Models, Prediction, and Quickly Changing

May 18, 2020

Models, no not the skinny kind walking the runway or even the plaid wearing types in an L.L. Bean catalog.

But the guesses that are given gravitas. “Our guess is that blah blah blah” sounds like, well, a guess. “Our model reveals that blah blah blah” sounds far better educated. And, “our predictive model” — sheesh, it must be true; give the scientist a Nobel prize.

But as anyone still alive knows, models are just guesses. Make an error in design, put the wrong data in, or hold your metaphorical finger on the “algorithm” so that it gives you the desired outcome, and sooner or later, the “guess” is revealed, Wizard of Oz style, to be less than all-seeing, perhaps even as a prediction by a panderer with a publicist.  

Are these the ramblings of an “anti-modelist”? No. 

The reality is that we all construct and use models. Often, these are mental models, concepts or representations about how the world works. These concepts can be grand, as in the concept of human rights, or specific, such as in the concept of hospital-centric healthcare. 

In fact, the more models we use to understand the world, the better. It’s like the difference between looking at a picture of someone only from the back, versus a series of shots from every angle, plus maybe even an MRI. 

But here’s the thing: Models are fine as long as we don’t confuse them for reality. Take, for example, a map, which is a type of model. It’s an error to confuse the map with the actual terrain. It’s an even bigger error to blame the terrain for not matching the map. 

Yet, we see modeling errors play out all the time in healthcare.

For example, many hospital CEOs bought whole hog into purest model of hospital-centric healthcare in which physicians would be “aligned” via direct or indirect employment. They ignored the input of other models. Billions were spent acquiring practices. And then, reality hit. These behemoth entities were almost universally unprofitable and were far more fragile due to their bloated overhead. Add a few months of coronavirus cancellation of both elective procedures and nearly all physician office visits, and, well, it’s either bankruptcy time or, how do they put it, oh, time to consider merging with a strategic partner. Sick hospital plus sick hospital equals Sears plus Kmart. 

Or, consider a far more everyday modeling mistake; the physician group leader who believes that hospital contracting, say for an exclusive contract, follows a certain process and that all negotiation takes place within its “traditional” bounds. It’s certainly true that hospital CEOs want physician group leaders to completely buy into that mental model, but you’re blind to possibility if you actually do so.

The takeaways for you involve understanding a few problems to avoid:

You need to understand that multiple models, not simply blind adherence to one, are useful to inform your analysis, decisions, and actions. 

You need to develop the ability to cycle faster through the analysis of your model as it hits the windshield of reality, and then to cycle faster through the process of deciding on and taking your next action, that is, a faster cycling through the OODA loop. 

Note that I said that those were needs. They may not be your wants. But not wanting to do something that needs to be done is simply an example of a model that has no utility in this real world domain.

P.S. We’re designing a small group program for medical group leaders like you who want to understand the secret sauce underlying opportunistic action. If you’d like to be on the invitation list, send a message to one of my assistants.

Act fast!

Comment or contact me if you’d like to discuss this post.

Mark F. Weiss

Leave a Reply