I enjoyed this article alot, and it's very germane to current hot topics at CMS (and AMA). Here's an AI picture, an AI summary, and a link to Gottlieb's open-access article.
###
https://jamanetwork.com/journals/jama-health-forum/fullarticle/2842893
Scott Gottlieb, JAMA Health Forum (2025):
How AI Will Help Solve Medicine’s Productivity Challenges
His core argument:
Health care appears “unproductive” not because medicine hasn’t improved, but because productivity is measured incorrectly—by counts of visits and procedures, not by health outcomes. This creates a classic case of Baumol’s cost disease, where human-intensive professions (medicine, education, performing arts) improve quality but fail to show output-per-hour gains, suppressing compensation growth and driving costs upward.
What AI actually changes (beyond hype):
Gottlieb’s key claim is that AI must move from augmentation to substitution. Merely making physicians “smarter” does not solve the productivity problem. True productivity gains require AI systems that safely replace discrete physician tasks—diagnostic interpretation, triage, routine decision-making—so that one clinician can effectively care for more patients per unit time.
Why regulation is the bottleneck:
FDA’s framework assumes static devices, not learning systems. Gottlieb argues for a regulatory model that allows controlled algorithmic drift, with predefined performance thresholds and re-review triggers. Without this, developers avoid FDA entirely by labeling tools as “research use only,” sharply limiting real-world deployment.
Why reimbursement is the bigger bottleneck:
CMS does not pay software vendors directly and ties physician payment to labor inputs. Ironically, AI that reduces physician labor can lower RVUs, triggering cuts to physician income. Physicians must buy productivity-enhancing AI out of their own reimbursement, often with no financial upside.
Policy solutions proposed:
-
Bundle AI into episode payments
-
Expand OPPS pass-through mechanisms
-
Create new benefit categories (e.g., via the Health Tech Investment Act of 2025)
The unifying goal is to align AI adoption with physician compensation growth, finally breaking the cost-disease trap.
Bottom line:
If AI is allowed to replace physician effort (not just decorate it), medicine could finally follow the same productivity curve as other tech-enabled industries—improving outcomes, stabilizing costs, and restoring physician earnings growth.
