The effects of financial incentives on diabetes prevention program attendance and weight loss among low-income patients: the We Can Prevent Diabetes cluster-randomized controlled trial

An estimated 353 million adults worldwide are at high risk of developing diabetes [1]. Numerous randomized controlled diabetes prevention lifestyle interventions have demonstrated strong efficacy in delaying or reducing the onset of type 2 diabetes among high risk adults yet penetration and participation of real-world diabetes prevention programs has been limited [2‐6]. In the United States this has been a particular challenge among low-income, less educated, and culturally diverse individuals [7, 8]. A meta-analysis of community-based DPP interventions and evaluation of the Centers for Disease Control and Prevention’s (CDC’s) National Diabetes Prevention Program (NDPP) found that program attendance and weight loss goals were much lower among black, Hispanic, and other racial/ethnic individuals compared to non-Hispanic whites [9, 10]. This is concerning because these individuals have poverty rates two times higher than those of non-Hispanic whites and, among those in the highest poverty level, diabetes prevalence is 18% compared to 8% for those in the lowest poverty level [11, 12]. Furthermore, low income individuals are increasingly insured through the public programs of Medicaid and Medicare, which account for 66% of the estimated $237 billion in annual direct medical costs attributable to diabetes in the United States [13, 14].

A potential solution is the use of extrinsic incentives to promote participation and outcomes in diabetes prevention lifestyle programs. Several studies have demonstrated that prompt and frequent delivery of incentives can lead to improved healthy behaviors [15‐22]. Many insurers and 90% of large employers in the United States are using such incentives to foster healthy behaviors and there is a growing body of evidence that financial incentives improve tobacco cessation, physical activity, healthy eating, and weight management [16, 18‐21, 23‐25]. The 35 million economically disadvantaged adult Medicaid recipients may be especially responsive to financial incentives [21, 26]. Medicaid is a federal health care insurance program. In the United States adults at or below 138% of the federal poverty level are eligible for Medicaid. Some States, like Minnesota, have expanded insurance coverage to include adults at or below 200% of the federal poverty level. Currently, gaps remain in understanding the optimal incentive design for specific behaviors and populations, especially low income or culturally diverse populations who are not necessarily reached by corporate wellness programs.

At present, among 164 funded NDPP sites, 150 use participant incentives with 32 using financial incentives [5]. The NDPP incentive targets, how they are delivered and their effectiveness is, however, unknown. Providing incentives for DPP attendance may be a good approach considering that for every DPP session attended participants lost 0.31% of their baseline weight and, a 1 kg reduction in weight results in a 16% decrease in diabetes incidence [9, 27]. Providing incentives for attendance over 16 weeks was shown to be effective by VanEpps and colleagues as part of the Centers for Medicare & Medicaid Services Medicaid Incentives to Prevent Chronic Disease (MIPCD) initiative but it did not result in differential weight loss [28, 29]. The MIPCD was an ambitious effort to design and evaluate incentive programs that encourage smoking cessation, weight management, blood pressure and cholesterol control, optimal diabetes care management, and type 2 diabetes prevention [28].

We present findings from the We Can Prevent Diabetes (WCPD) study, also part of the MIPCD initiative [30]. The study was conducted among Minnesota Medicaid beneficiaries with prediabetes who were recruited from their primary care clinics. Two financial incentives structures were designed to offer immediate incentives for DPP attendance and weight loss over a 12 month period [30]. Our incentive design, patient population, recruitment process, and study duration differ from prior studies, thus adding to the current literature. We hypothesized that participants receiving financial incentives would improve DPP attendance and weight loss compared with participants not receiving incentives.

From January 1, 2013 through April 15, 2016, the 13 WCPD primary care clinic systems identified 2871 potential participants, reaching 72% and enrolling and consenting 55.7% of those contacted (n = 1154) [30]. Of these, 43 participants were never assigned to a randomized class and 264 participants never attended a DPP session, with no significant differences between the study arms [30]. Our analyses included the remaining 847 participants who enrolled and attended at least one DPP session (73.4%). Ninety-three groups were started (IND = 33, GRP = 30, AC = 30) with, on average, nine participants per group.

Baseline characteristics of participants are detailed in Table 1. Participants were racially and ethnically diverse. Only 17% were White while 64% were African American. However, a large proportion of African Americans were Somali immigrants and whose primary language was Somali. Almost 54% had a BMI ≥35 kg/m². Less than 15% had more than a high school education and 40% had evidence of a prior mental health condition.

Weight Change Outcomes: Based on planned analyses, at 16 weeks the mean percent weight change was similar for the three study arms (IND = − 2.6% (95%CI -3.1, − 2.0); GRP = − 3.1% (95%CI: − 3.7,-2.4); AC = − 3.4% (95%CI: − 9.2, 4.9)) (Table 2). At 12 months, mean percent weight changes for IND, GRP, and AC participants were − 3.7% (95%CI: − 5.7,-1.6), − 4.4% (95%CI: − 6.5,-2.2), and − 7.1% (95%CI: − 9.2,4.9) and non-significant. However, these results reflect the differential attrition rates across groups, and thus the number of participants contributing with data in the maintenance phase: 142 (IND), 122 (GRP), and 71 (AC) participants with two or more weight measurements from week 17 to week 52.

During DPP core sessions, the 5% weight loss goal was met by 15.2% of AC, 21.5% of IND, and 24.0% of GRP (P < 0.05) participants (Table 3). Significantly more GRP participants also attained 7% weight loss than did AC participants. At 12 months, significantly more IND (20.3%, P < 0.05) and GRP (21.5%, P < 0.05) participants achieved the 5% weight loss than did AC participants (14%). A greater proportion of GRP participants also achieved the 7% (P < 0.05) and 10% weight loss goals (P < 0.05) than did AC participants (Table 3). There were no weight loss goal differences between the IND and GRP participants. Additionally, among all participants achieving at least 5% weight loss, 80% maintained the weight loss through the 12-month intervention period (data not reported).

Attendance Outcomes: Attendance rates by session declined steadily from about 90% to just over 50% at week 16 in the IND and GRP participants. Among the AC participants, attendance was just over 30% at week 16. Rate of decline was similar for IND and GRP (P = 0.742) but were significantly lower compared to the AC participants (IND vs AC, P < 0.001; GRP vs AC, P < 0.001) (Fig. 1). Sixty-one percent of IND and 64% of the GRP participants attended at least 75% of the DPP core sessions, compared with 39% of AC participants (P < 0.001). Attendance of at least 75% of DPP maintenance sessions was 23 and 26% for the IND and GRP participants, respectively, compared with 11% for the AC participants (Table 3).

Food and Physical Activity Monitoring: Weekly completion of food and physical activity logs was monitored in DPP core sessions 4 through 16. In general, completion of food and physical activity logs was low with, on average, 3.3 to 5.2 sessions. This represented 42 to 63% of sessions attended, depending on the study arm. Among beneficiaries completing their physical activity log, the proportion self-reporting at least 150 min of physical activity per week was 30, 47, and 36% among the IND, GRP, and AC participants, respectively (Table 3).

Incentives earned: A total of $116,205 in incentives was disbursed to the 847 participants (IND = $60,120; GRP = $49,110; AC = $6975). The median, lower quartile, and upper quartile of incentives earned per participant for IND participants were $215, $85, and $285, respectively. These were $215, $95, and $270 for GRP participants and $25 for AC participants.

In this pragmatic randomized controlled trial with Medicaid beneficiaries in the United States, we found that financial incentives substantially increased DPP participation during core sessions and modestly increased participation during maintenance sessions. All three study arms lost 2.6 to 3.4% from their baseline weight (5.5 to 7.3 lbs.) during the first 16 weeks. By 12 months, weight loss was 7.1% in the AC study arm, compared with 3.7 and 4.4% in the IND and GRP study arms; however, due to participant attrition, these differences were not significant. Thus, financial incentives did not differentially affect mean weight loss during DPP core and maintenance sessions.

Despite no differences in mean weight loss, a greater proportion of IND and GRP participants attained 5% weight loss during the first 16 weeks and at 12 months than did AC participants. Significantly more GRP participants also achieved 7 and 10% weight loss at 12 months suggesting that group-based incentives may enhance weight loss by promoting group support or accountability. It was notable that sustained weight loss was observed among 80% of all participants who were able to achieve at least 5% weight loss. This may represent a high level of intrinsic motivation to improve healthy behaviors among these participants, potentially facilitated by extrinsic financial incentives since a greater proportion of participants in the IND and GRP groups achieved at least 5% weight loss.

Similar MIPCD DPP financial incentive interventions were implemented among Montana and New York Medicaid beneficiaries [28, 29, 32, 33]. A composite evaluation across all three studies found increased DPP attendance in participants receiving incentives but with no significant differences in weight loss between participants with and without incentives [33]. In contrast to the WCPD, New York and Montana did not find differences in the proportion of participants achieving 5% weight loss. In fact, in New York, 19 to 43% of participants across all study arms, including the 36% of the control group, achieved 5% weight loss [29, 32]. In all these studies, weight was only measured at a DPP session and thus reflects participants retained in the study over time. This is important to consider. Participant attrition at 16 weeks was almost two times higher in New York compared with WCPD, even with similar incentive structures [29]. This suggests other study design factors such as eligibility criteria, recruitment approaches, and DPP delivery may impact retention beyond the financial incentive structures and could potentially ‘wash-out’ incentive weight loss effects by retaining primarily participants with high initial intrinsic motivation in the treatment and control study arms.

Among Minnesota Medicaid beneficiaries, receipt of the incentives was immediate, and the GRP design may have fostered peer accountability and support. These design characteristics have proven effective in several weight loss and physical activity studies [19, 24, 34]. However, most previous studies lasted fewer than 6 months, and behavior changes were not sustained after the incentives were removed [15, 18, 34]. Longer-term (> 6 months) incentive effects are likely diminished due to loss of motivation, difficulty in adhering to the behaviors, and lack of or less frequent incentives [17, 23, 34‐36]. We may have experienced this in the WCPD when, at 4 months, incentives changed from once a week to only once a month and attrition rates appeared to rise among the intervention arms.

The MIPCD studies used a standard-gain economic model for structuring incentives. Participants were guaranteed known financial rewards for participation, weight loss, and in Montana, completing food and physical activity logs. These are DPP leverage points within participant control. Future studies could explore other instrumental targets such as providing incentives for individually tailored step-wise improvements in healthy behaviors and outcomes. Another promising approach would be to explore incentive designs leveraging behavioral economic principles as these have demonstrated improved weight loss and physical activity compared to standard gain designs [15, 17, 18, 37]. Capitalizing on a person’s tendencies to expect future rewards based on past experience, desire for rapid feedback and rewards, aversion to loss, loss regret, and response to variable reinforcement may increase the effectiveness of financial incentives, particularly in low-income populations [22, 38, 39]. Currently, however, such studies in low-income or culturally diverse populations do not exist.

Participants in most DPP community-based translational studies have not been able to achieve a weight loss of 5% from baseline [9, 10, 40]. Furthermore, 25% of eligible NDPP participants did not even enroll in the program, and another 23% were lost to attrition [9, 10, 41]. We experienced similar issues, with 37% of those contacted declining to participate or never attending a DPP session after enrolling while among AC participants attending at least one DPP session, less than 40% attended 12 or more DPP core sessions, and only 11% attended 6 or more DPP maintenance sessions. Like the NDPP, the WPCD also found a 0.31% weight decrease for each attended session. But, even with the DPP provided at no cost, classes arranged for convenience (i.e., location, day, time, free childcare, and free transportation), cultural adaptations (i.e., culturally trained lifestyle coaches, interpreters, groups with culturally similar participants), and recruitment from trusted providers, sustained engagement among Medicaid beneficiaries without additional motivators like incentives was difficult [9].

Currently, eight States provide Medicaid DPP reimbursement. In the WCPD, accounting for programs costs and the incentives earned by Medicaid beneficiaries, the DPP was cost-effective and, therefore, should be more broadly considered by payers [42]. In addition, increasing organizational capacity to deliver the DPP in low-income communities is needed. This may require the CDC Diabetes Prevention Recognition Program to revisit the 5% weight loss recognition requirements; otherwise, DPP-delivery organizations serving low-income individuals will be in danger of losing their certification and reimbursement status, exacerbating diabetes health disparities [43, 44].

Limitations: Our study has some limitations. First, participants volunteered for this study and were mostly women, reducing generalizability to all Medicaid beneficiaries and suggesting novel strategies may be needed to engage men in the NDPP. However, from a pragmatic perspective, we believe Medicaid beneficiaries who are women would be more likely to participate in similar programs. Furthermore, our participants were racially, ethnically, and culturally diverse. They did, however, cluster within the same DPP groups which enhanced DPP delivery but reduced power for the analyses because the unit of randomization was the DPP group. In addition, we had initially powered on 44 DPP groups per arm with an average of 10 participants. At the end of the study, we had 30 groups per arm with an average of nine participants per group thus limiting our power to detect significant differences. Analytically, we offset the effect of these two factors by using repeated weight measurements within participants and covariate adjustment. Second, our primary outcomes were assessed at DPP sessions. If a participant did not attend a session, we did not have weight information, which was problematic, especially after 16 weeks. Third, though participant characteristics were associated with lost to follow-up these were similar across study arms, even though attrition was larger in the AC group. Overall, participants lost-to-follow-up at 16 weeks were younger, African American, had Somali as their primary language, and were less educated which suggests that future studies may need to tailor their designs to better engage and support participation of these individuals. Interestingly, even though 40% of participants had evidence of a mental health condition, this did not differentially impact attrition although it may have influenced adherence and weight loss. Finally, participants and lifestyle coaches were not blinded to the study assignments. Lifestyle coaches may have been more or less engaged given the intervention assignment, but we were unable to determine this potential effect. Nonetheless, all investigators, statisticians, and data analysts were blinded to study assignments until study completion.

The DPP is an intensive, 12-month, structured lifestyle program designed to help participants increase physical activity, eat healthier, reduce caloric intake, lose weight, and ultimately, prevent or delay the onset of diabetes. Modifying habits established over many years is challenging—and potentially more so for Medicaid beneficiaries, who are often less educated, un- or under-employed, struggling to maintain family and housing stability, and living in obesogenic environments where food and physical activity options are limited [45]. Despite these difficulties, our study demonstrated that, with active outreach from trusted primary care providers, a culturally diverse group of Medicaid beneficiaries were interested in an intensive, structured, and free lifestyle behavior change program. However, reward-based financial incentives were needed to sustain participation and promote achievement of weight loss goals. Financial incentives may be an avenue to overcome individual, socioeconomic, and environmental barriers experienced by low-income populations in the United States and provide a model for bridging primary care health systems with community-based health programs [46]. We did find the WCPD to be cost-effective, thus, it could improve the current and future health status of a large number of Medicaid beneficiaries while mitigating future health care costs [42]. At the same time, future incentive studies need to carefully consider design factors such as individually tailored goals, targeting other outcomes such as physical functioning, and using incentive structures informed by behavioral economics to maximize participation and improvements in cardio metabolic outcomes among low-income populations that are quite heterogeneous in terms of race, ethnicity, immigration status, and culture.