It goes without saying that I love working with John and Vinay on Sensible Medicine. I appreciate that they often write about topics I haven’t thought about. I like it even more when they address something I have thought about, maybe even something we have talked about, because I can trust that they will come at it with a different perspective.
Vinay’s recent article, Things you should have stopped doing already because of AI, falls into this second category. As much as I try to avoid all the chatter about AI in medicine, I do think about how it will change what we do. I loved the Prasadian take: Don’t waste your time doing this anymore!
One of our committed readers texted me before I had even read the article, suggesting that Vinay’s article was a good example of first-level thinking that called for some second-level thinking. It was said without disrespect, just, “If you shouldn’t do those things, what should you do?” Being the cheery optimist among the three of us, here is what you, doctor, should be doing in the days of AI.
Prepare for your patient visits. There is nothing you can do to make a patient’s visit more efficient and productive than being fully prepared. Patients often tell me they are grateful and feel well served when it is clear I have reviewed their history before I come into the room. I always like to enter with a clear agenda – one which I am willing to abandon if the patient has his or her own agenda.
The AI built into the EHR has made this prep work easier. You can usually get a pretty good overview of a patient’s history with little prompting. What is left is a careful review of test results and your colleagues’ assessments.
Spend time with your patients. AI should free us up to do what only we, as human doctors, can do. Osler famously told us, “The good physician treats the disease; the great physician treats the patient who has the disease.” If you’d like, or must, use AI tools to help you make the correct diagnosis and suggest available treatments. It is up to you to understand who you are caring for, discover what matters to them, and deliver personalized care.
Document your thinking. We have let the EMR and billing requirements ruin the medical record. It became filled with useless checkboxes and copied, pasted, auto-inserted mindless bloat. With AI able to summarize charts, all this is now more useless than ever. Let AI draft your note, formatted to meet billing requirements.
Then, sit down and write a paragraph or bulleted list of your thoughts. Five minutes of writing about your patient will sharpen your thoughts about your plan. It will help you remember the visit. It will also help your colleagues know what you are doing and why. It is what we were trained to do: think about medicine.
Talk to your patients between visits. The EHR has buried us in messages. AI has helped to manage some of this, and it could probably help us more. I hope we will see AI handle more of the mindless communication with insurers, home care agencies, and prior-authorization authoritarians. Supervised AI can help with routine communication with patients. AI will do more of the follow-up that we should do more of: check-ins regarding blood sugar, blood pressure, weight, and the like.
This should enable us to have the conversations that will make people better and more satisfied with their care:
Ms. Jones, I am calling to talk about your labs. They all look good except for your calcium level. Let me explain what that means and what the next steps should be.
Mr. Brown, I know you saw the EKG results; they might sound scary to you. Let’s go through them.
Mr. Smith, I gave you a lot of information during our visit on Tuesday. I wanted to see how you are doing and hear if any more questions came up.
Ask interesting questions. Vinay was particularly (and appropriately) harsh on the questions that no longer need humans to investigate.
We should develop the interesting, unanswered questions. Every day in the office, the emergency room, the procedure suite, or the operating room, we face decisions guided only by expert opinion or observational data. Spend your time identifying the questions whose answers will benefit our patients.
The next step, figuring out how to do the trial that will answer the question, is incredibly hard. So make sure the purpose of the research benefits patients, not a company’s bottom line or your CV.
Interview applicants. I think we do a terrible job selecting from the applicants to medical school, residency, and fellowship. We have weakened our assessments in every stage of training. Personal statements and letters of recommendation are being written or enhanced by AI. Let’s get back to putting two humans in a room together.
Interviewers can learn how applicants think and what they know. They can ask hard, creative questions to figure out whether this person will flourish in their program. Interviewees can determine if these are the people they want to work with, learn from, and emulate.
Teach the next generation of physicians. Trainees can acquire knowledge from textbooks, podcasts, and computers (with or without AI). They can only learn to be doctors from other doctors. Experienced doctors must mentor students, trainees, and junior colleagues. Training requires role modeling, observation, feedback, and deliberate practice. It requires trainees taking the best from the people they work with (while also learning what not to do). It requires experienced clinicians to model professionalism, clinical reasoning, and surgical technique; to forgive and remember; to hold their juniors to the highest of standards.
What is notable about these seven things? They are the things that have always been, and always will be, the core of medical practice: dedicated, personalized patient care; inquiry aimed at improving medical practice; training the next generation so that medicine will continue to improve.
If AI is as time saving as the EMR we’re all effed.