HugoOC retweetledi
HugoOC
6.3K posts


Using AI for healthcare? I'm hoping to connect Bay Area ePatients using #ChatGPT, #Claude, or #Gemini for medical visits + records. Looking to build a network for patient-led AI use. If you are in Bay Area please DM/reach out. #PatientsUseAI #ePatients #HealthAI #BayArea
English
HugoOC retweetledi

AI-assisted primary care in Kenya racks up impressive reduction in diagnostic and therapeutic errors. When do we fix primary care in Boston and Palo Alto?
Karan Singhal@thekaransinghal
📣 Excited to share our real-world study of an LLM clinical copilot, a collab between @OpenAI and @PendaHealth. Across 39,849 live patient visits, clinicians with AI had a 16% relative reduction in diagnostic errors and a 13% reduction in treatment errors vs. those without. 🧵
English
HugoOC retweetledi

@danamlewis Great diagram! A few more:
1⃣ Submit the same task to multiple models and compare outputs;
2⃣ Provide the AI with detailed, task-specific context;
3⃣ Use one model to help refine prompts for another;
4⃣ And my secret weapon: apply chain of thought reasoning across models.
✊👊
English
HugoOC retweetledi
HugoOC retweetledi

Engaging patients as co-authors is essential for patient-centered science. Yet little guidance now exists to explain or help navigate complex co-authorship processes. With @HugoOC @mcpoliti and others, I share some thoughts on how to make things better:
link.springer.com/article/10.118…
English
HugoOC retweetledi

@TheLizArmy @UCDavisHealth @UCSFHospitals @BeLikeLight @chethanr @trentrosenbloom @ePatientDave I found that it has trouble reporting accurate details. So, I started using PicnicHealth so I could easily check the timeline and compare dates, and also started feeding the LLM smaller PDFs. I find that it's more accurate that way.
English

Is the GPT accurate?
It got these hospitalizations right!
Can't wait to include data from @UCDavisHealth & @UCSFHospitals (2017-Now). 🤓
#healthIT #patientsuseAI

English

I'm about 1/3 of the way through creating a personalized GPT that includes:
-medical records from @kpnorcal @UCDavisHealth + @UCSFHospitals
-research grade genomic results
-personal narratives from >500 blog posts
So far just the KP data is uploaded. Ask me how... 😅


English

@zakkohane 5/5 #GenerativeAI has revolutionized how I manage my health and interact with the #healthcare system. It has enhanced my #autonomy and allowed me to better partner with my PCP and cardiologists. #HealthTech #AI #PatientCare
English

@zakkohane 4/5 For example, I find that asking one model to help me prompt another often yields the best results. I cannot imagine my life as a chronic patient without access to these AI tools. They are life-changing!
English

@zakkohane 3/5 I just started using Meta Llama 3.1-405B and have been impressed with its medical knowledge for my use case. Each LLM has its unique strengths and weaknesses, so I find that combining various models is the most effective way to get the answers I seek.
English

@zakkohane 2/5 I also conduct research, analyze data, seek medical advice (yes, that too), and even assist with diagnoses using AI. I primarily use GPT-4, followed closely by Perplexity Pro, Claude 3.5 Sonnet, Opus, Gemini, Cohere Command R+, and any other LLM I can get my hands on.
English

@zakkohane 1/5 As a patient and caregiver to an elderly parent, I use #GenerativeAI every single day for everything health-related. This includes writing succinct messages to my doctor, preparing for clinic visits, and translating records to Portuguese.
English

@asher_971 I live in the San Francisco Bay Area and got the same message. I use Gig car share and love the service. This is too bad.
English

I just published Howard, my healthcare agent. link.medium.com/3vYvs1W1zJb
English

@ePatientDave I'm only experimenting to see which of the three provides the most value to a patient seeking to understand their condition or that of a loved one.
Like I said, I have high hopes for Gemini Advanced, but GPT4 is really knocking it out of the park.
English

@HugoOC I'm thrilled that you're doing this work, Hugo. What can we learn from this comparison? Are there clinically useful differences for a patient (or clinician)? Or are you just experimenting to see what happens?
English

@ePatientDave I asked Claude.AI the same question but it needed a readable PDF. Claude missed the EF, a critical bit of information. #GPT4 provided the most comprehensive analysis.
What these LLMs do really well is translate medical lingo into language we all can understand.👏👏

English








