Telehealth app development is mostly everything except the call.

A visit is a record, not a meeting link

Write the states down before anything else. A visit is requested, triaged, scheduled, consented, checked in, connected, documented, closed and billed. Some of those states fail: the patient never joins, the clinician loses signal, the visit turns into an in person referral halfway through. Each failure needs its own ending. A visit left sitting at "connected" is a row nobody can answer questions about later.

The second thing to decide is which record each screen writes to. Intake answers belong to the visit, not to the patient profile, because the answer to how long this has been going on is true on Tuesday and stale in March. Allergies, medications and conditions belong to the patient. Mixing the two is the most common data model mistake here, and it only gets expensive once you have history.

Identity is third. Before the call connects, something has to confirm that the person in the waiting room is the person on the booking, and something has to record where that person physically is. Location matters because clinician licensure and prescribing rules generally follow the patient's location, not the clinician's. Rules differ by state and by what is being prescribed, so confirm what applies to your specialty. Store the location on the visit and the question is still answerable two years later.

The video call is the one part you do not write

Real time video between two people is WebRTC, and neither React Native nor the Expo SDK ships it. The video and camera packages in an Expo project record files and play them back. They do not carry a live two way stream to another participant. To get a call you add a library, usually react-native-webrtc or a vendor SDK that wraps it.

That library is native code, and the consequence is worth knowing before you plan the sprint. Expo's own config plugin for it states that the package cannot be used in the Expo Go app because it requires custom native code, so adding it means rebuilding the app rather than reloading it. On Android the same plugin raises the minimum deployment target from 21 to 24, which can collide with other packages. Plan for a native build step from day one.

This is true of generated projects too, and it is worth saying plainly: a Newly build does not give you a real time video call on its own. What it writes is a React Native and Expo project, so the call still comes from a third party SDK you add, and you pay that vendor by the minute. Signalling, TURN servers, reconnection and call quality are a subject of their own, so we have covered the video call itself separately.

Expo config plugin for react-native-webrtc

What the security rule turns into on screen

The HIPAA Security Rule's technical safeguards are shorter than people expect, and they read like a product spec. Access control requires a unique name or number for identifying and tracking each user, and an emergency access procedure for reaching patient data when normal access fails. Both are marked Required. Automatic logoff after a predetermined time of inactivity, and encryption, are marked Addressable. Addressable does not mean optional. It means you implement it, or you document why it is not reasonable and do something equivalent.

Audit controls are a standard in their own right: mechanisms that record and examine activity in systems holding electronic protected health information. In an app that means an append only log of who opened which record and when, kept apart from the record itself. Person or entity authentication requires procedures to verify that whoever is asking for access is who they claim to be. Transmission security covers integrity controls and encryption of data in flight.

Turned into build decisions, that list is concrete. No shared clinic login on the front desk iPad, because unique user identification rules it out. A session timeout on the clinician's phone. An audit row written on every read, not only on writes. A break glass path that logs loudly. And a business associate agreement with every vendor that handles patient data on your behalf, the video vendor included, signed before the first real visit.

Consent is the other paperwork the app has to carry: what the patient agreed to, which version of the wording they were shown, and when. That is the same problem a clinical trial app has to solve, and the same answer works here. Store the consent version and timestamp, not a boolean.

45 CFR 164.312, technical safeguards

Scheduling, waiting and the part after the call

A waiting room is a queue with a clock, and it is where patients decide whether your app is any good. It has to say who is ahead of them, roughly how long that is, and what to do if they lose signal. A screen that only says please wait produces more support calls than the visits do.

No shows and reschedules are the boring half of the build and the half that pays for itself. An evening reminder, a one tap join link and a written rule for what happens after ten minutes of nobody joining will move your completed visit rate further than anything on the call screen.

After the call there is a note, sometimes an order, sometimes a prescription, and often somebody else who has to know. Family carers and home care staff are frequently the people who act on what was decided, and giving them a scoped view is a separate product with its own permissions, closer to a mobile caregiver app than to the clinician's screen.

The one number worth working out before you commit to a vendor is minutes. Video is priced per minute, a visit has at least two participants, and some vendors count each participant separately.