviews with 14 (87.5%). The 14 in-depth interviews included 9 interviews with US stakeholders and 5 interviews with EU stakeholders. Of these, 7 were from pharmaceutical companies, 5 were currently or previously employed by payer/government organizations, and 2 were subject matter experts. We received 15 responses to our survey (response rate of 40.5%), including 10 responses from individuals from pharmaceutical companies, 3 from payers/pharmacy benefit managers, and 2 from other nonprofits and consultancies. The majority of respondents (93.3%) were from the United States.

In general, our interviews suggest that while there is interest in RSAs, there has been limited use in the United States due to difficulties in implementing and carrying out such agreements. The majority of interviewees (12 of 14) were cautiously to mildly optimistic about the future potential for outcomes-based RSAs to be adopted in the United States. To illustrate the basis of this position, selected comments and quotations from interviews are shown in 

, and are organized to represent comments ranging from more optimistic, to cautious, to more pessimistic. Roughly half of manufacturers and payers expressed interest in outcomes-based RSAs and see value in their use; almost all were optimistic about the use of financial-based RSAs. In general, due to the difficulty in implementing and executing outcomes-based RSAs, interviewees indicated an interest in more financial-based RSAs (eg, utilization or financial capitation) but less interest in clinical and health outcomes-based RSAs.

Outcomes-based agreements, while attractive to some, were perceived by interviewees to be difficult to execute and as having high transaction costs. Interview respondents were skeptical about being able to use outcomes-based RSAs, citing challenges in implementing and executing outcomes-based RSAs that would mitigate their potential in the United States, particularly given the fragmented payer system with patient movement across plans, as well as the current lack of data infrastructure that limits feasibility and, to some extent, interest in measuring long-term outcomes. Interview respondents indicated that simpler, financial-based agreements, on the other hand, have had demonstrated success. The survey responses corroborated the general support for financial-based RSAs seen in the interviews: 80% of respondents stated that the use of financial-based RSAs in the United States would grow.

The interviews suggest that the perceived value of an RSA versus a traditional discount mechanism depends on the product, disease area, and availability of the necessary data infrastructure. The manufacturers we interviewed stated that they would use RSAs as a way to differentiate and demonstrate the effectiveness of their product versus competitors. RSAs are typically not used in a US setting when a product is first to market or the market leader, but are more attractive to manufacturers when there is competition. The manufacturer seeks to secure beneficial formulary placement or gain market share. They can also be a mechanism to increase patient compliance.

Based on our interviews, US payers leverage—or would like to leverage&mdash;RSAs as a way to reduce uncertainty about a product’s clinical value, performance, or budget impact, as they allow payers and patients to gain experience with the medication. As payers and regulators are often interested in different types of data, at the time of launch, a product’s impact on costs and comparative effectiveness is often not well understood. Ultimately, payers expressed interest in RSAs as they allow payers to ensure that the price of a drug is more closely aligned to its value. During our interviews, payers indicated interest in RSAs for products that are more costly (eg, specialty drugs, biologics, combination products) and for disease areas for which cost consequences are substantial. Payers are not interested in engaging in RSAs for undifferentiated products where there is no real or perceived advantage.

 lists potential barriers to implementing RSAs that were identified during the interviews. The implementation and execution of RSAs is perceived to present a significant administrative burden that requires a substantial time investment. Further, outcomes-based agreements require significant payer and provider infrastructure: if payers lack the databases required to track individual patients, outcomes-based RSAs are difficult. Additional challenges that were mentioned include a) the need for adequately trained staff, b) the risk to pharmaceutical companies associated with being responsible for outcomes when they cannot control the way a drug is prescribed or used, c) the management of consequences in terms of changing preferred drugs and denying coverage for drugs, and d) the identification of outcomes that are meaningful but measurable within a reasonable time frame.

We tested the barriers identified during the interviews with a broader audience in the survey by asking respondents to rate and rank from the list of barriers provided in Figure 4 and any additional barriers not listed. The interview themes were consistent with the top barriers identified in the survey (

). The significant additional effort associated with RSAs was selected as the number 1 barrier to the use of RSAs in the United States by 33% of respondents and as the number 2 barrier by 27% of respondents. Inadequate data infrastructure was cited as the number 1 barrier by 27% of respondents and as the number 2 barrier by 33% of respondents. Other barriers that were selected as one of the top 3 hindrances to conducting RSAs in the United States included: federal (Medicaid) best price (40% ranked this as one of the top 3 barriers), significant resources/costs of adjudication (40%), challenges in measuring relevant outcomes (27%), and difficulty in reaching contractual agreement (27%).

While not necessarily barriers, there were several logistical challenges highlighted during our interviews that may be a consideration for both payers and pharmaceutical companies when evaluating opportunities for RSAs. The number of agreements in place with a given payer or for a given class of drugs is a key consideration. While payers expressed interest in having multiple agreements for competing products, pharmaceutical manufacturers indicated this becomes a complex arrangement that they are reluctant to enter. Further, only large payers are likely to have the capability for multiple, simultaneous agreements, given the burden of negotiation, data collection, monitoring, and adjudication of RSAs. Another key consideration is duration, as payers and manufacturers agree that short-term deals are not desirable given the considerable investment in evidence development. On the other hand, they state that long-term deals are also not desirable given the costs and risks involved. Both parties acknowledge that medium-term deals (18-36 months) are the right balance trading off the sizable up-front investment and a preference to execute agreements reasonably fast.

Access to clinical data was stated as a key barrier to the use of RSAs; scenarios where clinical data are more readily available make RSAs more feasible. For instance, interviewees indicated it may be easier to measure health outcomes for drugs that are administered in settings where there are more immediate clinical data available (eg, the hospital setting), where drugs are administered in person, or when data are already being collected as part of ongoing processes (eg, performance measurement via the Healthcare Effectiveness Data and Information Set). Complex outcomes might be more easily measured where there is an active provider (eg, patient-centered medical homes). Similarly, interviewees stated outcomes-based RSAs can be most successful where the infrastructure is robust to collect such data, such as in single payer or closed settings including integrated delivery networks. Drug cost as a share of total episode cost can also be a factor.

A key success factor is the need for manufacturers to understand the amount of risk they are undertaking when entering into an RSA. Manufacturers feel they should be able to reasonably predict plausible outcomes of the agreement and assess the level of risk; for example, what level of compliance is required and to what extent the clinical trial population differs from the real-world population.

There are a few scenarios that are considered prohibitive to successful RSAs. Both parties indicated that trust is a critical component of agreements between pharmaceutical companies and payers. It is imperative that both parties trust the data: clear agreements on data validation and analysis are important to create trust in the data.

Certain types of agreements can also be problematic. For instance, manufacturers expressed that population-based agreements are perceived to be risky because there are many unknowns around compliance, prescribing, and so forth. Manufacturers are reluctant to take on risk when they cannot predict how their product will be used in the population. Agreements that involve the manufacturer paying for nonpharmaceutical costs are difficult to execute, as payers do not generally have the systems or data to support such agreements. Both parties shy away from agreements in disease areas where there are many different treatment paradigms or the relevant outcome is an intermediary outcome, because it can be challenging to attribute the outcome to the product in question.

Future Prospects: Barriers, Health Reforms, and Policy

This study confirms that very few private sector RSAs have been implemented in the United States; while payers and pharmaceutical companies remain interested in these types of agreements, various challenges restrict the types of situations in which they can be feasibly employed. Also, because of the effort required, payer and pharmaceutical respondents are less enthusiastic about outcomes-based agreements and more optimistic about simpler, financial-based agreements.

Although the global growth in RSAs seems to have slowed, several countries maintain robust programs, particularly Italy and Sweden, and interest is growing in Latin America and Asia. There has also been a broad shift to financial-based agreements for several reasons. Our research suggests that public payers, especially outside the United States, are attracted to making arrangements with manufacturers, first and foremost, to obtain an effectively lower price in a world where reference pricing and the prospect of parallel trade make differential (or tiered) pricing difficult. And, of course, this can benefit patients by providing earlier and broader market access. Second, few countries have an adequate data infrastructure to operate outcomes-based schemes efficiently.

The history of RSAs in the United States is dominated by the Medicare experience with CED. Although the interest remains, it seems to have waned. The growth in visible private RSAs in the United States seems to have stalled, but our interviews suggest growing interest and some new activity. There are multiple barriers in the United States—most prominently, the additional effort needed to negotiate and maintain these agreements relative to traditional rebates. The challenge of measuring individual-level outcomes in a claims-oriented data infrastructure makes operationalizing RSAs difficult because of the lack of relevant clinical outcomes contained in these data. Medicaid best price (and the similar 340B Drug Pricing Program) provisions, which link mandated discounts for public sector programs to private sector prices, are considered complications, but would seem to be manageable in a well-designed contract; notwithstanding, several interviewees regarded this as a major barrier.

If a medicine covered by an RSA does not perform as anticipated by the manufacturer and a price discount ensues, the reduced prices could be perceived as a new best price and made available to all Medicaid purchasers. In the case of Medicare, we were told that provisions around the Medicare 5-star rating program, which ties payment to achievement of certain quality metrics in health plans, have made adherence programs—a kind of outcomes-based program&mdash;attractive to payers. The preference for a short-term horizon of 18 months to 3 years being the most practicable arrangement in the US private health system, because it reduces long-term liability for manufacturers and payers, and limits the feasibility of RSAs in some disease areas.

Our interviewees were generally pessimistic as to whether the growth of ACOs would have any positive impact in the near term, since many ACOs are in the initial development and market experimentation phase

; however, they may well bring a new orientation to risk-sharing that will support the future development of pharmaceutical RSAs.

The ISPOR PBRSA Task Force emphasized that better information about what works in medicine is a global public good,

 but it still appears that the RSAs being developed in the United States will be confidential ones, with the knowledge gained only shared inadvertently or sporadically, rather than by design. Both payers and manufacturers have competitors, and they have some incentive to be “free riders” rather than developing innovative pricing schemes that yield knowledge on best medical practice. One of our payer interviewees—a national medical director for a major US payer&mdash;made this revealing and somewhat encouraging comment:

“If we learn something about a pharmaceutical by looking at our own data and we decide that we are going to use that information to make coverage determinations, it is our policy that we make that public (publish in a peer-reviewed journal). If the coverage decision is based on an economic analysis, that may be kept private; however, if the decision is based on something clinical, it has to be available for physicians and patients to evaluate.”

This suggests an important dichotomy between clinical and economic evidence, though they clearly become related when combined in a cost-effectiveness ratio or other assessment of value. It is not clear if this distinction is sustainable given that the recent joint American Heart Association and American College of Cardiology Task Force on Value and Cost called for consideration of cost-effectiveness in establishing clinical treatment guidelines in cardiology.

The growing emphasis on comparative effectiveness research in the United States, including the establishment of the Patient-Centered Outcomes Research Institute,

 are testimony to the fact that the country is investing more in generating real-world evidence on clinical outcomes and their related economic implications. In principle, RSAs will generate some of this information, but if the results remain confidential, the new knowledge will, at best, slowly diffuse through a healthcare system that aspires to become a “learning healthcare system.” This raises the question of the need for greater public policy intervention. The following question was not part of our remit in this study, but we should be asking: Are there other policy tools or incentives that could be used to encourage and leverage US private sector RSAs to reveal their lessons for the greater good of the system and all patients?

Assessment of current trends, success factors, and challenges in the use of risk-sharing agreements in the US private sector.

Private Sector Risk-Sharing Agreements in the United States: Trends, Barriers, and Prospects

 Unintended consequences may result from changes in pharmacy benefit design. The objective was to determine the impact of increasing patient prescription copayments for guidelinerecommended, long-term asthma controller (LTC) medications on asthma-related medication use and healthcare services.

 We used 2005 MarketScan healthcare and pharmacy claims data to identify asthma (

International Classification of Diseases, 9th Revision, Clinical Modification [ICD-9-CM]

 diagnosis code 493.xx) patients aged 12 to 64 years who were continuously enrolled through 2006 with 

1 claim for an asthma medication in 2005. LTCs included: inhaled corticosteroid (ICS) (n = 10,251), ICS plus long-acting beta agonist (COMBO) (n = 27,407), and leukotriene receptor antagonist (LTRA) (n = 20,664).

Using multivariable models, we estimated the associations between changes in LTC copayments and LTC consumption and asthmarelated outpatient and emergency department (ED) visits.

$5 average increase in patient copayments per month of medication supplied (yes/no). The mean annual change (2005-2006) in copayments per month was $13.23 versus —$3.88 (ICS), $11.76 versus –$3.06 (COMBO), and $9.78 versus –$2.06 (LTRA). The 

$5 group experienced a significant decline in average annual days of medication supplied of —47.1 days of ICS (95% CI –43.5 to –50.8), –35.3 days of COMBO (–32.4 to –38.2), and –47.5 days of LTRA (–43.2 to –51.7). Among COMBO and LTRA medication users, the 

$5 copayment increase was associated with more asthma-related outpatient visits and ED visits compared with the <$5 group.

 The findings suggest that even small changes in average copayment for asthma medications can result in significant reductions in medication use and\ unintended increases in healthcare services.

$5 average increase in patient copayments per month of medication supplied (yes/no).

$5 increase in copayment was associated with more than 1 month’s supply less utilization of long-term controller medications over the course of 1 year.

		For certain classes of long-term controller medications, a 

$5 increase in copayment was associated with more subsequent oral corticosteroid bursts, increases in asthma outpatient visits, and increases in asthma emergency department visits.

		The study findings should be considered when designing drug benefit decisions that change formulary patient copayment structures.

	As the costs of healthcare rise, healthcare payers struggle to balance access and affordability for patients and plan sponsors. To mitigate overconsumption of pharmaceuticals, many US pharmacy benefits have built-in, out-of-pocket (OOP) payments for pharmaceuticals. Historically, the patient’s contribution to pharmaceutical purchases (copayments) has been quite low, on the order of 10% to 15% of the average market price. In recent years, however, payers have increased patient copayment substantially and have created tiered pharmacy benefits and step-edits to countermand increasing demand for expensive drugs.

 For example, a 3-tier pharmacy benefit may require a monthly patient (OOP) cost of $10 for less costly generic drugs, a $25 copayment for preferred branded drugs, and a $50 copayment for nonpreferred branded drugs. This type of system is intended to incentivize the patient to move toward less costly substitutes. Substitutes are not, however, always available at the lower copayment level.

	An unintended and possibly harmful consequence of changes in pharmacy benefit design is the impact that higher copayments have on clinically beneficial or optimal medication consumption.

 Several studies have shown that increasing OOP costs for prescription drugs reduces the rate of utilization of medications for patients with chronic diseases.

 However, the literature is limited on the impact that changes in health benefit design have on asthma controller medication utilization. Canadian researchers have found that increases in patient contributions to controller medications, such as inhaled corticosteroids (ICS), have reduced ICS utilization.

 In the United States, findings on the impact of changes in health benefit design have been mixed. Fung et al showed a reduction in ICS use with increased demand barriers on the Medicare population,

 whereas Crown et al suggested a limited change in patient-level treatment patterns due to higher patient contributions.

 The latter did, however, suggest that physician and practice prescribing preferences influenced patient treatment.

	An important unanswered question is to what extent changes in consumption resulting from alterations in pharmacy benefit design impact health outcomes. Our study objective fills a gap in the literature that relates changes in longterm asthma controller (LTC) medication patient copayment with LTC utilization and with asthma health outcomes. We tested whether or not an increase in patient-level average LTC prescription copayment per month would result in lower LTC utilization and higher rates of asthma-related health services, which are typically thought of as components of asthma exacerbations. Asthma is a convenient disease state to test these hypotheses using years 2005 and 2006 claims data because: (1) LTCs are guideline-recommended therapies; (2) LTCs prevent inflammation and reduce the risk of asthma exacerbations but many LTC users do not necessarily observe noticeable and immediate benefits from LTCs; (3) some consequences of LTC medication failure can be observed in health insurance claims data; and (4) during the study period, no known large changes occurred to the LTC market in terms of new LTCs or new generics.

	We performed a retrospective cohort analysis using the Thompson Healthcare MarketScan commercial database (29 million covered lives in all geographic regions of the United States). MarketScan captures medical (inpatient, outpatient, and emergency care) and pharmacy claims information including the claim-level patient copayment amount, as well as eligibility status, from employees of large corporations. MarketScan is known for its high standard of claims-based information that is representative of the privately insured US population.

	We used the 2005 and 2006 years of the MarketScan claims database to identify the study cohort and test our hypotheses. Inclusion criteria for the study (

) were: (1) a 2005 diagnosis of asthma defined by 1 of the following criteria: at least 2 outpatient claims with primary or secondary diagnoses of asthma (

 diagnosis code 493.xx) OR at least 1 emergency department (ED) or hospitalization claim with primary diagnosis of asthma (493.xx); (2) continuous enrollment during 2005 and 2006; (3) age 

12 and <64 years on January 1, 2005; and (4) at least 1 drug claim for an LTC medication (ICS, ICS combined with long-acting beta-2 agonist [LABA], leukotriene receptor antagonist [LTRA], mast cell stabilizers, or omalizumab) with 

30 days supply during 2005. Patients were excluded (Figure) if they had: (1) a diagnosis of chronic obstructive pulmonary disease (

491.2, 493.2, 496, 506.4), or at least 1 claim for an anticholinergic medication; (2) a diagnosis of emphysema (

 492.x, 506.4, 518.1, or 518.2); (3) a diagnosis of cystic fibrosis (

 277.0x); or (4) were enrolled in Medicare.

	We linked national drug codes in the Medstat data with classes as defined in the Healthcare Effectiveness Data and Information Set 2009 Final NDC table.

 Upon reviewing the LTC utilization patterns of the study cohort during the year 2005 (Figure), we included the following most common LTC classes for analysis: ICS, ICS plus long-acting beta agonist (COMBO), and LTRA. The primary exposure variable of interest was the average change in patients’ monthly prescription copayment from 2005 to 2006 for each LTC class.

	The average copayment per patient month for a given LTC class within a claim year was estimated by averaging all of the LTC claims filled for a patient and the patient OOP payment (copayment) divided by the number of months supplied (assuming 100% medication adherence). The months supplied were estimated using the days supplied field in the claims data. Patients who filled an LTC class in 2005 but not in 2006 were assigned the plan-level average monthly copayment for 2006. In this way, we assumed such patients would have faced this copayment had they filled the same LTC class in 2006. The change in copayment was calculated for each LTC class and patient by taking the average copayment per month supplied in 2006 and subtracting the average copayment per month supplied in 2005.

	The outcome measures captured from 2006 pharmacy and medical claims included the following for 2005 LTC users: LTC utilization (yes/no), LTC days supplied, LTC substitution and switching patterns, oral corticosteroid bursts, asthma-related outpatient visits (primary or secondary diagnosis of asthma with 

 493.xx), asthma related ED visits (primary asthma diagnosis with 

 493.xx) and asthma-related hospitalizations (primary asthma diagnosis with 

 493.xx). LTC utilization was characterized in terms of average days supplied and also adherence for those who had at least 30 days supplied. Adherence was defined as the medication possession ratio where the total days supplied in a year was divided by the number of days in 1 year.

	The distribution of patient copayment change (primary exposure measure) was split into those experiencing 

$5 average increase in monthly copayment versus <$5, based on an empiric analysis. We also had interest in evaluating an economically significant change of at least a $5 increase because previous evaluations suggested that this exposure variable is likely to not be linearly related to outcome measures.

 We used descriptive statistics to display the 2005 demographics and clinical characteristics and the 2005 versus 2006 LTC utilization by copayment change category for each LTC. Differences in continuous characteristics were tested using 

 tests and differences in the categorical characteristics and utilization proportions were tested using X

 test. For each LTC, we used linear regression to test for differences in the continuous outcome of 2006 days supplied and Poisson (negative binomial distribution for asthma-related hospitalizations) regression to test for differences in 2006 count outcomes, including oral corticosteroid bursts, asthma-related outpatient visits, asthma-related ED visits, and asthma-related hospitalizations.

	We adjusted all regression analyses for 2005 baseline and clinical characteristics including: age, gender, insurance plan type, comorbidities, LTC use in terms of 2005 annual days supplied, asthma-related outpatient visits, oral corticosteroid bursts, asthma-related ED visits, asthma-related hospitalizations, and total OOP patient expenditures. We applied a modified Charlson Index (Quan et al index) to generate scores on the level and burden of comorbidity for the study cohorts to adjust for the potential confounding of health status.

 The Quan et al index draws on diagnostic information from 

codes and procedure codes, resulting in conditions that are weighted based on the adjusted risk of 1-year mortality. The index score is the sum of the weights for all of a patient’s conditions, with higher numbers indicating increased levels of comorbidity.

	We conducted 2 sensitivity analyses on key exposure and outcome variables. First, we examined the 

$5 average increase in monthly copayment only in those who filled the same LTC in both 2005 and 2006 (a subgroup of the base case cohort). Second, we examined the effect of a 

$10 average increase in monthly copayment cutoff on the base case cohort.

	Of a possible 2,434,764 patients with claims in the 2005 MarketScan claims data set, our study cohort was a sample size of 40,784 (Figure). Of the overall study cohort, 10,251 (25.1%) patients filled at least 30 days supplied of ICS; 27,407 (67.2%) filled at least 30 days of COMBO; and 20,664 (50.7%) filled at least 30 days of LTRA. Note that the LTC categories are not mutually exclusive. A patient could be counted in more than 1 LTC category if he or she filled at least 30 days supplied in each category.

	The 2005 mean copayment per month supplied was $18.51 with standard deviation (SD) of $15.10 for the ICS cohort, $19.16 for the COMBO cohort (SD = $13.10), and $16.46 for the LTRA cohort (SD = $10.60). The mean copayment per month supplied from 2005 to 2006 for all 3 LTC class cohorts changed less than $1.00 in absolute value. Patients in all 3 cohorts with <$5 change in copayment per month supplied had, on average, a reduction in monthly copayment from 2005 to 2006 (

). Conversely, patients in all 3 cohorts with 

$5 increase in copayment per month supplied had, on average, an increase in monthly copayment from 2005 to 2006 (Table 1).

$5 increase in patient copayment per month supplied, compared with patients with a <$5 increase, were younger (

 <.0001), had a different distribution of insurance plan type (

 <.0001), had lower baseline LTRA copayment (

 = .008), and lower annual patient OOP expenses (

). The 2005 demographic and clinical characteristics for the ICS and COMBO cohorts were not displayed. However, statistical differences between copayment groups followed the same trend as in the LTRA cohort.

 displays the proportion of patients who filled the same LTC in 2006, comparing those with <$5 increase in average monthly copayment with patients with 

$5 increase in average monthly copayment. Among the ICS cohort, 66.7% (<$5 increase) versus 37.3% (

 <.0001) filled an ICS prescription in 2006. Among the COMBO cohort, 70.8% (<$5 increase) versus 58.3% (

 <.0001) filled a COMBO prescription in 2006. Among the LTRA cohort, 77.3% (<$5 increase) versus 59.8% (

 <.0001) filled an LTRA prescription in 2006. Further, patients in the ICS cohort with a 

Even small changes in average copayment for long-term controller asthma medications can result in significant reductions in medication use and increases in healthcare services.

Sean D. Sullivan, PhD

Articles

Impact of a Value-Based Formulary in Three Chronic Disease Cohorts

Private Sector Risk-Sharing Agreements in the United States: Trends, Barriers, and Prospects

Increasing Pharmaceutical Copayments: Impact on Asthma Medication Utilization and Outcomes

Colony-Stimulating Factor Prescribing Patterns in Patients Receiving Chemotherapy for Cancer