Effect of Process- and Outcome-Based Financial Incentives on Weight Loss Among Prediabetic New York Medicaid Patients: A Randomized Clinical Trial

Abstract

Purpose:

To determine whether different financial incentives are effective in promoting weight loss among prediabetic Medicaid recipients.

Design:

Four-group, multicenter, randomized clinical trial.

Setting and Participants:

Medicaid managed care enrollees residing in New York, aged 18 to 64 years, and diagnosed as prediabetic or high risk for diabetes (N = 703).

Intervention:

In a 16-week program, participants were randomly assigned to one of 4 arms: (1) control (no incentives), (2) process incentives for attending weekly Diabetes Prevention Program sessions, (3) outcome incentives for achieving weekly weight loss goals, and (4) combined process and outcome incentives.

Measures:

Weight loss over a 16-week period; proportion who completed educational sessions; proportion who met weight loss goals.

Analysis and Results:

No intervention arm achieved greater reduction in weight than control (outcome incentive −6.6 lb [−9.1 to −4.1 lb], process incentive −7.3 lb [−9.5 to −5.1 lb], combined incentive −5.8 lb [−8.8 to −2.8 lb], control −7.9 lb [−11.1 to −4.7 lb]; all P > .29). Session attendance in the process incentive arm (50%) was significantly higher than control (31%; P < .0001) and combined incentive arms (28%; P < .0001), but not significantly higher than the outcome incentive arm (38%).

Conclusion:

Process incentives increased session attendance, but when combined at half strength with outcome incentives did not achieve that effect. There were no significant effects of either process or outcomes incentives on weight loss.

Keywords

financial incentives weight loss Medicaid diabetes prevention process outcome

Purpose

In August 2010, the Centers for Medicare and Medicaid Services awarded Medicaid programs from 10 states a total of $85 million over 5 years as part of the Medicaid Incentives for Prevention of Chronic Disease (MIPCD) initiative. Authorized by section 4108 of the Affordable Care Act, these grants to states enabled investigation of ways to promote preventive health behaviors among Medicaid recipients at significant risk of developing chronic disease, aiming to decrease related morbidity and mortality and potentially save state Medicaid programs’ significant resources. Such cost savings are of particular importance, given the size of these programs. For example, as of 2016, New York’s Medicaid population is one of the largest in the United States, with over 6 million beneficiaries and annual expenditures of approximately $62 billion.¹ Furthermore, Medicaid covers more people than any other insurer in the country, placing greater urgency on the program’s need to reduce the health burden and financial costs of chronic diseases within Medicaid, where approximately 80% of resources are spent on people with chronic conditions, including chronic obstructive pulmonary disease (COPD), hypertension, and diabetes.²

Research has shown that financial incentives for healthy behaviors can significantly improve health outcomes for participants. For example, trials of financial incentives for smoking cessation in workplace settings have demonstrated rates of cessation nearly 3 times as large as those achieved by information about cessation programs alone,^3,4 and financial incentives have been shown to be effective in promoting compliance with medication regimens, medical advice, and medical appointments.⁵ A meta-analysis of both randomized and nonrandomized studies of financial incentives for medication adherence demonstrated significant improvements in adherence rates among incentivized groups,⁶ and other work has shown significant increases in weight loss with a variety of different incentive program designs.⁷

The effectiveness of incentives used within Medicaid programs, however, has been mixed.⁸ The Florida Medicaid program offered credits, redeemable for health-related products and supplies, to recipients for activities such as smoking cessation or attending dental visits; however, close to half of the available credits went unredeemed by eligible Medicaid beneficiaries. Furthermore, the majority of credits awarded were not associated with preventing or reducing risk of chronic disease through behaviors like smoking cessation or exercise, but rather were associated with relatively typical behaviors that may have occurred without incentives, such as wellness visits to the doctor’s office. The Idaho Medicaid program, aimed at increasing rates of annual wellness visits for children, did successfully improve adherence by increasing up-to-date compliance from 23% to 49%, but a simultaneous program designed to encourage weight management and tobacco cessation was less successful, reaching less than 1% of Medicaid enrollees in the state. The mixed empirical record of previous incentive programs within state Medicaid programs suggests that more research is needed to determine optimal approaches to encourage healthy behaviors among beneficiaries. Although Medicaid programs have the most direct access to a broad audience of potential patients, there may be infrastructure or programmatic limitations within Medicaid that reduce the effectiveness of incentives relative to external research programs that also enroll Medicaid beneficiaries.

The relative effectiveness of incentives based on process versus outcome also remains unknown, leading to the central research question of this study: Is incentivizing a process measure, an outcomes measure, or a combination of the two more effective? We examined this question in the context of the Diabetes Prevention Program (DPP), which has been widely used to prevent diabetes onset.⁹ This is an especially important problem in New York, where over 1.7 million residents have diabetes and an estimated 5 million have prediabetes, putting them at risk of developing diabetes in the near future.^10,11 Specifically, we compared the effectiveness of incentives for processes (eg, attending diabetes prevention counseling sessions), outcomes (eg, achieving weight loss goals), and a combination of processes and outcomes (eg, attending sessions and losing weight) in the context of a DPP not included in the current New York State Medicaid benefit. This design also provides an overarching framework for assessing the relative importance of process versus outcome incentives in different clinical contexts and for different populations.

Methods

Design

We conducted a 4-arm randomized controlled trial comparing a usual care DPP aimed at promoting weight loss with 3 incentive-based programs. All costs for study participants were covered by the Medicaid managed care plan in which the participant was enrolled. As part of the DPP, participants were asked to attend free weekly sessions at a location selected to minimize travel burdens, such as a community health center, their Medicaid managed care plan center, or a local YMCA. Sessions featured evidence-based lifestyle change content that addressed healthy eating, increasing physical activity, and losing weight, and were based on a previous clinical trial of lifestyle intervention for diabetes prevention.^9,12 Trained lifestyle coaches facilitated discussion among small groups of adults about behavior changes that can improve the health of participants. Participants were encouraged to set goals toward achievement of modest weight loss in the range of 5% of baseline body weight. This program also emphasized the need for long-lasting lifestyle changes, through features such as problem-solving for weight management setbacks (see Appendix A for greater detail). The protocol was approved by the institutional review board at the New York State Department of Health (NYS DOH).

Sample

Eligible participants in the MIPCD DPP included New York State residents, enrolled in a Medicaid managed care plan, between 18 and 64 years of age, and at risk of diabetes as defined by one of the following: (1) diagnosis of prediabetes; (2) diagnosis of obesity; (3) body mass index (BMI) >25 (or for Asian members, BMI >22); and (4) elevated glycated hemoglobin. All 19 Medicaid managed care plans within New York State, including 3 HIV Special Needs Plans, participated in the recruitment and enrollment of participants and implementation of the DPP. Participants were recruited through direct mailings, telephone outreach, and/or provider referral, with eligible participants identified using service utilization data and health records by the NYS DOH and Medicaid managed care plans, respectively.

Medicaid managed care enrollees were recruited from June 2013 through October 2015; DPP sessions were held from July 2013 through December 2015. In total, 703 participants were recruited (trial enrollment reported in Figure 1; demographics reported in Table 1). Nearly 70% of the study participants were female. Consenting participants were randomly assigned to one of 3 treatment arms (process incentives, outcome incentives, or combined incentives) or to a control arm in blocks of size 8 to maintain balance. Randomization occurred at the DPP group level; an entire group was assigned a specific treatment intervention rather than assigning individuals in each group to different interventions. Each of the 19 Medicaid managed care plans were encouraged to meet a minimum goal of participants, based on the proportion of their population enrolled in Medicaid. The plans established and offered the DPP to these participants. Randomization was stratified by managed care plan so that intervention arms were balanced within and across plans. No other stratification variables were used.

Figure 1.

CONSORT flow diagram. Note: Those who lost Medicaid eligibility could still attend sessions and are thus included in analyses.

Table 1.

Demographics of Participants in Diabetes Prevention Program.

	n (%) or Mean (SD)					Omnibus P Value
	Total (N = 703)	Control (n = 170)	Outcome Incentives (n = 181)	Process Incentives (n = 182)	Combined Incentives (n = 170)	Omnibus P Value
Age, mean (SD)	47.3 (11.5)	49.2 (11.7)	48.4 (11.4)	45.8 (11.6)	45.7 (11.1)	.0056
Female sex	489 (70%)	118 (69%)	134 (74%)	125 (69%)	112 (66%)	.413
Race						.0005
White non-Hispanic	57 (8%)	6 (4%)	23 (13%)	7 (4%)	21 (12%)
Black non-Hispanic	322 (46%)	82 (48%)	70 (39%)	85 (47%)	85 (50%)
Hispanic	154 (22%)	34 (20%)	49 (27%)	37 (20%)	34 (20%)
Other	51 (7%)	12 (7%)	7 (4%)	21 (12%)	11 (7%)
Unknown^a	119 (17%)	36 (21%)	32 (17%)	32 (17%)	19 (11%)

Abbreviation: SD, standard deviation.

^a Participants who did not complete the race questions.

Despite randomization, some participant demographic characteristics were unbalanced by arm. The mean age of participants was 47.5, but those in the control arm were slightly older (49.2, P < .01). Among all study participants, 55% were black, 26% Hispanic, and 10% white, but the control arm had slightly more black participants (61%) and fewer white participants (5%) than the other 3 arms (P < .01). Mean baseline weight among study participants was 208 lb, but was higher for participants in the combined incentives arm (222 lb) relative to the outcome incentives arm (206 lb), process incentives arm (202 lb), or control (207 lb; omnibus P = .03). We adjusted for these baseline variables in multivariate regression models assessing differences in outcomes.

Measures

The primary outcomes of interest were weight loss as evaluated on a weekly basis and over the course of the 16 weeks. Weight loss has been shown to be a dominant predictor of reduced diabetes onset; past work showed that for every kilogram (∼2.2 pounds) of weight loss, there was a 16% reduction in diabetes risk.¹³ Secondary outcomes included the number of completed education sessions in the 16-week program and the proportion of participants in each arm who successfully met weight loss goals at 16 weeks. Due to sample size limitations and the length of the follow-up interval, diabetes onset was not assessed.

Participation, weight, and attendance in the DPP were regularly recorded by the lifestyle coach hosting the DPP. These data were provided to the Medicaid managed care plans and subsequently to the NYS DOH for the provision of incentives. The Office of the New York State Comptroller, in collaboration with the NYS DOH, issued checks to participants completing required activities via direct mail.

Intervention

Participants randomly assigned to the process incentives arm received a $15 incentive for each weekly DPP session attended through 16 weeks. Those in the outcome incentives arm received $15 for each week their body weight decreased by 0.31% (1/16th of the overall goal of 5% by the end of the 16-week DPP) from baseline or received $240 (minus any weekly incentives already received) at the end of the DPP for reducing their body weight by 5% relative to their starting weight. Participants in the combined incentives arm received one half of the incentives for each of the activities above; this consisted of $7.50 for each DPP session attended, as well as $7.50 for reaching weekly weight loss goals or $120 (minus any weekly weight loss incentives already received) at the end of the 16 weeks for 5% reduction in body weight. Thus, all participants in the 3 treatment arms were eligible to receive up to $240 in financial incentives for the 16-week intervention period, in addition to the $50 enrollment payment. Enrollees in the control arm also received the $50 enrollment payment. Participants were informed that they would receive a check after the completion of a behavior (eg, after attending a session) within 2 months from the date of their most recent activity.

Analysis

We built participant-level random-effect models to estimate the effect of each experimental arm on weight change and on odds of meeting weight loss goals over 16 weeks, controlling for participant age, gender, race, and baseline weight. Differences in session attendance rate by arm were estimated using generalized linear models. These models incorporate all observed data on participants and implicitly assume that any missed outcomes are missing at random.¹⁴ Analyses were done separately for participants with higher baseline weight and for participants who did not experience significant payment delay. Since there was no true baseline measurement of weight, we identified baseline as the first weight data obtained in sessions attended within the first 3 weeks of study. To preserve the integrity of intention-to-treat analyses, the analyses included those participants who did not have the opportunity to complete the program (either because the vendor stopped the program early or because the participants lost Medicaid eligibility during the study; n = 14) but were still randomized to a treatment arm. Any eligible participant who attended at least 1 DPP session was included in analyses. All analyses were performed using SAS version 9.4 (SAS Institute, Atlanta, GA).

Results

Weight Loss

Moderate weight loss was observed across all arms at 16 weeks (Figure 2). However, participants in the intervention arms did not lose more weight relative to control (outcome incentives: −6.6 lb [−9.1 to −4.1 lb], comparison to control P = .48; process incentives: −7.3 lb [−9.5 to −5.1 lb], comparison to control P = .73; combined incentives: −5.8 lb [−8.8 to −2.8 lb], comparison to control P = .29; control: −7.9 lb [−11.1 to −4.7 lb]).

Figure 2.

Weight change over time by arm. Note: Weight change relative to baseline is measured as weight change relative to the first weight-in captured in weeks 1-3; thus, the measure of weight change begins in week 4.

To test the possibility that those with higher baseline weight were more likely to lose weight in response to incentives, we refined our analysis by evaluating differences in outcomes for those whose baseline weight was above 230 lb (the third quartile in baseline weight distribution), but found similar results. We further utilized weight data from adjacent weeks around study end points to reduce missing weight values at 16 weeks (using weight from week 15 if 16-week weight was missing), but the results remained unchanged.

Goal Achievement

Participants who were weighed at 16 weeks were no different in their likelihood of meeting weight loss goals, defined as losing at least 5% of baseline weight, across the 4 arms (control: 41%, process incentives: 43%, outcome incentives: 36%, combined incentives: 19%, omnibus P = .18).

Diabetes Prevention Program Session Attendance

Session attendance was relatively high at week 1 (above 75% for control, process incentives, and outcome incentives; 57% for combined incentives arm) but steadily declined throughout the study period (Figure 3), such that attendance at any given session was less than 50% in weeks 7 to 16 and cumulative attendance through the 16-week period was at or below 50% for all 4 arms (31% for control, 50% for process incentives, 38% for outcome incentives, 28% for combined incentives). Higher attendance was observed in the process incentive arm compared to control (P < .001) or combined incentives (P < .001) and for the outcome incentive arm than combined incentives (P = .009). Session attendance was negatively correlated with observed weight, r = −0.45, P <.0001, that is, we observed greater weight loss among those who attended a greater number of sessions.

Figure 3.

Session attendance by arm.

Additional Analyses

Actual payments were distributed from 3 to 372 days after eligible behaviors were completed, with a median of 36 days. The number of days between activity completion and payment distribution was not equivalent across arms (P < .01). Tukey post hoc tests (α = .05) revealed that average time taken to receive payments was significantly shorter for those in the process incentive (36.8 days) arm than those in the combined incentive (51.5 days) or those in the control (51.7 days) arms but was not significantly different from those in the outcome incentive (41.0 days) arm. Because the effect of incentivizing processes or outcomes may be particularly attenuated by the unexpected logistical delays in delivering initial payments to study participants, we replicated all primary analyses after excluding participants whose initial payment was received after 50 days (75th percentile) or longer and found similar results.

The proportion of participants who weighed in each week decreased across all arms through the 16-week period, indicating substantial attrition throughout the course of the study. Relative to the number of participants who weighed in during week 1, the control arm experienced the highest decline (62%) over the course of the study, while the intervention arms had smaller (though still sizable) absolute declines in weigh-ins (outcome incentive 47%, process incentive 42%, combined incentive 37%, omnibus P < .01).

Discussion

For New York State Medicaid Managed Care plan enrollees in a DPP, weight loss over a 16-week period did not significantly differ when participants received financial incentives for achievement of a targeted process (attendance at DPP sessions), a targeted outcome (achievement of weight loss goal), or the combination of the two, relative to control. On average, participants in all arms, including the control arm, did lose weight, suggesting that the program attracted highly motivated participants and that joining the program was associated with success in losing weight. Process incentives were effective at improving DPP session attendance; process incentive arm participants achieved a higher rate of attendance than the control and combined incentive groups throughout the entire 16-week period.

Although the present work suggests that a process incentive may produce the largest measurable changes in behavior, at least for attendance at a DPP session, it does not appear that increased attendance by the process arm participants leads to increased weight loss. It is worth noting that while there is an overall correlation between program attendance and weight loss, we cannot say how much of this reflects a “treatment effect” of the DPP and how much reflects a tendency for more motivated individuals to attend more sessions. However, the fact that the process incentive arm attended at higher rates but did not lose more weight suggests that the overall correlation between weight loss and program attendance may mostly reflect a selection effect, rather than that increased attendance in the process arm caused increased weight loss. In addition, given that all participants in the study were motivated enough to accept the invitation to participate in the DPP, it is possible that in this type of self-selected sample, the additional motivational impact of financial incentives was small. The weight loss achieved by those in the control arm suggests that there was considerable motivation and effort expended toward weight loss by all those who chose to participate in this study, each of whom received the DPP for free. The provision of the program itself, in fact, may have constituted a strong incentive for weight loss that could have reduced the additional impact of financial incentives due to diminishing marginal sensitivity to incentives.

The present study did have limitations. At the time of this study, the DPPs did not conduct sessions in languages other than English, Spanish, or Chinese, limiting member recruitment to English-, Spanish-, and Chinese-speaking members in the Medicaid managed care plans and HIV Special Needs Plans. Another central limitation concerned the infrastructure put in place to administer the incentives to participants. The initial protocol for the MIPCD program was to deliver money to participants within 2 weeks of a completed behavior—a time line that was met for only 6% of participants. Due to challenges in securing a private payment vendor, checks to participants had to be manually processed after the data were passed from the DPP to the managed care plan to the DOH, resulting in delays in distributing funds. In addition, the Medicaid population is particularly likely to have transient living situations, and several delays in program payments occurred because of changes in address. Some of the participant attrition that was observed over the course of the study may be attributed to these unforeseen delays in initial study payments, and past research shows that the motivational impact of promised financial incentives can be dampened when incentive delivery is not immediate.¹⁵ Third, the time interval of the study was relatively short, meaning that even though the program was advertised as a “DPP,” the current trial is unable to identify whether the program or different experimental arms successfully prevented or delayed the onset of diabetes for participants.

Fourth, the Medicaid managed care plans were required to establish and implement the DPP and assume the per-participant cost associated with DPP. Attempts to implement a DPP were protracted by many of the Medicaid managed care plans, due to the need to partner with outside vendors to conduct the program, underscoring how costs, timing, and other restrictions created barriers to implementation and ultimately reduced standardization across sessions in terms of participant recruitment and engagement. In New York, the DPP is not included as a Medicaid benefit, nor is transportation to DPP sessions directly covered by Medicaid. Although some Medicaid managed care plans voluntarily provided transportation to their enrolled members to address a reported barrier for attending DPP sessions, many participant barriers like transportation or lack of child care during sessions may have limited utilization of DPP, and participant attrition was a major concern across all experimental arms. To address this limitation, models incorporated a missing-at-random assumption, and when possible, missing values were imputed using temporally proximate weigh-ins for those who attended sessions in other weeks.

Although the current study did not demonstrate a significant effect of incentives on the primary outcome of weight loss over 16 weeks, many other studies have shown that incentives can promote healthier behaviors.^4,6,7,16,17 The degree to which these benefits of incentives can be replicated within Medicaid programs remains unclear, in part because the delays in administering payments in the current study—which may have reduced the effectiveness of the incentives—may be ameliorated if incentives are implemented within Medicaid managed care plans and more streamlined processes are developed outside the context of a research study. Incentive magnitude may also have been an important limitation here, as the incentives that were offered were relatively modest; larger incentives would likely have had larger motivational effects, but managed care plans are limited in the size of incentive they can offer to beneficiaries. Finally, weight loss and session attendance in the DPP were correlated in the present study, but using process-based incentives to directly increase session attendance did not lead to increased weight loss for those in the process incentives arm. Similarly, a weight loss study that provided monetary incentives for completing supervised exercise found that incentives increased attendance but did not improve long-term weight loss.¹⁸ Future research could explore not only the impact of process-based incentives on different outcomes but could also explore different types of processes to incentivize which might have a more causal impact on weight loss.

So What? Implications for Health Promotion Practitioners and Researchers

The provision of up to $240 in financial incentives did not lead to significantly greater weight loss in Medicaid managed care beneficiaries participating in a diabetes prevention program, regardless of whether incentives were tied to outcomes or to processes associated with weight loss or a combination of the two. Issues with participant retention and payment delays may have limited the effectiveness of the incentive interventions. Continued research is needed to better identify if and to what extent financial incentives could motivate Medicaid recipients to engage in behaviors that promote weight loss and prevent diabetes onset. Special attention should be paid to logistical concerns and minimizing participant attrition prior to further incentive provision expansion to ensure that any effects of incentives can be properly detected.

Footnotes

Appendix A

Authors’ Note

The contents provided are solely the responsibility of the authors and do not necessarily represent the official views of the Department of Health and Human Services or any of its agencies. The results presented have not been verified by the independent evaluation contract.

During the revision process, several of the authors changed institutions. The authors listed below completed some of their work for this article at their prior institution, as listed, and have provided this notice to acknowledge this earlier work.

Eric VanEpps was previously at the Corporal Michael J. Crescenz VA Medical Center and at the Center for Health Incentives and Behavioral Economics, Leonard Davis Institute of Health Economics, both in Philadelphia, PA. Joseph Anarella and Patrick Roohan were both previously at the New York State Department of Health, Albany, NY. All three authors completed part of their work for this article at these prior institutions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The project described was supported by Funding Opportunity Number CMS-1B1-11-001 from the Centers for Medicare & Medicaid Services.

References

The Kaiser Family Foundation State Health Facts. Data Source: Urban Institute estimates based on data from CMS, as of January 2017. http://kff.org/statedata?state=NY Accessed June 11, 2018.

Anderson

Chronic care: making the case for ongoing care. Robert Wood Johnson Foundation. 2010. http://www.rwjf.org/content/dam/farm/reports/reports/2010/rwjf54583. Accessed June 11, 2018.

Halpern

French

Small

. Randomized trial of four financial-incentive programs for smoking cessation. N Engl J Med. 2015;372(22):2108–2117.

Volpp

Troxel

Pauly

. A randomized, controlled trial of financial incentives for smoking cessation. N Engl J Med. 2009;360(7):699–709.

Giuffrida

Torgerson

. Should we pay the patient? Review of financial incentives to enhance patient compliance. BMJ. 1997;315(7110):703–707.

Petry

Rash

Byrne

Ashraf

White

. Financial reinforcers for improving medication adherence: findings from a meta-analysis. Am J Med. 2012;125(9):888–896.

Jeffery

. Financial incentives and weight control. Prev Med. 2012;55(suppl):S61–S67.

Blumenthal

Saulsgiver

Norton

. Medicaid incentive programs to encourage healthy behavior show mixed results to date and should be studied and improved. Health Aff. 2013;32(3):497–507.

Diabetes Prevention Program Research Group. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med. 2002;346(6):393–403.

10.

Centers for Disease Control and Prevention. Behavioral Risk Factor Surveillance System. Atlanta, GA: Centers for Disease Control and Prevention; 2009.

11.

Centers for Disease Control and Prevention. 2016. Diabetes At a Glance 2016. https://www.cdc.gov/chronicdisease/resources/publications/aag/pdf/2016/diabetes-aag.pdf Accessed June 11, 2018.

12.

Diabetes Prevention Program Research Group. The Diabetes Prevention Program. Design and methods for a clinical trial in the prevention of type 2 diabetes. Diabetes Care. 1999;22(4):623–634.

13.

Hamman

Wing

Edelstein

. Effect of weight loss with lifestyle intervention on risk of diabetes. Diab Care. 2006;29(9):2102–2107.

14.

Little

RJA

Rubin

. Statistical Analysis with Missing Data. Hoboken, NJ: John Wiley & Sons; 2014.

15.

Lussier

Heil

Mongeon

Badger

Higgins

. A meta-analysis of voucher-based reinforcement therapy for substance use disorders. Addiction. 2006;101(2):192–203.

16.

Sutherland

Christianson

Leatherman

. Impact of targeted financial incentives on personal health behavior: a review of the literature. Med Care Res Rev. 2008;65:(suppl 6):36S–78S.

17.

Cahill

Perera

. Competitions and incentives for smoking cessation. Cochrane Database Syst Rev. 2011;13(4):CD004307.

18.

Jeffery

Wing

Thorson

Burton

. Use of personal trainers and financial incentives to increase exercise in a behavioral weight-loss program. J Consult Clin Psych. 1998;66(5):777–783.