A Randomized Trial of an Avatar-Hosted Multiple Behavior Change Intervention for Young Adult Smokers

Abstract

Background

Young adulthood is a critical transition period for the development of health behaviors. We present here the results of a randomized controlled trial of an online avatar-hosted personal health makeover program designed for young adult smokers.

Methods

We conducted a three-group randomized trial comparing delivery of general lifestyle content (Tx1), personally tailored health information (Tx2), and personally tailored health information plus online video–based peer coaching (Tx3) as part of a 6-week online health program. Participants were asked to set weekly goals around eating breakfast, exercise, alcohol use, and cigarette smoking. Eligibility criteria included age (18–30 years) and smoking status (any cigarette use in the previous 30 days). The primary outcome was self-reported 30-day abstinence measured 12 weeks postenrollment.

Results

Participant (n = 1698) characteristics were balanced across the groups (72% women, mean age 24, 26% nonwhite, 32% high school education or less, and 50% daily smokers). Considering intention to treat, 30-day smoking abstinence rates were statistically significantly higher in the intervention groups (Tx1 = 11%, Tx2 = 23%, Tx3 = 31%, P < .001). Participants in the intervention groups were also more likely to reduce their number of days spent on binge drinking and increase their number of days eating breakfast and exercising. Overall, intervention group participants were much more likely to make positive changes in at least three or four of the target behaviors (Tx1 = 19%, Tx2 = 39%, Tx3 = 41%, P < .001).

Conclusions

This online avatar-hosted personal health makeover “show” increased smoking abstinence and induced positive changes in multiple related health behaviors. Addition of the online video–based peer coaching further improved behavioral outcomes.

The Internet is a promising channel to support smoking cessation among young adults (1,2). A growing body of evidence supports the efficacy of online smoking cessation interventions (3–8). Our own work with the RealU study demonstrated the efficacy of an online cessation intervention for college smokers (9). Although the results of the original RealU study were promising, important questions remained unanswered. First, the original intervention combined online cessation messages with peer e-mail support. Thus, it was not possible to identify the relative contribution of these components. Second, the initial study enrolled primarily white college students from a single campus, leaving unanswered the efficacy of this approach in a more diverse group of young adults. We conceived a follow-up study, the RealU2, to address these limitations.

As we approached the design of the RealU2 intervention, we could not help but note that major changes had occurred in the online environment since the original RealU study. A large part of these changes can be considered to be part of a continued evolution of the Web 2.0 paradigm, which broadly describes a shift in the creation and control of online content from a relatively small group of expert sources to a larger community of users (10,11). For example, between the launch of the original RealU study (Fall 2004) and the planned launch of the RealU2 (Spring 2011), Facebook had grown from a single campus start-up to a worldwide entity with more than 500 million users (12,13). At the launch of the original RealU study, the first video had yet to be posted on YouTube (14). The increased interactivity and richness of media in the online environment created a daunting challenge. We were concerned that a simple static or text-based site would not be perceived as sufficiently engaging by a target audience that had grown used to rich interactive online media.

We adopted several innovations in the design of the RealU2 intervention. First, we shifted the mode of online peer support from e-mail to delivery of personalized video messages. Second, we designed the RealU2 to have a more interactive user-centered focus by incorporating high-depth tailoring of health messages (15–23). We also sought to tie into the popularity of “reality” entertainment by presenting the RealU2 as an online personal health “makeover,” where an individual user’s health goals and progress would become a central part of the site experience (24). Finally, we adopted the use of an avatar (ie, digital character) to serve as the host of the online personal makeover show. Use of avatars and virtual characters is a common engagement strategy in the electronic and online gaming environments that are popular among young adults (25).

We report here the results of the RealU2 randomized controlled trial. The primary objective of this study was to determine whether providing individually tailored cessation messages with or without online peer support increased self-reported 30-day abstinence from cigarette use among young adult smokers. We also examined the effect of the intervention on several secondary health behaviors, including alcohol use, breakfast consumption, and exercise.

Methods

Overview

The RealU2 is a three-group randomized controlled trial. Partic ipants were assigned to receive one of three treatments: 1) untailored general interest messages (ie, Treatment 1, control); 2) individually tailored health messages (ie, Treatment 2); or 3) individually tailored health messages plus online peer support (ie, Treatment 3). All transmissions of data to and from the Web site were encrypted using Secure Socket Layer (SSL) protocols. The Institutional Review Boards of the University of Michigan and the University of Minnesota approved all study methods.

Setting and Participants

Participants were recruited via a national online sample of young adults provided by Survey Sampling International. Survey Sampling International invited members of their panel in the age group of 18–30 years who reported a history of cigarette smoking to complete an eligibility screening survey for this study. All participants who completed the survey were entered into a drawing for one of five $100 cash prizes or an Apple iPad.

Participants were eligible for the full RealU2 study if they 1) were between the ages of 18 and 30, 2) had smoked at least one puff of a cigarette in the previous 30 days, 3) had Internet access for the next 3 months, 4) used the Internet more than twice per week, and 5) lived in the United States. Eligible individuals were shown a brief video inviting them to participate in the RealU2 study.

Enrollment and Randomization

Interested individuals could then enroll in the RealU2 by providing online informed consent, picking one of 18 avatars (representing different genders and ethnicities) to serve as the “host” for their online makeover, completing a baseline survey providing more detailed information regarding the four target health behaviors (ie, cigarette smoking, alcohol use, exercise, and eating breakfast). Individuals who completed the baseline survey were called by study staff to confirm eligibility and consent; they were assigned to one of the three study arms using a blocked random number sequence generated by the study statistician. Individuals who completed the enrollment process received $10.

Intervention Groups

All participants received e-mail reminders and financial incentives ($10 per week) to make up to six weekly visits to the study Web site. The control program (ie, Treatment 1) consisted of six sessions of general interest (ie, not health-related) lifestyle content. Weekly topics included music, finances, relationships, living green, movies, and online dating.

The tailored health message group targeted four behaviors as part of a general wellness–framed Web site: cigarette smoking, alcohol use, exercise, and eating breakfast. The intervention strategies and content were grounded in social cognitive theory (26–32), the theory of reasoned action and planned behavior (33–36), and the self-determination theory (37–39). Week 1 focused on building social support for healthy lifestyles. Week 2 focused on eating healthy breakfasts. Week 3 focused on increasing exercise. Week 4 encouraged smoking cessation or reduction in smoking. Week 5 encouraged responsible drinking or abstinence from drinking. Week 6 addressed the “total lifestyle” by asking participants to consider all four of these behaviors. Each weekly session followed the same basic five-step process (See Supplemental Materials, available online, for details).

• Step 1: Check-In. Participants visit the site and report on their health behaviors.

• Step 2: Why Page. Participants receive motivational messages tailored to individual outcome expectancies for the target behavior.

• Step 3: Goal Page. Participants receive goal-setting messages tailored to their individual self-efficacy and social support for the target behavior.

• Step 4: How Page. Participants receive strategy messages tailored to address perceived barriers to changing the target behavior.

• Step 5: Home Page. Summarizes current and past week content.

The participant’s avatar “makeover host” was integrated through each of these steps. The avatar encouraged behavior change with dialog that was tailored to the individual’s experience the previous week. To add variety each week, the avatar was placed within a different magazine-themed environment with an often-humorous connection to the target behavior (eg, a National Geographic–themed “Finding Breakfast in an Urban Jungle”, a Yoga magazine–themed “Lung Power”). The avatar also augmented delivery of tailored messages by holding up signs with tailored text and making verbal references and gestures to highlight tailored content in the surrounding page.

Additional intervention components in the tailored health message group included a behavior tracking progress bar, a weekly interactive goal-setting activity, and behavior change testimonials from young adults making similar health behavior changes. Participants in the tailored health message group were also asked to set a goal each week for the target behavior. In keeping with the principle of autonomy in self-determination theory, individuals were provided free choice about their weekly behavioral goals (ie, goals not constrained to any predetermined ranges). Finally, the site incorporated photograph–diary testimonials from other young adults making positive behavior changes.

Participants who were assigned to the tailored health message plus online peer coaching group received the same message content as those in the tailored health message group. In addition, participants in the peer coaching group selected an online coach from a team of twelve peer coaches. Each week, the peer coach would view each of their assigned participants’ behavioral tracking progress charts and health goals and record a personal video message. This message addressed the participant’s degree of success with their health goal from the previous week, reviewed trends in the target behavior for the upcoming week, and reinforced the participant’s motivation and confidence for their current health goal. Approximately 1–2 days after recording and uploading these personal video messages, the peer coaches would make a brief phone call to each of their assigned participants to assess progress toward meeting their weekly health goal. Peer coaches were undergraduate students or recent graduates from the University of Michigan with interest in a health-related field. Peer coaches received a 3-day training in motivational interviewing by a recognized expert in the field (Ken Resnicow, School of Public Health, University of Michigan) and participated in weekly coach meetings. Ten percent of the video messages and phone calls were monitored by a counseling supervisor to ensure adherence to protocol.

Measures and Analysis

The screening and baseline surveys assessed a range of health behaviors, including self-reported cigarette use, alcohol use (any alcohol use and binge drinking defined as five or more drinks on one occasion), eating breakfast, and exercise (exercise of 20 minutes or more per day). Participants reported the number of days in the past 30 days they had engaged in each of these activities. We also collected demographic information, including age, gender, ethnicity, race, education level, and current student status.

Behavioral outcomes were measured at 7 and 12 weeks postenrollment. The main outcome measure was self-reported 30-day abstinence from cigarette smoking at 12 weeks postenrollment (calculated by intention to treat, with nonrespondents considered as continuing to smoke). Secondary outcomes included change in the number of days of alcohol use (days drinking and episodes of binge drinking), eating breakfast, and exercise in the previous 30 days, again measured at 12 weeks postenrollment. We also assessed the number of improved behaviors for each participant. An individual was considered as improving in their smoking behavior if he or she reported being abstinent for the past 30 days on follow-up, as improving their alcohol behavior if they decreased the number of days of binge drinking, and as improving their exercise or breakfast easting if they increased their number of days of these behaviors (from baseline to 12-week follow-up). Individuals received a $10 incentive to complete the 7-week evaluation and a $20 incentive to complete the 12-week evaluation.

Results

Figure 1 shows the recruitment flow for the RealU2 project. Of the 18 548 individuals who were directed to our screening survey by the Survey Sampling International, 9360 (50.5%) were eligible for the study. Reasons for ineligibility included not having smoked in the past 30 days (n = 5261), being younger than 18 (n = 69) or older than 30 (n = 348), living outside the United States (n = 20), not having Internet access for the next 3 months (n = 1215), and accessing the Internet fewer than two times per week (n = 204). Of the eligible smokers, 2136 (22.8%) consented online to be a part of the full study and 1698 (18.1%) were reached by phone and assigned to a study arm based on a blocked randomization scheme.

Figure 1.

RealU2 study flow.

The three study arms were comparable in terms of the demographic and behavioral variables (Table 1). There were no statistically significant baseline differences between the study groups.

Table 1.

Characteristics of the study population

Characteristic	Total	General lifestyle (control)	Tailored health message	Tailored health + peer coach	P
Gender (%)
Men	27.56	26.28	27.03	29.38	.4763
Women	72.44	73.72	72.97	70.62
Age (mean)	24.07	24.14	23.96	24.12	.6445
Ethnicity (%)
Hispanic/Latino	10.95	10.23	10.95	11.68	.7364
Non-Hispanic/Latino	89.04	89.77	89.05	88.32
Race (%)
White	73.91	71.96	76.86	72.92	.2466
Black or African American	10.36	11.64	9.01	10.44
Other	8.60	8.29	9.01	8.49
Multiple	7.13	8.11	5.12	8.14
Education (highest level, %)
High school or less	32.63	34.04	32.51	31.33	.5145
Some college + 2-year degree	50.35	48.50	52.29	50.26
4-year degree or more	17.02	17.46	15.19	18.41
Student status (%)
Currently enrolled in school	36.93	35.98	36.75	38.05	.7656
Not enrolled in school	63.07	64.02	63.25	61.95
Cigarette smoking
No. of days smoked in past 30 days (mean)	20.98	21.06	20.99	20.88	.9640
Proportion of daily smokers	49.71	49.74	49.29	50.09	.965
Average cigarettes/day (daily smokers)	19.82	19.63	23.37	16.52	.3570
Plans to quit in next 30 days (%)	46.94	46.74	45.05	49.03	.4053
Alcohol
No. of days consumed any alcohol in past 30 days	6.87	6.59	7.16	6.86	.4125
No. of days binge drinking in past 30 days	3.16	2.99	3.16	3.33	.5064
Breakfast
No. of days eating healthy breakfast in past 30 days	10.84	10.36	10.57	11.58	.0651
Exercise
No. of days exercising in past 30 days	10.51	10.38	10.04	11.10	.1355

Adherence to the weekly Web sessions was similarly high among all three study groups, with approximately 80% of participants completing at least 4 of the 6 weeks of the weekly sessions (80.78%, 81.98%, and 78.94% for the three study groups; P = .429). Average completion rate of weekly check-in surveys was 82.6% in the tailored health message group and 80.6% in the tailored health plus peer coach group (P = .254). For the tailored health message plus online peer coaching group, we estimated that 48% of video postings resulted in additional visits to the site within a given week to view the video. An average of 60% of weekly follow-up telephone calls were completed by peer coaches. Follow-up rates were 80% at week 7 and 75% at week 12 in all study groups (with no statistically significant difference between groups).

Figure 2 shows the 30-day abstinence rates at 12-week follow-up for the study groups. The percentages of participants who abstained from smoking for the 12-week evaluation were 11%, 23%, and 31% for Treatment 1 (general lifestyle), Treatment 2 (tailored health message), and Treatment 3 (tailored health plus peer coaching), respectively. These differences were statistically significant: P value < .0001 overall and for each tailored message group compared with the general lifestyle group. The difference between the tailored heath message groups (with vs without peer coaching) was also statistically significant (P = .0058).

Figure 2.

Smoking abstinence by study arm.

The mean changes in the number of days of drinking, eating breakfast, and exercise in the past 30 days from baseline to the 12-week follow-up are shown in Table 2. Examination of the 95% confidence intervals for these differences revealed positive changes in both tailored message treatment groups (Treatment 2 and 3) vs the control. Figure 3 shows the number of improved behaviors from baseline to the 12-week follow-up by study arm. This comparison reveals that participants in both tailored message groups (with or without peer coaching) were much more likely to make multiple improvements in the four targeted health behaviors compared with those in the general lifestyle group (P< .0001). For example, nearly 40% of participants in the tailored health message groups made a positive change in three or four of the targeted health behaviors (compared with only 19% of those in the general lifestyle group).

Table 2.

Change in drinking, breakfast, and exercise behaviors from baseline to 12-week follow-up*

	Mean	95% CI
		Lower	Upper
Number of days of any drinking
General lifestyle content	−0.79	−1.09	−0.49
Tailored health messages	−1.90	−2.20	−1.61
Tailored message plus peer coach	−2.34	−2.63	−2.05
Number of days of binge drinking
General lifestyle content	−1.00	−1.21	−0.79
Tailored health messages	−1.16	−1.35	−0.97
Tailored message plus peer coach	−1.87	−2.09	−1.64
Number of days of eating breakfast
General lifestyle content	2.64	2.20	3.07
Tailored health messages	6.51	6.04	6.99
Tailored message plus peer coach	7.58	7.10	8.06
Number of days of exercise
General lifestyle content	1.74	1.30	2.18
Tailored health messages	4.57	4.11	5.04
Tailored message plus peer coach	5.08	4.61	5.56

*CI = confidence interval.

Figure 3.

Count of improved behaviors by study arm.

Discussion

The main finding from this study is that an avatar-hosted online personal health makeover increased rates of smoking abstinence among young adults. This is consistent with our previous work demonstrating the efficacy of an online intervention for college smokers (9). There are, however, several notable features of the current study. This study engaged a more diverse sample of young adults in terms of education (only about one-third of participants in college), ethnicity (26% were nonwhite), and smoking behavior (approximately one-half were daily smokers). Taken together, these points support the notion that online intervention is an appropriate strategy to encourage and support smoking cessation among a wide range of young adults.

The three-arm design of this study also allowed us to examine the additive effect of online peer coaching based on the priciples of motivational interviewing. Our results demonstrate a clear benefit of the peer coaching intervention, which combined delivery of personalized videos with brief phone follow-up calls. Future studies could be designed to disentangle the effects of video messaging and telephone contact.

A key process finding was the high rates of intervention engagement across all three groups. Several aspects of the current intervention likely contributed to these high utilization rates. As in our previous work, we continued to incorporate issues of general lifestyle interest to our target audience [eg, music, movies, travel (2)]. The framing of the intervention as a “personal health makeover” also allowed us the opportunity to engage with participants on a range of health topics (eg, diet/nutrition, exercise, alcohol use), in addition to smoking cessation. This intervention incorporated high-depth tailored messages around motivation and strategies for change in various health behaviors. Our previous work has shown that high-depth tailoring increases perceived relevance of intervention content and intervention adherence (40). Finally, incorporation of an avatar to host the online personal makeover may have contributed to program adherence. Future studies could examine the separate and combined effects of these elements (eg, intervention with or without avatar) on intervention adherence and outcomes.

Finally, it is important to note that this intervention was associated with positive change in multiple health behaviors, something reported in only a handful of previous studies. Part of the reason for success in this intervention may be the potentially synergistic nature of change in the targeted health behaviors (41–44). Future studies can explicitly compare interventions focused on single behaviors (eg, smoking cessation) with interventions that target multiple behaviors (eg, smoking and related behaviors).

There are several limitations to this study. First, the findings of this study were based on self-report. Self-report of smoking status is generally considered accurate when there is not a high demand to report abstinence [eg, financial incentive or substantial social pressure (45–47)]. There were no high-demand characteristics in this study. Reporting of dietary practices (ie, eating breakfast) and exercise (ie, number of days with 20 minutes or more of exercise) in this study was limited to brief measures. For example, we did not assess the nutritional content of additional breakfasts consumed or the change in the total duration or intensity of exercise performed by study participants. Future studies could include more rigorous measures of diet and activity, including objective monitoring of the latter. Finally, it is important to acknowledge that our assessment of smoking abstinence was based on a relatively short-term follow-up (ie, 12 weeks postenrollment). Future studies could incorporate longer follow-up periods to examine the durability of intervention effects.

Several areas for future work are particularly intriguing. First, it is interesting to consider an expanded role for avatar characters. For example, in future studies, avatars might provide some of the video feedback delivered by live peer coaches in this study. Second, the continued Web 2.0 shift toward user-generated content suggests novel intervention strategies. For example, supportive video messages could be generated by users themselves (eg, as a form of self-affirmation), by members of a user’s online social network (eg, Facebook friends), or by a more distant network of individuals who might come together around a shared interest in a particular behavior change (eg, smoking cessation, weight loss, etc.). Third, future interventions could consider the use of mobile or “smart phone” technologies (eg, activity monitoring, geolocation) to enhance participant engagement. Finally, it is important to consider which research methods will be most appropriate for future studies, given the rapidly changing consumer and health information landscape. Multiphase optimization studies (ie, MOST) and sequential multiple randomized assignment trial (ie, SMART) adaptive designs can help to identify the “active ingredients” of complex interventions and the appropriate sequencing of different intervention components [eg, transition from avatar to live peer coach support (48,49)]. Pragmatic trials or cohort multiple randomized controlled trial designs may be used to assess the effectiveness of new eHealth programs as they are introduced to larger and more diverse groups of potential users (50,51).

Funding

This study was funded by a grant from the National Institutes of Health (RO1 HL089491) to LCA.