Walking down the halls of the research wing at Beth Israel Deaconess in Boston, it’s hard to miss the bright orange biohazard warnings on the door of nearly every lab — except one. Formed in 2018, the Division of Digital Psychiatry is focused on technology’s role in improving mental health care.
Led by Dr. John Torous, the division’s first digital tool, dubbed mindLamp, is focused on examining "digital biomarkers," or patient data gathered from both active and passive data together to improve care and patient-clinician conversations.
“mindLamp is a clinical care platform for digital medicine as a whole,” Aditya Vaidyam, a research assistant at the Digital Division of Digital Psychiatry, told MobiHealthNews. “What we do is take a look at active data that comes from participants — for example, answering surveys, cognitive test they may take, and the passive data that comes from sensors on your smart phone. Then we take all of that data and put it through the Lamp platform. The output is a nice clean dashboard.”
The app is customizable, allowing clinicians to adjust the survey questions they ask their patients and how often a patient is asked each question. For example, if a patient with anxiety wants to monitor their mood throughout the day, that might mean they are required to answer more surveys than a patient who wants to look at their memory over a long period of time. Researchers are able to see how long it takes a participant to answer any given question.
Through the platform, patients may allow clinicians, physicians, family and friends to view their data.
“You can add them here ,and then have a network of support and can have conversations that center around your data,” Vaidyam said. “So, it’s unlocking this door for talking about data that isn’t just collecting and showing data.”
The research team is also looking at how passive data, collected from a patient’s phone, can help care.
“The passive data is all collected in the background through GPS, accelerometer, call and text logs to give behavioral context to our participants and our patients,” Phil Henson, a research assistant at the Digital Division of Digital Psychiatry, told MobiHealthNews.
While the content of the texts and calls is not seen by a physician or researcher, the idea is that the overall use of the phone can give indications of mental wellbeing. This data is then combined with the active survey data.
“We kind of connect them together in the backend and try to see if there is any signal we can get,” Henson said. “Are there any biomarkers that can inform symptomatology? For the question: is it possible to use the passive data? We don’t really have an answer for the yet. It’s something we are still working on, and it depends what you are looking at.”
The data is always shared with the patient and they can look at it on their own time as well as with a clinician.
“I take the data and talk it over with them, and present what we find in a way that makes sense, and validate whether or not they agree with what we’ve seen with their data,” Hannah Wisniewski, research assistant at the Digital Division of Digital Psychiatry, told MobiHealthNews.
As for privacy, Torous assures that the information from the app stays with the research institution.
“Look, we are not a company,” he said. “We are not here to sell your data. We are a Harvard hospital. The data goes from you to us, and that’s it.”
Overall the research team said that patients have reconsented positively to the tool. Vaidyam noted that his patients often note they are OK with the data being collected as long as they can see it as well.
Predicting an event before it happens
mindLamp is currently used in the clinic to help facilitate conversations with patients and track behaviors. But looking forward, researchers are interested in using data as a predictive tool.
“The future direction for the stuff I’m working on is the question of can we predict relapse before it happens,” Henson said. “Are there any signals in the passive data that can clue us in to an imminent relapse event, … or if someone is taking a new medication, is there any way for us to predict if they are better or worse based on their passive signals?”
The team is looking at the relationships between different symptoms and creating models for how these symptoms could interact. For example: if someone is reporting that they are not sleeping, will that lead to anxiety the next day?
“We have all that data. I think one of the clinical questions of interest [is] how do those self-reported symptoms sort of interact with one another, or what might be the relationship between those. My recent work has been [on] how can we visualize that to see some of those relationships,” Ryan Hays, a research assistant at the Digital Division of Digital Psychiatry, told MobiHealthNews.
Hays said that at this point they are still trying to tease out correlation versus causation between these interlocking symptoms.
“It’s almost like preventative psychiatry in some ways,” Torous said. “The field is always reactive and you say 'What could be coming next for people with depression?' You kind of look down the road and say 'How can we prevent things?'”
Researchers are also tapping into machine learning as a vehicle to help predict potential adverse events down the pipeline.
“The problem we are trying to address here is that patients relapse and we can’t predict when they are going to relapse,” Dr. James Benoit, associate director of the Digital Division of Digital Psychiatry, told MobiHealthNews. “We want to use machine learning to predict those hospitalization events or significant changes in their symptom scores, and the way we are going to do that is through mobility, sociability and self-report data. And [when] they go above that baseline on a per patient basis, they are going to transition into a hospital or higher symptomatic state.”
Educating the patient
While the digital tools may be useful to facilitate patient-doctor conversations, there is a major issue if the patients themselves can’t use the tool.
“We run a digital health literacy group, … what is the point of handing someone an app if they are not able to open their phone?” Elena Rodriguez-Villa, a research assistant at the Digital Division of Digital Psychiatry, told MobiHealthNews.
In the group, facilitators touch on everything from how to set up a LinkedIn profile to how to set up medications reminders on the Calendar app. Rodriguez-Villa, who runs the group with fellow research assistant Eric Camacho, said she sees the literacy groups as a holistic approach to digital, covering ways to connect and use social media as well as to keep track of their health.
Another way the lab is looking to close the technology education gap is by developing a new role called a "digital navigator," a support staff position intended to help patients understand the tools.
“In a visit we want the clinician to be able to spend enough time with the patient giving clinical care," Wisniewski said. "With technology, a lot of the time patients might need time setting it up. … That is where the digital navigator comes in to augment that. Their job is to keep up to date on these technologies, and keep in mind what might be useful for one patient versus another and help set it up for them and how to use it and extract some data and understand what are we getting out of it."
Torous likened the role to a radiology tech who helps coordinate and free up the clinical team.
The team is also looking at educating patients and clinicians about how to choose external mental health apps. Historically, objective validation has been a major issue in mental health apps. Last March a study published in Nature Digital Medicine (of which Torous was an author) found that a majority of the apps studied do not provide evidence or peer-reviewed studies to back up their products.
Recently the digital division teamed up with the American Psychological Association to develop a tool that would vet apps on the market. The tool evaluates the apps on a number of factors including privacy, engagement, clinical validation and cost.
Torous said that many of the apps on the market today have poor study data, despite the increase of randomized trials.
“It’s tricky because now they are speaking the right language. It’s getting subtler. You really have to dive in and say ‘What is the evidence they are showing us? Does it matter? Is it generalizable?’” Torous said. “We are really not seeing high-quality validation. We are seeing companies with those small pilot studies [say] that they have cured everything and you read it and ... there isn’t that follow up or true validation. It’s becoming [difficult] to even pick a tool because they are getting so savvy in the marketing and language, [and] the validation part is completely just lacking.”
The push to go global
The bulk of the work in Torous’ lab takes place in Boston; however, the team is looking to bring mindLamp to India to see how it works in different locations.
“We are running a global mental health study and we are leveraging mindLamp, which we’ve talked a lot about,” Rodriguez-Villa said. “And the general gist of the study is we are teaming up with two sites in India. One is in [a] much more sort of urban center in Bangalore, and one is more rural in Bhopal, which is in central India. We are working with two mental health hospitals as well as a nonprofit over there.”
The team plans to zero in on how cultural and location factors contribute to the use of the app, and what needs to be adjusted based on these factors.
“We are hopefully gathering a lot of data and comparing it, thinking how does the data change in a rural Indian setting as opposed to here and what does that look like and what is different,” Rodriguez-Villa said. “It’s really thinking about that access idea and how can we change access to care using Lamp.”
While this pilot will be specifically looking at India, Torous said the key to using digital biomarkers in the future is making sure they reach across culture and settings.
"For these biomarkers to work they have to be global,” he said.
Assessing Emerging Technologies
In February, MobiHealthNews will be taking a closer look at how digital tools are validated and assessed by health systems, payers and investors.