Microsoft Built a Chat Bot to Match Patients to Clinical Trials
A chat bot that began as a hackathon project at Microsoft’s lab in Israel makes it easier for sick patients to find clinical trials that could provide otherwise unavailable medicines and therapies.
The Clinical Trials Bot lets patients and doctors search for studies related to a disease and then answer a succession of text questions. The bot then suggests links to trials that best match the patients’ needs. Drugmakers can also use it to find test subjects.
Microsoft won’t release the bot as its own product. Instead, the software giant is talking to pharmaceutical companies that it hopes will use the bot to find trial participants, while pitching it to other partners that could turn the technology into a tool for patients, said Hadas Bitran, group manager of Microsoft Healthcare Israel. Bitran declined to name possible partners because no deals have been agreed yet.
The project is part of a larger Microsoft (MSFT) health care bot initiative that’s helped partners build automatic chat programs for things like triaging patients and answering questions about insurance benefits. The clinical trials bot was accepted as part of the U.S. White House Presidential Innovation Fellows program and Bitran showed it on Thursday at a closed door event at the White House. She will demonstrate the tech publicly on Friday during a session at the U.S. Census Bureau.
Half of all clinical trials for new drugs and therapies never reach the number of patients needed to start, and many others are delayed for the same reason, Bitran said. Meanwhile patients, sometimes desperately sick, find it hard to comb through the roughly 50,000 trials worldwide and their arcane and lengthy criteria—typically 20 to 30 factors. Even doctors struggle to search quickly on behalf of patients, Bitran said.
“It was a big passion project for everyone involved,” Bitran said. “We heard stories of families who would sit for days and days looking at the trials.”
The technology uses a form of artificial intelligence called machine reading to ingest the selection criteria for each clinical trial. It uses this data to decide which questions to ask patients and how to match their answers to suitable trials.
Here’s how it works for patients: They type in a search, such as “trials for a 52-year old California female with breast cancer.” The bot responds with questions such as whether the patient had chemotherapy for metastatic disease—a cancer that has spread—and how far the patient can travel. It offers five choices that describe the patient’s current health and ability to be active and care for herself. As the patient selects from the multiple-choice answers, the software generates the next question and refines the list of available trials.