Melanie Cole, MS (Host) : Welcome to Better Edge, a Northwestern Medicine podcast for physicians. I'm Melanie Cole, and we have a Northwestern Medicine panel for you today. Our discussion focuses on research at the NEAR Lab, capturing mood dynamics through adolescent smartphone social communication. Joining me is Dr. Stuart Shankman, he's the Dunbar Professor of Psychiatry and Behavioral Sciences; and Dr. Lilian Li, she's a postdoctoral scholar for the Department of Psychiatry and Behavioral Sciences.

Doctors, thank you so much. I think this is such a great topic, and thank you for joining us today. Dr. Shankman, I'd like to start with you. As we look at the biggest challenges in diagnosing and treating adolescent depression today, I'd like you to speak about what those challenges are and really give us an overview. What's happening with our kids in the digital world? Do you feel it's making them smarter, more worldly? I feel that my kids know more than I knew at their age because we didn't know what was going on in the world and they do. But it could be positive and negative. Do you think that that is what's contributing to some of the challenges that we're going to discuss here today?

Dr. Stewart Shankman: It's a great question. I think in past generations, people didn't really talk about mental health. And I think particularly with young people these days in adolescence that they're more mindful of wellness initiatives and the importance of mental health. But with that said, there's still a very low rate of professional help seeking. I think a lot of youth maybe turn to their friends, turn to social media and TikTok for mental health support. But there's a lot less. The disconnect between the need and people seeking professional help is quite large.

Melanie Cole, MS: So then, based on that, Dr. Li, your recent study highlights the potential use of smartphone social communications to detect depression. What does that mean? Explain how that method works and its advantage over possible traditional screening methods that we've known for years.

Dr. Lilian Li: So for traditional methods, we typically just give them a questionnaire and ask for their feelings and depressive symptoms. So, it's been recommended to be done in primary care settings. And you can see it can be a little bit difficult to do just because of time constraints or reluctance from either the provider or the adolescents to talk about these mental health issues.

And for us, we think that monitoring their social communication on the smartphone could provide a good way to really understand and identify symptoms before they begin to escalate. And for that study, we focus on smartphone communication data that's passively collected using an app for 90 days. And based on these text data, we were able to extract some linguistic features that have been shown to really predict feelings and mood and depressive symptoms and set out to test whether they actually do predict symptoms and mood.

Melanie Cole, MS: That's fascinating. And, Dr. Shankman, expand on that. Some of the key linguistic markers that Dr. Li was just talking about that would indicate depression in adolescents' social media posts and text messages. Because I think that even for parents listening, these are things we're being taught more now than we used to, is to pay attention to some of these markers that we see in the social media posts of our children. Speak about these markers and how reliable they actually are.

Dr. Stewart Shankman: So, this is just the natural language that the kids are typing into their phones, whether it's texting with their friends or posting comments on social media, like TikTok and Instagram. And some of the really interesting methods that Lilian employs sort of pulls out particular markers.

So, what are some examples of those? Some of those are the use of first person singular pronouns. So, do they use a lot of I's? I feel this, I am this, me, mine, those types of words. And it's this indicator of a very self-focused attention. And we found that that relates to worse mood over time.

Other examples are things like sort of the general negative sentiment of their words. So, they're using a lot of negative words, like upset, angry, sad, fearful. And the more use of those types of negative words over time relative to their positive words, like happy, joy, elation, again relates to their mood over time.

Melanie Cole, MS: This is really an important study because it's giving us an eye view of the way that our kids are interacting with each other. Dr. Li, the study found that words related to friends and affiliation were significant predictors of positive mood on the other end of that spectrum. So, how can that insight be used to develop interventions that leverage social connections to improve adolescent mental health? Take us from bench to bedside. I think this is the most important key message here.

Dr. Lilian Li: So, first and foremost, to keep in mind that these words are sort of like created by psychologists to see, we think that, "hi," "help," or "hello" is about affiliation and for friends, when they say like, "Oh, dude," is part of the friend word dictionary. So partially, they kind of ignore the context under which that these words are used. But we do find that whether using more of these friends-related words and affiliation-related words, predicted feeling better on the next day. And like you said, it really highlighted the importance of social relationships on mental health, which has been extensively studied in the field.

And I think the study is just a first step to highlight that in a relatively novel, social communication platform on the smartphone. So, more work needs to really focus on how these time by time dynamics within our social relationships can be protective or even harmful towards our mental health.

So, example could be, you know, we think seeking social support and talking to friends about our problems is generally a good thing to receive support and advice about our problems, but there's been some studies showing that excessively discussing these problems with friends, on the other hand, could backfire and could relate to mental health issues down the line.

Dr. Stewart Shankman: And just to jump in about that, I mean, the hope down the road, as Lilian said, this is sort of early work in this area, but the hope down the road is that these types of methods that detect changes in the adolescent's mood can lead to just-in-timeintervention. So, you can imagine a situation or analogous, I guess, to diabetes, where people, their glucose level is being monitored continuously. And then, if they have a dip or increase in their sugar levels, they get a notification on their phone saying, "Hey, you should eat a candy bar" or "Sit down," you know, depending on the level of their glucose level. You can imagine something similar here with their mood. So if the algorithm detects a big increase in their first person use or a decrease in their affiliation or friend terms that maybe they get a notification, either checking in to see how you're doing or maybe giving some sort of advice. Again, these kind of just-in-timeinterventions can be really helpful, I think, for adolescents in the real world.

Dr. Lilian Li: One point on that is that because we can collect quite a bit of data from the smartphone, and these kind of interventions exactly can also be tailored to the person to create a really precision, individual-based, suggestions and recommendations.

Melanie Cole, MS: The technology is moving so quickly these days. And Dr. Shankman, one of the concerns when people think of this kind of monitoring or even AI, ChatGPT, because I know that many children, kids, young adults, my kids age in their 20s use ChatGPT as sort of an interim counselor. Like they'll ask them their questions or tell them their problems. And some of these markers you're talking about are in those. How do you ensure privacy when using smartphone data for mental health monitoring? I think this goes across the board for so many of these new ways that we are looking at adolescent depression.

Dr. Stewart Shankman: Yeah, the ethical challenges are definitely an important aspect of our research. So, I guess in the research context, people voluntarily, they're signing up, giving full consent, the parents and the kids, knowing what the risks are and the benefits. And we do everything we can to try to keep all their information secure, behind HIPAA-compliant firewalls and so that their research data doesn't get open to the public.

Now, down the road, in terms of clinical tools, I think it'll be important that the technology is again HIPAA compliant and there's privacy protections in place. And the hope is that people who are signing up for these interventions, I guess, who are aware of the risks of putting their information out there and it's possible also that the app, the information they enter into whatever intervention apps they're using, maybe don't end up into the broader internet because they're sort of behind a firewall, but it's definitely something that me and my colleagues who are doing digital mental health research are very cognizant of and prioritize highly.

Melanie Cole, MS: Dr. Li, expand on that a little bit about some of these ethical challenges with language-based depression detection tools. How do you address them?

Dr. Lilian Li: So, using these kind of detection tools, I think, at this moment is still kind of early. We're still doing more research to build more sophisticated and advanced models that could have a better prediction accuracy, for example. But it is, at the outset, very important to make it clear to them what we're doing, what kind of data we're collecting, and using very plain language to explain everything in detail. Especially the concept of digital phenotyping, things like that could be a little bit foreign to people. So, explaining that clearly is a very important step so that they have full knowledge to decide whether to enroll or not.

And I believe in the future if such tools are advanced enough to be able to predict an episode in advance, then I think there could be some other important issues to consider, such as the false positive rates or false negative rates. So if we miss an important negative event, like a suicide, that could have dire consequences. But if we predict that somebody was having a bad mood or on a risky period and then that were not, then that could have a lot of issues as well. So, these are the things that we really need to keep in mind while deploying these tools.

Melanie Cole, MS: I'd love to give you each a chance for a final thought here, because this is a really, really fascinating study, and I know the research is going to continue. But Dr. Li, how do you keep your findings relevant with the rapid changes in social communication apps? Where do you see this going, and how are you going to keep up with rapid rise in this technology?

Dr. Lilian Li: There are definitely so many different kinds of social communication apps. I think at the time when the study was done, TikTok wasn't really a big popular thing, but now it has gone super popular. And so, the landscape of different social communication apps just definitely changes all the time. And for us, I think one way to address that is to look at not just one app or two apps, for example. Instead, we looked at all of these social communication apps like, TikTok or Instagram or Facebook, rather than specifically only looking at Facebook or Twitter, which was what the common approach for past studies, so that partially, accommodate for these changing landscapes. And we did not really make too fine-grained distinctions between apps. And that could be a limitation because people certainly could behave differently depending on what kind of social communication app they're using. So, that is a limitation to this lump sum approach. But with enough data, future studies can certainly take a more fine-grained investigation on this.

Dr. Stewart Shankman: And just to add to that, I think, a lot of times parents and providers who work with kids are asked questions about, "Well, the kid's on their phone all the time, and they're on the screen all the time. I need to get them off." It's sort of like this knee-jerk reaction that they're on their phone and it's necessarily a bad thing. But for a lot of kids, it's a chance for them to connect with people in a community who they might not normally have in their everyday or in their "real life." And they make friends that are just as close, online as they have in "the real world." And so, the idea that the use of cell phones or social media is "bad," I think it's a more nuanced story than that. And I think it's important for parents and clinicians to be aware of that there's a lot of benefits that kids are drawing from their use of social media. And I guess the question for us researchers and clinicians, is how can we leverage the fact that the kids are on this to improve kids' well-being and overall health care.

Melanie Cole, MS: Beautifully said. And thank you both so much for joining us today and keeping us apprised of this study. It's important work, and I hope that you'll both join us again as the study continues and keep us updated. Thank you so much again for joining us, and to refer your patient or for more information, please visit our website breakthroughsforphysicians.nm.org/psychiatry to get connected with one of our providers. And that concludes this episode of Better Edge, a Northwestern Medicine podcast for physicians. I'm Melanie Cole.