Behavior change websites not only offer lower cost per participant than face-to-face contacts with counselors, but also hold promise as an unexplored new mechanism for supporting long-term behavior change. Our experience developing the WLM behavior change website shows that high rates of use can be maintained for at least 1 year. Participation in this study was much greater than in some other long-term behavioral interventions [40
], but similar to that seen in the Stop-Regain trial [43
We have identified several major lessons from our experience developing and implementing the WLM interactive website. First, it is essential to specify the theoretical foundation of the intervention program and the website objectives early in the design process. Website design does not iterate the same way as development of in-person, counseling-based interventions. Making clear decisions about intervention objectives and abiding by them during the design process helps eliminate costly rework.
Our second key learning was that detailed paper prototypes and specification documents should always precede programming. While the design team may want to jump directly to screenshots before the logic has been thoroughly outlined, the results tend to be better if they wait for the paper prototyping to be completed. In the WLM, we called paper prototype meetings “wall meetings” because hours were spent taping freehand paper “screens” to the wall and determining the outcome of every link. These wall meetings resulted in many modifications and intervention improvements that would have been difficult during later stages in the development process. The final set of freehand paper screens was also very useful when writing the detailed specifications (use case) for the programmer.
A final key learning was to not underestimate the essential role of a product manager. To have all groups doing what they do best requires a central team member to intersect with each group. We believe that the product manager must be able to manage in all three domains (concept, content, and code) to ensure an effective website design process. The product manager helps to keep the development team working toward a common goal while serving as a translator for those working at different levels of detail.
Additionally, we gained insight into participant use of an interactive website and what is required to keep participants engaged. Based on preliminary “hit” counts (data not shown), interactive features like the weight entry form and the bulletin board discussion appeared to be most popular. We were concerned that a website intervention would have problems maintaining the interest of participants for long-term follow-up. We found, however, that a high rate of participation can be achieved for at least 12 months with an interactive website. The automated email and telephone reminders were quite effective in prompting regular use of the website for at least 1 year. In an 8-week, stage-based physical activity website intervention, Leslie et al [39
] determined that emails prompted return visits to a website (77% returned after email prompts), but that the same emails were not helpful for encouraging new users to visit for the first time. In contrast to the Leslie et al study, participants in WLM were initially oriented to the website through a 1-hour individual visit with an interventionist. Personalized website orientation may be a critical factor in the effectiveness of subsequent email prompts that encourage returning to the website. Another study [44
] reported that well-constructed email messages can have a beneficial effect on diet and exercise health behaviors. The authors suggest, however, that the email messages may need further tailoring and grounding in health behavior change theory to strengthen their potential. The WLM email prompts were customized to the individual user, easy to read, provided choice rather than advice, and included a link for easy access to the WLM website. These factors were potentially positive contributors to the effectiveness of our email prompts. We were initially concerned about the number and frequency of email reminders sent to study participants. We only sent email prompts when the participant met the specified criteria of not logging on to the website within 1 week of the last log-in. However, during the first 6 months of the intervention, 83% of participants received a weekly email prompt to log on to the website, providing us with two important lessons: (1) participants are not bothered by reminders to return to the website, and (2) in general, participants do not set up outside reminders to log in, but simply wait to be prompted.
Finally, it is important to note the current limitations of Web-based programs. Developers of behavior change websites must be prepared to continually update the product and limit the use of available technologies in consideration of bandwidth limitations. Danaher et al [45
] urge developers to consider the bandwidth necessary to operate rich-media websites as a possible barrier to participant use. User frustration resulting from long page downloads presents a near terminal problem for researchers looking to test behavioral website use. Therefore, we limited the bandwidth requirements of the WLM website to accommodate those with limited bandwidth. For this reason, the WLM website is devoid of photos, moving text, video clips, and music. Also acknowledged by Danaher et al [45
], the scalability of a behavior change website must be considered at the time of development. The capacity of a website to “grow” beyond its current capacity is an essential consideration.
Given our study timeline, we developed and implemented this website in 12 months. Looking back on our experience, additional development time would have been beneficial in three key processes: (1) at least 6 months of general beta testing (we had 3 months), (2) an even richer understanding of the user experience from the pilot participants (ie, periodic individual interviews to understand how/if the user experience evolved over time), and (3) additional opportunities to test multiple prompting strategies for encouraging participants to continue using the site. Even without such additional development time, use of this website remained high throughout the first year, with over 80% of the participants continuing to be active users during the 12th month.