Intervention costs and cost-effectiveness for a multiple-risk-factor diabetes self-management trial for Latinas: economic analysis of ¡Viva Bien!

Abstract

Information on cost-effectiveness of multiple-risk-factor lifestyle interventions for Latinas with diabetes is lacking. The aim of this paper is to evaluate costs and cost-effectiveness for ¡Viva Bien!, a randomized trial targeting Latinas with type 2 diabetes. We estimated 6-month costs; calculated incremental costs per behavioral, biologic, and quality-of-life change; and performed sensitivity analyses from health plan and participant perspectives. Recruitment, intervention, and participant costs were estimated at $45,896, $432,433, and $179,697, respectively. This translates to $4,634 in intervention costs per ¡Viva Bien! participant; $7,723 in both per unit reduction in hemoglobin A1c and per unit reduction in body mass index. Although costs may be higher than interventions that address one risk factor, potential risks for longer-term health-care costs are high for this at-risk group. Given the benefits of ¡Viva Bien!, cost reductions are recommended to enhance its efficiency, adoption, and long-term maintenance without diluting its effectiveness.

Several studies have documented the costs and cost-effectiveness of lifestyle interventions for persons with diabetes [1, 2]. However, to our knowledge, none of these interventions has targeted women of Hispanic heritage or Latinas with type 2 diabetes. This issue is important, given that Hispanic Americans, the fastest-growing ethnic population in the USA—and particularly Hispanic women (Latinas)—have a greater prevalence of type 2 diabetes and more diabetes complications than non-Hispanic Whites (NHW) [3–5]. Diabetes is an independent risk factor for coronary heart disease (CHD), which is the leading cause of death among women in the USA [6]. Latinas have higher mortality from CHD [7–9] than NHW women and have more CHD risk factors (e.g., hypertension and obesity) than NHW and Latino men [10, 11]. In addition, a gap has been observed between ethnic groups indicating that self-management behaviors are performed with less frequency among blacks and Hispanics compared to NHWs [12]. Yet, CHD-risk-prevention research with women, especially Latinas, is limited [13].

Understanding the effectiveness of diabetes self-management programs relative to their associated costs is an important component of health-care policy making [2]. Relatively few behavioral lifestyle interventions report on intervention costs [14]; and, when they do, these costs are often not detailed, comprehensive, or transparent. In addition, only a few of these studies have included the potential costs to the participant to travel to, and participate in, these interventions [15–17].

Further, they seldom model the replication costs of conducting the intervention under different settings or conditions. The primary purpose of this paper is to illustrate how the costs of a complex multiple-risk-reduction lifestyle intervention, including participant-related costs, can be collected and reported in a way to enhance decision making by potential adopting organizations and policymakers.

The overarching aim of ¡Viva Bien! was to culturally adapt and test our established, efficacious lifestyle change program—the Mediterranean Lifestyle Program (MLP) [18] and the Women’s Lifestyle Heart Trial (WLHT) [19]—with an important underserved population at high risk for CHD: Latinas with type 2 diabetes. The WLHT and the MLP were social-cognitive and socio-ecological theory-based programs which demonstrated improvements in behavioral, psychosocial, quality of life, and biological outcomes in postmenopausal, mostly Anglo women with type 2 diabetes in Oregon [20]. The primary aim of this paper is to present the costs and cost-effectiveness of the main intervention component of a comprehensive lifestyle program called ¡Viva Bien!, which targeted Latinas with type 2 diabetes. The secondary aim is to present models for reporting on three important issues not frequently addressed in translational research: recruitment costs, replication costs, and participant costs. Included are details about recruitment, intervention, and participant costs, cost per participant, and the cost per incremental change in key intervention outcomes.

METHODS

Setting and overview of ¡Viva Bien! program

The ¡Viva Bien! study recruited 280 Latinas with type 2 diabetes. One hundred forty-two were randomized to the ¡Viva Bien! intervention arm and 138 to the usual care, control arm. Recruitment, participation rates, CONSORT diagram, reasons for declining participation, and recruitment costs are presented in detail in Toobert et al. [21]. As noted in considerable detail in both the appendix of Toobert et al. [21] and Osuna et al. [22], relative to this team’s earlier research associated with the MLP, a number of cultural adaptations were implemented to the intervention and recruitment components of ¡Viva Bien!. As noted in Osuna et al., most of these adaptations were derived from literature review along with data collected in a 3-month pilot that was conducted as a preliminary test of cultural adaptation procedures. These cultural adaptations are summarized in the sections below that describe recruitment and intervention resources and procedures.

Recruitment

In brief, 7,945 female participants with diagnosed diabetes were mailed recruitment letters from two EMR-based diabetes registries (one managed care and one community health center) over 12 months to enroll participants in four waves. From those mailed, 4,045 were telephoned, 83% were found ineligible due to not being Latino, and the recruitment goal of 280 Latinas (61% of those eligible) was achieved. Participants received their primary care from one of nine Kaiser Permanente (KPCO) managed-care clinics in the Denver, CO, metropolitan area, or from the Salud Family Health Center, located in Commerce City near Denver. KPCO serves approximately 450,000 participants in the Denver metropolitan area, about 17% of whom are Latino. Salud Family Health Center provides comprehensive primary health services to low-income participants who may lack private insurance and have little or no access to other health-care providers. About 65% of Salud’s participants are Latinos whose primary spoken language is Spanish.

Cost analysis procedures

We captured cost data attributable to resource use associated with the recruitment and intervention components (first 6 months) of the ¡Viva Bien! program. We also collected data associated with participant costs via assessments that were conducted at 6 months, and we used geocoding software to estimate participant travel costs. Using methods employed in several of our previous studies, we created a detailed accounting system that captured all implementation—and recruitment—attributable expenses [1, 23, 24]. While not reported here, our procedure also allowed for the capture of research and development costs that may be useful for future dissemination. In addition, we made considerable efforts to track recruitment costs, the importance of which is often overlooked by many studies [23].

To better track study staff’s efforts and associated costs, consistent with instruments described in Ritzwoller et al. [23], we created tailored cost-capture templates that were based on study staff responsibilities at each site. Each template consisted of relevant task categories representing each person’s involvement in the project based on their job category. To better understand project progression, data from templates were collected and analyzed monthly. To establish consistency in wage rates across different sites, we utilized 2009 data from the Bureau of Labor Statistics (BLS) [25]. Work units were recorded as hours, and were transformed into the corresponding dollar amounts based on a unit cost measure of hourly wages derived from the specific job classification of each staff person reporting hours (e.g., health educator, dietitian). These cost estimates were then aggregated across intervention categories.

We also accounted for other intervention-related expenses, including recording supplies, printed materials (including translation costs), and other project costs. To better understand everyday project activities, we conducted personal interviews, attended staff meetings and conference calls, and maintained close contact with key personnel.

In this analysis, we focused on the recruitment and intervention phases of the program, and excluded any tasks and expenses that dealt with intervention development, assessment and data collection, and other research-related activities that would not need to be replicated if the study were continued at KPCO or if it were replicated or implemented in another setting.

Recruitment and intervention resources

Recruitment

Recruitment resources included computer programmers to identify diabetes participants from registries, bilingual KPCO research staff, a contracted recruiting company, and mailing supplies. As described in Toobert et al. [21], achieving the enrollment target of 280 for the ¡Viva Bien! program proved challenging. In brief, letters were prepared in English and Spanish signed by the project’s Latino physician and were mailed to potential participants, along with self-addressed stamped postcards that recipients could return to decline further contact or request more information. Bilingual project staff initiated contact by telephoning prospective participants, describing the program in either Spanish or English, determining eligibility, and inviting qualified candidates to participate. At KPCO, the diabetes registry, which did not at that time include ethnicity, was used to identify women with type 2 diabetes, which meant that recruiters telephoned a large number of women who were ineligible due to non-Latino ethnicity. An external recruiting company was initially used, but was later replaced with KPCO staff. At Salud, the clinic provided a list of female diabetes participants aged 30 and older. The same recruitment procedures were used for Salud as for KP.

Intervention condition

Intervention inputs associated with the ¡Viva Bien! project consisted primarily of labor components. Intervention staff members had frequent direct contact with study participants, and their main efforts were directed toward conducting in-person intervention sessions, tracking participant progress, addressing participants’ questions and concerns, mailing program-related correspondence, and data collection, data entry and data cleaning. The ¡Viva Bien! program is described in detail elsewhere [22]. Briefly, the ¡Viva Bien! intervention (in addition to usual diabetes care) incorporated theoretically informed social-cognitive, social-ecologic, and problem-solving factors into a program designed to improve four behavioral risk factors known to affect coronary heart disease: diet, physical activity, stress, and smoking. The major intervention components included a culturally adapted “Mediterranean Diet”-style eating plan, physical activity and stress management regimens (with supportive equipment, instruction, and DVDs), group social support, and smoking cessation. The intervention featured a 2½-day non-residential weekend retreat, followed by 6 months of weekly, 4-hour meetings. The retreat schedule included instruction for the dietary protocol, grocery lists, and food preparation as well as consumption of the foods recommended on the Mediterranean diet with a Latin “twist” [26] (i.e., ingredients and foods common to Latin culture). Latin American recipes were altered by the Latina project dietitian to conform to the Mediterranean diet. Recipes from specific Latin American countries were modified using common staples and included a wide range of ethnic Latin American foods. Potluck dinners were key to the diet component. The Latina dietitian conducted cooking demonstrations to show new methods for preparing typical foods. Participants were shown how to modify their favorite recipes by incorporating the principles of the Mediterranean diet into their usual foods. Sessions at the retreat were structured so that information was conveyed in Spanish and English concurrently. Bilingual staff gave presentations in English and Spanish. English and Spanish PowerPoint slide shows and program pamphlets were provided, illustrated with photographs of Latina women.

During the retreat, daily group physical activity sessions included warm-up, walking or aerobics, strength training, and a cool-down led by an American College of Sports Medicine (ACSM)-certified exercise physiologist. The instructor led in both English and Spanish, with Latin (for example, salsa) style steps and music. Thus, women could determine for themselves which physical activities were culturally appropriate. Colorful pamphlets, available in English and Spanish, were created to support the physical activity component. The pamphlets included photographs of Latina women. Take-home exercise DVDs/CDs available in English and Spanish also were created for the program. Study participants also received instructions and had an opportunity to practice the stress-management techniques that included stretching, breathing, and meditation exercises, similar to yoga. Take-home stress-management CDs were created for the program in English and Spanish.

Weekly 4-hour meetings, which began in the week immediately following the retreat, provided 1 h each of physical activity, stress-management practice, Mediterranean diet potluck, and support-group-based problem-solving activities. Because the support-group structure was flexible, participants could choose topics that were culturally appropriate and personally relevant. Other cultural adaptations included additional presentations and question–answer sessions with the project’s Latino physician.

Control condition

Usual care (UC) at KPCO typically followed standards of care recommended by the American Diabetes Association, and was directed at diabetes control, management of complications associated with diabetes, and monitoring of other health factors such as cholesterol and hypertension. Participants received care, for nominal visit co-payments ranging from 0 to $25, through their primary care providers. Those with elevated hemoglobin A1c levels and other cardiovascular disease risk factors were assigned case managers. In addition, these members had access to free group diabetes education and support activities offered through KPCO. At Salud, usual care consisted of diabetes management directed by the primary care physician, nurse practitioner, or physician assistant. In addition, free, monthly diabetes classes were conducted in Spanish and English. At KPCO, medications were covered under HMO plans for a nominal co-payment. At Salud, medications were available through a non-profit pharmacy offering reduced rates. Specialty services were available through partnerships with community providers and offered at discounted rates.

Participant costs

Participant costs associated with the intervention were estimated from three sources: a participant survey adapted from the instrument employed in the Diabetes Prevention Program, Form Q12, Economic Evaluation Questionnaire [17], the use of Microsoft MapPoint^® software, and from intervention logs that noted the participant time associated with the intervention sessions (i.e., retreat and follow-up meetings). The participant cost survey captured employment status and assessed out-of-pocket costs related to health club and gym memberships, exercise equipment, and changes in the cost of food that participants experienced since they entered the ¡Viva Bien! program.

Travel time and costs were estimated by mapping the address of each participant and the address where the retreat and weekly support meeting took place with the use of Microsoft MapPoint^® software. This program generated participant-specific average drive time and distance to the intervention location. These time estimates were converted to dollars based on the amount allowed by the U.S. Internal Revenue Service for business expenses. Cost per mile in 2009 was $0.55.

We also estimated the participant time cost associated with attending the retreat and follow-up classes. In the absence of actual wage/income data, we used methods employed by Sharma et al. [27] to classify situations in which a participant gave up leisure time but not time from work. This time was valued at 30% of the average gross wage noted by the Bureau of Labor Statistics’ mean wage in Colorado for 2009, which was $24.40 (http://www.bls.gov/ncs/ocs/sp/ncbl1444.txt). Other potential participant costs attributable to the intervention were also captured at the 12-month survey, including food, health/fitness club membership, exercise clothes, and exercise equipment.

Outcome measures

Outcomes important to decision makers included hemoglobin A1c and body mass index (BMI). BMI was calculated from measures of height and weight taken, if possible, in the morning in the fasting state and in stocking feet on a sensitive digital scale (Detecto Electronics). Hemoglobin A1c assays were performed at the KPCO Regional Reference Laboratory in Aurora, CO, and measured on a Bio-Rad Variant II Turbo liquid by high-pressure liquid chromatography. Intention-to-treat analyses were used to report primary outcome results. The intention-to-treat analyses were conducted after missing data were imputed using multiple imputation procedures via the expectation–maximization algorithm with NORM software.

Costs and cost-effectiveness estimation

Using 2009 wages and prices, we estimated total recruitment, intervention, and participant costs, taking care to separate out any captured costs related to development or research. Incremental costs associated with the first 6 months of the ¡Viva Bien! program relative to UC were considered those costs associated with the intervention over and above those associated with the UC condition. For example, hemoglobin A1c testing costs were not considered specific to ¡Viva Bien! given that testing subjects every 6 months is considered standard practice and part of UC. We also calculated costs per ¡Viva Bien! participant. Our cost-effectiveness measure was estimated as the marginal costs of the ¡Viva Bien! program per incremental improvement in the primary outcomes attributable to the intervention. To calculate the incremental change in main outcome, we estimated the cost per participant in each arm, and applied these data to estimate the average change in hemoglobin A1c or BMI attributable to the intervention to form cost-effectiveness ratios.

Sensitivity analysis related to intervention

To address the issues of uncertainty with respect to adoption or dissemination of this intervention, we conducted a variety of sensitivity analyses to estimate the range of intervention costs given potential variation in the circumstances. In these analyses, we estimated the following: changes in the calculation of indirect participant costs associated with attending the retreat, knowledge of the ethnicity of the target population (for recruitment), and changes in the length of attendance time for the retreat and intervention meetings.

RESULTS

Recruitment and intervention costs

Table 1 details the distribution, recruitment, intervention, and participant costs associated with the ¡Viva Bien! program. Total costs associated with the recruitment effort were $45,896. The use of professional recruiters at a cost of $26,853 was the largest component of recruitment costs and many of the costs were accrued because of not knowing the ethnic identity of KPCO members prior to the telephone call. Total intervention costs of the program were $432,433. Costs were associated with the resources required to conduct the intervention components of the study, including retreat meetings, intervention meetings, and participant support. These costs included retreat and meeting preparation, along with study-related e-mail correspondence, supervision, debriefing, and time travel to meetings. Costs to translate intervention print materials were estimated at $4,455. Overhead costs, estimated at $165,663, included resources not directly associated with the number of participants involved or direct interaction activities associated with day-to-day operation of the project and the resources required for implementation of the intervention unrelated to participant contact or the delivery of the intervention components (e.g., personnel issues, unscheduled meetings, staff communication).

Table 1.

¡Viva Bien! project and patient cost components

Cost element	Total cost
Project cost
Recruitment
Project managers	$4,732
Research assistants	$6,113
Recruitment incentives	$3,550
Professional recruiters	$26,853
Overhead	$4,648
Total recruitment	$45,896
Intervention
Retreat	$36,595
Staff meetings and planning	$216,666
Staff training	$9,054
Translation	$4,455
Overhead	$165,663
Total intervention costs	$432,433
Recruitment and intervention costs	$478,329
Participant costs
Travel timea	$44,267
Retreat timeb	$20,789
Meeting timec	$97,707
Equipment
Exercise shoes	$7,895
Exercise clothes	$4,737
Fitness club membership	$4,284
Total participation costs	$179,679
Total recruitment, intervention, and participation costs	$658,008

^aTravel time = 26 meetings × 2 (round trip) × 10.9 (average miles) × 0.55 (price) × 142 (n) = $44,267

^bRetreat time = 20 h × 7.32 (30% of $24.40) × 142 (n) = $20,789

^cMeeting time = 94 h × 7.32 (30% of $24.40) × 142 (n) = $97,707

Participant costs

Employing an intent-to-treat perspective, ¡Viva Bien!’s 142 participants attended 2½ days of a retreat and 23 intervention meetings, for a total of 26 round trips to the intervention center. On average, participants traveled 10.9 mi (SD = 7.4) to the intervention center with maximum distance of 46.5 mi. Based on the distance and frequency of the travel, total travel cost for the intervention cohort as noted in Table 2 was $44,267.

Table 2.

Changes in primary outcome measures

Intention to treat (N = 280)	Baseline M (SD)	6 months M (SD)	Change baseline to 6 months	Difference in change (p)
Body mass index (kg/m2) change				0.6 (<0.05)
Usual care (N = 138)	33.2 (6.8)	33.1 (6.9)	0.1
¡Viva Bien! (N = 142)	35.4 (7.0)	34.7 (6.8)	0.7
Hemoglobin A1c (%) change				0.6 (<0.05)
Usual care (N = 138)	8.2 (1.7)	8.3 (1.7)	−0.1
¡Viva Bien! (N = 142)	8.4 (1.9)	7.9 (1.6)	0.5

Labor participation questions in the baseline survey demonstrated that employment status was not significantly different between ¡Viva Bien! and UC groups. At baseline, 53.3% of ¡Viva Bien! participants worked for wages outside of home. Of those, 35.7% reported being in a managerial/professional position and 32% in a clerical position, and the remaining were distributed across service, farm labor, craftsman, and other categories. At the follow-up 6-month assessment, participants were asked what they would be doing if they were not participating in the ¡Viva Bien! program, and only 27.3% responded that they would be working for wages. The other responses were distributed across work, leisure, and school activities. Given the relatively low labor market participation rate, and the high reported substitution rate between the ¡Viva Bien! program and other non-labor market leisure activities, indirect participant costs of participating in the intervention were calculated, consistent with Sharma et al. [27] based on 30% of the state of Colorado average gross wages. Using the intervention-related contacts noted above, we calculated that there were 20 h required to attend the retreat and 94 h for the follow-up intervention meetings, for an estimated total indirect labor/leisure costs of $20,789 for the retreat and $97,707 for the intervention meetings.

The follow-up participant survey demonstrated that food costs were not significantly different between ¡Viva Bien! and UC groups. However, as noted in Table 1, the ¡Viva Bien! participants reported spending more on exercise shoes, clothes, and health/fitness club memberships than did the UC participants, for a total of $16,916 participant costs attributable to the intervention.

While a more complete description of the changes in the primary outcome measures attributable to the intervention are described elsewhere [21], Table 2 describes the intention-to-treat estimates of changes in hemoglobin A1c and BMI. The estimated change in hemoglobin A1c and BMI attributable to the ¡Viva Bien! Program was 0.7 (p < 0.05) and 0.5 (p < 0.05), respectively.

Table 3 describes the marginal costs per incremental improvement in the primary outcomes attributable to the intervention. When costs are fully loaded (e.g., participant, recruitment and intervention costs), the costs per ¡Viva Bien! Participant was $4,634; with a marginal cost per incremental change in both hemoglobin A1c and cost per BMI unit change of $7,723. Taking only direct intervention costs into account, the costs per ¡Viva Bien! participant, and marginal cost per incremental change in hemoglobin A1c and BMI, were reduced to $3,045 and $5,076, respectively.

Table 3.

Per-participant costs and incremental cost per incremental change in hemoglobin A1c and incremental change in body mass index (N = 142)

	Total cost (patient + recruitment + intervention costs)	Recruitment + intervention cost	Intervention cost
¡Viva Bien! participant	$4,634	$3,369	$3,045
Outcome:
Hemoglobin A1c	$7,723	$5,614	$5,076
Body mass index	$7,723	$5,614	$5,076

Sensitivity analyses

Our original estimates of the indirect costs to participants of attending the intervention-related meetings were based on 30% of average wages. Using estimates of 50% of average wages would increase the participant retreat and meeting time costs from $20,789 and $97,707 to $34,648 and $162,846, respectively. Using estimates of 80% of average wages would increase participant costs to $55,437, and $260,553, respectively. If labor force participation rates of the targeted population were high, and if intervention participation interfered with hours spent at work, then the indirect cost of participation could increase substantially. Applying the average market wage rate to 80% of the population would increase the participant costs to $377,173 and per-participant cost of $6,025, and would increase the incremental costs associated with incremental changes in the primary outcomes, as noted in column 2 of Table 4.

Table 4.

Results from sensitivity analyses associated with changes in per-participant costs and incremental costs per incremental change in outcome

	Total cost current ¡Viva Bien! intervention (patient + recruitment + intervention costs)	Total cost from column A, with increases in patient costs	Total costs from column a with changes in recruitment effort	Total costs from column A with reductions in the length of the retreat and intervention meetings
¡Viva Bien! participant	$4,634	$6,025	$4,445	$3,096
Outcome
Hemoglobin A1c	$7,723	$10,041	$7,408	$5,159
BMI	$7,723	$10,041	$7,408	$5,159

In an alternative scenario, the ¡Viva Bien! program may not require recruitment of only Latinas, or it might be possible that the ethnicity of the target population was known (e.g., all Hispanic community health centers). Services of the professional recruiters may not be needed and there would be a significant reduction in project staff time used to identify Latino participants and translate written material. Thus, recruitment costs would decrease to $19,043. Alternatively, were the program fully integrated into diabetes care protocols, eligible participants could be identified on an ongoing basis through an EMR or direct physician referral, and registered in the next offered program at the same cost of enrolling participants in diabetes education classes. This would reduce costs to enrollment-related activities. As noted in column 3 of Table 4, overall cost estimates would decrease.

In addition, if intervention results could be achieved by reducing the intervention contact hours to a 1-day retreat and reducing the 23 follow-up meetings from 4 h to 2 h, then the total project cost could be reduced even further (column 4, Table 4). One-day retreats would result in $12,198 total retreat cost, and a 50% reduction in the time for intervention meetings would result in $108,333 intervention meetings cost savings. Patient costs associated with the intervention would be reduced by $85,697. As a result of these reductions, total intervention cost would be $299,703 compared to $432,433 of original intervention cost. Similar reductions in cost could be achieved if it were possible to reduce the salary of individuals delivering the intervention through use of nurses rather than physicians or community health workers rather than dietitians and sports exercise experts, or through existing community resources who could be trained and certified to deliver the ¡Viva Bien! program. This strategy would be similar to YMCA staff delivering Silver Sneakers^® [28] or Weight Watchers^® [29] or other evidence-based programs that follow specified protocols. Table 4 summarizes the incremental costs per change in incremental outcomes under alternative scenarios.

DISCUSSION

The primary purpose of this article was to illustrate comprehensive, transparent reporting of intervention costs from an intensive, coronary-heart-disease-risk-factor intervention designed for Latinas with type 2 diabetes. These calculations included costs of recruitment, overhead, training, and supervision, as well as participant costs, for a program adapted for the Latina population. These costs are not often reported for behavioral risk-factor interventions, but are important for translational research. Intervention participant costs, which for ¡Viva Bien! consisted primarily of participant time for transportation and participating in the meetings, are important considerations in any intervention. In these calculations, we followed some of the DPP procedures to capture costs, such as employment status and costs of exercise equipment associated with the intervention. We also employed more novel procedures to estimate the travel demands of the program, using the distance from participant home address to the meeting sites. If health care is to be patient centered as recommended by the Institute of Medicine, Agency for Health Care Research and Quality and the new Patient-Centered Outcomes Research Institute [30], greater attention must be paid to these issues, as well as patient preferences, goals, and satisfaction [31]. The sum of these factors accounted for approximately 60% of overall project costs.

There is a critical need for more comprehensive and transparent reporting of all intervention and replication costs in behavioral medicine research to address realistic economic issues for research translation. Recruitment costs are often considered as research costs and are unreported. However, any replication or implementation in real-world settings would require identification and recruitment of participants. Future research on the costs and impacts of different approaches to participant recruitment and their impact on the types of patients recruited, particularly for minority and high-risk participants, are sorely needed [32].

If different recruitment/enrollment procedures, such as physician referral or electronic outreach (e.g., email, automated phone messages) were used in implementation rather than in the research evaluation, then these costs could be modeled using sensitivity analyses, but they should not be simply omitted [23]. Other types of sensitivity analyses important to decision makers might include factors such as intervention intensity, modality, expertise, and wages of intervention staff. In this study, we conducted sensitivity analyses on intervention intensity and changes in participant costs, since meeting costs were the primary determinant of overall costs.

Our per-participant cost estimates are more than threefold (in 2009 dollars) higher than our previous estimates associated with the MLP program [1], and many times higher than the 2006 published estimates associated with the CDC-funded WISEWOMAN project [33]. While we believe these high cost estimates are directly related to efforts associated with cultural adaptations along with the intensity of the intervention, based on these reported estimates and the sensitivity analyses noted above, we intend to explore ways to reduce face-to-face intervention time and resources. Potential changes include employing telephone-based meetings, social media, and mobile health behavior-change technologies [34, 35] for parts of future ¡Viva Bien! group meetings and follow-up. In addition, an important issue for future research is the extent to which interventions such as ¡Viva Bien! can be delivered by different disciplines and even by community health workers or peer leaders vs. medical personnel [36, 37].

Our goal is to retain effectiveness and personal connections, but reduce intervention costs. Reducing the demands for in-person meetings and expanding the interventionist criteria to train and certify non-medical personnel could also enhance public health impact by increasing both the potential adopter pool and the percent of people who would participate or regularly attend. More generally, we recommend much greater use of simulation models [38, 39] to understand the sometimes counterintuitive impact of different program modifications. Sensitivity and replication analyses [23, 40] are key, as they provide the type of information that decision and policy makers need [41, 42]. Consistent with the Medical Research Council [43] recommendations for development and evaluation of complex interventions, we especially recommend that simulation analyses be conducted prior to beginning very expensive, lengthy multi-site trials to model both potential costs and outcomes [44].

This economic analysis was conducted primarily from the perspective of an adopting or sponsoring organization such as a health plan, although we included the participant perspective. However, a limitation of our study was that we did not capture all participant cost-sharing measures (e.g., pharmacy and visit co-payments). We also did not attempt to calculate full social cost analyses, as the Centers for Disease Control and Prevention would do [44]. Given the relatively short follow-up period, we also did not estimate relative (to UC) changes in health-care resource utilization and costs attributable to the intervention. In addition, we did not conduct cost–benefit or project long-term CHD reductions, given the large number of assumptions involved, or estimate cost per quality-adjusted life years. We believe such analyses are best done with longer-term data with outcomes based on results over at least 5–10 years.

This paper has many strengths, including relatively comprehensive, detailed, and transparent reporting of many cost components of a complex intervention [44], sensitivity analyses to help guide future program modifications, and inclusion of the participant perspective from a relatively large sample of high-risk Latina women in two different health-care settings. Although the ¡Viva Bien! intervention as conducted was moderately expensive, it may be that intensive interventions are necessary to produce lasting changes in complex multiple lifestyle behaviors and conditions resistant to intervention such as obesity and diabetes [18, 45]. Such intensive behavioral interventions, including those used in the DPP [15] or in Ornish programs [46], should be evaluated in the context of the considerably greater costs, risks, and potentially disabling side effects of invasive medical procedures such as gastric and coronary bypass surgeries, open heart surgery, and amputation. Our simulations and sensitivity analyses suggest it may be possible to decrease costs and increase cost-effectiveness by reducing intervention intensity, exploiting other modalities—such as phone or interactive voice response technologies—to retain effectiveness, reduce costs, and increase reach.