Factors Associated With Employee Participation in a Value-Based Insurance Design Initiative

AJPB® Translating Evidence-Based Research Into Value-Based Decisions®November/December 2013
Volume 5
Issue 6

We describe enrollment in a voluntary VBID initiative. Uptake was suboptimal, suggesting that barriers to enrollment may have significant implications for the overall effectiveness of VBID.

Value-based insurance design (VBID) programs are designed to reduce or eliminate cost barriers to evidence-based healthcare services. Accumulating evidence suggests that VBID initiatives are associated with modest increases in medication adherence with targeted therapies.1-7 The ultimate impact on cost and clinical outcomes is less clear, but decision-analytic models and preliminary evidence from VBID implementations suggest that reductions in subsequent complications and comorbidity may lead to overall cost-effectiveness and possibly cost savings.3,8-12

VBID initiatives are featured in the Affordable Care Act of 2010 and are generating signifi cant interest among large employers. Despite becoming increasingly common, the optimal design of VBID initiatives has not been determined. In particular, it is not clear whether these programs should be made universally available to all beneficiaries or whether more selective criteria should be used, such as requiring patients to actively enroll in order to qualify for the benefits. On one hand, requiring patients to actively participate (ie, by requiring simultaneous participation in a disease management program) may increase awareness and benefit at the individual level. However, this may have unintended consequences if the requirement to participate provides a barrier to enrollment and decreases participation in the VBID program. This may beparticularly harmful if the individuals who could potentially derive the most benefit are excluded from the program. In 2009 Christiana Care Health System (Christiana Care) implemented a VBID program that eliminated copayments for medications and supplies for glycemic control for benefi ciaries with diabetes. To receive the benefit, participants were required to complete a brief baseline evaluation and provide consent for review of administrative claims data. Our objective in this paper is to determine the participation rate and describe the baseline characteristics of eligible beneficiaries who did and did not choose to enroll in Christiana Care’s voluntary VBID initiative. Our findings provide the first empiric evidence to understand how participation requirements may impact the uptake of a VBID program and will have implications for implementing VBID more broadly.

METHODSStudy Design

We conducted a retrospective cohort study of all benefi ciaries who were eligible to participate in a voluntary VBID program for employees and dependents of Christiana Care Health System in Wilmington, Delaware. We compared demographic and medical characteristics and healthcare utilization patterns in the year before the program started between eligible beneficiaries who chose to participate and those who chose not to participate.

Study Setting, Program Description, and Requirements for Participation

Christiana Care is the largest healthcare provider and largest private employer in the state of Delaware. Christiana Care is self-insured with approximately 17,000 beneficiaries. Blue Cross Blue Shield of Delaware provided administrative management for Christiana Care benefits.

In March of 2009 Christiana Care implemented the Copayment Elimination Program (“the program”), an 18-month-long pilot program eliminating copayments for all diabetes medications and supplies. Details of the program have been published previously.13 Eligible beneficiaries were identifi ed using claims data and received invitation letters mailed out from Blue Cross Blue Shield of Delaware. Advertisements also appeared on the Christiana Care intranet portal and in key locations throughout Christiana Care facilities. Interested benefi ciaries completed an online registration followed by an onsite meeting with program staff. During that meeting benefi ciaries completed a baseline survey of diabetes history and health status; received standardized measurements of blood pressure, weight, and height; and had baseline laboratory tests including a fasting lipid panel and glycated hemoglobin (A1C). Participants were required to provide consent for release of administrative claims data for the purposes of evaluation. Once these steps were complete eligible copayments for participants were eliminated for 12 months. After 12 months, contingent on the participant returning for laboratory testing and survey completion, the copayment elimination extended an additional 6 months.

Identification of Study Population: This study used claims data to retrospectively identify benefi ciaries who would have been eligible for participation in the program at the time it was initiated. Eligible beneficiaries included all employees and dependents with at least 1 inpatient or outpatient visit with a diagnosis of diabetes mellitus (International Classification of Diseases, Ninth Revision, Clinical Modification codes 250.x) in the year prior to the start of the program. Because our goal was to compare health services utilization and medication adherence among eligible beneficiaries in the year preceding program start, we excluded patients not continuously enrolled in the employer-sponsored insurance plan during the entire study period. We also excluded beneficiaries who would turn 65 years of age prior to the end of the fi rst year of the program because the nearness of Medicare eligibility may have impacted the decision to participate.

Study Variables: We extracted demographic and clinical characteristics and utilization patterns for the study cohort from March 1, 2008, to February 28, 2009, the year prior to program initiation. All data were from administrative claims data deidentifi ed by a third-party organization (Healthcore, Wilmington, Delaware).

Disease and comborbidity variables included diabetes type, comorbid diagnoses, healthcare utilization, laboratory testing, and costs. We classified diabetes as type 1 if there was at least 1 claim for type 1 diabetes mellitus (T1DM) (250.1, 250.3). We classified comorbidities using all primary and secondary diagnosis codes from inpatient or outpatient settings in the 2 years prior to program initiation using the composite of Elixhauser and Charlson-Deyo classifications as described by Gagne et al.14 We measured hospitalizations and visits to an emergency department and ambulatory providers. We classified ambulatory providers as primary care, cardiology, endocrinology, nephrology, and ophthalmology based on primary designated specialty in claims files. We used billing codes to determine the frequency of testing of A1C and lipids. We generated total paid claims and total out-of-pocket costs for medical care and pharmacy claims.

Though the copayment program eliminated copayments only for diabetes-related medications and supplies, we assessed baseline medication use for medications for diabetes, hypertension, and hyperlipidemia based on prescription claims data. Diabetes medications comprised biguanides, sulfonylureas, thiazolidinediones, dipeptidyl peptidase-4 (DPP-4) inhibitors, glucagon-like peptide 1 (GLP-1) analogues, and insulin. Hypertension medications included diuretics, beta-blockers, and angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACE inhibitors/ARBs). Only statins were included for hyperlipidemia. We calculated medication adherence for fixed-dose medications as the proportion of days covered (PDC).15 The PDC is the number of days for which the patient has medication available during a selected time period. Prescription medications and supplies from the 6 months prior to the study period that overlapped into the study period were included, and we used algorithms to avoid double counting pill days. Prescriptions that were filled prior to the completion of the previous prescription were assumed to start on the day following the end of the previous fill. Supplies that were still available at the end of the study period were not counted. The observation window for each patient included 90 days after the last refill, after which time the medication was assumed to be terminated. We calculated the PDC separately for biguanides, sulfonylureas,and ACE inhibitors/ARBs because these were the most commonly used medication classes. Adherence measures cannot reliably be calculated for insulin.16 We additionally identified the proportion of beneficiaries who initiated or terminated a medication during the study period.

Data Analysis: We compared the baseline characteristics of participants and non-participants using t tests for continuous variables and χ2 tests for categorical variables. We used the Wilcoxon-rank sum test for continuous variables that were not normally distributed. We also conducted a sensitivity analysis on total and out-of-pocket medical costs based on a modifi ed inclusion criteria that required that patients had at least 2 billing codes for diabetes. All analyses were completed with SAS version 9.3 (Cary, North Carolina). The project was approved by the institutional review board at Christiana Care.


Program Participation: We identified 858 beneficiaries who met our administrative defi nition for eligibility. Of these, 200 (23.3%) enrolled in the program (“participants”) and 658 did not enroll (“non-participants”). An additional 42 beneficiaries enrolled in the program but did not meet our administrative defi nition for eligibility because of age (8 years) or because they lacked administrative data and/or continuous coverage during the baseline year (34 years).

Population Characteristics: Demographic and disease characteristics of participants and non-participants appear in

Table 1

. The majority of both groups were female, with similar proportions of sex between participants and non-participants. Participants were slightly younger and much more likely to have T1DM (16.5% vs 5.6%, P <.001). Additionally, participants had a higher burden of comorbidity, with 40.5% of participants having 3 or more comorbidities compared with 29.9% of nonparticipants (v = 0.008). Participants were more likely to have complicated diabetes (46.0% vs 20.2%, P <.001) and congestive heart failure (13.0% vs 7.9%, P = .03).

Healthcare Utilization: Healthcare utilization patterns appear in

Table 2

. Participants were more likely to have at least 1 emergency department visit (23.0% vs 15.4%, P = .01) as well as to visit an endocrinologist (43.0% vs 15.5%, P <.001) or an opthalmologist (39.5% vs 31.3%, P = .03). Participants were also more likely to visit a primary care physician, cardiologist, and nephrologist, but these differences were not statistically signifi cant. Participants were more likely to have documented A1C measurements (81.5% vs 69.1%, P <.001), but were no more likely to have measurements of low-density lipoprotein (81.5% vs 78.1%, P = .30). There were no differences in the proportion of patients with at least 1 hospitalization (17.5% vs 14.7%, P = .34).

Medication Use and Adherence:

Table 3

shows the proportion of study subjects using medications for diabetes, hypertension, and hyperlipidemia. Participants were much more likely to use diabetes medications overall (66.5% vs 35.4%, P <.001) as well as every class of diabetes medications. Participants were more likely to use ACE inhibitors/ARBs (48% vs 37.7%, P = .009), but the overall proportion taking at least 1 antihypertensive medication was not statistically different between groups (53% vs 46.2%, P = .09). Participants were slightly more likely to use a statin (45% vs 37.2%, P = .048). There were no differences in adherence to sulfonylureas, biguanides, ACE/ARB, or statins.

Total and Out-of-Pocket Costs:

Table 4

displays costs during the year prior to the program. Median costs were significantly higher for participants for overall and out-of-pocket costs for medical claims, total pharmacy claims, and diabetes-related pharmacy claims (P <.001 for all comparisons). The differences in costs were identical when we restricted the analysis to the 546 non-participants (83.0%) with 2 or more encounters with a diagnosis of diabetes (results not shown).


VBID is increasingly identifi ed as a mechanism for employers and payers to encourage uptake of evidencebased healthcare services by restructuring benefits to reflect the anticipated value of a given service or intervention. There is great variety in VBID programs and the optimal design and implementation of VBID to maximize benefit has not yet emerged. Our retrospective evaluation of employees and dependents who chose to enroll in a voluntary, employer-based program that eliminated copayments for diabetes-related medications and supplies is the first to systematically examine VBID uptake in a VBID program that required participants to actively enroll. Fewer than one-fourth of eligible beneficiaries chose to complete the necessary steps to participate.Importantly, the few benefi ciaries who chose to participate were more likely to have higher baseline costs, including out-of-pocket costs, type 1 rather than type 2 diabetes, and a greater burden of comorbid illness. These findings suggest that barriers to enrollment in VBID programs influence uptake and may lead to enrollment of higher cost beneficiaries with higher burden of existing illness.

The association between nonadherence to evidencebased therapies and costly adverse events in diabetes is well supported.17-20 Therefore the evidence suggesting that VBID programs are associated with increased medication adherence and the potential for lower medical expenditures is promising.2,6,7 However, in order for VBID programs to optimize impact they must be available to those who would benefit from the high-value services that the VBID programs are designed to promote. Therefore VBID programs need to avoid barriers on entry that may limit uptake. Importantly, our program had significant requirements for participation, including that beneficiaries have bloodwork drawn, complete a questionnaire, and release medical records to their employer. We are not able to determine which individual requirement presented the strongest barrier to participation, if any, but we acknowledge that requiring participants to release medical records to an employer could be a particularly powerful barrier due to mistrust and fear, including fear that making medical data available to an employer could infl uence employment.21,22 Some VBID programs have mandated participation in disease management programs, and the relative strength of such a barrier is unknown.23

In our population we found that those who chose to participate in our VBID program were more likely to have T1DM, have complications of diabetes and comorbidities including heart failure, and have higher medical and pharmacy costs. Although participants appear to be at higher risk for complications than those who did not participate, they also appear to be already engaged in care of their diabetes. Participants were more likely to take diabetes medication, have visited an endocrinologist, and have evidence of monitoring of A1C. Participants may therefore be choosing to participate to offset costs being incurred as opposed to providing an opportunity to access under-utilized treatments or services, which is the premise of VBID. Similarly, non-participants may choose not to enroll because current costs are lower, making the program less immediately valuable to them. However, this eliminates the possibility that the program could ultimately enhance the use of evidence-based services in this population. Taken together, it appears possible that rather than increasing the use of high-value services in our population, the program may have simply shifted the burden of costs for currently engaged patients from the beneficiary to the employer.

While a VBID design should not leave in place copayments for those already practicing appropriate prevention or treatment behaviors, to realize the greatest value VBID programs must reach those for whom the barrier of medication copayments is contributing to poor glycemic control and clinical outcomes. While cost barriers are a signifi cant driver of underuse for many patients,24,25 patients who experience cost-related underuse may be more likely to also face barriers other than cost, including lack of family support,26 lack of knowledge,27 or cultural or literacy barriers.28,29 These non-financial constraints may make these patients particularly susceptible to barriers to entry to VBID programs; and it is important that VBID program design avoids such barriers.


Our findings should be understood in light of study limitations. First, our program had several requirements for participation. We do not know the specific reason why each individual chose not to participate and are unable to assess the relative importance of each component of the registration process. Second, our program was implemented in a single large regional healthcare provider, and it is not clear that employees in other industries would react similarly to program requirements. Third, our administrative defi nition of diabetes includes an intentionally broad spectrum of patients, but may identify patients without diabetes who would therefore appropriately not participate in the program regardless of programmatic barriers. A sensitivity analysis restricted the overall population with diabetes. Thus, to optimize the success of VBID programs, barriers to enrollment should be minimized.

Related Videos
Practice Pearl #1 Active Surveillance vs Treatment in Patients with NETs
© 2024 MJH Life Sciences

All rights reserved.