Evaluating the Impact of a Digitally-Based Diabetes Prevention Program in an Employer Population

Wednesday, June 26, 2019: 9:30 AM
Taft - Mezzanine Level (Marriott Wardman Park Hotel)

Presenter: Jean Abraham

Co-Authors: Helen Parsons; Bryan Dowd; Lucas Higuera; John Nyman

Discussant: Rebecca Myerson

Current estimates indicate that 30.3 million or 9.4% of Americans have diabetes mellitus, and that 86 million are pre-diabetic, based on elevated glucose levels. The National Diabetes Prevention Program is an evidence-based initiative sponsored by the Centers for Disease Control and Prevention (CDC) to reduce diabetes onset through lifestyle modification and weight loss. Translations of the Diabetes Prevention Program (DPP) protocol have been tested to determine their effectiveness in promoting weight loss in various “real-world” settings, including community-based organizations, healthcare provider organizations, and employer worksites. Online or digitally-based DPP translations represent an important development as they offer greater flexibility to individuals who may have access barriers due to work schedules or transportation.

This study uses a pre-post with comparison study design to evaluate the impact of Omada Health’s Prevent program, a digitally-based DPP launched in April 2015 for the State of Minnesota’s health benefits-covered population. Our evaluation uses two sources of administrative data: 1) 2014-2017 Health Risk Assessment data to capture information about individuals’ health status and health behaviors for the time period before and after Prevent was launched, and 2) medical claims and eligibility data to measure Prevent engagement through payments to Omada Health using unique CPT codes and modifiers at four points in time: 1) when an individual’s program enrollment is initiated, 2) when the participant completes 9 of 16 weekly lessons in the core educational phase; 3) when the participant achieves 5% body weight loss from baseline; and 4) when the participant achieves 10% body weight loss from baseline. Claims data are also used to measure pre-existing clinical diagnoses and onset of diabetes in the post-period as well as medical and pharmaceutical spending.

We begin by assessing the demographic, health status, and attitudinal factors associated with employees’ engagement levels with Prevent. Next, we use a difference-in-differences (DiD) regression approach to investigate how individuals’ engagement with Prevent change eating habits, exercise frequency, health status, medical care and drug spending relative to a comparison group of employees whose age and body mass index align with the program’s eligibility criteria. Finally, we examine Type 2 diabetes incidence in the post-period to assess whether participants had lower incidence relative to the comparison group.

Among eligible employees, 2,577 (19%) initiated enrollment. Of these, about two-thirds completed the core phase of the program. And among that subset, 51% achieved weight loss of at least five percent of their baseline weight. Heckman probit selection models reveal systematic differences by age, female status, job type, and clinical indications in employees’ probabilities of enrollment, completion of the core phase, and weight loss goal attainment. Our DiD results demonstrate statistically significant improvements in eating habits, exercise frequency, and self-reported health status in the initial post-period (approximately five months after core program completion). Notably, the magnitudes of these changes are modest relative to baseline values, they exhibit declining effects by the second post-period (17 months after the program), and they are concentrated among employees who had achieved their weight loss goal.