Built in Rwanda, for health systems across Africa
MedLink attaches a coded, countable reason to every clinical encounter — inside a country's own systems, on its own hardware. It requires no new form, no new hardware, and no change to how anyone works.
Records that follow the patient. AI that helps a clinician decide what to prescribe. Follow-up after a patient goes home. Health systems across Africa are taking all of them on — and each will be asked the same question: did patients get better?
Today that answer depends on asking: sampling patients, waiting for replies, trusting what people remember or choose to say. The record already holds a better one. The next visit shows whether a treatment worked, and the clinician usually knows why it did not — the medicine was not taken, was not what it should have been, or no longer works. That reason is lost, because the record has nowhere to put it.
MedLink gives it a place. One coded reason at the return visit, counted like any other field — so every programme, every tool and every prescription can be measured by what happened to the patient.
How many were treated and came back. How many referrals were raised and never completed. Where a diagnosis did not hold up at the next visit.
Aggregated by facility and by district, that is a map of where care is breaking down — built from encounters already being recorded, with no new inspection and no new form.
MedLink shows where outcomes are worse. It tells you where to look.
Illustrative — the shape of the data, not real figures.
A total. Encounters counted, and nothing inside the count. Every mark is real and already recorded — but they are indistinguishable from one another, so nothing about the quality of care can be read off them.
The same marks, now separable. Treated and recovered. Returned. Referral raised and never completed. Diagnosis did not hold at the next visit. No new data was collected to get here — the reason was attached to records that already existed.
Named by function, so the argument holds in any health system that keeps records. Rwanda's institutions are given as the worked example.
One patient, several facilities. Every visit can be recorded and visible, and a referral can still be raised without anyone learning whether it was completed — or why not. Linked by reason, the same encounters become a path that can be followed.
A patient is not getting better. Today that is one fact, and it reads the same whichever of three things is behind it. Each calls for a different response, at a very different cost.
Treatment is not working.
This line does not change when you choose a different cause. That is the whole problem: all three arrive looking the same.
Doses were missed. The medicine would have worked.
The responseFollow-up and a refill the patient can reach. The cheapest of the three, and the one most often mistaken for something worse.
Who actsThe facility, and the patient's own clinician.
The medicine underperformed. The question is about a batch and a supply chain, not about the patient.
The responseTrace the batch. Check storage and distribution against other facilities that received it.
Who actsThe ministry and the medicines regulator.
The organism has changed. It will not answer this regimen, and it can be passed on.
The responseGenotyping, a change of regimen, and surveillance around the cluster. The most expensive of the three, and the one that must not be missed.
Who actsThe national laboratory, and the researchers who do this work.
Confusing them is expensive in both directions. Treat resistance as an adherence problem and it keeps spreading while everyone waits. Treat adherence as resistance and a country pays for second-line regimens it never needed.
Resistance surveillance today means taking samples: a subset of patients, at selected sites, periodically. It is careful, expensive work, and between rounds the national picture is an inference drawn from a fraction of the country.
A reason attached to every encounter is a different kind of input. It covers the whole country, continuously, because it is assembled from encounters that are already being recorded — arriving as a by-product of care rather than as a commissioned study.
The laboratory capacity does not grow. It stops being spread evenly, and starts being pointed.
A subset of sites, revisited periodically. Everything else is inferred.
The whole country, continuously — and a small number of places worth sending a test.
MedLink is not a test, and it does not replace one. No genotype is read from a record. What it does is read patterns across records that already exist, and say where a real test is worth running. The sample is still taken by the people who take samples, and read by the people who read them — they are simply told where to look.
Year one explains records that are already complete — text only, on the analytics tier, with no regulated device in scope.
eBuzima already exists. The national digital health layer is unusually mature, and there is a real counterpart to build against in the National Health Intelligence Centre. A method that attaches meaning to existing records needs existing records worth attaching to — and Rwanda has them.
The gap is not Rwanda's alone. Every health system that counts encounters has the same blind spot, and the same records sitting underneath it. Because the method rides on what is already being written down — no new form, no new hardware, no change to clinical practice — what has to be rebuilt for another country is the vocabulary, not the system.
The reasoning runs on-premise, inside the country that owns the records. Personal data never leaves. The trained adaptation, the vocabulary and every reason produced remain the country's own — which is the condition on which a ministry can adopt this at all, and the reason it cannot quietly become someone else's dataset.
Elvis Mutangana
Co-founder
Miriam Meles
Co-founder
Ezio Munyengango
Co-founder
Heritier Bagumire
CTO
Munyaneza Peace
CTO
Ineza Rwigema Gabin
Head of Growth
If you hold national or facility-level records and want to see what a reason field would show you, we will walk you through it against your own data structures — not a generic demonstration.
We are building this in Rwanda first, for health systems across Africa. If your work touches health outcomes, data sovereignty or digital public infrastructure, we would like to talk.
The brief was prepared for the National Health Intelligence Centre and covers what changes, and for whom, in a single page.