AI wants to know, how does your doctor get back to you with a reminder?
It sounds like a simple question, but it’s important. Digital advisors are going to generate a lot of recommendations. Those recommendations won’t fit into your once-per-year, 15 minute doctor’s appointment. In fact, it will drive doctors crazy if we jam this stuff in on their time. So, our AI advisors need a way to get back to you in a different time and space.
How are healthcare providers getting back to patients with recommendations? Do they have full case management? Just texts and emails? Or, do they skip it and wait until the patient shows up? Maybe we can make this process more streamlined, connected, smarter, outsourced. Let me know if you can suggest anyone that I can speak with about it.
A faulty assumption
Providers often make the assumption that they can only talk to patients that are in front of a doctor of clinician. Patients slip into a remote and invisible universe of “not my problem” as soon as they walk out the clinic door.
There is a software architecture that formalizes this idea, called CDS hooks. CDS is “clinical decision support”, advice from software to a doctor and patient. A CDS hook is an API that gets called by an EHR when it realizes that you are seeing a doctor. Behind the API are rules that scan a human’s health record and can remind you to get a test, look for problems from drug/drug interactions in your prescription list, etc. It returns “cards” that a clinician is supposed to pay attention to.
CDS hooks is a good architecture, but the idea that we have to “hook” reminders into a doctor’s visit starts to look like a problem when you think about simple reminders for diagnostics. I get comments from decision support experts like “You don’t want to remind someone to get a pap smear if they show up with a gunshot wound” or “you don’t want to remind them about a colonoscopy when they go to see a dermatologist.” Good points. But, back up. Why are they getting these reminders when they see a doctor? In theory, the reminders should come around whether they see a doctor or not.
Jamming all advice into doctor visits starts to look outdated when you consider that in the near future, computers will watch and measure everything that we do. These systems will generate alerts whether you go to the doctor or not.
It starts to look dangerous when you consider clinician overload. One expert in deploying CDS rules notes that the main obstacle is that “many sites don’t want 4000 different rules firing on every patient load. It just irritates docs.”
Setting up a conflict between managers and doctors
That 4000 rule list is going to grow, because there are good reasons for the managers of the healthcare system to want automated rules. Software doesn’t have the same communication talent, empathy, and creativity that doctors have. But, it is:
- always available
- frequently updated
- able to handle real time and complex data
- and aware of reimbursement policies.
That’s a safety net that healthcare managers want to wrap around the semi-random behavior of doctors.
We may be able to resolve the conflict between managers that want CDS, and doctors that need to stay focused during their time with a patient. One idea is to separate the rule/model/CDS distributions into a small number of “hot” rules that should come up during a conversation with a doctor, and a larger number of “cool” rules that look at patient records during some other time. Then we can figure out the workflow for approving relevant findings, and delivering them to the patient. Much of this followup communication and “case management” could be outsourced and automated.
Please let me know if you can introduce us to people that are figuring out these workflows.
Here is a diagram of the concepts that we worked through to arrive at this question.