Healthcare AI Guy Weekly Newsletter | 4/23

Our Picks ✨

Highlights if you’ve only got 2 minutes…

1/

Scaling real-world adoption of medical AI

As medical AI continues to evolve and improve rapidly, real-world adoption remains a key obstacle. In the world of healthcare, adoption mainly rests on one thing: reimbursement. In a recent NEJM AI article, the authors explore the pros and cons of current paths to reimbursement, while offering a new framework for medical AI developers.

• Fee-for-Service: treats medical AI similarly to how new drugs or medical devices are reimbursed, and is a viable path for AI that can clear the hurdle of demonstrating improvements to clinical outcomes, health equity, clinician productivity, and cost-effectiveness (e.g. AI for diabetic eye exams).

  • Pros: FFS might be appropriate for AI because health systems are adept at assessing the financial impact of new technologies under it, and reimbursement through a CPT code is common.

  • Cons: among the 692 FDA-authorized AI systems, few have been able to meet the criteria above. The approach carries substantial risk in terms of time and resources for AI developers.

• Value-based Care: provides reimbursement based on patient- or population-related metrics (MIPS, HEDIS, full capitation).

  • Pros: obtaining authorization for medical AI to “count” toward closing care gaps for MIPS and HEDIS has been shown to be considerably more straightforward than attaining a CPT code.

  • Cons: if a given measure is not met the financial benefit is typically zero, potentially disincentivizing AI adoption.

• New Framework: the authors offer an approach derived from the Medicare Part B model, which reimburses drugs administered in an outpatient setting based on a “cost plus” markup. This model essentially splits revenue between AI creators and users.

  • Pros: providers could acquire the rights to use AI, then get reimbursed based on the average cost of the service plus a specified margin, contingent upon CMS coverage of a particular CPT code.

  • Cons: risk of overutilization and requires the creation of new CPT codes.

Without sustainable reimbursement, widespread medical AI adoption just isn't possible. Though there's no quick fix, exploring different pathways is a crucial step forward. (link)

2/

Hugging Face releases health-focused LLM benchmark

Hugging Face, a hub for AI models, just released a health-focused LLM benchmark. The new test, called Open Medical-LLM, aims to standardize the evaluation of generative AI models’ performance on a range of medical-related tasks. The benchmark wasn’t built from scratch but rather a stitching together of existing test sets — MedQA, PubMedQA, MedMCQA, and so on — designed to probe models for general medical knowledge and related fields, such as anatomy, pharmacology, genetics, and clinical practice.

Open Medical-LLM aims to help researchers and practitioners identify strengths and weaknesses in AI models, enhancing advancements and patient outcomes. Of course, the quality of the test matters, but having more domain-specific AI benchmarks makes a lot of sense and we’re happy Hugging Face decided to tackle this. (link)

3/

Docs more comfortable with AI than patients? 

A recent survey by Wolters Kluwer Health highlights the need for transparency and trust among doctors and consumers regarding AI in healthcare. The survey shows that while many physicians have changed their minds about AI, patients remain wary of it. Here's a summary of the key findings:

• Both physicians, 91%, and consumers, 90%, emphasize the need for transparency about the data and development of AI tools in healthcare.

• 40% of doctors are ready to implement AI this year, while 80% of patients express concerns about its use in diagnostics.

• 8/10 physicians think AI will improve interactions with patients.

• +50% of doctors believe it will save them substantial time by quickly searching medical literature or summarizing patient data from EHRs.

The survey shows that physicians are growing more comfortable with AI than patients and reflects a complex landscape where the potential of AI in healthcare is recognized, but its adoption is contingent on clear communication and stringent safeguards. (link)

Tools & Partnerships 🔧

Latest on business, consumer, and clinical healthcare AI tools and partnerships…

TOOLS

• GPT-4 nears expert eye doctor analysis: A new study from the University of Cambridge showed that OpenAI's GPT-4 model performs close to the level of expert ophthalmologists in analyzing eye conditions and suggesting treatments. GPT-4 scored 60, higher than the average scores of trainees (59.7) and junior doctors (37) but slightly lower than the expert’s average of 66.4. (link)

• VR for Cancer Care: Researchers at Georgetown University and MedStar Health randomized 128 cancer patients to undergo either a 10-minute VR session or a “two-dimensional imagery experience” on a tablet, the VR group saw twice the drop in pain levels on a 10-point scale and pain severity remained just as low after 24 hours, while the guided imagery group returned to baseline. (link)

• AI for AFib: Cedars-Sinai-led team developed an echocardiography AI model that accurately detects patients with atrial fibrillation – including asymptomatic patients – potentially revealing a new “opportunistic screening” method to drive earlier AFib detection. (link)

• Seed Health: Seed Health launched an AI-powered microbiome mapping platform. (link)

• Technogym: Technogym unveiled an AI-powered wellness kiosk Checkup, competing with EGYM’s healthspan-tracking platform BioAge. (link)

  

PARTNERSHIPS

• Abridge + MemorialCare: Abridge, the Ambient AI startup, and MemorialCare, SoCal-based system with +225 care locations, announced a partnership to make Abridge’s generative AI tool for clinical documentation (50 specialties & 14 languages) available to MemorialCare’s physicians. (link)

Deal Desk 💸 

Spotlight on latest capital raises, M&A, and investments…

FUNDING

• Two Chairs, a hybrid behavioral health provider with a proprietary 300-variable algorithm for therapist-patient matchmaking, closed $72M in a series C equity and debt financing round. The round was led by Amplo with Fifth Down Capital, with participation from other investors. Bridge Bank provided the debt financing. (link)

• Kontakt.io, a NYC-based platform designed to provide insight into how patients and staff physically move through hospitals, raised $47.5M in Series C funding from Goldman Sachs. (link)

• Auxa Health, a NYC-based AI-powered benefit navigation platform, raised $5.2M in seed funding. Zeal Capital Partners led the round and was joined by existing investor AlleyCorp and new investor K50 Ventures, Laconia Capital Group, and Chaac Ventures. (link)

MERGERS & ACQUISITIONS

• DrFirst + Myndshft, medication management platform DrFirst this week closed its acquisition of Myndshft Technologies, an AI-enabled benefits and prior authorization verification platform. (link)

• ABOUT Healthcare + Edgility, ABOUT Healthcare, hospital transfer software and patient flow systems service backed by Rubicon Technology Partners, acquired Edgility, a Tampa, FL-based AI-powered operations management platform for healthcare systems. (link)

Other Relevant News 🔍

News, podcasts, blogs, tweets, resources, etc…

• Vi reveals AI’s impact on healthcare, pharma, and wellness in new reports (link)

• How Epic, Oracle customers are using EHR AI tools (link)

• Why health system AI predictions can fail (link)

• WHO’s AI chatbot providing incorrect medical information (link)

• AI that drafts more empathetic responses helping UC San Diego physicians (link)

• AI helping collaboration between physicians, nurses at Stanford Health (link)

• Health systems increasingly using AI assistants to reduce burnout (link)

Visuals of the Week 📸

Funny memes, cool pics, and interesting data from around the web…