Several health promotion and disease prevention programs designed for older adults have been developed and evaluated for their health benefits and resource use (1-5). These programs focus on improving older adults' general health and encouraging self-management of chronic medical conditions. Specific aspects of health improvement, such as improving mental well-being or increasing physical activity, are often the focus of health improvement and are pursued because of a client's interest and motivation. These programs connect clients with information and resources that help them address their personal health concerns, build confidence in health care decision making, and increase physical activity. Such health promotion programs for older adults improve health outcomes, and they have demonstrated decreased use of health care resources resulting from participation, which results in decreased health care costs (2,3). However, to our knowledge these studies used self-reported data rather than actual health care costs.

EnhanceWellness (EW, formerly known as Health Enhancement Program, or HEP) targets older people at risk for functional decline. Nurses and social workers meet with community-living elders to help increase physical activity, promote social activity, improve mental health, and enhance self-management of chronic conditions to improve health and functioning. In a 1998 randomized controlled trial, EW participants increased their physical activity, decreased their use of psychoactive medications, and decreased their number of hospital days (2). After that study, senior centers in the Seattle, Washington, area began implementing EW, and enrollees of Group Health Cooperative (GHC), a consumer-governed, nonprofit health care system that provides both health care and medical coverage, started participating in the program. A follow-up study conducted in 2002 evaluated the program as it operated in the community, outside the controlled setting of a randomized trial (5). This study also demonstrated significant benefits, including a reduction in disability risk factors, improvement in health status, no decrease in functional status, and no increase in self-reported health care use.

EW has been confused with the EnhanceFitness Program (EF) because of their similar names and the fact that both have been studied in a similar older adult population (6,7) However, the programs are distinct: EF is a group exercise program, whereas EW is a comprehensive, participant-centered wellness program that includes a health assessment, a tailored health plan, and motivational support to achieve a self-chosen goal. EW participants, if desired, may include regular physical activity and join EF, a covered benefit for GHC members. Less than 10% of GHC members typically participate in both programs, although not necessarily simultaneously (M. Thompson, oral communication, December 2008).

Although health benefits and a reduction in hospital days have been demonstrated, EW's effect on health care use and costs has not been previously analyzed. The availability of comprehensive cost and use data for GHC members made studying these questions with GHC members who had participated in EW attractive. We hypothesized that participation in EW would lower overall health care costs, via reductions in costly forms of health care use (especially hospitalizations).

Most EW participants (88%) spent 6 months or longer in the program; approximately 50% spent 12 or more months, and 20% were in the program for more than 2 years. EW participants were identical to nonparticipants in terms of average age (79 y) and sex (72% female) (Table 1). We noted several significant differences between the groups, including a larger comorbidity burden among EW participants, as measured by a higher Charlson comorbidity index and a larger proportion enrolled in the GHC diabetes and heart disease registries. The preventive services score was significantly higher for EW participants, suggesting a stronger tendency to receive preventive services.

Total costs, inpatient costs, percentage hospitalized, and number of hospital days were not significantly different between participants and nonparticipants, either at baseline or in the year after the index date (Table 2). The only significant difference was in unadjusted primary care costs, which were higher by $325 in the EW group (P < .001) at baseline and $177 higher in the year after the index date (P = .04). After adjusting for age, sex, prior year total costs, preventive services score, Charlson comorbidity index, and presence on the GHC diabetes or heart disease registries, total health care costs in the year after the index date were $582 lower for EW participants than for nonparticipants, but this difference was not significant. The results were unchanged when we used log-transformed costs. There were no differences in inpatient use or primary care use between the 2 groups at baseline or the year after the index date.

We found that, compared with nonparticipants, EW participants had nonsignificantly lower total health care costs and no difference in hospitalizations during the year following EW enrollment. This finding may have resulted from the fact that EW participants in our sample had a significantly larger comorbidity burden than did nonparticipants. Comorbidity is a major driver of hospital costs and total annual costs (18,19). Furthermore, the methods we used to adjust for comorbidity, although widely used, may not have allowed us to fully control for comorbidity differences between study groups (10,11,13)

Many health promotion programs, some designed for the older adult population, have been associated with decreased health risks and decreased health care use (1-3,20-23). Health promotion programs evaluated by Lorig et al and Holland et al most closely resemble the EW program (3,4). These studies evaluated health outcomes and health care use, but neither assessed health care costs. Lorig et al found a significant decrease in hospitalizations and hospital days during their 6-month randomized controlled trial. The average age in this study was 10 years younger than in ours, and the 2 study groups had balanced comorbidities. Conversely, Holland et al did not find a difference in health care use between study groups during their year-long randomized controlled trial of the Health Matters Program in Sacramento, California, a program modeled after EW (4,24). Similar to our analysis, the mean age of participants in Holland's study was 73 years.

There are several differences between our analysis and the original randomized controlled trial that was used to evaluate EW (2). The original trial lasted 12 months, and outcomes were evaluated for the 12 months of program enrollment. After the original trial, EW evolved into a 6-month program, so our analyses included EW participants with varying duration of program participation. For most EW participants, 6 months of EW participation confers favorable effects on disability risk factors (eg, depression, physical inactivity) that are comparable to 12 months of participation (25). However, such reductions in disability risk factors do not appear to translate into lower overall health care costs.

A strength of our analysis was that we reported actual health care costs. To our knowledge, no other analysis of a health-promotion, disability-prevention health resource has used actual cost data. Health care use is often used as a proxy for costs, or alternatively, costs are estimated from claims data (1,22). Furthermore, our health care use data were derived from automated data sources, which are more accurate in assessing health care use and costs than are self-reported data (26,27).

Our study has several limitations. Our study had an observational design, which can result in residual confounding and selection bias. Residual confounding in relation to the comorbidity differences we observed between study groups is likely, although we attempted to adjust for them. Research published after our study ended demonstrated that total annual costs increase with increasing comorbidity and that 4 conditions — hypertension, depression, use of warfarin, and skin ulcers/cellulitis — should be added to the Charlson comorbidity index to accurately predict total annual costs (19). We used the preventive services score to address selection bias related to the potential tendency of more prevention-oriented people to participate in EW. This score has been used in prior research with GHC members but may not have fully accounted for this form of selection bias (14). We also considered using propensity scores to adjust for selection bias but lacked enough covariates to independently predict program participation.

Another limitation was a lack of detail on health care use. In particular, we could not distinguish increased use that may have been prompted by participation in EW (eg, more visits related to health problems identified by EW). Also, apart from EF, we had no information about exercise and other health promotion programs that EW participants may have pursued as a result of their participation in EW. Finally, our sample size of just over 200 EW participants may have been too small to detect meaningful cost differences given the large variances associated with health care cost and use data.

EW improves the health of older adults at risk for functional decline (28). However, we did not find that overall health care costs were significantly reduced by EW program participation.