What You Missed in Healthcare IT: September Edition

1. Medical Language Models Under the Microscope

It didn’t seem like the large language model (LLM) or medical language model (MLM) markets could get much hotter, but we definitely saw an uptick in stories here across fundraising announcements, product updates, and general (and sometimes juicy) news.

• Abridge and DeepScribe have both released papers (here and here, respectively) discussing how they measure and control quality in their AI ambient scribes. Our head of research, Patrick, discussed in a post one interesting aspect to this: These healthcare vendors outperform the frontier models for now, but they’re in a race to successfully distribute before the OpenAIs and Googles of the LLM world can catch up and better tune their models for healthcare.

• In another analysis, Patrick shared his thoughts on a paper released by GenHealth.ai about the models they are building for predicting conditions and risk both in patient populations and in specific patients to help model costs more accurately across a variety of use cases. Separately, Dandelion Health released a paper where they were able to build a virtual cohort via their datasets to expand on the Novo Nordisk SELECT trial and find a stronger cardioprotective effect for GLP-1s. Putting these ideas together, we imagine that as these datasets grow and models become multi-modal with the ability to interpret imaging and labs, real world evidence will increasingly start to look at these model distributions to understand causality and effectiveness.

• Not to be left out, OpenAI also recently released o1-preview, which is focused on improved reasoning. The Oscar team did a great job breaking down the healthcare use cases this new model might unlock.

• While not strictly an MLM business, AI clinical summaries solution, Pieces, found itself in some hot water. The vendor reached a settlement with the Texas Attorney General resolving an allegation that the vendor promoted deceptive claims about the accuracy of its capabilities. As part of the settlement, Pieces is required to disclose its products’ accuracy and ensure staff understand how much they can rely on it, which creates an interesting precedent for AI vendors: That they must be able to quantify a level of certainty or reliability in their products.

• Hippocratic AI announced that it raised an additional $17 million as part of a Series A extension round—including participation from NVIDIA’s investment arm—bringing its total funding amount to $137 million. They also partnered with WellSpan Health to build a conversational AI phone agent for patient outreach.

• With all these happenings, you may find yourself looking for a refresher on this space. Don’t miss the LLM market map we shared this month.

2. Medical Imaging Is Getting the AI CDS Treatment

AI solutions around medical imaging aren’t new, but we did see more major vendor updates this month than we have in the past, particularly in the clinical decision support space.

• AI imaging CDS and care coordination vendor Viz.ai added the heart disease diagnostics tool Cleerly to its Viz.ai One platform

• Oncology-focused Paige released a new AI clinical summaries platform.

• Harrison.ai announced the launch of a radiology-specific vision language model, Harrison.rad.1, which is currently being made available to select users.

• Qure.ai closed a $65M series D funding, which it plans to invest into further work on foundational models and potential acquisitions.

• Segmed, an imaging data startup, announced it closed a $10.4M Series A led by iGan Partners and Advocate Health.

• RapidAI partnered with the Alaska Stroke Coalition on a three-year project to improve health equity for Alaskan stroke patients.

3. Hospitals & Vendors Are Pursuing Solutions for Nursing Workflows

According to a nursing shortage report from the AACN, “52% of nurses are considering leaving their current position due primarily to insufficient staffing, work negatively affecting health and well-being, and inability to deliver quality care. In addition, 60% of acute care nurses report feeling burnt out, and 75% report feeling stressed, frustrated, and exhausted.” Solving nursing workflows is a major priority for many of the execs we speak to, and we’re starting to see that reflected more and more in product updates and adoptions.

• LifeBridge Health announced the implementation of the virtual nursing solution Care.ai Virtual Nursing at one of its hospitals for discharge, documentation, and other patient support needs.

• AvaSure, Oracle, and NVIDIA are collaborating on a new “virtual concierge” for bedside patient engagement and to allow bedside nurses to request support, translation, and more.

• While not announced this month, Epic did recently announce a partnership with Abridge and Mayo Clinic to work on AI documentation for nurses.

• We also spoke to Bon Secours Mercy Health Chief Clinical Innovation Officer Mark Townsend in our Executive Insights series, who shared BSMH’s plan to co-develop a nursing ambient scribe solution.

4. Epic Gets Taken to Court

Particle made news this month when it filed its antitrust lawsuit against Epic, alleging anticompetitive behavior in the payer platform category. Here’s a quick overview of their claims:

• Epic required manual approval for new Particle customers or expansions to access Epic-stored data—a practice exclusive to Particle’s customers—which often delayed onboarding by over a month.

• Particle alleges that Epic told some of its customers that their access to Epic EHR data would be restored only if they ceased their relationship with Particle. As a result, XCures, previously a Particle customer, ended its contract.

• Particle claims that Epic damaged its reputation by implying that Particle posed security and privacy risks.

• Finally, Epic initiated a dispute against Particle within Carequality, alleging misuse of the “treatment” PoU by Particle’s customers. Carequality found no wrongdoing by Particle but imposed a “corrective action plan,” potentially as a result of Epic’s influence.

In response, Epic issued a statement calling on Particle to ask Carequality to release its resolution.

We sat down with Elion advisor and interoperability expert Brendan Keeler to unpack the claims in the suit as well as potential ramifications. Watch their conversation and see a full breakdown of events here.

5. AI Fundraising Updates

This month, Flare Capital Partners published an in-depth analysis of healthcare AI funding over the past 10 years, exploring where these investments have shown the greatest ROI. Despite almost 50% of health system AI funding going toward clinical use cases, “clinical decision support solutions have yielded amongst the lowest maturity rates amongst all health system AI startups (6.8%) while imaging AI solutions have fared slightly better with a 9.9% maturity rate.” Financial, patient engagement, and operations startups have had better returns, but clinical use cases could still outperform them given their enormous potential. Here are this month’s AI fundraising announcements by category:

Other resources you may have missed

In case you missed it, here’s a quick roundup of other resources we shared this month:

• This month’s market maps: Smart EHR UI, Clinical Data Abstraction, Clinical Pathways, and Large Language Models

• This month’s conversations with healthcare leaders: Steve West of Healthliant Ventures, Adam Carewe of NerdMDs and formerly of Kaiser Permanente

• Our analysis of Oscar’s new payer claims system