clinicians.build · July 9, 2026

The Payable Sliver

This week CMS proposed the first Medicare payment lane for “Software as a Medical Service.” Before you build for the door that's opening, look at the one that's already open: the software-shaped services Medicare pays for today — remote monitoring, chronic-care management, e-visits. Every dot below is one real billing code.

Primary source: CMS proposes interim payment for Software as a Medical Service (CY2027 OPPS)
Data: CMS Medicare Physician & Other Practitioners, by Geography & Service (2020–2024) · via MIMI Labs

The setup
34
Software-adjacent HCPCS codes
$934M
Total Medicare paid, 2024
$0
Paid for a raw consumer wearable feed

The SaMS debate sounds abstract until you notice Medicare has spent a decade quietly building code-shaped lanes for non-face-to-face care: remote physiologic monitoring, chronic-care management, behavioral-health integration, online E/M. Each dot below is one of those codes in 2024. Its reach (patients) runs left–right on a log scale; what Medicare pays per service runs bottom–top; the size is total dollars. Color is the code family.

The explorer

What a “software service” actually pays — and reaches

Minimum beneficiaries reached: 034 of 34 codes shown
↑ Drag right to hide the niche codes almost no one bills. Watch the map collapse to a handful of real lanes — and notice how few of them exist as software rather than clinician minutes.
34
Codes shown
—
Median Medicare pay / service
—
Total Medicare paid (shown)
The critical lens

These aren't software payments. They're clinician-minute payments.

Read the code descriptions and the illusion breaks. The dominant lane, 99490 (chronic-care management), pays ~$60 for 20 minutes of clinical staff time. RPM's 99457 pays for 20 minutes of management using the monitor's data. The device-supply code 99454 is the closest thing to “paying for the software” — and it explicitly requires an FDA-defined device with daily transmissions, which a consumer Apple Watch is not.

The gap the SaMS rule is trying to close. In every lane above, Medicare is paying a licensed human to interpret an algorithm's output, not paying the algorithm to do clinical work. That's why “we have an AI that reads X” has had no home on the fee schedule: there's no code for the read itself. The proposed “Software as a Medical Service” status indicator (O1) is CMS's first attempt to price the algorithm, not just the minutes around it. This scatter is the before picture.

Toggle the year buttons: RPM (red) barely existed in 2020, then 38×'d by 2024 — proof a software-shaped lane can scale once it has a code. The question SaMS asks is whether the next lane can skip the “requires 20 clinician minutes” step.

Why it matters for builders
The newsletter's thesis — “Medicare just grew a door shaped like an algorithm” — is easy to over-read. This map is the reality check: today, the only software that bills Medicare does it by renting a clinician's attention. If the SaMS indicator survives comment, the winning build isn't a slicker dashboard bolted onto 99457. It's the evidence layer that proves your algorithm did clinical work on its own — the thing that lets a read become a billable service instead of a feature waiting for a human to sign it.
Read the SaMS payment-lane analysis → Or read today's newsletter — “Did Medicare just build the first payment lane for clinical AI?” →