Bob Lindner, left, and Meghan Gaffney Buck are the co-founders of VEDA Data, a machine learning company that focuses on improving health care provider directories.


As a post-doctoral researcher at the University of Wisconsin-Madison, Bob Lindner used to develop artificial intelligence to better understand the stars, programming ways to analyze the data collected by massive telescopes.

Now, his startup VEDA Data applies those same mechanics to something smaller in scale: the directories of doctors found on health insurance websites.

VEDA Data uses machine learning — in which computers are programmed to learn behaviors on their own — to improve the accuracy of health care provider directories. According to Lindner, computers are great at parsing not just complex astronomical data, but complex health care data.

“The same way of thinking is being applied,” Lindner said. “These data amounts are so huge.”

Improving provider network accuracy has proven a mounting issue. According to a report released by the Centers for Medicare & Medicaid Services earlier this year, 45 percent of entries in the directories provided by insurers contracted with Medicare are inaccurate. There are incorrect addresses, wrong phone numbers, and even incorrect information about whether a doctor is actually within the network — the group of providers that an insurer has decided to include in their plans.

Listing accurate information in what’s essentially a provider phone book seems like it shouldn’t be that hard. But it’s a problem that insurance companies have been struggling to solve, said the company's co-founder and CEO Meghan Gaffney Buck.

“There are some really stupid billion-dollar problems, and this is one of them,” she said.

Inaccurate directories are also a barrier to health care in the U.S. As a New York Times report from last year found, patients have accidentally sought care with out-of-network providers due to inaccurate directories, finding themselves slammed with unexpected bills.

Buck said the flawed directories can also result in people not getting care at all, especially those with lower incomes or who are older.

"You call, and (the directory) says they're accepting new patients and they're in network, and lo and behold they're not," she said. “You might call three providers, and then give up. You just say, ‘I'm not feeling well, but I give up.’”

Sometimes glitches or the sheer volume of data to trick can be an issue. Buck also suggested that health care providers have an incentive to provide confusing information — a health system may list a physician as operating from multiple offices so as to more easily process claims, for example.

"There is actually profit motivation for health systems to list every doctor at every location in their system,” she said.

In 2015, the Obama Administration enacted new rules and penalties for insurers offering inaccurate directories. VEDA Data markets to insurers looking to dodge fines and improve their listings. 

The system that Lindner has constructed pulls data about providers from disparate sources ranging from the self-reported lists doctors give insurers to the publicly available National Provider Identifier database. Even Yelp reviews are included. The system picks out chunks of information — a phone number, whether a provider is in the network, whether they're taking new patients — and determines whether it’s accurate or not.

Humans grade the computer program on how it’s doing by fact-checking small portions of its work. The program then takes that feedback, learns from it, and becomes better at determining what's accurate.

The startup has forged a partnership with Humana and another client. Lindner said that they're not quite at 100 percent accuracy with their software, but have been able to improve databases on average to accuracy ranges of over 70 percent.

It’s a complex task. But as Buck tells potential clients, it’s not exactly astrophysics.

"There's a lot more data in the galaxy than in this health record, so we're good," she said.

As for Lindner, he's happy with the change from astrophysics to health care. He was always more interested in the big data methodologies than the field itself.

"While the universe never ends in fascinating me ... sometimes I felt disconnected from society on the Earth," he said. "It is refreshing and compelling to tackle problems that can affect the lives of millions of people for the better."

Erik Lorenzsonn is the Capital Times' tech and culture reporter. He joined the team in 2016, after having served as an online editor for Wisconsin Public Radio and having written for publications like The Progressive Magazine and The Poughkeepsie Journal.