Text Messaging for Weight Loss

Text messaging has been shown to be effective for weight-loss in very limited testing. DH has experience with both text message based intervention and weight management intervention. This study aims to enhance Denver Health's (DH) existing Patient Relationship Management (PRM) text message (SMS) infrastructure to include culturally-appropriate outreach to promote weight loss.

Subjects will be randomized into one of two arms:

• Usual care: Patients randomized to this arm will not receive text messages or any other weight-loss support besides usual care provided at DH. They will receive a weight loss educational packet and be asked to follow up with their primary care provider to further discuss their efforts at weight loss with additional follow-up as directed by their provider. They will be contacted periodically to be weighed. Providers will not be made aware that their patients are participating in the study's control arm.
• Text Message-based weight loss support. Patients will receive the same weight loss educational packet as those randomized to usual care. Patients will be assisted in choosing a self-management goal related to exercise and to eating behaviors. They will receive text messages at a frequency up to 1x daily; input from focus groups will guide text message frequency.

SMS content is similar to the Diabetes Prevention Program (DPP) curriculum and will fall into the following categories:

• Outgoing "tips of the day"
• Interactive messages that solicit a simple response for the day from the participant
• Outgoing reminders to inform participant about events in the community and about clinic appointments with their primary care provider.

Basic descriptive statistics for categorical variables will be generated to describe the baseline demographic and clinical characteristics. Univariate analyses will be performed to determine whether there are differences between patients in the three arms of the intervention. For clinical outcomes, general linear mixed effects models will be used with intervention and time (and their interactions) specified as fixed effects to determine whether change over time in outcomes differs significantly for the three groups. A random subject effect will be specified to model the correlation of observations taken on an individual. Contrasts will be constructed to test the difference in interventions at six months and 12 months. If the data do not follow an approximate normal distribution or cannot be normalized using log transformation the ranks of the outcomes will be analyzed in the mixed effects model. All statistical analyses will be performed using Statistical Analysis Software (SAS) version 9.2.

The study is powered to detect a difference at six months between the text message group and the usual care group. Previous research suggests that, with 60 participants per group, researchers would have 89% power to detect a difference between groups of 1.7 kg. The investigators think that this is a reasonable estimate, as subjects in the control group in this study lost 0.4 kg, and it is believed that subjects in the control group will be weight stable. While investigators anticipate that patients who participate in the DPP program will lose weight, it is believed participation across the usual care and SMS groups will be comparable as i) randomization is stratified to the two groups by past participation in the DPP and ii) relatively equal participation in the two groups is anticipated throughout the intervention period. This stratification is intended to equalize the effects of class-attendance heterogeneity across the two study groups. The study aims to have 90 participants per group to improve the power for sub-group analyses.