Social influence as a driver of engagement in a web-based health intervention

Abstract

Background: Web-based health interventions can drive behavior change, but their effectiveness depends on participants' usage. A well-recognized challenge with these interventions is nonusage attrition or weak engagement that results in participants receiving low doses of the intervention, negatively affecting outcomes. We present an approach based on the theoretical concepts of social influence and complex contagion in an effort to address the engagement problem in a specific, commercial, online behavior change intervention.

Objective: To examine the relation between social ties and engagement within a specific online intervention. The aims were (1) to determine whether experiencing the intervention socially influences engagement, such that individuals with social ties show higher engagement than those without ties, and (2) to evaluate whether complex contagion increases engagement-that is, whether engagement increases as the number of ties an individual has in the intervention increases.

Methods: We analyzed observational data from 84,828 subscribed members of a specific Web-based intervention, Daily Challenge. We compiled three measures of engagement for every member: email opens, site visits, and challenge completions (response to action prompts). We compared members with and without social ties within the intervention on each measure separately using 2-tailed independent-sample t tests. Finally, we performed linear regressions with each simple engagement measure as the dependent variable and number of social ties as the independent variable.

Results: Compared with those without social ties, participants with social ties opened more emails (33.0% vs 27.2%, P < .001), visited the website more often (12.6 vs 6.7 visits, P < .001), and reported completing more of the actions they were prompted to perform (11.0 vs 6.1 actions, P < .001). Social ties were significant predictors of email opens (beta = 0.68, P < .001), site visits (beta = 1.52, P < .001), and reported action completions (beta = 1.32, P < .001).

Conclusions: Our initial findings are higher engagement in participants with social ties in the program and are consistent with the view that social influence can drive engagement in a Web-based health intervention.

Figure 1

Screenshots of the Web-based intervention Daily Challenge. Top left: A challenge with its Done button, as seen on the site. Top right: An email reminder with social proof. Bottom: Activity stream in which members can follow their connections’ activity, including challenge completions.

Figure 2

Distribution of the number of connections social participants formed within Daily Challenge.

Figure 3

Distribution over time of friendship formation for participants’ first month in Daily Challenge. X-axis: Participants’ days in the program. Y-axis: Percentage of connections formed.

Figure 4

Email opens, site visits, and completion rates per number of personal connections in Daily Challenge. For illustrative clarity, aggregate data are shown. The category 15+ encompasses less than 0.2% of the participants.