EXOME

all the information, none of the junk | biotech • healthcare • life sciences

Catalia’s Big-Eyed Robots Soon Will Nudge Patients To Take Their Meds

Xconomy San Francisco — 

Cory Kidd discovered during a graduate school project at MIT Media Lab that patients are more likely to follow doctor’s orders if they get the nudge from a talking robot, rather than the same reminder from an animated character on a computer screen.

That research, and his further health robotics work, prompted Kidd to found Catalia Health in 2014. He’s been touting the persuasive powers of the startup’s prototype robot, Mabu, at conferences for several years now, based on field tests with patients. An initial production run of 500 yellow Mabus with big amber eyes will begin in China soon, because Catalia has landed its first three commercial contracts to send them into patients’ homes on behalf of clients such as drug companies and healthcare systems.

San Francisco-based Catalia is one of the tech companies that stand to benefit from provisions of the Affordable Care Act (also known as Obamacare) that can make reimbursement rates for medical care contingent on health outcomes for patients. For example, a hospital might get paid less if its patients tend to be readmitted after treatment for a heart condition.

That policy raises the incentives for healthcare providers to make sure patients comply with follow-up care instructions, such as taking medicines or exercising. So healthcare companies have been more willing to pay for measures that encourage compliance, such as mobile apps, smart pillboxes, and social networks that link caregivers and family members with the patient.

This outcomes-based reimbursement model—a potential boon for businesses like Catalia—was not a new idea, but the Affordable Care Act “made that happen at some scale,” Kidd says. Despite the failure of Republicans in Congress in their first attempt to repeal or substantially modify the ACA, the fate of the health insurance law is still in doubt as President Trump ponders his next moves.

“It’s something we think about a lot,” Kidd says. “It’s definitely challenging right now.”

Meanwhile, Kidd continues trying to build the business. Under Catalia’s first commercial contracts, the startup has a chance to prove the worth of its robot personal assistants by two measures—better health for patients and better financial rewards for health-related companies.

In May, the robots will be going to live with patients suffering from late-stage kidney disease, in one client-backed project; or rheumatoid arthritis, in another. A third project, for a healthcare system, will dispatch robots to the homes of people with congestive heart failure. (Kidd says he’s not yet free to disclose the names of the clients.)

These contracts provide a testing ground for the robots, but also a test of the human capacity to develop a personal relationship with a machine.

After early trials, Kidd concluded that face-to-face interaction is a powerful motivator, even though patients know that Mabu is just a robot. In the field tests, the individual robots weren’t called Mabu for long, he says. They became something else, like Ingrid, or Casper. “One hundred percent of patients name their units,” Kidd says. They often dress up the robot with hats or scarves, he says.

Mabu has a much-simplified humanoid face; its small, rudimentary “body” holds up tablet screen. But the robot has enough expressive range to convey reactions to what the patient is saying or doing. It speaks in a calm female voice; its eyes can blink and move to maintain eye contact; and its head can nod or bow.

While patients will be reading these simple robot expressions, Mabu is being adapted to read the patient’s much more complex facial expressions with software supplied by Boston-based emotion-sensing company Affectiva. That company says it can detect seven different emotional states, including anger, sadness, joy, fear, and contempt, by observing combinations of muscle movements and contractions on the human face.

Catalia is using this technology to get to know patients better so it can personalize their conversations with Mabu. Early in the relationship, Mabu cracks a corny joke and scans the patient’s face to see if he or she likes humor coming from a robot. “A majority of our patients think it’s hilarious,” Kidd says. “Some of our patients hate it.”

By building up psychological model of the patient, the system figures out how best to deliver content and motivate patients. Some people like detailed explanations, Kidd says. Others prefer just the basic facts.

Catalia, and its clients that specialize in certain diseases, have studied the reasons why treatment adherence rates tend to be low. Among the top four or five reasons why patients don’t take their medicine, Kidd says, “forgetting is never on the list.” So compliance aids have to go beyond reminders, he says.

Some patients can’t cope with medication side effects. But there can also be psychological issues, Kidd says. For example, for a patient with no symptoms but a potentially lethal cancer, the pill may remind them they’re sick, he says. They may take an ill-advised holiday from the medicine.

Mabu is designed to ask open-ended question such as, “How are you feeling?’’ to draw out information that could be relevant to treatment adherence. It may also encourage a chat about unrelated events of the day, such as a relative’s visit.

Kidd understands why this might seem creepy and manipulative to some people. “We’ve made a lot of conscious decisions to avoid that,” Kidd says.

Using the robot is never mandatory under Catalia’s contracts, he says. Patients must opt in. The company follows HIPAA (Health Insurance Portability and Accountability Act of 1996) privacy regulations.

The robot is not in continuous listening mode—a patient can initiate an interaction by tapping the tablet screen, or touching the robot’s head. Some patients set a schedule to wake it up at certain times of the day, Kidd says.
“None of those conversations are ever recorded or sent anywhere,” he says. Through real-time speech recognition, the patient’s answers are converted into relevant data.

Currently, Mabu doesn’t report back to clinicians if a patient’s face reveals … Next Page »

Single PageCurrently on Page: 1 2