Vikram Jha

We’ve just released the definitive guide to the AI-driven prior auth space. Prior auth is one of those areas that’s both ripe for AI and deeply frustrating for everyone involved: • The hardest parts of the process are exactly what AI is good at • If done well, it can drive real ROI for health systems and providers • And yet, it remains painful, slow, and opaque The reality is that prior auth is incredibly complex and fragmented. Requirements vary dramatically by payer, plan, procedure, state, and site of service. Submission rules are often difficult to even find, let alone interpret. And once a request is submitted, turnaround times are long while visibility into status is limited at best. It's no surprise then that dozens of vendors have jumped in to try to fix it. The problem? It’s become almost impossible to make sense of the landscape. Who does what? Which pain points do they really address? How do they actually solve the issues? That’s what this guide is for. We explain how the prior auth process works across both medical and pharmacy benefit. We highlight where AI is being applied today. And we map the vendor landscape so you can see how the pieces fit together. This is the resource I’ve wanted for a long time to understand the space. Link is in the comments, and we'd love to hear your thoughts. Special shoutout to Colin DuRant for his incredible work putting this together 🙌

447

36 Comments

"Health practitioners are becoming increasingly uneasy about the medical community making widespread use of error-prone generative AI tools. One glaring error proved so persuasive that it took over a year to be caught. In their May 2024 research paper introducing a healthcare AI model, dubbed Med-Gemini, Google researchers showed off the AI analyzing brain scans from the radiology lab for various conditions. It identified an "old left basilar ganglia infarct," referring to a purported part of the brain — "basilar ganglia" — that simply doesn't exist in the human body. Board-certified neurologist Bryan Moore flagged the issue to The Verge, highlighting that Google fixed its blog post about the AI — but failed to revise the research paper itself. The AI likely conflated the basal ganglia, an area of the brain that's associated with motor movements and habit formation, and the basilar artery, a major blood vessel at the base of the brainstem. Google blamed the incident on a simple misspelling of "basal ganglia." It's an embarrassing reveal that underlines persistent and impactful shortcomings of the tech. Even the latest "reasoning" AIs by the likes of Google and OpenAI are spreading falsehoods dreamed up by large language models that are trained on vast swathes of the internet. In Google's search results, this can lead to headaches for users during their research and fact-checking efforts. But in a hospital setting, those kinds of slip-ups could have devastating consequences. It's not just Med-Gemini. Google's more advanced healthcare model, dubbed MedGemma, also led to varying answers depending on the way questions were phrased, leading to errors some of the time. "Their nature is that [they] tend to make up things, and it doesn’t say ‘I don’t know,’ which is a big, big problem for high-stakes domains like medicine," Judy Gichoya, Emory University associate professor of radiology and informatics, told The Verge." What happened with Med-Gemini isn’t the exception, it’s the rule. It is the perfect metaphor for the state of AI today: confidently wrong, beautifully phrased, and fundamentally hollow. LLMs architecture cannot be evolved. No amount of scale, fine-tuning, or prompt engineering will turn a glorified guesser into a system that truly understands. It is structurally incapable of grounding itself in reality. It cannot model the world, cannot self-correct or admit when it doesn’t know because it doesn’t 'understand' anything to begin with. The world deserves better, systems that are grounded in the real world, that learn and understand the way we do, not just mimic the surface of thought, but replicate the underlying process. Systems that know when they don’t know. We call this performance theater as progress. We listen to LLM hypesters spinning foundational flaws into features, insisting that the next version will somehow cross the chasm, deep down even they know that it won't. Build Cognitive AI → Unlock Real Intelligence.

505

112 Comments

Epic announced 160+ AI projects at its recent UGM conference, while rival Oracle Cerner just launched an AI-backed EHR. But can Epic or Cerner really “run care operations”? A hospital executive I spoke with (whose hospital runs Epic) was clear: “No, Epic cannot run operations.”   I wrote down a few thoughts:   1️⃣ 𝐋𝐞𝐠𝐚𝐜𝐲 𝐄𝐇𝐑𝐬 𝐰𝐞𝐫𝐞 𝐧𝐞𝐯𝐞𝐫 𝐛𝐮𝐢𝐥𝐭 𝐭𝐨 “𝐫𝐮𝐧 𝐨𝐩𝐞𝐫𝐚𝐭𝐢𝐨𝐧𝐬.”   Hospital EHRs began as systems of record: tools for admin tasks and billing.   Running operations is a different challenge: it’s about process optimization and automation.   EHRs were the accountants; now they’re willing to become COOs.   2️⃣ 𝐍𝐞𝐰 𝐂𝐚𝐫𝐞𝐎𝐩𝐬 𝐩𝐥𝐚𝐭𝐟𝐨𝐫𝐦𝐬 𝐚𝐫𝐞 𝐞𝐦𝐞𝐫𝐠𝐢𝐧𝐠.   Provider to provider: Sword Health, a leading digital & AI rehab scale-up from Portugal, launched Sword Intelligence, sharing the tools it built to optimize its own operations (see visual)   Oil & gas to #healthcare: DNV Imatis, a CareOps player from Norway, has its roots in the oil and gas industry. The company originally optimized processes to transport oil from offshore platforms in the North Sea to continental Europe. A physician on its board pointed out that similar process optimization was urgently needed in hospitals, leading DNV-Imatis to expand into healthcare more than 10 years ago.   Airports to hospitals: Singapore’s ST Engineering, originally managing airport flows, was brought into hospitals during COVID to improve patient flow.   3️⃣ 𝐓𝐡𝐞 𝐂𝐚𝐫𝐞𝐎𝐩𝐬 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐜𝐚𝐬𝐞 𝐢𝐬 𝐜𝐨𝐦𝐩𝐞𝐥𝐥𝐢𝐧𝐠.   ST Engineering: A 600-bed hospital typically spends ~$6M/year on 60 senior nurses to manage operations. With ST Engineering’s 300+ standardized processes and command center model, the same work can be done by 3 non-clinical staff for ~$150k/year (and much better results)   DNV-Imatis: Clients report 20% productivity gains through automation and better resource use. Sword Health, the digital rehab scale-up, delivers 3.7× ROI per member per year in MSK care with its AI Agents. Key drivers: ◾14× clinician capacity ◾74% lower session cost ◾71% less clinical workload Now spinning off its #CareOps tech into Sword Intelligence, powered by AI Care Manager Agents: ◾Missed calls cut from 40% → 0% (24/7 coverage) ◾84% of patients scheduled instantly (eligibility auto-verified) ◾$100k new monthly revenue in 4 weeks 4️⃣ 𝐓𝐡𝐞 𝐜𝐫𝐢𝐭𝐢𝐜𝐚𝐥 𝐞𝐧𝐚𝐛𝐥𝐞𝐫: 𝐨𝐩𝐞𝐧, 𝐦𝐨𝐝𝐮𝐥𝐚𝐫 𝐭𝐞𝐜𝐡 𝐬𝐭𝐚𝐜𝐤𝐬.   Legacy EHRs aim to vertically integrate everything: from databases to CareOps to care pathways.   But advances in interoperability and AI now allow hospitals to build open, modular stacks.   Unfortunately, most hospital IT teams are underfunded and tasked with maintenance, not transformation.   To shift, budgets would need to rise from ~2% to 10%+ of hospital spend, and hiring would need to shift from traditional IT to AI and tech talent. --- Follow HGM Advisory for daily reposts of the most relevant #HealthTech news

124

7 Comments

The Single Patient Record is finally moving from concept to evidence, and NHS England has kicked off the Test and Learn phase. Three technical proofs of concept are being tested in the Test and Learn phase: 1. A Hub & Spoke model to link regional records via shared APIs 2. A Central Integration Model for real-time national access 3. A Virtual Data Layer that joins data on demand, not by duplication Each approach asks the same question 👉 Can we join up care safely, securely, and simply - without starting from scratch? For years, we’ve talked about interoperability like it’s a technology problem. It isn't. It’s a design and delivery challenge - and this phase is a chance to test what "good" really looks like at system scale. The interesting part is the approach is open, collaborative, and grounded in reality - discovery with frontline users, not just design in meeting rooms. If it works, it could redefine how health data flows across the NHS. Not a single system. A single understanding of the patient. Finally.

215

43 Comments

💡 I spent 10+ years at Google to learn how to change medicine in 45 minutes. Here's how I did it ⬇️ Last week I posted about BrachyStories, Mission-Driven Tech's outcome improvement research. I had an incredible response, now I'm sharing the behind-the-scenes on we did it. ☢️ Brachytherapy cures cervical cancer. It also brutalizes and forever changes a person's life. At MDT, we know it's possible to treat *and* achieve high quality outcomes. 🟠 Enter, BrachyStories. In under and hour, BrachyStories takes clinicians from 0 to 1 on 3 key patient safety principles: 1️⃣ The patient experience 2️⃣ Pain control management 3️⃣ Treatment recovery resources + referrals ⁉️ How did MDT do it? ✅ Feature real survivors: for the first time on camera, international cervical cancer survivors from the Cervivor, Inc. community + beyond bravely contributed their real, raw and unfiltered stories surviving brachytherapy. In some rooms, it’s the first time clinicians have heard patients recall their experience surviving surgery while awake. 🏆 Work with the best partners: Leyla Tatiana Rosario, Jadalyn Ortiz and Jessica L. Ransom of @The Gig Media, Andrew Strickman & Kim Cortese came together to support MDT's vision of bold storytelling to create change 🎬 Produce a short doc film like a top-performing YouTube video: ⏲️ grab attention in the first 5 seconds 💥 use bold supers to underscore key points + aid recall ➡️ feature a clear call to action that moves people to act 🔬 Get in front of leaders responsible for change: MDT debuted BrachyStories at the 2024 AMERICAN BRACHYTHERAPY SOCIETY World Congress to an audience of global radiation oncologists and medical physicists. It was the first time survivor voices have been featured at the conference. 📊 Measure in real time: Google surveys powered simple pre & post surveys to quantify lift against BrachyStories' 3 key safety principles. This flexible measurement format quantified competency lift and quickly built proof points. Want to learn more? Check out BrachyStories here: 🔗 https://lnkd.in/eEjsENiP

64

13 Comments

Regulation is finally catching up to AI in healthcare. The FDA just released its first comprehensive draft guidance for AI-enabled medical devices. Over 700 AI-enabled devices have already been cleared by the FDA, but until now, there’s been no unified framework that tells companies how to build, validate, and monitor these tools across the lifecycle. This guidance finally lays that foundation. It calls out the need for: - Clear model transparency (what goes in, what comes out) - Performance validation on diverse populations - Usability testing for clinicians - Post-market monitoring, especially for continuously learning systems In short, it’s a shift from static product approvals to dynamic, ongoing oversight. For builders, it means two things: - There’s more clarity now on what the FDA expects. - The bar for rigor and accountability just went up. The bigger question how quickly can startups and health systems operationalize this lifecycle-first mindset. #aiinhealthcare #regulation

32

2 Comments

OpenAI's #OpenSource #GPTOSS was released with strong claims of performance on #healthcare tasks. However, our benchmarks show that it disappoints on benchmarks that weren't built by OpenAI... from medical question answering and reasoning to information extraction and summarization. JSL-MedReasoning-32B, John Snow Labs reasoning model, achieved an average accuracy of 84.99% across public medical benchmarks, surpassing OpenAI’s gpt-oss-120B which scored 77.18%. Both models are optimized for inference on a single 80GB GPU, proving that performance gains don’t require added compute costs. Even JSL-MedS-14B, not a reasoning model, outperformed gpt-oss-20B, and almost matched gpt-oss-120B despite being one-eighth its size. #ai #generativeai #llm #llms #healthcareai #healthai #datascience

58

4 Comments