New Eleos Research Links Homework Follow-up to Improved Therapy Outcomes
Homework has always been part of the therapeutic process. Whether it’s a formal, “This week you’ll practice mindfulness every morning”—or a casual, “What if you learned to paint?”—homework serves as a bridge between a client’s therapy sessions and their daily life.
As Amy Schroll, licensed counselor and clinical director at Trilogy, puts it, “Most therapists that I know use homework in a variety of ways. You’re meeting with someone for 55 minutes a week, and there’s so much going on in between sessions. But if the person is only working on it and talking about it for one hour per week, there’s probably not going to be a lot of change that happens.”
In other words, homework keeps clients engaged between therapy sessions—which can be critical to their overall progress. But what if we’ve only scratched the surface of homework’s potential?
That’s the topic of the latest Eleos-enabled study, “Action recommendations review in community-based therapy and depression and anxiety outcomes: a machine learning approach,” which reevaluated the role of homework in mental health care—and found it to be even deeper and more nuanced than previously understood.
Using Eleos Health’s massive dataset of real-world behavioral health sessions, researchers examined the ways therapists “assign” homework, whether or not they follow up, and associated symptom improvement outcomes.
Starting with a Huge, Uniquely Authentic Dataset
This study used the same massive data set that informed the team’s previous paper investigating the ability of behavioral health-trained AI to accurately detect the assignment of therapeutic homework. (Spoiler alert: It could. You can review the results of that study here.)
“We thought, ‘So we know therapists give therapeutic homework—now let’s look at more data, because we have a lot more data,’” said Amit Spinrad, Eleos data scientist and lead researcher for the study. “We have tens of thousands of sessions in our data set. So we looked at more data from diverse organizations and many more—hundreds—of different therapists.”
The size of the data set is unique in the world of mental health research, especially considering that it consists of real-world sessions provided at everyday treatment settings, such as community-based clinics.
“Most studies testing psychological interventions are relatively small scale and involve few study sites and a relatively low number of therapists,” Spinrad explained. “This creates a situation where you don’t actually have large amounts of data-driven psychological evidence.”
According to Spinrad, most formal therapy research is conducted in an academic environment, which might skew the results. “The majority of treatment data was gathered from university students or individuals receiving care in programs affiliated with a university, where therapy is recorded for the primary purpose of research,” he said. “But here, we had data from 14 different organizations, hundreds of patients, and 2,400 sessions. If these data overcome all these factors—various organizations and various therapy techniques—and you still see a statistically significant correlation, then that means something.”
Defining “Homework” in Therapy
Although this study was based on the same dataset used for the previous homework study, Spinrad and his fellow researchers went back to the drawing board when it came to deciding what counts as “homework.”
“In the last paper, therapeutic homework was defined in a very structured way,” he said. “For example, ‘You should do breathing techniques. Do mindfulness every night.’ This kind of direct, structured instruction.”
Of course, therapy rarely fits a particular mold. What happens in a session is highly nuanced and contextual—and that definitely extends to the nature of homework assignments.
“Most of the time, therapists—in the real-world settings Eleos has been developed for—tend to give homework in a much less structured way,” Spinrad explained. “For example, instead of recommending that the client practice breathing techniques by saying, ‘I’d like you to incorporate these skills into your routine every evening until our next session,’ the therapist would say something along the lines of, ‘Okay, maybe something you could do for your mental health is to go back to cooking—that was good for you.’ It’s not really well-structured, is it?”
Seeing in the data that many “homework” assignments took the form of these casual, less direct suggestions, Spinrad and his team decided to expand the definition of homework. Instead of using the term “homework,” however, they decided to label these assignments “action recommendations.”
“This includes those structured homework assignments, but it’s a larger perspective of homework,” he said.
In the previous study, researchers found that 60% of therapy sessions contained a well-defined “homework” assignment. But under the wider definition of homework, Spinrad and his team found that an incredible 95% of sessions included at least one action recommendation—and usually more.
“The average was four to five of these action recommendations in each session,” he shared.
Following Up on Action Recommendations
Think back to your school days for a moment. When your teacher assigned homework, you assumed you’d be asked about it during a future class session. In therapy, though, that’s not a given.
According to the new study, while 95% of therapy sessions included at least one action recommendation, only 50% of the sessions included a review of the recommendations from previous sessions.
“Even when we defined ‘review’ very broadly—like just mentioning ‘what we talked about last time’—we found that only 50% of the sessions included some form of review,” Spinrad said.
Based on her experience as a clinical director, Schroll says some clients are easier to work with than others when it comes to homework—and that may lead to a lower incidence of homework review, because the client isn’t completing the homework in the first place.
“It depends on the person you’re working with,” she explained. “Some people are more willing to do homework outside of sessions. Some people want something to do. They’re thinking, ‘What should I be doing now, and what should I do next time?’ But for some people, it’s about figuring out what is getting in the way of practicing. That can be worked on, too. You can ask, ‘I want you to think more between now and next time about why you’re not wanting to try or do these things in between sessions.’”
In many cases, the consistent integration of homework follow-up into sessions is an effort that rests with the provider. But while homework itself is well established as a core component of modalities like cognitive behavioral therapy (CBT)—for which homework engagement has been shown to improve outcomes for certain clients—the impact of homework follow-up on treatment efficacy is much less clear, which may help explain why it seems to be less of a priority.
The Impact of Homework Follow-up
It would have been enough to leave the study here, simply stating that most therapists give “action recommendations,” but only half of them follow up. With the data right there in front of him, though, Spinrad couldn’t resist the opportunity to dig deeper.
“We wanted to take it even further and look at the impact,” he said. “We had a model to extract action recommendations from sessions, and we had another model to extract reviews of previous action recommendations. So we asked if there was a correlation between how much the therapist reviews the action recommendations and whether they are getting better outcomes for their patients.”
The data, it turns out, says yes—in some cases, anyway.
“We didn’t find any type of correlation among clients with anxiety symptoms, but we did find one with clients reporting depression—especially mild depression,” Spinrad explained. “We found that the more the therapist reviewed homework with clients with mild depression, the better their treatment outcomes in terms of depression reduction.”
The Unanswered “Why”
With the limitations of the current study, Spinrad could only guess why homework review seemed to be so noticeably connected to better results for those with mild depression.
“Maybe it’s not that more homework review reduces depression,” he speculated, considering the possibility that perhaps therapists don’t seek new strategies because they begin to see progress, which explains the increased homework reviews.
Researchers also had a theory on why mild depression was the most impacted diagnosis. “We hypothesize it’s because clients with mild depression are more likely to respond to their provider’s encouragement to engage in activities,” Spinrad said. “So when therapists evaluate their past discussions more, they get the clients to do more health-promoting activities between sessions.”
Dr. Shiri Sadeh-Sharvit, Chief Clinical Officer at Eleos, floated another idea about the link between mild depression and homework review.
“Many of the manualized interventions that include homework assignments have been studied primarily on people with relatively mild to moderate symptoms,” Sadeh-Sharvit said. “Therefore, it’s simpler to apply the same manualized interventions with this subset of clients.”
In other words, therapy approaches that use homework were, in some ways, “designed for” mild depression, making them easier to apply for clients with that particular diagnosis.
Ultimately, though, like most research, this study opens up as many questions as it answers.
“We were very careful in the clinical implications section of the paper to emphasize that we found a positive association—we found a correlation between the review magnitude and the better depression outcomes,” Spinrad said. “And this calls for further investigation to learn more about this association.”
Opening New Research Doors with Purpose-Built AI
Just a few years ago, a study of this magnitude would have been unimaginable in mental health. But thanks to the power of Eleos Health’s AI platform—and the large volume of real-world data accessible through it—researchers like Spinrad and his team can study therapy practice across multiple diverse settings, drawing connections to inform better therapeutic practices.
This kind of research is invaluable. It provides a solid foundation of evidence that therapists can rely on so their practice can evolve and improve based on a blend of professional expertise and concrete data. And as Eleos continues to enable this kind of in-depth study, we’re moving closer to a world where therapy is consistently informed by the best available evidence—so that every therapy session is as effective as it can be.
With innovative technologies like Eleos leading the charge, the future of mental health care looks promising. We’re on a path to a more informed, evidence-based approach to therapy, where the decisions clinicians make are supported by a wealth of data. And that’s a future worth striving for.
