Oystar
Close the loop.
A clinician sends a complete case to a specialist. The specialist sends the clinical answer back. And the patients who never arrive get flagged — instead of lost.
Early access · real accounts, seeded data, no real patient records
Breathlessness on exertion for 3 months, ankle swelling. BP 158/96. Murmur at apex. ECG shows left ventricular hypertrophy.
- Sent
- Loop closed
Hypertensive heart disease with moderate mitral regurgitation. Continue amlodipine at the district hospital, recheck BP in 4 weeks, repeat echo in 12 months.
Never arrived · REF-0258
Urgent cardiology. Booked, patient did not attend. Transport from Kabacuzi is the likely barrier.
The problem
A referral is a question. Most systems only send it.
Routing a referral is the easy half. The clinical work is everything that has to come back — the specialist’s answer reaching the clinician who asked, and somebody noticing when the patient never made the journey.
The question never arrives whole
A transfer form goes up to Kigali without the ECG, without the last set of bloods, without the medication the patient actually stopped taking. The specialist spends the appointment reconstructing what the health centre already knew.
The answer never comes back
The patient is seen. A decision is made. The letter travels home in the patient's hand, or not at all — and the clinician who asked the question never reliably learns what the answer was.
Nobody notices the patient didn't go
This is the one that harms people. A referral is sent, an appointment is offered, and the patient cannot make the journey. Because no signal travels back, the referring clinician has no way of knowing. The case simply stops.
The loop
Six states. A referral is not finished until it reaches the last one.
Every referral on Oystar is a record of what has actually happened to it. Status isn’t typed in by anyone — it is derived from the events, which means it cannot quietly disagree with reality.
- 01
Sent
A frontline clinician sends the case complete — history, observations, results, images. Structured, not a letter. The specialist has nothing to chase before they can think about it.
Referring clinician
- 02
Accepted
A specialist at the receiving facility picks it up and either accepts, asks for what's missing, or redirects. If nothing comes back inside the target, the referral is flagged to the sender rather than sitting in a queue.
Specialist
- 03
Appointment set
The specialist proposes a time; the referring facility confirms it and tells the patient at their next contact. Where the specialist offers remote review, nobody travels at all.
Both facilities
- 04
Patient attended
Attendance is confirmed against the booking. This is where referrals normally vanish: nobody at the sending end ever learns the patient didn't go. Here, silence becomes a flag.
Receiving facility
- 05
Answer returned
The specialist returns findings, what was done, and what the referring clinician should do next. Not a discharge letter that arrives in three weeks — a structured answer to the question that was asked.
Specialist
- 06
Loop closed
The referring clinician confirms the plan is in place. Only then is the loop closed. Everything before this point is still open, and still visible to everyone accountable for it.
Referring clinician
The patient
The patient needs nothing.
Every digital health product eventually asks the patient to do something — install this, confirm that, tap here to keep your appointment. The people most likely to fall out of a referral pathway are exactly the people least able to do any of it.
So Oystar asks them for nothing. The clinician drives the referral, and the loop closes between the two facilities already accountable for it.
No app to download
Nothing to install, nothing to keep updated, nothing that stops working when the phone is replaced.
No account to create
No password, no email verification, no portal that asks for a reference number nobody has kept.
No message to answer
The loop is closed between clinicians. A patient who never replies to anything is tracked exactly as well as one who replies to everything.
Where a facility does choose to contact a patient — after a missed urgent appointment, for instance — Oystar surfaces who to call and why. It does not make the call, and it does not send the patient anything on the facility’s behalf.
Capacity
Routing a referral is worthless if there is no one to receive it.
Rwanda has roughly one doctor per 8,500 people, far fewer specialists, and most of them in Kigali. So the question isn’t really how referrals move. It’s whether a scarce national resource goes further.
Complete cases
A specialist who receives the full picture spends the appointment deciding rather than reconstructing. Fewer wasted slots, fewer referrals that turn out to have been the wrong one.
Remote review
One cardiologist in Kigali can clear cases from several district hospitals without anyone travelling. Scarce capacity serving many facilities instead of one waiting room.
Nothing lost in transit
Capacity that gets used should reach the patient. Flagging the ones who never arrive is how a booked slot stops being a wasted one.
What teams see
Completion rate, by district, without asking anyone to report it.
Because the loop is the record, the measurement is free. No audit cycle, no spreadsheet returned by a district hospital, no retrospective register review. The number is simply what happened.
| District | Sent | Completed | Median |
|---|---|---|---|
| Muhanga | 84 | 85% | 6d |
| Kamonyi | 67 | 82% | 8d |
| Ruhango | 58 | 76% | 11d |
| Nyanza | 49 | 78% | 9d |
| Kicukiro | 41 | 80% | 5d |
| Gasabo | 32 | 63% | 14d |
Illustrative figures from the seeded network. Real facilities, fictional patients and clinicians.
Security
Built for records that belong to someone else.
These are the design commitments the platform is built against.
The facility is the account
The facility is the primary entity, not the individual — referrals are facility-to-facility, and Law 058/2021 needs a coherent data controller. Registration creates a facility and its first administrator, who then invites staff. Licence verification against the Rwanda Medical and Dental Council register is designed in and not yet switched on; facilities are marked unverified until it is.
Consent captured, every read logged
Patient consent is recorded at referral against a named clinician. Access to a record produces an audit entry naming the person, the facility, the record and the time — append-only, and visible to the referring facility.
Access follows the referral
A specialist can see a case because it was sent to them, and their access ends when the loop closes. Permission derives from the pathway rather than from a role alone.
Built to plug in, not to sit beside
Designed to resolve patients by national ID through the RHIE and to read and write the shared record, rather than running a parallel system. The integration contract is NHIC's to define; this is built to meet it.
Walk a referral end to end.
Register your facility, send a case to a specialist, and watch the answer come back. Paste your notes or photograph the paper form — Oystar does the typing.