As I anxiously await my copy of Deep Medicine from Dr Eric Topol
, there are several specific topics that I hope to learn more about in an Artificial Intelligence standpoint. While I am sure there will be plenty discussed in the realms of pharmacology and pharmacogenomics, and oncology, the two places I hope AI will advance medicine is in patient experience and the outpatient primary care space.
My initial thoughts revolve around what will add benefits to the medical experience for the patient and relieve some of the burden on physicians. There are definite parts of medicine that I hope AI will improve and some I hope AI stays away from.
What Artificial Intelligence should do:
Make it as easy as possible to make an appointment: Anyone can schedule a car service and dinner and airplane flights and hundreds of other tasks from their phone at any time. Why is making an appointment with a physician still a 19th century model, mostly calling a phone number, waiting on hold to speak to a person or just leaving a message? This can be easier, but I fear the rate limiting step is not the technology but the willingness of clinicians to change.
Find the right clinician and the right clinic at the right time: Secondary to the above issue, can we use machine learning to find a patient in need of care the right doctor? Can we create a system that can leverage decision making to determine which is more important: a timely appointment or a quicker appointment. While most times continuity is important there may be certain components of a patient’s complaint or a specific timing that lends itself to having the quicker appointment. Right now we leave that decision up to the patient and scheduler. Shouldn’t an algorithm at the direction of a clinician be a better alternative?
Make understanding a patient’s bill easier: Even the EOB’s and bills for my own families insurance are like a foreign language to me. In my own clinic I would often have patients bring their bills in hoping I could shed some light on their costs. Most frequently it was hospital or ER visits and numerous codes, charges, adjustments that needed to be interpreted. Wouldn’t an interactive AI guide help a patient through this morass of bills with infinite knowledge and patience?
Automate the clinical note: Anytime you see anything related to the causes of physician burnout, it is likely that the EMR is at the top of the list. It is not a stretch to say that the majority of physicians loathe, if not outright hate their EMR. EMRs are here to stay and the investments that hospitals and physician groups have made in wide spread use entrenches them even more. AI should be able to ease the ability of a physician to navigate a chart, create their note and review results. Can AI listen to my exam and have it ready to to review when I want to complete my note? Can I speak to my patient and tell them the lab tests I want to order and have AI recognize it and order it for me intuitively?
Automate paperwork: As a PCP, I feel like I could complete an FMLA form with my eyes closed. It is a fairly labor intensive task without much brain power needed. Could AI speed up this process for both the patient and the physician?
Triage symptoms to the right location of care: How do patients make the best decision on when to seek care, how urgently and to whom? They use their own intuition, their past experiences or they call their doctor. If it is after classic business hours, the most likely decision to be made is to go to the ER or UC. But this means that some people are using a higher level of care than needed and some are delaying an important diagnosis. AI could help drive patients to the appropriate type of care with the appropriate urgency.
What artificial intelligence should not do:
Diagnose in the PCP space: There are many AI services out there that will provide a diagnosis based on a set of symptoms. As every physician learns in their medical training, the patient’s history will reveal almost every diagnosis. Creating a list of potential diagnoses, a differential, is basically what physicians train for. We really don’t need anyone else deciding that. There isn’t a PCP out there that doesn’t have a built in ready to go brain-based algorithm for a “cough” or “stomach pain” or “fever.” We got this part. Help us with the mundane, burdensome tasks that limit our ability to talk with a patient.
Look for social determinants of health: We know it when we see it, we just can’t do anything about. No one needs an algorithm to tell alert them that their patient needs better food, a better living environment, a better job, more rest, a larger support network. All of that is obvious (in general). Just finding it isn’t enough, and an AI system that only finds issues with social determinants of health but can not help solve the issue is of no help. One Social Worker could do significantly more than any AI service.
These are my initial thoughts on the role of Artificial Intelligence in the outpatient medicine world. Would love to hear your thoughts and will have an update once I finish Dr Topol’s book.