FHIR-mediated, not FHIR-direct

There is a sentence in our investor brief that sits under the compliance posture table and does most of the regulatory work in the product:

Stoa is not a medical device. It is a practice management and self-regulation tool whose output happens to be FHIR-compatible.

This is not a marketing line. It is an architectural decision, and the whole regulatory strategy rests on it. If we explain it once, in public, it is harder for us to break later, which is the point.

The two ways to integrate with a state health system

A consumer health app that wants its data to live alongside the patient's medical record has two choices.

Direct integration. The app authenticates against the national system (Latvia's e-veselība, Estonia's e-Health, Germany's ePA), gets credentials, and writes records directly. The patient's clinical data appears in the state system because the app put it there.

Therapist-mediated integration. The app produces a structured export (FHIR R4 Bundle: Patient, Practitioner, Encounter, Observation, CarePlan, DocumentReference). The therapist downloads that bundle and imports it into their own certified EHR. The patient's clinical data appears in the state system because the therapist put it there, using a tool they were already certified to use.

The first path looks shorter. It is also a one-way door into being a medical device.

What changes when you write directly to a state system

Once an app writes clinical data to a national health system, it is part of that system. In most EU regulatory frames, including the EU Medical Device Regulation (MDR), an application that "stores or transmits information used to make a clinical decision" is potentially a medical device — and once it is making changes to a regulated medical record, that potential becomes a near-certainty.

Medical device classification is not bad in itself. Plenty of products are medical devices and do their work well. But becoming one is a 12-to-24-month process. ISO 13485 quality management system. Clinical evaluation. Risk classification (Class IIa minimum for anything touching diagnosis). Notified Body audit. Post-market surveillance. CE mark. Continuous re-certification for every meaningful change in the software.

You can do all of that. But you cannot do it in parallel with running a consumer beachhead and a therapist B2B wedge with a five-person team and a 24-month runway. It is a different company.

What changes when the therapist mediates

When the therapist is the one importing the data, the architecture flips. Stoa becomes a tool the therapist uses, not a system the patient is enrolled in. The therapist's EHR is the certified medical-record system. Their certification is the regulatory wrapper around the clinical information. We are below the line — practice management and self-regulation tooling, not a clinical decision system.

This is the architectural firewall. We export. The therapist imports. The patient's clinical record stays inside a system that was always designed to hold it. We never touch the state health system directly.

What this costs us

It costs us some friction. A therapist has to download a bundle and import it, instead of the patient's app being magically synced. Stoa cannot say "connected to your e-veselība profile" in marketing copy. Some therapists in some workflows will find it less convenient than a direct sync would be.

What we get in exchange is that Stoa stays the kind of company that can ship two consumer features per quarter without re-certifying its quality management system. We can launch in five EU countries in two years without becoming a Class IIa device in five jurisdictions. We can keep our 24-month runway pointed at building, not at regulatory affairs.

The trade is correct, and it is locked into the product. The export endpoint is implemented. The direct-integration endpoint does not exist and will not be built.

What this is not

This is not a story about avoiding regulation. The reverse — we built five granular GDPR Article 9 consent surfaces, an immutable audit log, 15-year retention, a privacy-first identity layer (anonymous Stoa IDs), and a disguise mode for users in unsafe environments. Stoa runs more compliance than a standard wellness app, not less.

What we are avoiding is the wrong category of regulation. Medical-device classification is for products that make clinical decisions. Stoa does not make clinical decisions. The therapist does. The architecture says so. The codebase says so. The investor brief says so. And now, this post says so.

The next time somebody asks "is Stoa a medical device?", the answer is — no, by design, and here is the architectural reason. Link them here.