This week I'm excited to share the results of a study that my colleagues and I just published in AIDS and Behavior. Here's the new article: https://link.springer.com/article/10.1007/s10461-026-05317-w The randomized trial in this paper is the first direct test of an intervention that was directly based on Two Minds Theory, and was conducted as part of a broader study on exercise to improve physical functioning in people with HIV (NIH grant #1R01AG066562, co-principal investigators Drs. Kristine Erlandson and Allison Webel). The first phase of that study involved patients exercising 3 times per week, with either continuous moderate-intensity exercise or high-intensity interval training (HIIT) exercise, for a total of 16 weeks. Then, at the end of the supervised exercise conditions, all patients in both groups were re-randomized to one of two behavioral intervention groups to help them maintain their exercise habits for another 3 months beyond the first exercise phase of the study. The two groups in this second phase of the study were (a) daily tailored text messages based on TMT, plus 4 sessions of motivational interviewing, or (b) daily educational text messages that weren't based on a specific behavioral theory, with some patients in this control group also receiving motivational interviewing [more on that below]. The outcome variable was minutes per day of moderate to vigorous physical activity, aggregated into a monthly average for each participant.
The marquee finding was that people in the tailored-messaging group started with a slightly lower average number of minutes of physical activity per day (solid line on the graph above), and ended up about where they started; meanwhile, people in the comparison educational-messaging group (dotted line on the graph) had a steady decrease in daily minutes of physical activity over the 3 months after the structured exercise protocol ended. There was some variability within each group (error bars on the graph), but overall the trend was statistically different between the two groups, showing an advantage for TMT-based tailored messages in helping people to sustain their exercise habits during the 3-month trial. We tried the analysis several different ways, including a conservative intent-to-treat approach for handling missing data (shown above) as well as an analysis where we only looked at people who had data throughout the 3-month follow-up period. And we found similar results using self-reported exercise minutes (again, that's what's shown above) and activity data collected by having participants wear a wrist-based movement sensor for 1 week per month.
The theory component of our tailored-messaging intervention involved two different things: First, we identified 10 different reasons why people might not exercise, and validated that list with two Community Advisory Board groups made up of patients with HIV who were receiving services at one of the two clinics involved in this study. The survey that participants completed each day on their phones allowed us to identify one of those 10 reasons as the most important concern that day, or to say that there were no current concerns, and the tailored messages' content was matched to whatever our algorithm identified as the most important reason. Second, the random element in our tailored-messaging intervention allowed us to pick one of the strategies for behavior change that are suggested by TMT -- half focused on the narrative mind and the other half on the intuitive mind.
Of course, the theory-based tailored messages weren't the only active intervention component here. There was also motivational interviewing (MI), and if our study had gone according to plan we wouldn't have been able to disentangle MI's effects from those of the daily text messages. But a fortuitous accident led to half of our control-group participants also getting MI, so of the total study population, 50% got MI + tailored messages, 25% got MI + educational messages, and 25% got educational messages alone. That distribution allowed us to make some educated guesses about the relative importance of MI versus tailored messages, even if the study design for doing so wasn't perfect. We couldn't do a direct 3-group comparison because the MI-delivery error was at one of the two clinical sites, and the sites had significant demographic differences, so those two factors were perfectly confounded with one another. We tried some creative sub-analyses instead.
At the site where all participants received MI, there was still a significant advantage for tailored messaging; in fact, the difference between groups was even greater in this subset of participants than it was in the sample as a whole. And in the educational-messaging comparison group, the subset who received MI had better results than the subset who didn't. In our study, the effects of MI also were comparable to those seen in some meta-analytic reviews of MI for other populations. Together, these findings suggested that MI and tailored messaging had independent effects, with each one contributing some benefit.
To me, what's most exciting about this study is a demonstration that behavior-change strategies grounded in TMT did more to help people than education alone. A methodological feature that I'm particularly proud of in this study was the use of a daily survey in both groups, and a random component so that people wouldn't get exactly the same message each day. The inclusion of these features in the educational-messaging group was intended to control for "pseudotailoring" effects, a phenomenon where people are more likely to change their behavior when they think that someone has crafted a message just for them. In the educational-messaging group there was no actual personalization, but completing a survey and receiving a random message made the intervention feel more personalized to people in the comparison group. We tried a number of different analyses to rule out potential competing explanations, which makes me more confident that TMT is really what made the difference in our results.
Comments
Post a Comment