The honest blocker on clinical AI was never accuracy — it was that Medicare had no way to pay for software. So watch what happened once one software-shaped service got a code: Medicare's remote-monitoring spend went from $6.75M to $256M in five years. Then ask why your smartwatch still can't get in.
Remote physiologic monitoring (CPT 99453–99458, 99091) got real Medicare codes at the turn of the decade. Here's total Medicare RPM spending, every year since. Press play and watch a “software-ish” service scale the moment the fee schedule made room for it.
$6.8M2019
Medicare paid for remote physiologic monitoring
38×
Growth in Medicare RPM $, 2019→2024
7.4M
RPM services billed in 2024
$256M
Medicare RPM payments, 2024
A code is a demand curve. Nothing about the underlying technology changed dramatically between 2019 and 2024 — connected blood-pressure cuffs and glucose monitors already existed. What changed is that CMS gave the service a billable slot. That is the entire argument for why this week's “Software as a Medical Service” proposal matters: the lane, not the sensor, is the unlock.
The catch
So why isn't your watch in this chart?
Because that soaring lane has a toll gate: RPM code 99454 requires an FDA-defined medical device transmitting daily — which a consumer Apple Watch generally is not. The result is a strange gap the AMA's new six-country survey just measured. Doctors see the data. They can't bill it.
Physicians who review wearable data
97%
…who say it has clinical value
large majority
…with it integrated into workflow
≤6%
Across all six advanced economies surveyed, no country reported wearable-data integration above 6% — and the gap tracked reimbursement, workflow feasibility and regulation far more than physician interest.
The finding that complicates the easy story. You'd think “just add a payment code” fixes this. The AMA data says it's necessary but not sufficient: France had reimbursement pathways and still posted the lowest integration rate — plus the highest liability concerns. The UK had high physician wearable use but no payment path and low feasibility. A billing code opens the lane; workflow, liability cover and trustworthy data plumbing are what actually get cars moving. The RPM curve above worked because it arrived with all four, not just a code.
Why it matters for builders
Put the two halves together and the newsletter's thesis sharpens: Medicare growing a door shaped like an algorithm is real leverage — but a door is not a road. RPM proves a software-shaped service can 38× once the code, the device standard, the workflow and the liability answer all land together. If you're building on the new SaMS lane, don't ship the model and wait for the code. Ship the evidence-and-workflow layer that lets a health system bill it, defend it, and fit it into a Tuesday — the parts that turned RPM from a 2019 rounding error into a quarter-billion-dollar line.