For investors, digital health data outweighs algorithms

Venture capital firms first look for useful and unique data when they invest in digital health companies. At the SoCalBio Digital Health Conference, representatives of six Southern California-based firms agreed that having access to good data outweighs even the most sophisticated algorithms.

Both technology and health care firms are investing in companies using AI to solve health care’s most costly problems, said Milo Bissin, director of life science and health care for Silicon Valley Bank. The amount of investment in health tech companies has grown significantly in the last five years, from $3.47 billion in 2015 to $6.28 billion as of September 2019.

While investors are interested in their solutions, companies first need to start with a strong foundation of data.

“Whenever I talk to companies that have AI or machine learning in their strategy, I tend to focus first and foremost on what is your data acquisition strategy?” said Kevin Zhang, a partner with Los Angeles-based Upfront Ventures.

That includes talking to founders about where they source their data, and what information they can bring to the table to create a unique dataset.

“Since we see so many companies that claim to have an aspect of machine learning, AI or deep learning… the ones that tend to be interesting to use are the ones that have approached the market in such a way that they have unique data,” said Richard Weil, a partner with Pasadena-based Mount Wilson Ventures.

Of course, finding good-quality health data can be difficult. Information pulled from electronic medical record systems can be a jumbled mess of unstructured text. Cleaning that data is time-consuming: After acquiring an oncology electronic medical records system, one biotechnology company hired 600 people just to clean and annotate those medical records so they could make use of them, Zhang said.

For companies that don’t currently have access to a high-quality dataset, get creative, suggested Luke Hayes, managing director of newly-formed firm Torrent Ventures. For example, one company Torrent worked with licensed a dataset from a nonprofit organization.

“We are more concerned with the quality of the underlying dataset than the technique used to analyze it,” Hayes said.

Other companies have turned this challenge into an opportunity. Yiwen Li, director of strategy and business development for Culver City-based NantHealth, said her company had invested in a firm that pulls data out of medical devices into electronic medical record systems.

“It sounds very simple,” she said. “We work with so many different medical devices and EMRs. … A lot of the data are still stuck in the medical devices.”

Prior to that, physicians had to write or copy the information manually from the medical device into the health record system.

“It’s incredibly useful,” she added. “Before AI, we have a lot of things to do.”


Related posts