8 reasons health AI companies fail

Artificial intelligence in healthcare "doesn’t really exist right now," even though it should, according to Aike Ho, a partner at venture capital firm Acme Capital. 

In an April 12 Twitter thread Ms. Ho outlined the key reasons she thinks health AI companies fail from her vantage point as a digital health investor. Here are the eight obstacles she detailed:

  1. The fragmented nature of U.S. healthcare: Health AI companies must work hard to obtain data to train their models. This often means interfacing with health systems and universities that house disparate datasets, a process that can take years.

  2. Bad data kills companies: Even after a health AI company acquires the data it needs, the data might be poor quality and require extensive cleaning. Bad data will lead to bad machine learning models, something Ms. Ho called "game over" for health AI startups.

  3. Deciding which application can secure funding: The total addressable market for most health AI applications is small, as these applications often provide ancillary services.

  4. Getting adopted by health systems: Ms. Ho said it is "brutal" for health AI companies trying to get their services adopted by providers unless they're also delivering care vertically. She said getting adopted while delivering care vertically can also be tough, as companies are "essentially building [their] own provider group and care delivery infrastructure."

  5. Demonstrating superior clinical outcomes: Health AI companies often struggle to win over risk-averse providers by failing to adequately communicate how their services lead to better care outcomes.

  6. Payers usually don't reimburse: Once a provider becomes convinced of a health AI company's clinical benefit, it will ask who is paying. Most health AI companies are not reimbursed by insurance, and providers are operating on thin margins, so it can be difficult for companies that have a value proposition of cost savings rather than adding revenue to get their services adopted by providers.

  7. Onboarding: Adoption will be minimal unless a health AI company's offerings integrate seamlessly with a providers' existing workflows, Ms. Ho said.

  8. Determining the right business model: It is easier for a health AI company to sell providers on a model where their company receives money based on the adoption of its services, but the company will go under if no one uses its products. Flat fee contracts are less risky for startups, but they're harder to sell to providers.

Copyright © 2024 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Whitepapers

Featured Webinars