Take Your Health Into Your Own Hands: Record Your Doctor
The healthcare system is overworked. AI has got better. Now is the time to own your own path.
5 years navigating the NHS and thousands of pounds wasted on tests, private appointments and potions for a chronic health issue I’m still yet to solve has convinced me of 3 things:
We are each responsible for our own health, and for the results we get from healthcare systems
We would all benefit from using the AI technology available today to build our own health sidekick
We should record all interactions we have with doctors/practitioners and feed them into our health sidekicks - probably the most controversial (for now) take
I’ve been navigating a chronic sinus issue for the last 4-5 years. Most days I’m running at 6/10. When it’s bad, I’m running at 3/10. I’ve felt exhausted, frustrated and lonely being sent down dead end after dead end.
I’ve driven hours for tests ordered by a new practitioner only to be told on arrival that they’re unnecessary. I’ve waited 9 months for a referral to a specialist, turned up to my first consultation with them full of hope and been told by that specialist that they’re not the right person to help. I spent £300 on a private appointment that lasted 10 minutes and ended with the doctor suggesting I try a bit of heartburn medicine (?!) and come back in a couple of months if things hadn’t improved.
The NHS is under huge pressure. It’s underfunded and under-appreciated and the excellent people within it are overworked. In an ideal world, I’d be receiving joined up care from practitioners with deep context about me, a clear view of my history, and enough time to work through all of it. Whether we like it or not, that world doesn’t currently exist for most of us.
We can pretend that’s not the case, or we can moan about it, or we can take our care into our own hands. Over the last year or so, I’ve decided on the latter option.
The most popular LLMS (Claude, Gemini and ChatGPT) all score above 90% on MedQA - a benchmark made of thousands of multiple-choice questions in the style of the US medical licensing exams that tests medical knowledge and clinical reasoning. c.60% is deemed a ‘passing’ result, with c.87% deemed a ‘human expert’ result.
I’ve set Claude, Gemini and ChatGPT up with full context on me - including medical history, symptom diaries, blood test results, data from my home, and extensive back-and-forth question and answers. They have become my guides to taking ownership of a resolution to my challenges, including my path through the NHS.
Recording my conversations with practitioners has been the final step. I now politely let doctors and specialists know that I would like to record our session for my own review (including with AI) later on. I use Granola, then paste the transcript into the LLMs afterwards. I ask for general reviews, and ask pointed questions. The conversation becomes more insight and data for the LLMs to use to help me get to the bottom of things.
For NHS doctor and hospital visits, we all have the legal right to record whether you ask or not, but it feels better asking. Recordings are easily transcribed or fed whole into an LLM with context about you to help understand what the practitioner has told you, expand on and explain any technical terms, build on their thinking, spot things that might have been missed, and suggest next steps.
This approach is working well for me. Late last year I saw a breakthrough with my own health journey when Gemini - armed with my history, latest appointment transcriptions, test results and symptom logs - spotted a potential correlation and suggested an experiment to run to confirm a potential cause of the issues.
LLMs have also helped me go into sessions with practitioners with a plan, insightful questions to ask and suggestions to make to health practitioners. Occasionally, they’ve been disappointed with the standard of care shown during appointments - and occasionally, they’ve been incorrect in their interpretations of things. They’ve been available 24/7 with deep context and no wait times.
There is pushback, of course - one ENT specialist scoffed that I was ‘another person using doctor Google’ - but faced with the alternative of going around and around without direction in a system too overwhelmed to help me solve my issues, and living life with the handbrakes on, that’s a small price to pay.
Other doctors have been far more positive - interested in the approach, or thankful for the depth of context provided. We’re aiming here for a collaboration - you, the healthcare system and AI vs the problem - and avoid conflict and confrontation.
There are pitfalls to avoid with this approach, but none are insurmountable. Don’t outsource all of your thinking. Think first, AI second, think third. Question the results you get. Use the LLMs with practitioners, not instead of them. Politely ask to record. Give as much context as possible. Use multiple LLMs.
To get started:
Start a new project in ChatGPT, Claude or Gemini
Upload anything relevant to your personal health situation that you’re comfortable sharing (check out each company’s privacy stance first). This might include test results, symptom logs, transcripts or summaries of appointments, etc.
Ask the LLM to interview you about your issue. Use a prompt like:
‘I know you are not a doctor. I want you to act as one to help me resolve a health issue i’m facing. I will also be consulting human specialists including doctors. Below is a brain dump of everything I can easily remember and consider relevant to my situation. My aim is [enter your aim - mine was ‘resolve this situation permanently within the next 3 months so that I never feel these symptoms again’]. You may be able to help me by suggesting potential diagnoses to explore with my doctor, suggesting experiments I can run at home, suggesting further testing I can undertake, or suggesting ways to reduce my symptoms. You can also suggest ways to help me outside of these.
Begin by thinking and researching deeply and then asking me questions that will help you to understand my situation, lifestyle, history, potential causes and the journey I’ve been on with this issue so far. You may also request more data, and I will share it if I have it.
Do not try to diagnose the issue until you feel you have enough context and information to do so. I am interested in your thoughts - even not fully formed or confident ones - but please explain your level of confidence for anything you share, and explain your thinking.
[Brain dump of everything relevant you can think of - history, symptom descriptions, hunches you’ve had, things you know it isn’t and why, interactions with doctors so far, etc]
Go back and forth answering the questions that pop up. Be critical of and do deeper research on anything it suggests.
You can return to the health coach each time you’re preparing for an interaction with a practitioner - just explain who you’re seeing and the aim, and ask it to be as helpful as possible in your preparation.
This feels like the way things will be going forward. OpenAI and Apple are both working on huge AI for Health bets. Doctors are writing think pieces like the one below about how their role ‘has changed forever’. There are thousands of startups working on becoming your AI Health Copilot. It’ll become easier to do - but it’s already absolutely possible.