Designing Financial Incentives to Maximize Participation of Target Populations in Weight Loss Programs

Despite decades of interest in financial incentives as a means of improving weight loss outcomes, few studies address systematic design of the incentives, the effect of incentive on participation levels and representativeness, and heterogeneity in individual preferences for incentive designs. Our study utilizes a choice experiment in order to elicit preferences for different incentive configurations systematically, by varying five aspects of the configuration according to an experimental design. Our study focuses on participation rate rather than efficacy, and oversampled males and low-socioeconomic status individuals, to ensure representation of subgroups which are most vulnerable to obesity and have been historically understudied in weight loss. Our model allows  a high degree of heterogeneity, including individual-specific covariates, so that the findings can be targeted to vulnerable subgroups or customized to the demographic mix within a certain geographic region.

Our study utilizes a choice experiment, conducted via mail survey to patients of a healthcare network in southwest Virginia, on preferences for different aspects of a financial incentive design (accompanying a minimal weight loss program). The financial incentives were described in terms of reward magnitude, payment form, payment frequency, reward condition, and program location.

A choice model specifying main terms and interaction terms between design- specific and individual-specific covariates was estimated, and the results are presented in terms of predicted choice probabilities, using a nationally representative sample (NHANES) and geographic-specific samples (BRFSS) as the demographic base on which the predictions are made. We performed subgroup- specific analysis, focusing on populations vulnerable to obesity: African Americans and low-income women, as well as the full population.

We found significant preference heterogeneity, which resulted in multiple ‘best’ designs: the best design differed depending on the characteristics of the individual. Looking at all the different best designs, we found that the most popular attribute levels differed by subgroup, with low-income women overwhelmingly preferring the clinic location, while the other groups also preferred the community center. African Americans preferred the weight loss reward condition while the other groups preferred the attendance condition. African Americans also preferred the weekly and monthly payment frequencies whereas the quarterly frequency was more popular with the other two groups. Finally, all groups preferred the debit card payment form exclusively.

The existence of multiple best designs suggested that offering a menu of multiple designs would increase participation. We constructed menus of multiple designs and predicted the probability that a given individual would choose at least one design from the menu, and found that as expected, the choice probability increases with the number of designs offered. Furthermore, the variability in choice probabilities across the population decreases with the number of designs.

Increasing the reward magnitude, on the other hand, does not increase the choice probability across all individuals. Some individuals actually react negatively to increasing reward magnitude, and so variability in choice probability actually increases as the reward magnitude increases. However, the number of such individuals is relatively small.

See more of: Roles of Financial and Nonfinancial Incentives in Health Promotion
See more of: Oral Sessions

Previous Abstract | Next Abstract >>