A Rural Mission Hospital on an Unreliable Connection

Mission hospital with outreach clinics · Rural Kenya

This is an anonymised, composite account drawn from deployments of this type. The facility is not named because we do not have permission to name it, and no performance percentages are quoted because we have not measured them. What is described here is the change in how the work is done, which is the part we can stand behind.

The constraint

The hospital had considered and rejected two previous systems for the same reason: both required a live internet connection to register a patient. At this site that is disqualifying. The connection drops several times a week, sometimes for hours, and it is worse at the outreach clinics — which run out of a vehicle, two hours away, on phones.

A system that stops when the line stops does not get used. Staff revert to paper immediately and correctly, because the queue does not care about the router, and the paper day is then never entered.

There was a second constraint, equally binding: there is no IT department. The system administrator is a clinical officer who does it in addition to their actual job. Anything requiring a server, a scheduled backup or a patching routine was not viable.

How it works here

Offline-first, not offline-tolerant. AfyaConnect installs as a progressive web app. Registration, consultation notes, vitals, dispensing and orders are written locally and queued. When the connection returns they synchronise. Staff do not do anything differently during an outage and largely stop noticing them.

Phones at the outreach clinics. The outreach team registers patients and records consultations on ordinary Android phones with no connection at all, in a village two hours from the main site. The data synchronises when the vehicle is back in coverage that evening. Before this, the outreach day was recorded on paper and, realistically, entered into nothing.

No server. Cloud-hosted, so there is nothing on site to back up, patch or replace. The clinical officer administering it manages users and little else.

What we got wrong at first

We introduced the offline behaviour on day one, and it confused people badly. A queued record that has not yet synchronised looks, to a new user, exactly like a record that failed to save. Staff started re-entering patients they thought had been lost, and we spent the first fortnight cleaning up duplicates that we had caused.

On later rural deployments we changed the approach: run online-only for the first two weeks so staff build a mental model of normal behaviour, then explain and enable the offline mode. Nothing about the software changed — the sequencing did.

What changed structurally

  • The outreach clinic is part of the medical record instead of a stack of paper in a vehicle.
  • A patient seen at outreach and later at the main hospital is one patient with one history.
  • The pharmacy knows what was dispensed on outreach, so stock reflects reality.
  • An internet outage is no longer a clinical event.

Questions this raises

Does AfyaConnect work without internet?
Yes. It installs as a progressive web app and continues working through an outage — registration, notes, vitals, dispensing and orders are stored locally and synchronised when the connection returns. This is designed in rather than added on, because in most of Kenya an unreliable line is the normal condition rather than an exception.
Can outreach clinics use it on phones with no connection at all?
Yes. The outreach team works entirely offline on ordinary Android phones and the data synchronises when the device is next in coverage. Records created offline join the same patient history as records created at the main site.
Do we need an IT department to run it?
No. It is cloud-hosted, so there is no server, no backup routine and no patching. Most small facilities have one person — often a clinical officer or an administrator — who manages user accounts, and that is the extent of the ongoing administration.
Should we turn on offline mode from day one?
We would advise against it, and this is a lesson we learned the hard way. Run online-only for the first fortnight so staff learn what normal looks like, then enable offline. A queued record looks identical to a failed save if you have no baseline, and staff will re-enter patients they think were lost.

What this facility runs

All case studies