Munjal Shah announced Hippocratic AI will partner with several providers for a beta test of its LLM-powered products for nondiagnostic services like chronic care management and post-discharge follow-up.
Munjal Shah, an entrepreneur whose previous artificial intelligence companies were eventually sold to Google and Alibaba, entered the health care AI space when he co-founded Hippocratic AI in 2022. The company is developing a large language model to power nondiagnostic products such as chronic care management and post-discharge follow-up. It’s building on the recent drastic improvements in LLM technology and has prioritized safety and effectiveness through a rigorous testing program.
It’s now ready to beta-test its product, working with partnerships with over 40 providers, health systems, and digital health companies, including Memorial Hermann Health System, University of Vermont Health Network, Fraser Health, and Side Health.
The goal is to subject Hippocratic AI’s product to thorough internal testing and feedback from health care professionals to ensure that it’s safe and effective before it launches.
“Generative AI holds huge potential for improving health care access, equity, and outcomes, but its impact has thus far been limited to back-office applications. At Hippocratic AI, we believe that the industry can — and should — aim higher,” Hippocratic AI co-founder and CEO Munjal Shah said in a statement.
“We’re on a mission to not only address existing staffing shortages but go above and beyond to unlock an age of true health care abundance. Imagine a world in which each one of us has unlimited access to safe, high-quality, personalized care. Our partners will be invaluable in bringing this vision to life as they test the technology internally and ensure its safety.”
Hippocratic AI’s LLM
The initial focus for the products built upon Hippocratic AI’s LLM is voice-based, patient-facing, nondiagnostic tasks. The widespread popularity of ChatGPT has raised awareness of how effective generative AI can become at answering questions and engaging in back-and-forth conversations with human users. Munjal Shah’s insight was to apply these capabilities to nondiagnostic medical conversations, training the model on evidence-based medical research and feedback from health care professionals.
More general-purpose LLMs like GPT lack a high degree of detailed specialization and expertise. These broader models may also have a higher risk of “hallucination,” passing off inaccurate statements as true. The sheer volume of information they learn from means their sources aren’t necessarily limited to accurate, peer-reviewed, rigorous medical research.
On the other hand, by curating the training set for its LLM, Hippocratic AI is hoping to use the incredible learning and communicating skills of this technology but not overextend its applications, building a more focused model while reducing the possibility of hallucinations.
The company has reported that its LLM surpassed GPT-4's performance across over 100 health care certifications. This achievement is attributed to the model's health care-specific vocabulary and reinforcement learning with human feedback training, which integrates direct feedback and insights from health care professionals testing the model.
Such technological advancements highlight the potential of generative AI to transform health care delivery by overcoming significant health care professional staffing shortages to increase the volume of high-quality patient interactions at sustainable cost levels.
Beta-Testing
The beta-testing phase will focus on several priority areas, including chronic care management, post-discharge follow-up for conditions like congestive heart failure and kidney disease, and wellness assessments. These are low-risk services — the AI isn’t diagnosing any conditions — but they’re nonetheless vital for improving patient outcomes, particularly for chronic conditions that require continued follow-up and care.
The beta will aim to validate the technology's accuracy, safety, and ability to integrate into existing health care workflows.
“We believe this technology can help with workforce shortages by taking on some of the more basic tasks to support workflow. For example, calling a patient to provide preoperative instructions or following a procedure for post-discharge checkups,” said Feby Abraham, Ph.D., executive vice president and chief strategy and innovations officer for Memorial Hermann.
“We are always looking for new ways to further engage our employees and make them feel better supported. The unique challenges of recent years have only heightened the need for innovation in this area.”
Munjal Shah and Hippocratic AI believe their product can enhance patient engagement and education, particularly for individuals with chronic conditions. They also believe that it will facilitate more frequent outreach and support between provider visits to improve patient outcomes and make care delivery more efficient.
Feedback from front-line health care professionals, including physicians and nurses, will be instrumental in refining the product. It will be evaluated for criteria including conversational ability, medical accuracy, script adherence, empathy, and listening skills. Testing partners will be able to track performance across all participants.
“We are excited to partner with Hippocratic AI in this beta-testing phase. While our patients will not experience the technology yet, our clinical staff will have the opportunity to learn about generative AI and the possibilities it offers for improving care, while helping to make the tool effective, efficient, and safe for patients nationwide,” said Jessica Moschella, senior VP for high-value care at The University of Vermont Health Network.
“We are especially proud to participate as a rural health care provider — often, innovations are developed in urban centers and later retrofitted to make do in rural areas. Being involved in this phase allows us the opportunity to lead the way for rural health care by aiding in the development of new technology that will support our patients’ unique needs and reduce health disparities.”
Hippocratic AI's commitment to collaborating with medical experts is further evidenced by its recent formation of several advisory councils composed of physicians, nurses, and administrators.
As the company advances through its beta-testing phase and beyond, its efforts will serve as a leading example of the potential role of generative AI in health care. The company is supported by significant investment, including funding from General Catalyst and Andreessen Horowitz. It has received a total of $67 million in funding, speaking to the enthusiasm for LLM-powered health care solutions.
Hippocratic AI's focus on safety, collaboration, and innovation aligns with broader trends in health care technology, where the potential for AI to improve care delivery and patient outcomes is increasingly recognized. Nonetheless, the successful integration of AI into health care will require ongoing scrutiny, adaptation, and dialogue among all stakeholders to ensure that technological advancements serve the best interests of patients and providers alike.
