Those of you with inpatient specialties are perhaps less affected by these trends but as technology is getting more intertwined with healthcare, more and more patients are choosing to receive part of their medical care online. Current trends in virtual medicine involves artificial intelligence.
Artificial Intelligence In Medicine
The term AI is used a bit loosely but the idea is that the software can learn from the different input that patients and clinicians provide, feeding back into the system.
This is very different from what is actually called AI in current practice. For now, a person sits there and inputs information into a software, then changes that information as more and more clinical encounters take place, hardly intelligent.
Imagine a patient mentions a chief complaint of diarrhea and includes various symptoms such as
- abdominal pain
- blood in stool etc.
A real AI platform would learn from such input and ask the next patient the same questions, “Do you have blood in the stool?”. On the backend, it would cross reference the final diagnosis assigned to this patient’s symptoms and learn how to make the diagnosis. The “how” is what needs to be programmed into the software – the rest is just data, which is abundant.
Extrapolate this to the diagnosis the doctor makes, the medications or further testing they order, and you’ll have a really neat system that’s self-contained.
You may still need to do an exam on your patient to determine if they are a right candidate for pre-op, sit across from them and have a good discussion regarding options – but imagine all the pertinent data having been collected already, even things you may have forgotten to ask.
Misconception About Artificial Intelligence And Medicine
AI isn’t here to replace doctors or decrease the face-time with doctors. Sure, businesses are trying to develop AI so that more patient-dollars get funneled into their pockets, but any such profits will be temporary.
If you’re a family medicine doctor or urgent care doctor you have to admit that probably 80% of what you see in the urgent care could be managed by your mother, by a nurse and most definitely AI.
Someone with abdominal pain, diarrhea, chills, nausea and vomiting that started 2 days ago and is able to keep down some fluids, has no blood in the stool and is otherwise healthy… this person would not benefit from an urgent care or primary care visit, in fact there is often more harm than good.
Such patients who “clog the system” are likely going to get medications they don’t need, testing that is unnecessary and get exposed to others who are ill in the waiting room. It’s a sad term we use for such patients but it’s the nature of a society so dependent on external help as opposed to cultivating self-sufficiency.
AI would siphon such patients from the top of the queue, ask the right questions, put the patient to work answering all the pertinent positives and negatives, then package the information for the clinician to review with all the appropriate medications and recommendations already pended for the clinician to sign off on.
If you’re a colorectal surgeon then your AI system will be different from mine. Your post-op patient could go on your website at any given time, discuss their obstipation with the AI software and get an immediate recommendation.
Your patient wouldn’t push off seeking help because they are worried they would disturb you and they wouldn’t have to wait to talk to someone on the phone. Your AI software would know the right questions to ask, the same ones you have asked 1,000 times in similar situations.
What Companies Are Doing With AI
I came across bright.md which seems to be developing this concept well. They have the patient do all the work, answering drop-down menu questions, selecting checkboxes and filling in relevant information to their chief complaint.
This information is then filled in by the AI software to include a differential diagnosis, recommendations as to how the patient can self-manage, medications which are pended for the doctor to sign, return precautions which are included in the after-visit summary.
This package is emailed to the doctor who with 1 click can select the right option or simply agree with the assessment/plan.
Taking Care Of The Most Common Issues
AI won’t replace the need for an in-person visit for things such as:
- eye pain
- knee pain after trauma
- fever of unknown origin etc.
And I think being able to pick and choose the best situations for AI to get deployed is what will set the good implementers apart.
Someone with cold URI’s would be an ideal candidate for this. And since in America patients present to their doctor for cold and flu symptoms, this market would be ideal for AI technology.
The technology is still lagging and the visionaries have their eyes set on home runs instead of more tangible goals. They want the software to assess, diagnose and manage the patient, when in fact the HPI is perhaps the most mundane of all to obtain for the doctor and easiest to for the AI. Second on the list would be ordering the same referrals, medications and after-visit instructions.
Imagine being able to have a little bit of idle chit-chat with the patient instead of diving right into the color of the mucous coming from their nostrils.
Why AI Is Needed
There are a ton of companies out there who are trying to fuse AI and virtual medicine. I really like Startup+Health for a source of up-and-coming companies in the medical field.
Doctors in the front-line are still in massive shortage. We don’t have enough providers who can deliver the right kind of care to patients. There is a huge mismatch of what patients expect and what doctors can provide. Patients want more medications and more medical expertise, they want it right away and they often have more faith in technology than a doctor’s opinion.
Patients want more face-time with their own PCP and want more objective data to reassure them. A doctor saying “it’s a virus” isn’t often reassuring. But a software that’s written for this which is then blessed by the doctor could, sadly, do the job.
The following symptoms can be easily addressed by an AI tool:
- back pain
- medication refills
- asthma exacerbations
- sore throat
- ear pain
- ankle pains etc
Proper reassurance can be provided, minimizing time taken away from patients who are truly sick and also decreasing patient costs.
The Legal Landscape
One of the companies I consult for, is doing just this with their AI backbone. They even have an adorable name for their AI robot. The layout is friendly and the questions asked are pertinent.
What constitutes a “visit” in our current medical system? I will only focus on the US because each country is different. That said, each State has their own laws regarding telemedicine.
You can go on your medical board’s website and their practice guidelines for telemedicine are often incredibly vague. Therefore, each new company that comes on the scene has their own lawyers who interpret the guidelines differently.
I am doing work with multiple different companies and each has their own unique “rule” depending on which state they are offering their services in.
In some States there must be a “video” interaction between doctor and patient in order for it to be a legitimate patient-doctor relationship. In others a telephone visit is sufficient. Based on whether an actual relationship was established, you may or may not be “allowed” to order a treatment for that patient.
Having worked a few virtual medicine gigs now, I would like to offer my impression of the direction of virtual medicine. I think there are ways for physicians to profit in various ways.
Very little of what I do in the clinic is based on my supposed “intelligence”, it’s mostly based on flowsheets, community standards, resources available and some gut-instinct. I suppose this last 6th sense could be considered a sort of intelligence but that wouldn’t be an accurate term.
The point is that intelligence has very little to do with how AI can play a role in helping clinicians manage their patients.
I used to work in an urgent care nearly a decade ago where MA’s would obtain the entire HPI for me, order the tests that they knew I was gonna order, set me up with the right tools I needed to perform a procedure, to the point that all I had to do is walk up to the exam room, grab the paper with the HPI on it, read it and proceed to my assessment/plan after performing an exam or clarifying some information.
Now, sadly, in the unionized world of our current medical groups, we are seeing RN’s with bachelor degrees who are as clinically inhibited as a google search. It’s partly due to their training, the medical-legal scene and union rules.
Technology will replace the majority of the MA’s, LPN/LVN’s, and most definitely the majority of the RN’s. CNA’s and MA’s will carry out orders from clinicians. RN’s will be placed in positions where they have to “babysit” technology, whether it’s physical technology or binary.
Information gathering will be done at the time of check-in, instead of the patient filling out idiotic, cellulose, health questionnaires, their time spent in the waiting room will be cleverly used by the medical group to answer all the relevant questions a doctor will ask in the exam room.
I expect this will look like touch screens with pictures and short videos, drop-down menus and checkboxes which will initiate a specific order, such as UA, blood test, finger-stick test, throat swab etc.
Orders will be placed, medications sent to the pharmacy and videos will be available for patients to view based on the diagnosis the doctor chooses after stepping into the room and confirming the AI management.
I think doctors will spend less time explaining to a patient why their URI doesn’t need a medical intervention. The patient will be explained through a short video how to manage their lateral epicondylitis and how to measure their own improvement.
They will be able to watch a video on how to manage their own iron deficiency anemia, clicking through an interactive video based on which OTC medication they are interested in taking.
I am not so positive-minded to think that all this is beneficial to doctors. That situation is far more complicated. As we are vying for higher pay, we are selling ourselves out a bit by rushing through patient care, surgeries and hospital rounding.
We are getting distracted more and more by various emails, texts, inbasket messages, pages and phone apps, all of which are being used to funnel more patient-care towards you.
Over the next few years you will see more and more of your knowledge being accessed by your medical group, software companies, surveys etc. Your expertise and time will be used, without reimbursement, to create larger systems such as the AI models discussed above.
Now, I’m not trying to sound like a rebel here. But I look at my Kaiser Permanente phone and I have no less than 5 apps which I am constantly watching directly related to my work.
Sure, I don’t have a physical pager anymore, I’m technically not on call, but I feel that more of my attention is taken up than when I did have to wear a pager.
On the profiting side, if you are someone who gets this AI concept and thinks it’s a likely direction for medicine, then I encourage you to invest in such businesses. It’s easier than ever through websites such as Angel.co to invest in startups.
You can use companies such as HealthTap to drive more traffic to your established medical practice. They build the entire backend for you, all you need to do is see the virtual patient. They even have a clever “Dr. AI” who will gather all the HPI for you – technology in action.