How MedConnect connected electronic prescriptions to EGISZ and a doctor’s QES

“We do not make the diagnosis remotely; the prescription is signed by a doctor.”

For the patient, it looks simple enough: the consultation takes place in MedConnect, the doctor issues an electronic prescription, and the pharmacy accepts it as a legally valid document. No paper form and no separate trip to the front desk for a stamp.

For the doctor, this also has to remain a routine action: check the data, make a decision, sign the prescription with a QES. But behind that action sits an integration with EGISZ — the Unified State Health Information System. And it is this integration that defines most of the rules, from the document format to who is allowed to sign.

Where the hint ends and the medical decision begins

MedConnect has an AI component that helps draft recommendations. For electronic prescriptions, this became the most sensitive part.

By law, a primary diagnosis cannot be made remotely. That means AI cannot write as if it conducted an in-person appointment, confirmed the diagnosis, and prescribed treatment instead of the doctor. We had to keep the wording within telemedicine boundaries: recommendations, clarifications, cautious options, and no statements that turn a remote consultation into an initial diagnosis.

The most active debate was not about the API, but about the text. At the data level, everything can be mapped into fields and statuses, but a single sentence in a recommendation changes the legal meaning. The difference between “the patient has X” and “X is possible, an in-person assessment is required” is small for the interface. It is not small for the medical process.

So we separated the roles:

• AI helps prepare wording within the allowed boundaries;
• the doctor makes the medical decision;
• the electronic prescription is signed with the doctor’s QES;
• the signed prescription, not the AI text, carries legal validity for the pharmacy.

This way, the patient sees a clear recommendation, the doctor does not get an “auto-diagnosis” in the interface, and the system does not step outside the legal limits of telemedicine.

An electronic prescription as a legal document

In this case, a prescription is not just a database record saying “the drug was prescribed.” The pharmacy has to accept it. That means the document must be signed with the doctor’s qualified electronic signature.

Here, QES is not a decorative attribute. It binds the prescription to a specific doctor and gives the document legal force. After signing, the prescription can be passed into the flow where the pharmacy will treat it as a valid electronic document.

Because of that, the signature in the integration cannot be treated as a final step to “bolt on later.” It affects the entire flow:

• the doctor must understand exactly what they are signing;
• the prescription data must be ready before signing;
• after signing, the meaning of the document cannot be changed unnoticed;
• AI recommendations must not replace the medical order being signed.

In the interface, this looks like a careful confirmation step. In the system, it is the boundary between a draft, a suggestion, and a legally significant prescription.

Integrating with EGISZ without blurring responsibility

In this setup, EGISZ is the government system through which the electronic prescription passes. For MedConnect, that means not only generating data, but also fitting the process into the rules of medical document flow.

The main risk was not that “the integration won’t send the JSON.” The risk was that product logic could accidentally mix three different entities: the consultation, the AI recommendation, and the signed prescription. The user sees one flow, but legally these are different actions.

We kept these entities separate. AI does not issue the primary diagnosis. The doctor signs the prescription with their QES. EGISZ receives the electronic prescription as part of the medical exchange. The pharmacy works with a document that has legal force.

This kind of flow is less exciting than “AI did everything in a minute,” but it fits the real constraints of telemedicine. The next question in integrations like this is usually not about a nice button, but about which exact data the doctor sees at the moment of signing and what the system prevents from being changed after that.