Employers feel increasing pressure to address rising health care costs. One option to help reduce employer costs is to shift from a defined health care benefit, in which the employer provides and subsidizes one or more health plans, to a consumer-directed health plan (CDHP), in which the employer provides a defined payment linked to one plan option, and the employee selects a health plan, either paying any incremental premium difference or receiving credit for a lower-priced option (Bureau of National Affairs 2001). In theory, a CDHP model of health benefits encourages greater employee accountability, offers more flexibility in plan design options, and gives employees greater choice (Employee Benefit Research Institute 2003). It may also reduce cost growth (Nichols 2002).
Many types of CDHP options are emerging. The designs vary in the degree of employee responsibility, from health plans at one extreme that are Internet-based, in which the employees construct their own panel of care providers, to personal care accounts with a high deductible, to traditional plan choices in which only the financing method is changed (Christianson, Parente, and Taylor 2002; Robinson 2002; Jacob 2001). Although these various CDHP options have received extensive publicity, we know little about employees' responses to them (Kelly 2003; Halterman, Camero, and Maillet 2003; Reinhardt 2001).
In June 2001, Humana Inc. offered a new health care benefit program for the nearly 5,000 employees in its corporate headquarters in Louisville, Kentucky. Humana's rationale for the change of health care benefit coverage was three-fold: to provide employees with a greater choice of plans, to give them greater financial responsibility for their choice, and to contain costs to the employer. This new benefit structure had a CDHP design in which the corporation paid a fixed amount--79 percent of the reference plan. The reference plan was a preferred provider organization (PPO), the most popular health plan option with the highest premium. Employees could apply the corporate contribution to one of six health care options, keeping the difference if they selected an option other than the PPO. All the other options had less expensive premiums than the PPO. (1)
The two CDHP plan options were similar to health reimbursement arrangements (HRAs) (Gabel, Lo Sasso, and Rice 2002). Gabel defines health reimbursement accounts as plans that "establish an account from which consumers draw to make health care purchases. When the account is exhausted, enrollees must typically pay out of pocket until the annual deductible is met, after which the plan becomes a traditional major medical plan." One of these CDHP options provided an allowance of $500, then 80 percent coinsurance until $2,000 in further out-of-pocket charges were incurred, and finally 100 percent coinsurance. The second CDHP option was similar to the previous one with a $500 allowance, then a $2,000 deductible, and finally 100 percent coinsurance. These options were offered in lieu of HRAs because the tax-sheltered status of HRAs was unclear when the plans were being designed and implemented.
The provider networks overlapped widely across these options. The HMO Plan had the most restrictive network and was also used as the first tier of the Tiered PPO, PPO Standard, and the two CDHPs. Although the enrollment process was supported with web-based information and decision-support tools, there was no ongoing Internet support to monitor expenses or evaluate care choices for employees who enrolled in the CDHP options.
All health care coverage options covered the same benefits, including pharmacy benefits. Concomitant with the change in structure, however, were two significant changes in benefits. The pharmacy benefit was restructured from a three-tier to a four-tier program (Tier 1: $10 copayment: included lower-cost generic drugs and some brand name drugs; Tier 2:$20 copayment: included higher-cost generic drugs and some brand name drugs; Tier 3:$40 copayment: included higher-cost, mostly brand-name drugs that may have generic or therapeutic equivalents in Tier 1 and 2; and Tier 4:25 percent coinsurance for high-technology drugs with a $2,500 out-of-pocket maximum). The other major benefit change was the addition of a $100 per day inpatient hospital copayment for both the Tiered PPO and HMO options.
THE ENROLLMENT PROCESS
Employees had no systematic comparative information on the quality of the options, such as a report card. For the first time, they had access to an online decision support tool that queried the employee about their coverage needs and preferences. This tool then ranked the plan options according to the employee's responses.
The enrollment design originally called for all employees to enroll electronically (positive enrollment). The design was revised, however, to include a default option, in which employees who did not enroll online were assigned to the new plan option most similar to their previous plan option. Employees could also decline coverage.
The designers of the health plan options had estimated that 5 percent to 10 percent of employees would select one of the two new options that would save them $15 per pay period ($400 per year) for employee-only coverage and upward of $45 per pay period for family coverage (approximately $1,200 per year). (The employee plus spouse rate was roughly 2 times the employee-only rate; the family rate was 3.2 times; and the employee plus child[ten] rate was 1.9 times the employee-only rate.)
EVALUATION QUESTIONS
The evaluation focused on two questions: (1) How did employees who chose the CDHP options compare with those who did not? (2) Which employee characteristics were related to their perceived ratings of the enrollment process?
OUTCOMES OF INTEREST
The primary outcome was the employee's self-reported choice of health plan option, specifically selection of either of the two CDHP options. For comparative analyses, employees were combined into two categories: (1) employees who selected the CDHP options, and (2) employees who selected any of the other plan options. The four secondary outcomes related to satisfaction with the enrollment process included: helpfulness of corporate communications, adequacy of time to review materials and enroll, understandability of benefits information, and ease of finding needed information.
DATA AND METHODS
We used a cross-sectional study design and surveyed all benefit-eligible employees (N = 4,680) immediately after the open enrollment period, excluding those who helped with the questionnaire. The survey content areas covered: sociodemographic characteristics, health insurance coverage, health care utilization, importance of plan characteristics for plan choice, health information-seeking behavior, the employee's relationship with a primary care physician, and the employee's perceptions of the new online benefits information and enrollment process. Relevant questionnaire items that had been tested and used in previous surveys (Braun et al. 2003; Fowles et al. 2000; Knutson et al. 1998) were incorporated into the questionnaire. New items were pretested by cognitive testing of the questionnaire with a convenience sample of five Humana employees by telephone. The authors developed the questionnaire with advice from Humana project staff in the human resources department. The study protocol and questionnaire were reviewed by an Institutional Review Board for the protection of human subjects.
The survey was conducted between July 27 and October 1, 2001. Following an initial letter from the Humana chief executive officer alerting employees to the forthcoming survey, the evaluator mailed the survey and followed it with a postcard reminder and two additional complete mailings to nonrespondents.
The Plan Choice Model The model predicts that plan choice will be dependent on four domains: sociodemographic characteristics (including coverage type), health status, previous and anticipated health care utilization (including relationship with primary care physician), and the perceived importance of various plan characteristics. As described by Scanlon and colleagues in their review of health plan choice (Scanlon, Chernew, and Lave 1997), we used logistical regression analysis to model dichotomous plan choice. We tested for collinearity among the health status variables and found none. The results, using a phi coefficient as a measure of correlation between dichotomous variables (Fleiss 1981), can be found in Appendix 1.
