Artificial intelligence and healthcare are hardly strangers, as machine learning programs have been aiding in the diagnosis of disease, drug discovery, the streamlining of clinical workflows and the analysis of medical data for decades. But their relationship may rapidly accelerate and deepen significantly with the rollout of large language models (LLMs) like ChatGPT and GPT4.
Being hailed as a revolution on par with the invention of the computer or internet for their potential to innovate as well as disrupt, LLMs are estimated to be a $6 trillion opportunity, offering massive productivity gains to virtually every industry. Digital health is no exception, and tech leaders and startups alike have begun leveraging LLMs—or building their own—to bring digital health and AI together in ways that were unthought of just a few years, or even months, ago.
While some analysts say as much as 28% of all working hours in healthcare overall could be automated thanks to GPTs, this is a change that won't come overnight.
LLMs still struggle with accuracy (also known as AI hallucinations), which is more consequential in healthcare and digital health than in other sectors. There are also ethical concerns whenever machines are used to augment human judgement and expertise. Industry leaders and regulators will have to reckon with these, and a host of other concerns, before LLMs can make a major impact on the delivery of care.
But these concerns have not stopped tech companies with deep R&D budgets from launching new healthcare AI initiatives:
Microsoft is teaming up with OpenAI, the developer of GPT4, on automating paperwork and member communications around health benefits, and is also working with electronic health record (EHR) company Epic to generate responses to patients submitted in EHRs.
Google is creating Med-PaLM 2, a new LLM, to analyze mountains of healthcare data to provide answers to a range of medical and treatment questions. It has already scored at the "expert" level on questions patterned after the U.S. Medical Licensing Examination (USMLE).
Amazon Web Services is teaming up with 3M Health Information Systems to combine generative AI with clinical documentation.
However, tech leaders are not the only ones aiming to bring LLMs into the healthcare system.
Startups Making Moves
ChatGPT has already shown its worth when it comes to basic healthcare literacy and responding to some patients' questions. Digital health startups are aiming to extend those capabilities.
Huma.AI, a leading healthcare AI company, in February announced the launch of its industry-first AI platform for life sciences, designed to accelerate the development of life-saving drugs through better usage of their data. It analyzes private enterprise data from multiple sources and its "expert-in-the-loop" approach leads to the highest accuracy possible.
Arine, a medication intelligence company that specializes in technology that optimizes medication therapy, utilized its AI-powered risk stratification to enable a client to significantly reduce hospitalizations, improve formulary adherence and lower the cost of chronic disease management for its members by implementing data-driven interventions.
Cornerstone AI, developer of an artificial intelligence-based healthcare software platform designed to analyze clinical data, raised $5.7 million in seed funding last year.
And startups are applying GPTs in other areas aside from directly interfacing with patients, including using generative AI to manage a range of health-related tasks for older patients and using LLMs to help providers and payers better manage information.
No surprise, venture investors see massive opportunities, with companies like Ferrum Health and Laudio making funding announcements recently. While venture investing overall is still far below levels seen during the last peak, according to Pitchbook, investments in generative AI overall are expected to hit $42.6 billion by the end of 2023. We are likely to see many more exciting potential use cases created by new and existing digital health startups in the months and years ahead, and many will get funded.
Looking Ahead
Clinical decision-making, risk prediction and pandemic preparedness, personalized care and drug discovery and development: These are the areas outlined by the World Health Organization as those where LLMs can make the greatest impact. Startups and public companies alike will offer novel solutions in these areas, as well as in other areas that are yet to be imagined.
But the challenges ahead are at least as massive as the opportunities. Regulation is likely to precede any significant rollout of LLMs in healthcare, and there are currently no proposals on the table in the U.S. Regulations must tackle everything from the accuracy of medical information to inherent biases that can be found in data. New rules will have to be written to govern who can use generative AI in healthcare, and under what circumstances. For more on digital health regulation see our past posts on the topic here and here.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
