What could possibly go wrong?
For one, bias propagation, reflecting clinicians' habit of copying forward notes in EHR, which may contain stereotypes based on race, ethnicity, gender, etc
For one, bias propagation, reflecting clinicians' habit of copying forward notes in EHR, which may contain stereotypes based on race, ethnicity, gender, etc
“OpenAI said today that it is integrating ChatGPT Health with Epic’s electronic health record (EHR) system, which holds data for over 325 million patients, to let clinicians import patient data and use AI to ask questions.”
ChatGPT Health adds Epic integration for clinicians to import patient data | TechCrunch
OpenAI said that the integration provides read-only access to health records for clinicians.
techcrunch.com
September 2, 2026 at 1:25 AM
What could possibly go wrong?
For one, bias propagation, reflecting clinicians' habit of copying forward notes in EHR, which may contain stereotypes based on race, ethnicity, gender, etc
For one, bias propagation, reflecting clinicians' habit of copying forward notes in EHR, which may contain stereotypes based on race, ethnicity, gender, etc
@infinitescream.bsky.social gonna need you on this one
I hate everything. I hate that I now have to wonder if my cloud-based EHR is in cahoots with any problem companies and they probably are, right? ugh ugh ugh
I hate everything. I hate that I now have to wonder if my cloud-based EHR is in cahoots with any problem companies and they probably are, right? ugh ugh ugh
April 28, 2026 at 10:40 PM
@infinitescream.bsky.social gonna need you on this one
I hate everything. I hate that I now have to wonder if my cloud-based EHR is in cahoots with any problem companies and they probably are, right? ugh ugh ugh
I hate everything. I hate that I now have to wonder if my cloud-based EHR is in cahoots with any problem companies and they probably are, right? ugh ugh ugh
We tried to release a “light” web-based EHR platform for practices that just needed the basics. It fell flat on its face. Every practice said it would be great if it just had 3 additional buttons, but each one wanted 3 different buttons. Soon you end up where you started, with a giant, complex EHR.
September 27, 2026 at 4:16 AM
We tried to release a “light” web-based EHR platform for practices that just needed the basics. It fell flat on its face. Every practice said it would be great if it just had 3 additional buttons, but each one wanted 3 different buttons. Soon you end up where you started, with a giant, complex EHR.
Are electronic health records (EHR) more predictive of disease onset than polygenic scores? Can we transfer EHR-based prediction models between countries? Our study on these questions using 3 biobank-based studies with N>845K, is out in @natgenet.nature.com today:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Cross-biobank generalizability and accuracy of electronic health record-based predictors compared to polygenic scores - Nature Genetics
Comparison of electronic health record-based phenotype risk scores (PheRS) and polygenic scores (PGS) across 13 common diseases and three biobank-based studies indicates that PheRS and PGS may provide...
www.nature.com
August 27, 2025 at 2:15 PM
Are electronic health records (EHR) more predictive of disease onset than polygenic scores? Can we transfer EHR-based prediction models between countries? Our study on these questions using 3 biobank-based studies with N>845K, is out in @natgenet.nature.com today:
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Idea: ATProto based EHR where the PDS concept serves as a Personal Health Record, ensuring portability.
*mutters* okay so can we integrate FHIR?
*mutters* okay so can we integrate FHIR?
November 16, 2024 at 4:28 PM
Idea: ATProto based EHR where the PDS concept serves as a Personal Health Record, ensuring portability.
*mutters* okay so can we integrate FHIR?
*mutters* okay so can we integrate FHIR?
Second: you can’t overstate how complicated EHR systems are. Not only do they have to handle a huge variety of complex workflows, but they have to do it at the pace demanded of a FFS system with the metrics required of a value-based system. The learning curve (for devs and users) is staggering.
September 27, 2026 at 4:11 AM
Second: you can’t overstate how complicated EHR systems are. Not only do they have to handle a huge variety of complex workflows, but they have to do it at the pace demanded of a FFS system with the metrics required of a value-based system. The learning curve (for devs and users) is staggering.
Over the last three years, I’ve been part of projects that have developed a number of computable phenotypes using NLP for to extract a features based on unstructured EHR data. This is important both for healthcare and public health. The methods used are very rigorous.
I'm not going to comment on the quality of this analysis, but if you are against the analysis of health data, idk what to tell you. They process text in clinical notes as the LLM part. They test a bunch of ML models to select the best one. Do you want prediction techniques in healthcare or not
March 13, 2026 at 10:02 PM
Over the last three years, I’ve been part of projects that have developed a number of computable phenotypes using NLP for to extract a features based on unstructured EHR data. This is important both for healthcare and public health. The methods used are very rigorous.
Do vaccines protect against dementia?
A new EHR-based study of >430,000 older adults found that AS01-adjuvanted vaccines (Shingrix for shingles & Arexvy for RSV) were associated with a lower 18-month risk of developing dementia whether given alone or together.
www.nature.com/articles/s41...
A new EHR-based study of >430,000 older adults found that AS01-adjuvanted vaccines (Shingrix for shingles & Arexvy for RSV) were associated with a lower 18-month risk of developing dementia whether given alone or together.
www.nature.com/articles/s41...
Lower risk of dementia with AS01-adjuvanted vaccination against shingles and respiratory syncytial virus infections - npj Vaccines
npj Vaccines - Lower risk of dementia with AS01-adjuvanted vaccination against shingles and respiratory syncytial virus infections
www.nature.com
July 22, 2025 at 7:52 PM
Do vaccines protect against dementia?
A new EHR-based study of >430,000 older adults found that AS01-adjuvanted vaccines (Shingrix for shingles & Arexvy for RSV) were associated with a lower 18-month risk of developing dementia whether given alone or together.
www.nature.com/articles/s41...
A new EHR-based study of >430,000 older adults found that AS01-adjuvanted vaccines (Shingrix for shingles & Arexvy for RSV) were associated with a lower 18-month risk of developing dementia whether given alone or together.
www.nature.com/articles/s41...
New preprint 👇
Does diagnostic misclassification in EHR-based psychiatry GWAS suppress genetic discovery?
We use simulations and FinnGen data to expand the SuperControl framework and propose a new ML approach (PRISMA) to boost signal discovery.
Led by Lisa Eick & Zhiyu Yang!
Does diagnostic misclassification in EHR-based psychiatry GWAS suppress genetic discovery?
We use simulations and FinnGen data to expand the SuperControl framework and propose a new ML approach (PRISMA) to boost signal discovery.
Led by Lisa Eick & Zhiyu Yang!
December 29, 2025 at 2:51 PM
New preprint 👇
Does diagnostic misclassification in EHR-based psychiatry GWAS suppress genetic discovery?
We use simulations and FinnGen data to expand the SuperControl framework and propose a new ML approach (PRISMA) to boost signal discovery.
Led by Lisa Eick & Zhiyu Yang!
Does diagnostic misclassification in EHR-based psychiatry GWAS suppress genetic discovery?
We use simulations and FinnGen data to expand the SuperControl framework and propose a new ML approach (PRISMA) to boost signal discovery.
Led by Lisa Eick & Zhiyu Yang!
NEW PREPRINT ALERT ⚠️
Are electronic health records (EHR) more predictive of disease onset than polygenic scores (PGS)? Can we transfer EHR-based prediction models between countries?
We studied these questions using 3 biobank-based studies with >845K people.
A thread 🧵
Are electronic health records (EHR) more predictive of disease onset than polygenic scores (PGS)? Can we transfer EHR-based prediction models between countries?
We studied these questions using 3 biobank-based studies with >845K people.
A thread 🧵
October 9, 2024 at 7:08 PM
NEW PREPRINT ALERT ⚠️
Are electronic health records (EHR) more predictive of disease onset than polygenic scores (PGS)? Can we transfer EHR-based prediction models between countries?
We studied these questions using 3 biobank-based studies with >845K people.
A thread 🧵
Are electronic health records (EHR) more predictive of disease onset than polygenic scores (PGS)? Can we transfer EHR-based prediction models between countries?
We studied these questions using 3 biobank-based studies with >845K people.
A thread 🧵
ich bin ja ehr in der Luftfahrzeugtechnik und weniger in Air Power, vibe based würde ich dasselbe sagen, aber dopple und reiche weiter an @katho.bsky.social
April 7, 2026 at 11:22 AM
ich bin ja ehr in der Luftfahrzeugtechnik und weniger in Air Power, vibe based würde ich dasselbe sagen, aber dopple und reiche weiter an @katho.bsky.social
One of the 'niches' that I now fill is working with clinician researchers to help them avoid implicitly 'correcting for race' when studying clinical outcomes using EHR linked with area-level data (based on patient residence).
I think it's well past time to see "correcting for race" as the racist logic it is.
Methodological individualism is well-suited for white supremacist logics because it ignores the fundamental, relational aspects of power. Here, outcomes borne of *racism* are encoded as individual characteristics.
Methodological individualism is well-suited for white supremacist logics because it ignores the fundamental, relational aspects of power. Here, outcomes borne of *racism* are encoded as individual characteristics.
February 9, 2025 at 4:08 PM
One of the 'niches' that I now fill is working with clinician researchers to help them avoid implicitly 'correcting for race' when studying clinical outcomes using EHR linked with area-level data (based on patient residence).
Can EHRs prevent hospital readmissions? 🏥
Systematic review of 116 RCTs shows EHR-based interventions significantly reduce 30- and 90-day readmissions. A step toward safer, cost-effective care.
Full article: doi.org/10.1001/jama...
#DHPSP #DigitalHealth
Systematic review of 116 RCTs shows EHR-based interventions significantly reduce 30- and 90-day readmissions. A step toward safer, cost-effective care.
Full article: doi.org/10.1001/jama...
#DHPSP #DigitalHealth
August 24, 2025 at 10:33 AM
Can EHRs prevent hospital readmissions? 🏥
Systematic review of 116 RCTs shows EHR-based interventions significantly reduce 30- and 90-day readmissions. A step toward safer, cost-effective care.
Full article: doi.org/10.1001/jama...
#DHPSP #DigitalHealth
Systematic review of 116 RCTs shows EHR-based interventions significantly reduce 30- and 90-day readmissions. A step toward safer, cost-effective care.
Full article: doi.org/10.1001/jama...
#DHPSP #DigitalHealth
Medical transcriptionists - mostly female - are out of work by the thousands because of speech-to-text applications like Dragon Medical, and now LLM-based charting applications that are baked into every EHR system.
February 12, 2026 at 2:56 PM
Medical transcriptionists - mostly female - are out of work by the thousands because of speech-to-text applications like Dragon Medical, and now LLM-based charting applications that are baked into every EHR system.
For clinical prediction on structured EHR, do complex LLM pipelines beat simple count-based models?
A new preprint by @simocristea.bsky.social et. al. shows wins are split. Count-based methods (like LightGBM) remain a strong, simple, and interpretable baseline.
#MedSky #MLSky #MedAI
A new preprint by @simocristea.bsky.social et. al. shows wins are split. Count-based methods (like LightGBM) remain a strong, simple, and interpretable baseline.
#MedSky #MLSky #MedAI
Count-Based Approaches Remain Strong: A Benchmark Against Transformer and LLM Pipelines on Structured EHR
Structured electronic health records (EHR) are essential for clinical prediction. While count-based learners continue to perform strongly on such data, no benchmarking has directly compared them…
arxiv.org
November 6, 2025 at 4:41 PM
For clinical prediction on structured EHR, do complex LLM pipelines beat simple count-based models?
A new preprint by @simocristea.bsky.social et. al. shows wins are split. Count-based methods (like LightGBM) remain a strong, simple, and interpretable baseline.
#MedSky #MLSky #MedAI
A new preprint by @simocristea.bsky.social et. al. shows wins are split. Count-based methods (like LightGBM) remain a strong, simple, and interpretable baseline.
#MedSky #MLSky #MedAI
Prevalent Metformin Use in Adults With Diabetes and the Incidence of Long COVID: An EHR-Based Cohort Study From the RECOVER Program
diabetesjournals.org/care/article...
diabetesjournals.org/care/article...
September 18, 2024 at 11:08 PM
Prevalent Metformin Use in Adults With Diabetes and the Incidence of Long COVID: An EHR-Based Cohort Study From the RECOVER Program
diabetesjournals.org/care/article...
diabetesjournals.org/care/article...
Observational study based on EHR and not a blinded interventional study, but still very promising. Seems to be the adjuvent because the rsv vaccine with the same adjuvent has the same association
May 6, 2026 at 4:02 AM
Observational study based on EHR and not a blinded interventional study, but still very promising. Seems to be the adjuvent because the rsv vaccine with the same adjuvent has the same association
Editorial: While the SCREEN trial shows promise in using EHR-based sepsis alerts to reduce mortality, further studies are needed to understand the effectiveness of these alerts.
#CCRdownunder @criticalcarereviews.com
ja.ma/4g9yyuK
#CCRdownunder @criticalcarereviews.com
ja.ma/4g9yyuK
December 11, 2024 at 4:55 AM
Editorial: While the SCREEN trial shows promise in using EHR-based sepsis alerts to reduce mortality, further studies are needed to understand the effectiveness of these alerts.
#CCRdownunder @criticalcarereviews.com
ja.ma/4g9yyuK
#CCRdownunder @criticalcarereviews.com
ja.ma/4g9yyuK
How AI-Based Tools Are Improving the EHR Experience at the University of Iowa Healthcare #EHR #HIT #HIE #AI #ArtificialIntelligence #Healthcare
www.hcinnovationgroup.com/clinical-it/...
www.hcinnovationgroup.com/clinical-it/...
How AI-Based Tools Are Improving the EHR Experience at Univ. of Iowa Healthcare
Jason Misurac, M.D., associate chief health information officer, discusses the impact of implementing clinical data intelligence solution Evidently and ambient AI assistant Nabla...
www.hcinnovationgroup.com
August 1, 2025 at 7:03 PM
How AI-Based Tools Are Improving the EHR Experience at the University of Iowa Healthcare #EHR #HIT #HIE #AI #ArtificialIntelligence #Healthcare
www.hcinnovationgroup.com/clinical-it/...
www.hcinnovationgroup.com/clinical-it/...
New JAMA Network Open study:
📊 Long COVID prevalence estimated at ~16%
🏥 58 U.S. hospitals
🤖 AI-based EHR phenotyping
⚠️ Diagnosis codes alone missed many cases
The burden of #LongCOVID is likely much larger than administrative data suggest.
jamanetwork.com/journals/jam...
📊 Long COVID prevalence estimated at ~16%
🏥 58 U.S. hospitals
🤖 AI-based EHR phenotyping
⚠️ Diagnosis codes alone missed many cases
The burden of #LongCOVID is likely much larger than administrative data suggest.
jamanetwork.com/journals/jam...
Long COVID Persistence and Surveillance Gaps Across 58 US Hospitals
This cohort study quantifies the gap between true postacute sequelae of SARS-CoV-2 burden and diagnostic code–based estimates, determines the proportion representing chronic disease, and characterizes...
jamanetwork.com
May 31, 2026 at 9:05 PM
New JAMA Network Open study:
📊 Long COVID prevalence estimated at ~16%
🏥 58 U.S. hospitals
🤖 AI-based EHR phenotyping
⚠️ Diagnosis codes alone missed many cases
The burden of #LongCOVID is likely much larger than administrative data suggest.
jamanetwork.com/journals/jam...
📊 Long COVID prevalence estimated at ~16%
🏥 58 U.S. hospitals
🤖 AI-based EHR phenotyping
⚠️ Diagnosis codes alone missed many cases
The burden of #LongCOVID is likely much larger than administrative data suggest.
jamanetwork.com/journals/jam...
From the US:
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
journals.plos.org/digitalhealt...
Image is from latest Science for ME weekly update
#LongCovidKids #LCKids
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
journals.plos.org/digitalhealt...
Image is from latest Science for ME weekly update
#LongCovidKids #LCKids
April 16, 2025 at 1:06 AM
From the US:
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
journals.plos.org/digitalhealt...
Image is from latest Science for ME weekly update
#LongCovidKids #LCKids
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
journals.plos.org/digitalhealt...
Image is from latest Science for ME weekly update
#LongCovidKids #LCKids
US research
Characterizing sleep disturbance phenotypes in long COVID: an EHR-based study from a safety-net hospital
Screenshot from science for ME weekly update
www.academia.edu/3070-3530/3/...
#LongCovid #COVIDBrain #NeuroPASC
Characterizing sleep disturbance phenotypes in long COVID: an EHR-based study from a safety-net hospital
Screenshot from science for ME weekly update
www.academia.edu/3070-3530/3/...
#LongCovid #COVIDBrain #NeuroPASC
June 12, 2026 at 12:53 AM
US research
Characterizing sleep disturbance phenotypes in long COVID: an EHR-based study from a safety-net hospital
Screenshot from science for ME weekly update
www.academia.edu/3070-3530/3/...
#LongCovid #COVIDBrain #NeuroPASC
Characterizing sleep disturbance phenotypes in long COVID: an EHR-based study from a safety-net hospital
Screenshot from science for ME weekly update
www.academia.edu/3070-3530/3/...
#LongCovid #COVIDBrain #NeuroPASC
Insights from an N3C RECOVER EHR-based cohort study characterizing SARS-CoV-2 reinfections and Long COVID
#LongCOVID risk is higher in 1st infection than reinfection, some people are more sensitive to symptomatic disease, cohort reinfection rate 6.9%
www.nature.com/artic...
1/5
#LongCOVID risk is higher in 1st infection than reinfection, some people are more sensitive to symptomatic disease, cohort reinfection rate 6.9%
www.nature.com/artic...
1/5
Insights from an N3C RECOVER EHR-based cohort study characterizing SARS-CoV-2 reinfections and Long COVID
Communications Medicine - Hadley et al. characterize COVID-19 re-infections utilizing electronic health record study cohort data of over 3 million patients. They find severe initial COVID-19...
www.nature.com
July 11, 2024 at 2:43 PM
Insights from an N3C RECOVER EHR-based cohort study characterizing SARS-CoV-2 reinfections and Long COVID
#LongCOVID risk is higher in 1st infection than reinfection, some people are more sensitive to symptomatic disease, cohort reinfection rate 6.9%
www.nature.com/artic...
1/5
#LongCOVID risk is higher in 1st infection than reinfection, some people are more sensitive to symptomatic disease, cohort reinfection rate 6.9%
www.nature.com/artic...
1/5
A new study found 6 subtypes of pediatric Long Covid:
• Respiratory
• Pain
• Neuropsychiatric
• GI
• Headaches
• Fatigue
It’s not one thing.
It’s a mosaic.
And kids deserve to be believed.
#LongCovid #PediatricLongCovid #DisabilityJustice
• Respiratory
• Pain
• Neuropsychiatric
• GI
• Headaches
• Fatigue
It’s not one thing.
It’s a mosaic.
And kids deserve to be believed.
#LongCovid #PediatricLongCovid #DisabilityJustice
Pediatric Long COVID Subphenotypes: An EHR-based study from the RECOVER program
Author summary Long COVID in children can have many different symptoms and clinical presentations. Understanding these differences may be important for both Long COVID research and patient care. We us...
journals.plos.org
April 16, 2025 at 6:10 AM
A new study found 6 subtypes of pediatric Long Covid:
• Respiratory
• Pain
• Neuropsychiatric
• GI
• Headaches
• Fatigue
It’s not one thing.
It’s a mosaic.
And kids deserve to be believed.
#LongCovid #PediatricLongCovid #DisabilityJustice
• Respiratory
• Pain
• Neuropsychiatric
• GI
• Headaches
• Fatigue
It’s not one thing.
It’s a mosaic.
And kids deserve to be believed.
#LongCovid #PediatricLongCovid #DisabilityJustice
Editorial: While the SCREEN trial shows promise in using EHR-based sepsis alerts to reduce mortality, further studies are needed to understand the effectiveness and implementation of these alerts.
ja.ma/3QJ6qmU
#MedSky
ja.ma/3QJ6qmU
#MedSky
March 4, 2025 at 5:37 PM
Editorial: While the SCREEN trial shows promise in using EHR-based sepsis alerts to reduce mortality, further studies are needed to understand the effectiveness and implementation of these alerts.
ja.ma/3QJ6qmU
#MedSky
ja.ma/3QJ6qmU
#MedSky