Josh Mandel, MD

Now with the full set of nearly a thousand comments on the CMS RFI: joshuamandel.com/regulations....

YouTube video by Josh Mandel W3C Digital Credentials Demo: SMART Health Wallet!

New live demo+deep dive! "What's next for sharing SMART Health Cards & FHIR bundles?" We explore W3C Digital Credentials API, browser-mediated presentation, & path beyond "show me your QR".

Demo+Deep Dive: youtu.be/pgoaZjYMGJY

Article: linkedin.com/pulse/better...

Source: github/jmandell/shl...

GitHub - jmandel/fhir-diff Contribute to jmandel/fhir-diff development by creating an account on GitHub.

FHIR is evolving! I prototyped an LLM-driven analysis to diff R4 vs. R6 resources and study downstream effects on IGs like US Core. The goal: explore automated methods for spec insight.

Article: linkedin.com/pulse/evalua...

Output and code: github.com/jmandel/fhir...

YouTube video by Josh Mandel MCP for EHR: Discussion with Aledade Team

Great discussion with Farzad Mostashari and the Aledade, Inc. team on applying Model Context Protocols (MCPs) for LLM-EHR interaction! We touc on data access, authorization & the future of AI in healthcare.

Write-up: linkedin.com/pulse/mcp-pr...

Video: youtu.be/AMPuz56qhx4 (sorry for poor audio!)

YouTube video by Josh Mandel MCP postMessage Transport: Secure, Local AI Tools In Your Browser (Demo)

New Proposal: "MCP postMessage transport" using browser sandbox for security, privacy + zero-install MCPs. See my EHR data demo & read the proposal.

Article: www.linkedin.com/pulse/mcp-yo...

Video with Demo: youtu.be/_VuMRotKbV8

I used 50 leftover o3 credits to test ChatGPT’s vision capabilities on cryptic crossword grids. Pure vision failed; combining vision with code got there.

Web app: chatgpt.com/canvas/share...

Blog post: www.linkedin.com/pulse/o3-mee...

New Perspective: The EHR Association's push to remove foundational data access is misguided & harmful for health innovation. www.linkedin.com/pulse/ehr-as...

YouTube video by Josh Mandel Live Demo: Conversational Interop for Prior Auth (LLMs, A2A, and MCP)

What if, instead of pre-specifying every data field and every possible workflow step, we could enable conversations between capable systems? Prior auth and beyond!

youtu.be/BRX7HUBlEqw?...

www.linkedin.com/pulse/conver...

... then Claude hit a token output limit while finishing transcribing the solution into a web app. So I spent a third turn saying "Continue," and it emitted the final HTML tags. I'm counting this as a moral success, but it technically violates my pre-specified terms.

Day 4: ⭐ ⭐ (*) claude.site/artifacts/f5...

On Part 2, Claude took 4 code analyzer blocks to fully debug its algorithm. (It's rare in my experience that broken code gets fixed after three failed attempts, so this was a pretty impressive feat.)

My running total score: 2 ⭐ (100%)

claude.site/artifacts/d1...

Fine print: web app artifacts allowed, w/ paste + one click only.

Per adventofcode.com/2024/about, wait until the daily adventofcode.com leaderboard is full before submitting LLM-generated solutions!

Of course, feel free to use ChatGPT custom instructions, static prompts, etc :-)

New Challenge: "Advent of Claude"!

Write a Claude(dot ai) custom style to solve Advent of Code puzzles within Claude's UI.

Score: # adventofcode(dot com) stars earned in 2 daily conversation turns.

turkeydown: a markdown-shaped, LLM-powered prompt flow for meal prep and more 🦃 🎉

My Thanksgiving 2024 template, defining processing logic:
github.com/jmandel/turk...

Output including prep plan, shopping list, & full recipe set for everything we're cooking this Thursday:
github.com/jmandel/turk...

Outputs.

All recipes: gist.github.com/jmandel/5ace...

Shopping list: claude.site/artifacts/2a...

Prep plan: claude.site/artifacts/57...

Prompt 5.

Now prepare a combined shopping list organized by grocery department and all duplicate ingredients combined to create totals, but listing the dishes and subquantities where they're used in parens. use check boxes interact app that is formatted for easy printing.

Prompt 4.

Now output an artifact containing each recipe with detailed ingredients including quantities, with prep steps written clearly and with quantities repeated inline. Don't bother to confirm with me first, just use your judgment and output all recipes. Separate with ---

Prompt 3.

Output a web app with this in a giant table

Prompt 2.

Now output a table where the rows are days and each column represents a different dish, so that we can see what prep steps should happen at which cell. (Notes: Focus Sunday on Turkey alone.)

Begin by outputting a recipe sketch for each dish. When a recipe is supplied, you should use it as the basis of your sketch. If no recipe is supplied, begin with the recipe sketch that you output so that we understand the overall process.

We're going to make: wild rice soup, salad, green beans, broccoli rabe, turkey, herbed gravy, rutabaga stuffing, sage biscuits, apple tart, pumpkin Bundts.

Sequential prompts.

Prompt 1. Our eventual goal is to create a prep schedule starting Sunday night through mealtime Thursday, emphasizing make ahead where it improves or doesn't detract from quality -- while also consolidating prep steps smartly.

Thanksgiving prep with LLMs. Here's my recipe:

0. Decide what to cook
1. Put all recipes (or dish ideas) in project context
2. Request
* 4 day prep plan
* All recipes with consistent formatting
* All ingredients in a combined shopping list

This graph shows birth outcomes data after 33 weeks of pregnancy, broken down by delivery method and induction status. The data spans weeks 34-42, with filters for age groups (15-50 years), BMI categories, and education levels. Key observations: 1. Peak births occur around week 39 (approximately 32% of total births) 2. Most common delivery method is Vaginal Non-Operative (both induced and non-induced) 3. C-section rates are relatively consistent across weeks, with slight increases in later weeks 4. Birth rates drop significantly after week 40 The weekly statistics show a progressive increase in births from week 34 (1.7% of total) to week 35 (2.5%) and week 36 (4.7%), with most deliveries being vaginal non-operative. C-section rates remain relatively low but show slight increases in later weeks. The graph shows both elective and medically indicated procedures through the color-coding of induced vs. non-induced deliveries for each method.

❤️ CDC's NVSS: It's *amazing* that anyone can download 200MB .zip with ~all 3.6M USA births in 2023, including dozens of row-level attributes to answer questions like "when and how did women deliver, given specific assumptions for age, educational status, BMI?"

joshuamandel.com/birth-analysis

@bsky.app feature request: include a "following back" button inline in the "New user followed you" notification screen.

(Right now, new follows are grouped into a drop-down, but as you click each one to review — and follow back — the drop-down group is re-filtered and needs to be re-expanded.)

Princinpled (Q)HIN Design: Convenience AND Control - HackMD All stories feature Elena, a 42-year-old marketing executive whose healthcare providers participate in AcmeConnect, a Health Information Network (HIN).

Healthcare data access vs. privacy tensions aren't zero-sum. Better-designed Health Information Networks can bend the tradeoff curve, enabling convenience AND control.

No silver bullets, but real progress is possible.

User stories & principles: hackmd.io/@jmandel/hin...

ChatGPT, act as physical therapy consultant && web developer to design neck exercises, write a web app to track daily progress, & ramp up difficulty.

Quick enough to complete btw boarding + takeoff to #FHIR #DevDays

Output: joshuamandel.com/nexercise
Chat: chatgpt.com/share/7fac01...

EHI Export insight: >100 individual users (i.e., provider org staff) contributed to some aspect of my EHI in the past 6 years of ambulatory care.

This diagram shows only users & tasks mentioned at least 10 times. (Task labels generated by GPT-4, based on Epic's EHI schema.)

Sure thing! Are you game for a conversation that we record and share openly? I find these topics can be of much broader interest than they initially appear.

Integration of AI in healthcare requires an interoperable digital data ecosystem - Nature Medicine Electronic health information, including from electronic health records, is needed to develop AI tools for health, but the seamless flow of data will require standards and interoperability.

Just out: www.nature.com/articles/s41...

Health AI integration requires workflow & data access. @SMARTHealthIT experience: it takes products, pragmatics, open standards... and regulations.

Better alignment & precision on "jobs to be done" => (much) faster timeframes!

@mandl @gotdan