I suspect that regardless of the theoretical medical benefits of these ML models or the actual underlying science, things will trend toward whichever medical decisions make hospitals the most money, until/if another big non-healthcare stakeholder like the Justice Dept or whatever figures out what's going on and intervenes.
Dig: The organizations that make more money by selling more saltwater and antibiotics (for example) will use their money to get bigger and bigger. They will in time make enough money to ditch their cheap little Meditech from 1985 and buy Epic, which will allow them to fire more of these BS sepsis triggers, thus (via subconscious repetition both electronic and verbal) warping the common knowledge of terms like "severe sepsis", and thus resulting in more and more saltwater and antibiotics getting sold.
The biggest ones, like HCA, will become big enough to, by subtle and ingenius means, effectively
create public policies that let them sell
more saltwater and antibiotics. Eg, One-Hour Sepsis, which is now a protocol recommended by the insurer of my current little rural hospital, which has nothing to do with HCA. (Although One-Hour Sepsis is not, amazingly to my knowledge, yet recommended by my CMG USACS! At least, not until the
equity to debt conversion cometh...)
The organizations that do not go this way, either because they only hire ethical and sensible docs or they have ethical and sensible policies, will not make as much money. They will also go against the grain of the big insurance and CMS and other bureaucracies, whose marching orders will be more and more created by... you guessed it... the big saltwater+antibiotics sellers.
Ultimately, those "sensible" organizations will die, so in retrospect they were not so sensible. And so the feedback will continue. Tail, dog, wag. Rich get richer, money goes to money.
I hope I am unrealistically cynical and naive about all this and someone will correct me that this is not how our medical system really operates these days.
Source: I'm pretty good at operating small-scale flywheels. I fed myself by writing and publishing scientific papers for 10 years before and during med school. In fact, by far the most important consequence of my publishing all these papers was that I was able to craft a clutch narrative that got me
into med school. All expenses paid, in fact!