The priority population for this study was Hispanic adults aged 18 years or older with type 2 diabetes who were patients at the Mercy Clinic in Laredo, Texas, from October 2009 to January 2010. Laredo is in south Texas on the Mexico border, in Webb County. About 96% of the population of Webb County is of Hispanic origin, and median household income is $37,140 (19). Almost half of adults aged 18 to 64 in the Laredo metropolitan statistical area (MSA) lack health insurance coverage (20).

Analysis inclusion prerequisites were a diagnosis of type 2 diabetes, availability of baseline and follow-up A1c readings, and evidence of program participation. Forty-six participants met these criteria. Of the 46, 16 had baseline A1c levels at or below 7%, indicating good glycemic control, and were therefore excluded from the study group. The sample analyzed consists of the remaining 30 participants.

The University of Texas Community Outreach (UTCO) intervention is an ongoing community-based diabetes education and self-management program that uses community partnerships and trained CHWs to reach participants. Four Texas counties, including Webb County, host UTCO programs. We used first-wave evaluation data (follow-up data collected through January 15, 2011) from the Webb County UTCO program for this analysis.

Mercy Clinic, the primary Webb County partner in the UTCO program, provides primary health care, health education, and social services to the financially disadvantaged and medically underserved population. Intervention components include home-based CHW visits, classroom health education classes, nutrition classes, exercise classes, and counseling sessions. The nutrition classes teach new diabetes management skills. Exercise classes provide venues and social support for physical activity. Counseling sessions provide targeted guidance on individual issues participants have in managing their diabetes. Home visits provide reinforcement of themes covered in classes and support in creating cues to action. CHWs bring schedules of upcoming classes to home visits, provide guidance on which sessions participants should give priority to, and assist participants in overcoming barriers related to effective diabetes management.

Five state-certified CHWs and a nurse educator are the primary UTCO service providers. Four of the 5 CHWs were previously employed by Mercy Clinic, and external funding of the UTCO program covered the project director’s salary and enabled hiring of a fifth CHW and a nurse practitioner. CHWs received training through the UT Health Science Center School of Public Health, Brownsville Regional Campus, on leading nutrition and exercise classes and on data-gathering protocols. All CHWs, who are called promotoras in the Hispanic community, are certified through the Texas Department of Health and Human Services’ CHW training program. The CHWs’ primary role is to develop relationships with participants to provide individualized guidance, one-on-one diabetes education, and counseling aimed at increasing participants’ ability and self-efficacy to manage their diabetes. Additionally, CHWs and the nurse educator teach participants in group environments how to manage their diabetes, improve their nutrition, and lose weight. Several volunteers, including a dietitian, a certified Zumba instructor, and students from Texas A&M International University (TAMIU), assist with teaching nutrition classes, leading exercise classes, and facilitating group counseling sessions.

To be included in the analysis, participants had to have received at least 1 home visit in addition to having a baseline A1c reading above 7.0. We set these criteria for several reasons 1): many people attended classes only, and our primary focus was the combination of CHWs in clinical (classes) and community (home) settings; 2) home visits are the most expensive component of the intervention, and we wanted to include these costs in our analysis; 3) we hypothesized that these in-home activities would be key components to the program’s success and to sustained outcomes; and 4) the agreement between Mercy Clinic and participants included attendance at both classes and home visits. All participants attended nutrition and health education classes (mean number of classes attended, 8.3); 77% attended exercise classes (mean number of classes attended, 4.2); and 33% attended counseling sessions (mean number of sessions attended, 4.3). Follow-up A1c readings were taken periodically, ranging from 37 to 565 days (mean, 75 days) for the sample.

We assessed the cost-effectiveness of the UTCO intervention in this location from a societal perspective. The societal perspective includes all measurable opportunity costs, representing all groups affected by a program. To assess program costs, we included staff time and valuation of volunteer and participant time for home visits, educational classes, counseling sessions, exercise classes, class-related materials, and mileage related to home visits.

Mercy Clinic paid staff salaries and for program supplies. We used the Bureau of Labor Statistics Metropolitan and Non-Metropolitan Area Occupational Employment and Wage Estimates for the Laredo MSA (21) to value volunteer and participant time. Where volunteer occupation was known (eg, dietitian), we used the mean wage for that occupation, and we included the value of fringe benefits and taxes for clinic staff and volunteers. We used the Texas minimum wage ($7.25 hourly) to value student volunteer time and the average hourly wage rate for the Laredo MSA ($16.14) to value the time of study participants. This average rate was likely higher than what study participants were earning in their jobs. We added only Federal Insurance Contributions Act (FICA) taxes to participant base wages. To calculate the average cost per home visit, we used information gathered from CHW interviews about the typical length of a home visit, the time needed to schedule visits, and travel time to and from appointments. Estimated average mileage expense was assessed at $0.51 per mile. Initial home visits took, on average, twice as long as subsequent visits but involved the same amount of scheduling and travel time; we factored this into total home visit expenses and assigned a value of $80.59 for initial home visits and $48.16 for follow-up visits.

We calculated a composite fixed cost per class from information gathered from Mercy Clinic records on time and resources needed to deliver each type of class. We weighted the cost for each type of education class by the frequency with which the class was offered and summed them to derive an average class cost. We similarly developed composite per-session fixed costs for exercise classes and counseling sessions. Total costs for each activity category were the sum of the fixed costs of the classes and aggregate attendee time costs. This composite fixed cost was multiplied by the total number of classes or sessions; to this product we added the average attendance and the per-participant time cost of attendance to derive the time costs of attendees. Ten percent of the total CHW training expense was included in the intervention costs. The time costs for staff, volunteers, and participants; the class and session time; mileage costs; resource costs; and the CHW training expenses were summed to derive a total program cost for the sample, resulting in a per-participant program cost of $1,175.63 for the 18-month period. These cost calculations ignore economies of scale that would accrue from the program being run at a capacity more reflective of real-world situations. To create a real-world cost structure, we used the original fixed and variable costs for each activity, but increased average attendance levels to reflect data collected on actual class attendance (ie, which included participants not used in the sample); we then calculated intervention costs based on the prevalence of diabetes in the Archimedes Model (Archimedes, Inc, San Francisco, California) (n = 6,551) (Table 1).

Finally, because diabetes management is a long-term process, we estimated ongoing postintervention costs assuming continued participation in educational classes, counseling, and exercise activities at 50% of intervention levels over the 3 periods of our projections (5, 10, and 20 years). This resulted in annual post-intervention costs of $140.63 per participant. We used data from the 2006 Medicare Current Beneficiary Survey for medical costs (22); lifestyle change costs were taken from a systematic review of costs of major US commercial weight loss programs (23).

We used the Archimedes model to project the intervention’s incremental lifetime health outcomes and related expenditures based on changes in A1c levels. It simulates the effects of interventions by generating multiyear forecasts of predicted disease outcomes, health behaviors, and health care costs relative to a hypothetical control group with similar demographic characteristics. Studies have independently validated the model for diabetes outcomes research (24).

We applied data from the Archimedes Cardio-Metabolic Risk (CMR) dataset, which includes data from a simulated US representative sample of 100,000 people aged 30 to 85 years (25) to our study sample of 6,551. The CMR dataset includes the results of 19 simulated controlled clinical trials comparing standard care (defined as current levels of care in the US population) with a set of health management interventions targeting cardio-metabolic risk, including glycemic control (25).

For the primary analysis, we grouped study participants into 2 cohorts: those whose diabetes was brought under control (ie, follow-up A1c reading was ≤7%) and those who were still out of control at follow-up (A1c >7%) (Table 2). During the intervention, the diabetes of 60% of the subjects was brought under control. We assumed that the sample’s levels of glycemic control would fluctuate over time. When A1c becomes uncontrolled, the intervention in the CMR dataset is designed to bring A1c under control. On the other hand, A1c in the dataset’s control group is uncontrolled at baseline and can vary over time; however, there is no specific intervention apart from usual care to reduce A1c to a target level. We discounted program costs, ongoing participation costs, medical costs, and quality-adjusted life years (QALYs) at a 3% rate. Costs are given in 2010 dollars. Archimedes uses disutility weights for diabetes-related disorders and illnesses from 2 studies. One (26) used responses from the 2000-2002 Medical Expenditure Panel Survey (27) and the EQ-5D questionnaire (28) from a nationally representative sample of 38,678 adults to develop health-related quality of life scores for 95 chronic diseases with ICD-9 codes. The other, by Coffey and colleagues (29), used the Self-Administered Quality of Well-Being index to calculate health utility scores for 2,048 people with type 1 and type 2 diabetes. A life-year with none of the 95 chronic diseases present receives a health utility score of 1; the presence of 1 or more of the chronic diseases reduces the score in accordance with its additive contribution to health disutility.

We applied the discount rate to the program and ongoing costs to derive the net present value of intervention costs. We used the Archimedes simulation model to calculate the long-term differential medical costs and QALYs of people with type 2 diabetes who do not have access to the program, who we used as controls. We summed the long-term medical costs and net present-value intervention costs, subtracted the standard-care medical costs, and divided by the difference in QALYs between each intervention and the simulated control group to derive incremental cost effectiveness ratios (ICERs).