The excitement over artificial intelligence among healthcare venture capital investors increasingly comes with a caveat.
Amid a challenging venture capital market for digital health, AI remains a bright spot and an area of enthusiasm. According to a survey published in October from venture capital company GSR Ventures, 87% of healthcare investors are altering their strategies due to ChatGPT and other generative AI models. Investors see the potential of AI to improve efficiencies in healthcare and increase access for underserved patients.
But VCs also recognize that AI can exacerbate issues of health equity. As a result, they want to know if portfolio companies are using unbiased AI datasets.
Read more: Olive AI’s failure: Digital health investors predict what's next
“When we started this [VC company], one of the goals we had was to make sure that the next generation of products that were developed didn't have bias in them,” said Eileen Tanghal, founder and general partner of Black Opal Ventures. “Maybe it’s OK if a sensor doesn’t turn on when a person of color washes their hand...but it’s not OK if a trained diagnostic misses your breast cancer diagnosis because you're Asian American and they use mostly European women to train their data sets.”
Black Opal, led by Tanghal and co-founder Dr. Tara Bishop, announced its inaugural $58 million fund Nov. 7 with backing from Eli Lilly and Company, Bank of America, Morris Plains, New Jersey-based Atlantic Health System Venture Studio and Detroit-based Henry Ford Health. The fund has invested in multiple AI companies, including conversational AI developer Hyro, obesity-focused Intellihealth and an undisclosed company aiming to improve representation in clinical trials.
Tanghal likened AI in 2023 to the internet in 1995 in terms of potential impact. She and Bishop are looking at AI companies that can improve diagnoses, speed up drug discovery and reduce repetitive tasks. However, the technology's capabilities can’t overshadow the potential harm, they said.
“When I'm looking at a company and evaluating their product, [I’m wondering] are they actually thinking about the bias of the datasets and how they're going to combat this,” Tanghal said. “And the more advanced companies are not just solving the easier problem…they are thinking about the next challenges of ensuring healthcare access is equitable. That is a harder problem to solve. It means you have to think about the data you are using.”
Another VC company, Material Impact, on Wednesday said it raised a $352 million fund, its third since its 2016 inception. The company has focused its initial healthcare investments on women’s health-focused companies.
For the latest fund, co-founder and managing partner Adam Sharkawy said the VC will invest in companies that increase access for underserved groups and gather more representative data for AI models.
“AI may help get us to a point where we can decentralize healthcare and increase access to those that currently don't have it. But we're basing it on data sets that exclude underrepresented groups,” Sharkawy said. “You can cause more harm than good if you start creating whole paradigms of treatments in healthcare that are based on data that is skewed and not representative.”
LRVHealth, a healthcare-focused VC company, closed a $200 million fund in May and partner Ellen Herlacher said it's looking to invest in companies using AI for improved operational efficiency, driving better clinical flow and transitioning providers to risk-based payment models.
For the latter, Herlacher said companies need to ensure their data models take social risks into account, particularly when it comes to patients on Medicaid.
“You can't take risks on these populations without taking health equity or social determinants of health issues into consideration because those ultimately determine clinical and financial outcomes,” Herlacher said.
