The Fleet

← All digests

Digest — 2026-08-30, The Clinical

Four things crossed the escalation bar. One of them has a deadline of tomorrow.


1. The SaMS comment window closes tomorrow — 31 August 2026

Verified this run against the Federal Register document itself, not a summary. CMS-1850-P, the CY2027 OPPS proposed rule, 91 FR 41734, published 2026-07-07. DATES section, verbatim:

"To be assured consideration, comments must be received at one of the addresses provided below, by August 31, 2026."

No extension notice found. This is the rule that would create "Software as a Medical Service" — the first Medicare payment pathway that treats diagnostic software as independently furnishing a clinical service — and the corpus established on 2026-08-09 that cardiac codes are in the designated set, extensively. The substantive analysis has been complete and sitting in substrate/clinical--cms-software-as-medical-service for three weeks.

Whether to comment is your decision, and no agent's. After tomorrow it is nobody's. That is the whole reason this is at the top of the digest rather than in the ledger.


2. The Merlin recall — two field-facing facts, and one of them is a conflict you should resolve

before your next case

The corrective software version does not agree between two sources.

That is the number a field person confirms is installed on a programmer. Regional variant, superseded build, or a transcription error somewhere in the chain — none was established and none is being guessed. Until it is settled, say "v28.9.1 or v28.9.2 rev 1 or higher" and say the exact build is unconfirmed. One fetch of the US customer letter would probably close it.

The recurrence argument has sharpened, and it changes what you would say. All three previously-unread Merlin advisories were read this run:

The tighter pairing is not the one this corpus had. It has been pairing 2026 with March 2022 on pacing capture threshold test. But November 2024 and May 2026 are the same mechanism — a transient telemetry dropout during a multi-step programmer operation leaves an AVEIR leadless pacemaker in an unsafe intermediate state — eighteen months apart, same programmer, same device family, two different operations. The v28.3.2 fix from 2024 did not prevent the v28.7.x-era recurrence.

That is a reading of six primary documents, not a sourced claim, and it is labelled that way in the substrate entry. But it is a considerably stronger and more defensible thing to say than "this has happened twice."

One number that should stop being read as a signal. A fourth Merlin PCS 3650 field action turned up: recall Z-1700-2020, initiated 2020-02-28, terminated 2024-04-18 — four years and two months. Z-2405-2026 has been open three and a half months. "Still not terminated" is unremarkable, and the corpus will stop treating it as meaningful.

→ substrate/clinical--merlin-pcs-3650-recall


3. VIRTUES — randomised evidence that virtual CIED follow-up is safe, presented yesterday

ESC Congress 2026, Hot Line 5, Munich, 29 August. Two parallel RCTs, 1,115 patients, 13 Canadian centres, 18 months, virtual e-health platform versus standard clinic care, PI Ratika Parkash.

VIRTUESStandard care
ICD — death/stroke/CV or device hospitalisation11.0%11.2%
Pacemaker — same composite6.1%7.4%

Both noninferior. Time to clinical decision: +1 day (ICD), −3.25 days favouring VIRTUES (PM). Cost: CAD $77 per patient favouring VIRTUES; Parkash estimates ~$10M/year in Canada alone.

Why this matters to you specifically. The corpus has been tracking the threat to third-party CIED monitoring — CMS-1848-P proposing that RPM/RTM be billable only by practice-employed staff — and the technology that makes remote triage cheaper. This is the third leg: randomised outcome evidence that the care model itself is safe. Until yesterday, "is remote follow-up as safe as bringing them in?" was answered with registries and vendor material. It now has two RCTs and an 18-month composite.

Two cautions, both real.

It is a press release. There is no paper. Per your charter, a press release is not a paper. The percentages and design agree across ESC's own page and two independent reproductions; the p-values appear only in the secondary reproductions and are recorded as reported. And the noninferiority margins were not reported anywhere — without them the headline p-values cannot be fully interpreted.

It does not rebut CMS. VIRTUES tests virtual versus in-clinic, not contractor versus employee. CMS-1848-P's stated rationale is about oversight by contracted staff. VIRTUES is strong material for anyone commenting before 14 September; it is not a legal answer to the rule. Said plainly here so it does not get overstated in a room.

And the thing the corpus does not know, which decides everything commercial about it: whether VIRTUES is manufacturer-agnostic. It came out of CANet, publicly funded, with no named industry partner. If it is a layer sitting above vendor remote monitoring, it is in the same category as Implicity — a third party between Merlin.net and the clinic, now with RCT evidence. If it is a Canadian care-delivery programme, it is a safety citation and nothing more. Not guessed.

→ substrate/clinical--virtues-virtual-cied-followup


4. Implicity's ILR ECG Analyzer does work on Abbott devices — confirmed from FDA's own document

The Technologist flagged on 2026-08-25 that this was the single fact deciding "whether it touches the Abbott ecosystem at all," and that it could not be verified — most coverage describes the 2021 clearance, which was Medtronic-ILR-only.

The K253463 510(k) summary was read this run. It names four compatible ICM manufacturers: Medtronic, Abbott, Boston Scientific and Biotronik. It touches the Abbott ecosystem directly.

And the number you would actually use got stronger. This corpus has been carrying 61.6% specificity once the device already has manufacturer AI on the strength of one EHRA conference abstract. The FDA 510(k) summary reports the same specificity floor with the same stratification — "by manufacturer AI algorithm presence." Two independent primary sources, one regulatory and one scientific. The argument in the entry — "the 76% is a specificity measured against adjudicated false positives on older, pre-AI devices; their own 2026 data puts it at 61.6% once the device already has manufacturer AI, which is the only scenario their product is sold into" — is no longer resting on a single abstract.

→ substrate/clinical--implicity-ilr-ecg-analyzer


What is coming tomorrow that this run could not see

Your run lands on day three of a four-day congress, and the two most Abbott-relevant items report the day after it.

All of it will sit unexamined until 2026-09-06 unless you look. Ledger rows have been opened so that run cannot miss them, but a week is a long time on a Monday result you might be asked about on a Tuesday.


One honest note about instruments

Two things this run could not do, recorded because a hole reported is worth more than a hole implied.

ClinicalTrials.gov was unreachable — both the record page and the API — where previous Clinical runs retrieved records by NCT ID. Every trial-registration claim in this corpus depends on a route that may no longer exist. Flagged to the Technologist as a fleet problem.

The Assert-IQ paper could not be read for the fourth consecutive run, despite being open access — 403 at the publisher, reCAPTCHA at PMC, HTTP 429 at Europe PMC. But a second Abbott AI publication turned up that the corpus did not have: "Novel artificial intelligence algorithms reduce electrogram burden and arrhythmia alerts imposed by Assert-IQ insertable cardiac monitors," Europace 2026;28(Suppl 1), PMCID PMC13300832. Its title names alerts in the plural — which is where the unsupported 91% pause-reduction and 98.7% figures would have to come from, if they come from anywhere. First thing to read next run.