Challenge is to analyze data stored in multiple locations by multiple providers

By Brian Gormley
Jan. 6, 2019 9:00 a.m. ET

The medical field’s lofty dreams of unleashing the power of artificial intelligence have set off a race to rework the way health-care specialists make use of their data.

Although technology exists to make AI a potent tool, there is a snag. Data relevant to answering specific questions often reside in various locations, from hospitals to diagnostic labs to pharmaceutical companies. These information silos are typical in the health-care field, leaving scientists and other medical professionals at a disadvantage to harness the full predictive power of AI.

Small businesses such as PatientMatters LLC and Prognos Health Inc. are overcoming the data-gathering obstacles to provide insights to medical customers including health plans. “For Prognos to do what we do, you need to have large data sets,” said Sundeep Bhan, co-founder and chief executive of Prognos, which helps insurers predict their members’ disease risks.

Prognos, based in New York and launched in 2017, has teamed up with diagnostic labs to accumulate diagnostic data on 200 million patients, which it marries with information from health plans to answer questions such as which members are likely to develop a specific condition.

Diagnostic labs, which hope their data will be used to solve medical problems, share the entrepreneur’s dreams of seeing medicine take a leap forward, Mr. Bhan said. “At the end of the day, in health care, that’s what we care about,” he added.

Prognos, which has about 100 employees, aids pharmaceutical companies’ marketing efforts by helping them identify health-care providers that have patients who could benefit from their drugs. Prognos has raised $42 million in venture capital from investors such as Cigna, which is one of its customers, and Merck Global Health Innovation Fund.

Companies such as PatientMatters have crafted services with the aid of publicly available data. The company aims to help health systems gauge patients’ ability to pay bills and identify financial-aid candidates.

PatientMatters, based in Orlando, Fla., and formed in 2012, analyzes medical information from clients as well as publicly available financial data from credit-reporting agencies and other sources, according to founder Sheila Schweitzer, who also is a managing partner of Blue Ox Healthcare Partners LLC. Blue Ox and other firms have invested an undisclosed amount in PatientMatters.

Increased use of high-deductible insurance plans means health-care providers must collect more of their revenue from patients. By gathering data and applying AI and machine learning, PatientMatters, which has 235 employees, can help clients anticipate people’s payment behavior, according to Ms. Schweitzer.

Researchers are also finding solutions to data fragmentation. They include Leo Anthony Celi, a scientist at Massachusetts Institute of Technology and a physician who sees intensive-care-unit patients at Boston’s Beth Israel Deaconess Medical Center. He has persuaded Beth Israel to share electronic medical-record data on intensive-care-unit patients.

De-identified data from 60,000 such cases now are freely available in a database called Medical Information Mart for Intensive Care. Scientists regularly publish papers based on this repository, he said.

A new version coming out early this year will add emergency- and operating-room data as well as medical images, Dr. Celi said. He added that he wants to bring more institutions into the database but often encounters resistance.

“Even now, it’s still a battle to convince hospital leaders—they’re afraid you’ll uncover some poor quality of care,” he said. “You shouldn’t be trying to hide that, but trying to address those deficiencies.”

Read more from WSJ Pro Artificial Intelligence at
Write to Brian Gormley at Twitter: @BrianPGormley.