Jeff Leek

Not a lot of small n jobs out there, this is a good one that has unexpected degrees of freedom.

9

MOREL (Multi-Output REgression and Learning) is a machine learning framework we developed to tackle the long-standing challenge of missing data in DNA methylation arrays. By combining multi-output regression with ensemble and deep learning models, MOREL can accurately restore systematically missing CpG sites caused by differences between array versions or even across platforms like RRBS and arrays. Implemented in the open-source ultra-impute package, MOREL not only outperforms traditional imputation methods but also preserves biological complexity, enabling seamless integration of legacy and modern methylation datasets and improving downstream analyses such as epigenetic clock predictions. The MOREL methods are available in our Ultra-Impute package: https://lnkd.in/gy47h_R6

21

2 Comments

Interesting paper alert - a commentary in Nature Medicine highlights a pivotal tension in healthcare today: while #AIagents promise transformative impacts — from autonomous triage to multi-modal clinical decision support — regulatory frameworks remain built for a different era (a recurring scheme in the world of Health AI which I do not see going away anytime soon: technical progress in AI will always outpace regulatory actions). Here are the take aways from the paper: 🔹 Current state of regulation of AI agents in health and medicine: Most regulatory pathways, like the EU Medical Device Regulation and the U.S. FDA’s 510(k), are optimized for narrow, static, non-adaptive tools. Approval of AI-driven devices with broad scope and high autonomy remains rare and challenging. 🔹 The problem (why existing regulatory frameworks fall short of capturing agentic AI applications): AI agents orchestrate multiple tools, access sensitive data, and make goal-directed decisions autonomously. Their adaptive nature, coupled with non-deterministic outputs and complex multi-component architectures, pushes them outside the assumptions underpinning today’s medical device regulation. 🔹 Proposed solutions (how can we explore and experiment safely with AI agents in clinical settings in the absence of comprehensive regulatory guidance): The paper explores several emerging regulatory strategies, including: Voluntary alternative pathways → tailored approvals for high-risk AI agents Adaptive oversight → dynamic, real-world performance monitoring Regulatory sandboxes → controlled environments to experiment safely Outcome-driven evaluation → focusing on patient-centered results rather than static technical compliance 💡 The takeaway (so where does this leave us): While regulation is still playing catch-up with the risks and complexities of AI agents in healthcare, it is possible to safely experiment and sandbox their use — provided there are clear guardrails, strong oversight, and rigorous evaluation frameworks. This balanced, sandbox-first approach could enable innovation without compromising patient safety. For first hand insights on how this is all playing out in the real world right now, have a read of these two recent papers from Eric Topol, MD (https://lnkd.in/gEycp__S) and Harsha Nori (https://lnkd.in/gKNxpxsA) John Lim Christopher Bain ZongYuan 宗元 Ge 戈 Deval Mehta #AI #Healthcare #Regulation #AIagents #DigitalHealth #NatureMedicine

136

13 Comments

GenAI is a game-changer for healthcare analytics leaders - but with a massive looming hidden cost to organizations. As a biostatistician and epidemiologist with 25 years of experience, I'm seeing first-hand how GenAI is revolutionizing healthcare analytics. It's already boosting my efficiency: - Generating synthetic data - Streamlining literature reviews - Aiding analysis plans, code & reports Great starting points, but it's still just a tool. Expert review and human oversight remains essential - for accuracy and quality (at least for now). I've got 100+ quality improvement initiatives on 30M+ patients and $25B+ of value creation to fall back on. But, when I was just getting started, I didn't. What happens when GenAI simplifies tasks traditionally assigned to entry-level analysts? For example, data processing and preliminary analysis. These foundational experiences are vital for junior staff to develop into skilled, high-quality analytical professionals. Automating these roles creates a cascading effect: - Unemployed grads - Crucial career path gaps - A shortage of competent leaders Depriving our industry of experienced analytical minds who can critically evaluate, guide, and validate complex AI-driven strategies. That opportunity cost could be catastrophic. The solution? Bridging the academia-industry divide. What was once a "nice-to-have" is now a "must." - healthcare organizations need intentional, direct connections with academia. At the RRI, we've built an internship program to address this head-on. Our year-long program offers: 1) Substantive analytical experience Intern gains practical experience on real-world projects. 2) Novel statistical research They research and develop new methods for their thesis, filling critical gaps where solutions don't currently exist. A win-win-win scenario: - Students gain invaluable practical skills and a competitive edge in the job market. - Universities gain insights into industry needs, enhancing their curriculum and attracting top talent. - Industry builds a pipeline of highly-skilled candidates while fostering innovative solutions to real problems. Experience has instilled me with humility. I certainty don't have all the answers. As a community we grow together. Let's discuss. What are your thoughts on the long-term impact of GenAI on healthcare analytics? What crucial aspects am I missing? How will your team support the future of talent development? Comment below or shoot me a DM.

33

6 Comments

[EN] ✨ Data Quality and Validation in HL7® FHIR® Mapping with AI ✨ The application of LLMs in clinical ETL (Extract, Transform, Load) has great potential to accelerate the mapping and harmonization of HL7® FHIR® data— provided that data quality and validation are treated as core requirements of the process. In a recent study, we explored the use of LLM, integrated with the MCP (Model Context Protocol), to support automated ETL and variable mapping for HL7® FHIR® repositories, using HAPI FHIR as the underlying technological foundation. 🔍 Technical approach used: ➡️ Mapping structured into threads. ➡️ Prompt-driven creation of HL7® FHIR® resource instances. ➡️ Automated generation with technical and semantic validation. ⚠️ Lessons learned: ➡️ LLMs accelerate ETL, but referencing and conformance errors still occur. ➡️ Validation remains indispensable, especially in clinical and regulatory contexts. 📚 Evidence from the literature: ➡️ The literature reinforces that automation in ETL and FHIR® mapping requires systematic validation. ➡️ Metadata- and template-based pipelines (e.g., MDRBridge) ensure data quality through automated checks for completeness, uniformity, and uniqueness prior to data persistence. ➡️ Complementarily, FHIR-oriented pipelines incorporate multiple validation stages throughout the ETL, ensuring structural and semantic compliance before transforming data into formats more suitable for AI and LLM consumption (“AI-friendly”). ♾️ Key challenge: The multilayer architecture of FHIR® is not always directly consumable by AI. In many scenarios, an additional abstraction layer is required to make data more usable by LLMs without compromising clinical semantics, traceability, or data governance. 🎯 Conclusion LLMs are powerful allies in ETL and HL7® FHIR® mapping, but they do not replace data governance, rigorous validation, or domain expertise. Data quality is the true critical success factor. I invite you to follow M-CareTech for updates and insights. Note: Images generated by AI. #HL7FHIR #ETL #HAPIFHIR #LLM #Data #Interoperability #DigitalHealth

14

Great commentary on an outstanding article. Both are must-reads for anyone developing or using or regulating or otherwise interested in prediction models in healthcare (particularly, medicine)

3

2 Comments

In our latest research, we looked at a core challenge in clinical AI: connecting information across a fragmented record.   Key details like medication changes, dosing, and outcomes are often spread across thousands of pages, documented months apart. If those relationships aren’t reconstructed correctly, the output breaks down.   This is what led to the development of C-RLM.   Instead of summarizing, the system builds a structured clinical state, enforces schema constraints, and links every output back to its source. What stood out wasn’t just accuracy. It was the consistency and traceability of the outputs. You can see how a regimen was assembled and verify where each element came from.   That’s where clinical AI needs to go, especially in environments where decisions must be explainable and defensible.   👉 Sharing a short carousel with key takeaways. 🔗 Read the blog to learn more about the research: https://lnkd.in/eXesRAZi   #HealthcareAI #ClinicalAI #AIResearch #ResponsibleAI #HealthIT #AIinHealthcare #DigitalHealth

14