Clinician Economics, Part 2

In Clinician Economics, Part 1 I briefly wrote about the idea of unit economics and tried to setup the stage for how that may be applied in a healthcare setting, starting with two distinct types of clinicians - the in-patient and out-patient physician. 

Now let’s see how those economics influence their decision making when it comes to at the adoption of new products and services. It’s not a secret that the healthcare industry is under enormous pressure, driven by aging population, rising costs and labor shortages. AI automation is promising to alleviate many of those problems, but what is the right entry point if you are an up and coming company (aka startup) and how do you get clinicians to actually use your tools?

Imagine startup Z - an AI app that helps clinicians save time in documentation and patient processing. The tech is working, certifications are in place, so the objective now is to deploy it in a real clinical environment.

First, the CEO approaches our in-patient physician, who’s a senior radiologist employed at a university hospital. He sees up to 100 patients per day and often needs to do 1-2 hours overtime, in order to document all the cases in the system. He’s surely going to benefit from having part of his workflow streamlined, right? Depends. Let’s see what a radiology workflow looks like:

1. Imaging request

2. Protocol selection

3. Scheduling

4. Patient preparation

5. Scan

6. Quality check

7. Results interpretation

8. Internal review (optional)

9. Results/Reporting

10. Billing

This is a highly structured, high speed workflow, with little space for creativity and many touch points with various departments and systems. Maybe Z-app can speed up one of the steps - granted it is integrated with the rest of the hospital IT systems - but does that really have a material impact (and here I don’t mean financial) or is it just moving the bottleneck to a different step? 

Then there is the question of having to learn a new interface and a new process - this might be okay, if it was just one app, but what if there were ten of them, one for each of the workflow steps? Is exchanging 10 old interfaces for 10 new ones actually a compelling argument?

Furthermore, even if we assume that all of this actually works as desired and patients suddenly start moving through the radiology department at 2x speed, this will simply create bigger queues in other parts of the system that are not that streamlined, causing bottlenecks.

So you might think - puh, radiology is tough - but it is still the most obvious first choice for AI automation. The situation is even more complex in other departments such as oncology, ER or surgery, where the interdisciplinarity and the orchestration is considerably higher. In hospitals physicians are like small wheels - hundreds of them - that all work together and move at a similar speeds in order to move the patient forward. Changing how one wheel works, in isolation, will cause for the whole system to break.

Next we’ll look at the out-patient setting.