Skip to content
Healthcare

Telemedicine Platform Built for Low-Bandwidth Consultations

Telemedicine where the video stack was designed for the network the patients actually have, not the one in the specification.

Client

A telehealth provider serving patients across Pakistan (name withheld under NDA)

Internal project codename

CarePulse Telemedicine App — an internal delivery codename, not a commercial product name.

The problem

The existing product assumed broadband. Consultations dropped, and a dropped consultation in a clinical context is not an inconvenience — it is an incomplete encounter with a patient who may not reconnect.

Prescriptions were free text, which made them unusable for any downstream pharmacy integration and unsafe for interaction checking.

Constraints we had to design within

  • Median patient is on a mid-range Android device on a congested mobile network.
  • Clinical documentation must be retained and auditable.
  • App store review for the health category, with the extra documentation that entails.

Approach

Degrade the video, do not drop the consultation

The WebRTC stack was configured to degrade progressively — resolution first, then frame rate, then to audio-only — rather than failing. An audio consultation that completes is clinically far more valuable than a video consultation that ends halfway through.

Structured prescribing

Prescriptions moved from free text to structured medication records, which makes downstream pharmacy integration and interaction checking possible at all. This required clinical input on terminology and was slower than a text field; it is also the difference between a record and a note.

Remote vitals with provenance

Vitals captured from patient devices are stored with their source and capture method attached, so a clinician reading them later knows whether a reading came from a validated device or a patient's self-report. An unlabelled measurement is a liability.

Engineering notes

Designing for degraded connectivity as the normal case

Telemedicine built on the assumption of stable bandwidth fails exactly where it is most needed. Sessions degrade progressively — video drops to audio, audio drops to structured asynchronous exchange — rather than terminating on a connectivity dip.

The clinically important consequence is that a consultation interrupted mid-way does not lose the record of what was already discussed. Notes and orders persist independently of session state, so a reconnect resumes rather than restarts.

Clinical safety constraints on the interface

The interface treats identity confirmation and consent as blocking steps rather than as dismissible prompts, because a consultation conducted with the wrong record open is a patient safety event, not a data quality issue.

Prescribing and referral actions are deliberately harder to complete than everything else in the product. Optimising them for speed would be optimising the wrong variable.

Handling of recordings and sensitive media

The default is not to retain consultation media. Where retention is required, it is scoped by explicit consent, encrypted under a distinct key context from general application data, and access to it is individually auditable.

Data not held cannot be breached or mishandled by a future integration nobody has designed yet — the strongest control available, and the cheapest.

Outcome

Consultations complete on networks where the previous product failed. Prescriptions are structured and integrable.

Clinical documentation carries provenance, which is what makes remote-captured data usable in a clinical decision rather than merely present.

500K+
patients served
Graceful
video degradation to audio-only
Structured
e-prescribing with provenance

Figures above are drawn from delivery records held under NDA and are pending independent confirmation. Where a figure cannot be evidenced it will be removed rather than qualified.

Technology and standards

SwiftKotlinWebRTCNode.jsHIPAA Security RuleHL7 FHIR

Services applied on this engagement

Related case studies

Similar problem?

Bring us the architecture, not the brief

A free 45-minute call with the kind of engineer who worked on this. If the honest answer is that you should not build it, we will say so.

Schedule a free technical review
Book a technical callAll case studies