A few random thoughts on a Sunday as I leave the DOCSF meeting, get ready to spend next week on-site at Commons in LA, and prepare for HLTH at the end of the month.
We talk a lot about the “garbage in, garbage out” problem in healthcare. But maybe the issue isn’t just what we’re putting in or what we’re getting out — it’s how we process it.
Maybe we don’t just need better data, better technology, better payment models, or better incentives, maybe we need a better garbage can.
AI, Hallucinations, and the “Garbage In” Problem
We rightfully worry a lot about hallucinations when it comes to using AI in healthcare. But let’s not pretend that modern EMRs didn’t already create a massive “garbage in” problem.
Templates, point-and-click notes, and copy-paste habits have all encouraged and perpetuated documentation errors. Once they’re in the record, they’re almost impossible to get out.
Two wrongs don’t make a right — and AI still needs careful clinician oversight — but it might turn out that hallucinations are the trade-off for more accurate documentation.
The Last-Mile Problem of Healthcare Innovation
When you get really smart people working on really hard problems using technology, you get really cool solutions.
But even the best tech and the brightest teams inevitably bump up against one of healthcare’s biggest hurdles: the last-mile problem.
Tech solutions often sound great in theory and work beautifully in controlled environments. But real-world healthcare is messy — full of inertia, silos, resistance to change, and the friction of the human element. That’s where most promising ideas struggle.
Payment Models Leading Innovation by the Nose
Many efforts to “innovate” care delivery fail because payment models keep leading us by the nose.
Innovation simply follows where reimbursement points.
Under fee-for-service, that means chasing volume. Under value-based care, it often means chasing arbitrage. Allow Medicare Advantage risk arbitrage? Watch MA plans and startups explode. Mandate downside risk? Cue cherry-picking and lemon-dropping. Set arbitrary metric targets? Care shifts, but value remains elusive.
Instead of asking, “What does good care look like, and how do we pay for it?” we keep asking, “What can we get paid for, and how do we make that look like innovation?”
We should be building care models that demand new payment models — not the other way around.
The Common Thread
True innovation starts with understanding the real problems and building solutions that create their own incentives.
Too much of what we do simply retrofits new tools onto broken processes.
We keep putting garbage into the same trash can — and wondering why we get garbage out.
Ben Schwartz, MD, MBA
Editor-in-Chief, The Surgeon’s Record
SVP, Care Services & Strategy
ben@commonsclinic.com
I was discussing this just the other day with a senior colleague. They were struggling wrapping their heads around how to position exactly this conundrum regarding priorities for technology investment..
I love a good analogy and I tired to help them out with understanding the power dynamics you describe in the discussion above.
Clinicians and care providers are basically just the “tools of production” of the healthcare delivery system. Patients and others needing care are simply “raw materials” for the system and have minimal power or agency in the relationship. Then the ‘tools of production’ machines receive signals from the true control systems…the payers. Whether that is the private insurance conglomerates or the federally funded agencies and their politicos with their misaligned incentives, ultimately the ‘tools of production’ are going to follow the signals they receive. That applies to models of care behaviors and choice in investment aligned with the signals (i.e. money)
It’s just as sadly true in other healthcare model systems around the world. BTW. I’ll be at HLTH, would be fantastic to stop and chat if you ever have time.