Medication Adherence and Enrollment in a Consumer-Driven Health Plan

February 1, 2010
Song Chen, MS

Regina A. Levin, MPH

James A. Gartner, RPh, MBA

Volume 16, Issue 2

Medication adherence was lower for some drug classes among CDHP patients who enrolled in consumer-driven health plans compared with patients who continuously enrolled in traditional managed care plans.


To evaluate the impact of enrolling in a consumer-driven health plan (CDHP) on adherence to maintenance drugs.

Study Design:

Two-year retrospective cohort study.


Consumer-driven health plan patients were enrolled in a traditional managed care plan in 2005 (pre-year) and a full-replacement CDHP in 2006 (post-year). Traditional-plan patients voluntarily enrolled in a traditional plan in both years. Adherence measures included (1) post-year continuation rate among continuous users, (2) time to refill the first prescription in the post-year, (3) change in the compliance rate from the pre- to the post-year, and (4) total days with continuous drug supply in the post-year. Analysis was conducted on 8 drug classes.


The CDHP patients had a slightly higher illness burden and used more medication in the pre-year. The continuation rate was relatively high for all drug classes, although the CDHP cohort had a lower probability of continuing cardiac and cholesterol drugs. Consumer-driven health plan patients took slightly longer on average to refill their first prescription in the post-year for cardiac, hypertension, cholesterol, and thyroid drugs. The compliance rate dropped over time in both cohorts, but the reduction was bigger among CDHP patients for 3 drug classes (adjusted ratio of the odds ratios was 0.77, 0.78, and 0.69 for asthma, cardiac, and cholesterol drugs, respectively). The CDHP patients also terminated their continuous drug supply 21 days earlier for epilepsy drugs and 27 days earlier for cholesterol drugs.


Adherence was lower for a few drug classes among CDHP patients.

(Am J Manag Care. 2010;16(2):e43-e50)

The impact on adherence to maintenance drugs of enrollment in a consumer-driven health plan (CDHP) was assessed in a 2-year retrospective cohort study.

  • The compliance rate dropped over time in both the CDHP and traditional managed care cohorts, but the reduction was bigger among CDHP patients for asthma, cardiac, and cholesterol drugs.
  • The CDHP patients also terminated their continuous drug 21 days earlier for epilepsy drugs and 27 days earlier for cholesterol drugs.

Poor adherence to drug regimens is a costly problem. It increases the risk for negative health outcomes as well as increased healthcare costs.1-3 Yet nonadherence is common; approximately 40% to 60% of patients do not take medications as prescribed.4-10 High-deductible health plans have the potential to discourage drug adherence by placing additional financial burdens on members. In these plans, members may face larger out-of-pocket expenses at the beginning of the plan year in the form of higher deductibles, which could contribute to delays in refilling prescriptions or stopping medication altogether. Using enrollment and claims data from a national insurer, this retrospective cohort study tested whether patients who enrolled in a consumer-driven health plan (CDHP) that fully replaced a traditional plan became less adherent to medications for chronic conditions compared with those who were continuously enrolled in a traditional managed care plan.


During the past 5 years, CDHPs have experienced fast growth. It has been estimated that in 2008 there were 10 million to 12 million CDHP enrollees nationally.11 The study insurer’s CDHP membership was about 750,000 in 2005. Membership almost doubled in 2006 and exceeded 3 million in 2009. Initially a CDHP was most commonly offered as an option alongside traditional plans, but recently an increasing number of employers have fully replaced their traditional plans with CDHPs. This insurer experienced a 67% increase between 2006 and 2008 in the number of employers that offered full-replacement CDHPs. Only fullreplacement CDHPs were used in this study to eliminate the selection bias that might have occurred with optional CDHPs.

Consumer-driven health plans typically consist of a personal care account and a deductible (usually $3000-$4000). Under a basic CDHP model, members receive an annual allocation of money into an account and can use this money to pay for medical and pharmacy services. Unused funds can be rolled over to the next year and be added to the next year’s fund deposit. Once the account is exhausted, members are financially responsible for the medical and pharmacy services until a deductible is met. After the deductible is met, members are partially financially responsible for the services until the annual out-ofpocket maximum is met. There are 2 types of accounts: health reimbursement accounts (HRAs) and health savings accounts (HSAs). The primary differences are that (1) both the employer and the member can contribute to the HSA, whereas only the employer can contribute to the HRA; and (2) the employee owns the HSA, whereas the employer owns the HRA (ie, after discontinuing employment, the employee loses the money in the HRA account). The HRA and the HSA have had roughly equal market penetration in recent years.11

Prescription drugs are paid for differently in CDHPs than in traditional managed care plans. In CDHPs, members pay the full amount for which the insurer has contracted with the pharmacy until members have satisfied the deductible. Once the deductible is satisfied, members are responsible for a copayment. In traditional managed care plans, members are responsible for prescription copayments regardless of whether the deductible is satisfied. (In some cases plans will have separate deductibles for the pharmacy benefit.) Typically, traditional plans use a tiered pharmacy design with different copayment tiers for brand versus generic drugs.

At the time this study was conducted, research on medication adherence in CDHPs was limited and the results were mixed. Some researchers reported increased use of maintenance medications and decreased pharmaceutical expenditures among CDHP enrollees.12 Some researchers reported that CDHP enrollees and traditional-plan enrollees had similar rates of brand utilization,13 generic utilization,14 and medication adherence.14,15 Yet other studies found that CDHP members tended to use fewer prescription drugs,16 were less compliant with drug regimens,17,18 were more likely to discontinue chronic illness medications,14 and were more likely to skimp on needed medications because of cost.19 These studies have limited generalizability with respect to medication adherence because they examined the experience of only 1 or 2 employers, they measured general medication utilization rather than adherence, or they looked at cross-sectional differences without controlling for baseline differences.


Setting and Design

The study insurer substantially increased membership in CDHP in 2006. The number of contracted employers increased by 300% to 14,500 and the total number of enrollees doubled to 1.5 million. That year 55% of the employers fully replaced the traditional plan with a CDHP. These full-replacement employers represented 35% of the new member population.

This retrospective cohort study identified a CDHP cohort from 33 employers and a traditionalplan cohort from 47 employers. All employers were middle to large size (≥100 employees). These employers offered traditional plans that had been provided by the same insurer in 2005 and started offering CDHPs to employees in 2006. Both self-insured and fully insured employers were included. Employers in the CDHP cohort offered CDHP as the only choice (full-replacement plan), whereas those in the traditional-plan cohort offered a traditional plan as an option in addition to a CDHP. The traditional plans included preferred provider organization, point-of-service, and exclusive provider organization plans. The CDHPs included HRAs and HSAs. Preliminary examination of the data showed that HSA and HRA members had similar adherence patterns. The HRA and HSA members were combined in the CDHP cohort; they will be analyzed separately in a subsequent study. Tiered pharmacy designs were used in all traditional plans.

Study Subjects

The unit of analysis is per patient per drug class. Patients were eligible for the CDHP cohort if they enrolled in a traditional managed care plan in 2005 (pre-year) and a fullreplacement CDHP in 2006 (post-year). Those who were eligible for the traditional-plan cohort voluntarily enrolled in a traditional plan in 2005 and 2006. Other inclusion criteria included continuous enrollment, under age 64 years on the last day of the 24-month study frame between January 1, 2005, and December 31, 2006, both pharmacy and medical benefit coverage, and no other source of health insurance. (All but 2 employers had January-start coverage, for whom the study frame was from January 1, 2005, to December 31, 2006. The study frame for the 2 non—January-start employers also was 24 months but was shifted to the applicable starting month.) Patients who used insurance from another carrier were excluded from this study due to the unavailability of insurance claims.

For a given drug class, a patient was identified if he or she had at least 1 prescription in that drug class during the first quarter of the pre-year and at least 1 refill during the last quarter of the pre-year. Those who did not have a script in the preyear were excluded from the analysis due to the unavailability of baseline prescribing experience. Eight therapeutic classes for conditions including asthma, cardiac, diabetes, epilepsy, hypertension, cholesterol, rheumatoid arthritis, and thyroid were selected because the drugs in these classes are rarely used to treat other conditions and should be taken continuously. Patients were allowed to switch drugs as long as they continued with a drug in the same class. Specified Therapeutic Class code, a drug classification system developed by the First DataBank, was used to define drug classes. Study subjects who had prescriptions in multiple drug classes were treated as separate individuals.

Risk Assessment

Using enrollment data and medical and pharmacy claims, patients’ health risk scores in the pre-year were computed and used in multivariate modeling. Health risk score measures the relative resources that are expected to be required for healthcare. Health risk was assessed with Episode Risk Groups, a derivative work based on Episode Treatment Group methodology, a widely used software product for illness classification and episode building (a product of Ingenix, a subsidiary of UnitedHealth Group). The literature reports that Episode Risk Group scores highly correlate with other risk-adjusted measures of practice efficiency such as Adjusted Clinical Groups, Burden of Illness Score, Clinical Complexity Index, Diagnostic Cost Groups, and General Diagnostic Groups.20

Outcome Measures

Appendix A

Adherence with regimens was measured from 4 perspectives. Continuation rate measured the proportion of eligible patients who had a positive number of prescription(s) in the post-year, when there was evidence that these patients had a minimum of 2 refills for a drug in that class in the pre-year. For patients who did fill prescriptions in the post-year, time to fill the first prescription measured the number of days between the last date of drug supply for the last refill of the pre-year and the beginning date of the first filled script of the post-year. Compliance rate measured the proportion of patients who had a medication possession ratio (MPR) greater than 0.8 (see ). Time with continuous drug supply measured the number of days with continuous drug supply until discontinuation occurred or until the end of the post-year, whichever occurred first. A discontinuation was identified when the gap between 2 consecutive refills was longer than a predefined grace period (see Appendix B).

Statistical Analysis

Descriptive statistics were used for demographic variables, the t test for continuous variables such as age and baseline year risk score, and the X2 test for sex distribution. To test whether patients in the 2 cohorts were equally likely to continue refilling behavior in the post-year, a logistic regression model adjusting for demographic characteristics, the pre-year risk score, and the pre-year total days of drug supply was applied. An analysis-of-variance model adjusting for demographic characteristics and risk score compared whether patients in the 2 cohorts took an equal amount of time to refill the first prescription in the post-year.

A difference-in-difference logistic regression model was used to evaluate whether the changes in the compliance rate from the pre-year to the post-year were the same for CDHP and traditional-plan enrollees. In this model, the coefficient of the interaction between cohort and year represented the change in the odds ratios between cohorts in a natural logarithm. This method had the advantage of evaluating the true cohort effect by controlling for the time effect. The odds ratios of the pre-year compliance rate over the post-year compliance rate for each cohort and the estimated ratio of the 2 odds ratios were reported.

To determine whether enrollees in the 2 plan types were equally likely to discontinue in the post-year, a Cox proportional hazard model adjusting for demographic characteristics, the pre-year risk score, and the pre-year total days with continuous drug supply was used. Patients who had a continuous drug supply through the end of the post-year were considered censored observations. The proportional hazard assumption was verified by using Kaplan-Meier survival curves. The adjusted hazard ratio of discontinuation for CDHP enrollees was reported using traditional-plan enrollees as the reference group.


Descriptive Analysis

Table 1

summarizes the demographic characteristics, pre-year risk level, and general drug utilization for unique patients who had filled prescriptions for the chronic condition (n = 33,393). Compared with patients in the traditional-plan cohort (n = 30,455), those in the CDHP cohort (n = 2938) had a slightly higher risk score in the pre-year (2.7 vs 2.6, P <.05), a higher proportion of females (55% vs 52%, P <.01), and higher drug utilization for all drug classes together in the pre-year (374 days vs 365 days of supply, P <.001; 9.8 vs 9.4 scripts, P <.001). Total days of supply decreased in the post-year in both cohorts, and the drop was larger for CDHP patients (29 fewer days in the CDHP cohort vs 14 fewer days in the traditional-plan cohort, P <.001). The total number of scripts also decreased (0.8 fewer scripts in the CDHP cohort vs 0.6 fewer scripts in the traditional-plan cohort, P <.001). The average age and proportion of patients who filled prescriptions in more than 1 drug class were approximately equal in the 2 cohorts.

Continuation to the Post-Year

Table 2

The continuation rate was relatively high for all drug classes (). Results from the multivariate logistic regression model suggested that CDHP patients were equally likely to continue their prescriptions in the post-year for 6 conditions and were less likely to continue for cardiac and cholesterol drugs, although the differences were small (cardiac: 97% vs 98%, P <.05; cholesterol: 96% vs 98%, P <.01).

Number of Days to the First Refill in the Post-Year

Table 2

Among those who did continue in the post-year, the average time between the last refill of the pre-year and the first refill of the post-year ranged from 1 week to 1 month (). This gap was slightly longer for CDHP patients for cardiac (4 days more), hypertension (4 days more), cholesterol (5 days more), and thyroid (3 days more) drugs. The gap also was longer for CDHP patients taking diabetes, epilepsy, and RA drugs, although the difference was not satistically significant.

Medication Possession Ratio

Table 3

The MPR dropped slightly in the post-year (−0.05 in CDHP and −0.03 in traditional plans, P <.001), although it should be noted that the change of MPR was big for a small group of patients. In both plans, the bottom 5% of patients had a more than 50% drop in MPR and the top 5% had a more than 50% increase in MPR (). Table 3 also suggests that the CDHP cohort overlapped highly with the traditional-plan cohort on both the original MPR and the change in MPR.

Compliance Rate

Table 4

The dichotomous measure for compliance was derived from the MPR. For a given drug class, a patient was compliant if his/her MPR was greater than 0.8. Among patients who filled a prescription in the post-year, compliance became worse in the post-year compared with the pre-year for all drug classes (). If the adjusted ratio of the odds ratios (the comparison of the compliance reduction from the pre-year to the post-year between the 2 cohorts) was less than 1, it indicated that the reduction was larger for the CDHP cohort. The CDHP population had a significantly greater reduction in compliance for 3 drug classes. The adjusted ratio of the odds ratios was 0.77 for asthma drugs (P <.05), 0.78 for cardiac drugs (P <.05), and 0.69 for cholesterol drugs (P <.001). Cholesterol drugs showed the greatest reduction in compliance and the greatest difference between CDHP and traditionalplan patients. For this drug class, compliance rates for CDHP members dropped from 77% to 62%, compared with a change from 75% to 68% for patients who remained in a traditional managed care plan.

Regardless of the type of health plan in the post-year, the compliance rate was relatively low for rheumatoid arthritis drugs in both the pre-year and the post-year, which suggests that patients with rheumatoid disease might not receive sufficient treatment.

Number of Days With Continuous Drug Supply

Table 5

Among patients who continued in the post-year, the total time with continuous drug supply was relatively long for both study groups, ranging from 7 months for rheumatoid arthritis medications to 9 months for thyroid medications (). A test on the likelihood of supply termination suggests that CDHP patients were more likely than traditional-plan enrollees to discontinue refills for 2 drug classes: epilepsy (21 days earlier, hazard ratio = 1.25, P <.05) and cholesterol (27 days earlier, hazard ratio = 1.25, P <.001).


There has been increasing concern that patients may discontinue taking their maintenance drugs for chronic condintions if they are required to pay more for them. Unlike members with a multitier pharmacy design who pay a portion of the full price through coinsurance or copayment, those in CDHPs pay the full price for drugs until the deductible is met. Maintaining control of chronic illnesses can delay or prevent complications and can therefore lead to savings on future healthcare.

This study tested the hypothesis that enrolling in an account-based high-deductible CDHP leads to adverse prescription drug refilling behavior. Using 2 years of claims data from 80 employers, this analysis compared patients who enrolled in a full-replacement CDHP with those who voluntarily were continuously enrolled in a traditional managed care plan. Eight drug classes were studied, and patients were identified if they had at least 2 scripts in the pre-year for a given drug class. The utilization for all drug classes together decreased in both cohorts, and the decrease was slightly more in the CDHP cohort.

Four aspects of regimen adherence were examined in the post-year: overall medication continuation rate, time to refill the first prescription, the change in compliance rate from the pre-year to the post-year, and the number of days with continuous drug supply. The CDHP patients were equally likely to continue their prescriptions for 6 conditions but were less likely to refill cardiac and cholesterol drugs. The CDHP patients took slightly longer to resume their first prescriptions in the post-year for all therapeutic classes except for patients taking asthma drugs. The CDHP patients also had poorer compliance for asthma, cardiac, and cholesterol medications. In addition, CDHP patients terminated their continuous drug supply earlier than traditional-plan patients. The greatest concern, however, is poor compliance with cholesterol medications. Adherence was consistently and significantly lower for CDHP patients taking medication for hypercholesterolemia by all measures.

These results are slightly different from those of a recently published study that found no difference in the decline in compliance between the high-deductible plan and the traditional plan for all 5 drug classes studied14 and another study that found no difference in medication persistence for all 3 drug classes studied (this paper is under review; the author information is not disclosed).


This study did not evaluate the impact of specific pharmacy benefits. Benefit tiers and the amount of copayment or coinsurance associated with each tier were not examined; nor was the specific benefit design at the employer level, such as insurance premium or amount of contribution made to the healthcare account. Such information was not available to the researchers at the time the study was conducted. The lack of this information, however, should not change the conclusions of this study. Although the available demographic characteristics were similar for the 2 study cohorts, the baseline medication utilization was lower among members who chose to continuously enroll in a traditional plan, indicating that there may be unmeasured differences between the populations.

Author Affiliations: From United Health Group (SC), Hillsborough, NJ; United Health Group (RAL), Sewickly, PA; and Ingenix (JAG), Eden Prairie, MN.

Funding Source: None reported.

Author Disclosure: Ms Chen and Ms Levin are employees of United Health Group and report owning stock in the company. Mr Gartner is an employee of Ingenix, a research unit of United Health Group.

Authorship Information: Concept and design (SC, RAL, JAG); acquisition of data (SC); analysis and interpretation of data (SC, RAL, JAG); drafting of the manuscript (SC, RAL, JAG); critical revision of the manuscript for important intellectual content (SC, RAL, JAG); statistical analysis (SC, RAL); administrative, technical, or logistic support (RAL); and supervision (RAL).

Address correspondence to: Song Chen, MS, United Health Group, 15 Runyon St, Hillsborough, NJ 08844. E-mail:

1. Johnson JA, Bootman JL. Drug-related morbidity and mortality. A cost-of-illness model. Arch Intern Med. 1995;155(18):1949-1956.

2. Noncompliance With Medications: An Economic Tragedy With Important Implications for Health Care Reform. Baltimore, MD: The Task Force for Compliance; April 1994:1-32.

3. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43(6):521-530.

4. Nichol MB, Venturini F, Sung JC. A critical evaluation of the methodology of the literature on medication compliance. Ann Pharmacother. 1999;33(5):531-540.

5. Ryan AA. Medication compliance and older people: a review of the literature. Int J Nurs Stud. 1999;36(2):153-162.

6. Haynes RB, McKibbon KA, Kanani R. Systematic review of randomised trials of interventions to assist patients to follow prescriptions for medications [published correction appears in Lancet. 1997;349(9059):1180]. Lancet. 1996;348(9024):383-386.

7. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353(5):487-497.

8. Fedder DO. Managing medication and compliance: physicianpharmacist- patient interactions. J Am Geriatr Soc. 1982;30(11


9. Sabaté E, for the World Health Organization. Adherence to Long-Term Therapies: Evidence for Action. Geneva, Switzerland: World Health Organization; 2003.

10. National Community Pharmacists Association (NCPA) and Pharmacists for the Protection of Patient Care (P3C). Take as directed: a prescription not followed [press release]. December 15, 2006. Accessed January 18, 2010.

11. Fronstin P. Findings from the 2008 EBRI consumer engagement in health care survey. EBRI Issue Brief. 2008;(323):1-42.

12. CIGNA HealthCare. CIGNA Choice Fund® Experience Study. October 2007. Accessed June 3, 2008.

13. Parente ST, Feldman R, Chen S. Effects of a consumer driven health plan on pharmaceutical spending and utilization. Health Serv Res. May 13, 2008. Epub ahead of print.

14. Greene J, Hibbard J, Murray JF, Teutsch SM, Berger ML. The impact of consumer-directed health plans on prescription drug use. Health Aff (Millwood). 2008;27(3):1111-1119.

15. HealthPartners. HealthPartners Consumer Directed Health Plans Analysis. October 2007.

Accessed June 3, 2008.

16. Wilson AR, Bargman EP, Pederson D, et al. More preventive care, and fewer emergency room visits and prescription drugs—health care utilization in a consumer-driven health plan. Benefits Q. 2008;24(1):46-54.

17. Fairman KA, Sundar H, Cox E. What Happens to Prescription-Drug Use After Consumer Directed Health Plan Enrollment? Express Scripts. April 2007. Accessed June 3, 2008.

18. Wang W, Khalid M. Chopra-Mehta S, Sungela G, Aubert R. Impact of consumer directed health plan selection on medication adherence. Abstract presented at Academy Health, June 2008; Washington, DC.

19. Fronstin P, Collins SR. Findings From the 2007 EBRI/Commonwealth Fund Consumerism in Health Survey. Employee Benefit Research Institute Issue Brief No. 315. March 2008. Accessed June 25, 2008.

20. Thomas JW, Grazier KL, Ward K. Economic profiling of primary care physicians: consistency among risk-adjusted measures. Health Serv Res. 2004;39(4 pt 1):985-1003.