“Mood disorder” is a mild term for a major problem in public health, which includes depression, bipolar disorder and other chronic conditions that worsen someone’s emotional state. These conditions are surprisingly common—for example, each year about 11 million U.S. adults suffer at least one major depressive episode—and the clinical monitoring that helps to avoid such episodes has been surprisingly unchanged for decades. CompanionMx, an MIT STEX25 company, takes on this problem with Companion™, a mobile app-based service designed to help clinicians make smarter decisions for their mood disorder patients. Using AI-based algorithms built from MIT research, Companion measures and analyzes voice and cell phone metadata to quantify symptoms of mental health. These symptoms are then reported to clinicians.
Early clinical trials have shown that incorporating Companion into regular care for people with diagnosed mood disorders can make a statistically significant improvement in their care outcomes, says Subhrangshu (Sub) Datta, chief executive officer of CompanionMx. “We’re able to provide the right care at the right time from the right level of care provider for people with these disorders,” he says. A second-generation spinoff Companion’s core technology is derived from research on automated conversational analysis carried out at the MIT Media Lab’s Human Dynamics Laboratory, led by Professor Alex (Sandy) Pentland. In 2007 the technology was spun out into Cogito Corp., which has pursued two main opportunities. The first is in assisting agents in customer service and call centers, and the second is in behavioral health with Companion.
Much of the early funding for what became Companion came from the Department of Defense. In a first clinical trial at a Veterans Administration clinic in 2013, a prototype system performed well in monitoring veterans with depression and post-traumatic stress disorder (PTSD). The National Institute of Mental Health supported further work with the next-stage Companion product, including a clinical trial at two Boston hospitals in which 73 participants with at least one symptom of PTSD or depression completed a 12-week field trial. Reported in 2017, the trial showed that Companion successfully predicted symptoms of depression and PTSD, and that participants were reasonably comfortable using the app. When CompanionMx spun out of Cogito to fully commercialize the product in 2018, Datta joined as CEO with a personal mission. After Datta himself experienced an episode of major depressive disorder, he had been surprised and disappointed to see that caregivers were making their ongoing decisions based only on intermittent, qualitative self-reported scoring by patients. In his career as an executive with medical device business units within large manufacturers, “the healthcare system that I knew is completely driven by data that is continuous, repeatable and reproducible,” Datta says. “I just couldn't believe that system didn’t exist for mood disorders. There's no reason for anybody to have episodes like the one I had. We should be able to manage it early so things don't escalate to that point.”
Details on disorders Companion passively collects two broad types of data from the user’s cell phone, Datta explains. One is voice; the user is regularly prompted to record a quick “audio diary,” the program does not pull voice data from calls or other communications. The other type is passively collected cell phone metadata: call log, text logs and geolocation. Next, Companion’s cloud service brings AI-based algorithms to quantify behavior health symptoms. Examining the collected audio diaries, “Companion is not looking at words but at speech patterns, which makes it extremely powerful, difficult to game and language-agnostic,” says Datta. “Similarly in phone metadata, it’s looking for patterns on the frequency, diversity and timing of interactions.” The system’s proprietary AI driven algorithms convert this data into quantitative measures for four symptoms of behavioral health grounded in standard clinical practice: interest, social isolation, mood and energy level. This information then is summarized in a dashboard delivered to clinicians (as well as the user). “The dashboard gives greater visibility to the user’s behavioral health symptoms on a continuous, repeatable and reproducible basis, so clinicians can detect worsening of symptoms early and do something about it,” Datta says. In a recent clinical study at a Harvard teaching hospital, clinicians used insights from the Companion dashboard to predict mania up to two weeks in advance—advance notice that could help to avoid hospitalization or worse outcomes. Companion’s clinical component is designed to fit within a healthcare provider’s current workflow, with a social worker or behavioral health counselor periodically checking the dashboard. The social worker pulls in a psychiatrist at the appropriate time when needed, both ensuring appropriate care and improving resource utilization, Datta says. Building partnerships for behavioral health Because Companion acts solely as an advisory service for clinicians, it hasn’t required Food & Drug Administration approval. CompanionMx is now rolling out the service commercially. In March 2019, CompanionMx was selected for the MIT Startup Exchange’s STEX25, an accelerator for “industry-ready” startups that are ready for significant growth. The company seeks to partner with technology firms, pharmaceutical companies and large organizations with employee wellness programs. Companion’s ability to provide proactive remote monitoring of behavioral health symptoms has shown significant appeal for technology companies, Datta says. At Veteran Affairs facilities, for instance, “we could help open up a whole set of opportunities to really make a difference to address the scourge of suicide among veterans that's happening today,” Datta says. His company already is working with one pharmaceutical firm, in a clinical study in which Companion acts as a measurement of efficacy for their firm’s treatment. Among large employers, “mood disorders are a huge challenge that many are trying to address,” Datta says. “They’re looking for partners who can give them the ability to proactively help individuals who have already been diagnosed.” Organizations must, however, consider the organizational culture and commitment to help individuals with mood disorders, “because there often is stigma around it,” he says. Over time, CompanionMx plans to expand and sharpen its software tools for addressing anxiety and stress related disorders. The company also wants to tailor Companion for both geriatric patients and adolescents—both populations in which depression is a major problem. “Suicide has become the number-two cause of deaths amongst college students today,” he points out. Additionally, CompanionMx may provide services for people for whom behavioral health concerns are heightened by comorbid health conditions such as heart disease or diabetes. “There are many opportunities on that front where we could make a big difference immediately,” Datta says. Overall, such behavioral health services fit well in today’s quest for “value-based” healthcare among insurers, providers and government agencies. “There are clear outcome improvement and cost savings opportunities when you bring in true evidence-based care driven by objective measures grounded in a strong clinical foundation,” he emphasizes. “This we firmly believe will help us in our mission to ensure no one has a significant mental health episode ever again.”