Cathy Gao-Howard MD MS

Exploring How AI Is Responsibly Deployed in Medicine Artificial intelligence (AI) is already revolutionizing healthcare delivery. From training and research to diagnosis and treatment, Northwestern Medicine experts are shaping how AI is responsibly...

At the recent AI in Healthcare symposium at Northwestern Memorial Hospital, participants heard from NM experts who are shaping how AI is used responsibly in health care. I.AIM Director Abel Kho served as course director.
@catgaohow.bsky.social
ow.ly/kpQb50X4w4U
#AIinHealthcare

Huge kudos 🙌 to #HBMI #HSIP MD/PhD student @vijeethg.bsky.social , plus Marjorie, Melissa, @keyboardbiologist.bsky.social & the whole #SCRIPT team! 🎉
@iaim-nu.bsky.social
@nufeinbergmed.bsky.social
@nuintmed.bsky.social

3/ Then came the LLMs 🤖✨
GPT-4o, GPT-4o mini, Llama 3.1 🦙, Mixtral-8x7B.
Clear winner: LLMs by far 🚀

To validate externally, we labeled MIMIC notes 📚 and published theses annotations so others can build on this:
🔗 physionet.org/content/immu...

Paper code: github.com/NUSCRIPT/gug...

2/ Most studies lean on structured data (ICD codes 📊) to capture comorbidities.
We tested 10 thresholds → decent performance, but not perfect.

Next: classic NLP 📝
6 bag-of-words classifiers (NB, lasso, ridge, decision tree 🌳, random forest 🌲, SVM).
Results? Also not bad!

Large language models accurately identify immunosuppression in intensive care unit patients AbstractObjective. Rule-based structured data algorithms and natural language processing (NLP) approaches applied to unstructured clinical notes have limit

1/ What started as a “hey, wouldn’t it be cool…” 💡 side project turned into a full study on how to identify immunocompromised patients.
Gold standard = research coordinators manually labeling statuses ✅
Everything else tested against that.
🔗 academic.oup.com/jamia/advanc...

Some important findings in this paper:
1) Working with AI boosts the performance of people solving math, science & ethics questions
2) The biggest boost is for the hardest problems
3) High performers remain highest performing, but low performers gain more
4) People who are good with AI gain most

Awesome work from @vijeethg.bsky.social - LLMs for the win! Hopefully this is one step towards doing less chart review. #LLMs #GPT @iaim-nu.bsky.social @nufeinbergmed.bsky.social @nuintmed.bsky.social #HBMI #HSIP

Large language models accurately identify immunosuppression in intensive care unit patients AbstractObjective. Rule-based structured data algorithms and natural language processing (NLP) approaches applied to unstructured clinical notes have limit

#LLMs outperformed structured data algorithms and natural language processing in identifying various types of immunosuppression in ICU patients. Read more about our approach here: doi.org/10.1093/jami...; ft. @keyboardbiologist.bsky.social @catgaohow.bsky.social

We hope it helps fellows as they begin their own searches.🌱 Grateful to Emily, Matthew, Pat, @nick-ingraham.bsky.social Clara, Nandita, and everyone who helped take and share the survey! <3 #earlycareer #PCCM

We found striking variability: clinician-educators 👩‍🏫 often received larger offers than physician-scientists 🔬, and both non-clinical FTE ⏱️ and startup packages 💼 were all over the map.
Our preprint is here 📄: www.medrxiv.org/content/10.1...

During my own job search, I was struck by how wildly academic offers vary—from $0 startup 😬 to multi-million-dollar packages 💰. Inspired by online datasets (ex: www.reddit.com/r/AskAcademi...), we surveyed PCCM physicians about their offers 📊.

CCCF Datathon The Canadian Critical Care Forum presents the inaugural 2025 Datathon supported by the ATS Critical Care Assembly. The CCCF Datathon will bring together clinicians and scientists in a unique,…

Join the Inaugural CCCF Datathon 2025! Participants will collaborate/compete in teams to tackle this year’s theme: “Overuse in the ICU” using MIMIC-IV! @atscommunity.bsky.social

Saturday, November 29th (limited to 25 participants): forms.gle/mj42ErBdxmKr...
Apply by 9/30!

Please join us on Wednesday, September 17th at Noon for our Next Medical Grand Rounds! We will be joined by...

-Ankit Bharat, MD
-Alexander Misharin, MD
-Mrinalini Venkata Subramani, MD
-Catherine Myers, MD
-Ambalavanan Arunachalam, MD

from @nufeinbergmed.bsky.social !

YouTube video by Institute for Artificial Intelligence in Medicine AI Essentials: Pretty Timelines with Claude

Check out this "AI Essentials" on creating pretty timelines with Claude. @catgaohow.bsky.social, Associate Director of the I.AIM Center for Collaborative AI in Healthcare, walks through the process, starting with ChatGPT and finishing with a polished Gantt chart in Claude.

youtu.be/sF_0VTNwAs0?...

Join us and Dr. Calfee TOMORROW for our first grand rounds of AY25-26!

Claude's new ability to work with Excel is the best I have seen so far

I have given it existing spreadsheets to work with and asked it to create new ones. Good use of formatting, formulas, etc.

It created all of this, including 406 formulas, from one prompt (& it would be solid work from my MBAs).

Eosinopenia as Predictor of Disease Severity in Patients With Community-Acquired Pneumonia Lower respiratory tract infections, including community-acquired pneumonia (CAP), are a frequent cause of death worldwide.1 A previous study2 found that 38% of patients with CAP requiring mechanical v...

3/ 🧪 Bottom line: Eosinopenia ≤50/μL = red flag 🚩 for severe CAP. Read paper in CHEST Infections journal.chestnet.org/article/S001...

@accpchest.bsky.social @chriscarrollmd.com

2/ 📊 German study 6,748 patients with CAP; 4,060 eosinopenic and 2,688 non-eosinopenic:
☠️ Mortality: 13.8% vs 9.1%

💨 Ventilation: 19.2% vs 14.3%

🦠 Sepsis: 7.5% vs 5.0%

🏥 Longer stays: 8.4 vs 7.6 days

1/ Something I don’t often look at in CAP patients, but might start to based on this paper: eosinopenia (eosinophils ≤50/μL), which has already been tied to poor outcomes in COPD & COVID-19.

American Thoracic Society | Pros and Cons of Private Practice: A Case… The American Thoracic Society is the world's leading medical society dedicated to accelerating the advancement of global respiratory health through…

🚀 Thinking about private practice? Join the Career Transitions Group on 9/8 at 4:30 PM EST for an interactive session! 📢 Facilitated by @pmdlc.bsky.social !

🔗 site.thoracic.org/events/pros-...

@ats-assemblies.bsky.social
@atsearlycareer.bsky.social
@atscommunity.bsky.social

Ever wonder about entering private practice after fellowship? We’ll discuss RVU targets, interviews, contracts, negotiations, and everything else you’ve wanted to know!

Mapping the Epidemiologic Features of Endotracheal Tube Obstruction Moderate ETB was more prevalent in patients in the ICU, with significant factors including coagulopathy, closed suction practice, and mechanical ventilation duration. Ppeak alarms lacked clinical impa...

2/ 🔑 Risks: coagulopathy, longer ventilation, closed suction
⚠️ Ppeak ↑ but not clinically useful as many blocked tubes did not have this elevation!
Full paper: www.chestcc.org/article/S294... @hbgmd.bsky.social
@chriscarrollmd.com
@accpchest.bsky.social

1/ 🫁 Recently bronched a patient with a gunked up ETT and wonder how often this happens? From #journal_CHESTCritCare:
369 ETTs cut and measured degree of blockage (see pic)
📊 Moderate blockage common (CICU 28%, MICU 17%)

Generative Medical Event Models Improve with Scale Realizing personalized medicine at scale calls for methods that distill insights from longitudinal patient journeys, which can be viewed as a sequence of medical events. Foundation models pretrained o...

Introducing Cosmos AI—a family of generative models trained on 115 billion medical events to simulate clinically probable futures and deliver actionable insights to care teams across the continuum of care.

Read the paper: arxiv.org/abs/2508.12104

Emmie hands things off to Art—AI that helps clinicians.

Designed to help doctors better understand their patients, eliminate busy work, and bring insights to help inform decisions, Art anticipates what’s needed for the visit and tees up tailored summaries for each agenda item.

Introducing Emmie—AI that helps patients.

Informed by your chart and connected devices, Emmie can explain results with context, answer open-ended questions to guide patients through healthcare, and take actions like scheduling visits.

@hey.epic.com first time at #UGM #EpicUGM conference and the campus is so cool! Also lots of exciting AI tools coming down the pipeline!

3) Variation persisted even with ARDS documented.
Predictors of proning: COVID status, code status, lower P/F ratio.

Take-home: Who your attending is may matter more than your oxygen level.
Also, prone your hypoxemic patients!

1) 514 ICU pts eligible for proning (P/F ≤150, FiO₂ ≥60%, PEEP ≥5): only 17% were actually proned. (why are we still so bad at this?)

2) 48 attendings analyzed → huge variation:
📊 Adjusted rates: 14.9%–74.2%
Greater effect than a 30 mmHg drop in P/F ratio.

Does it ever feel like some attendings prone everyone and some attendings prone no one? You're not imaging things! From #CLIFconsortium rockstar Anna Barker and #UMichMed:

#journal_CHESTCritCare
@hbgmd.bsky.social
@chriscarrollmd.com
@accpchest.bsky.social
www.chestcc.org/article/S294...