For patients and caregivers
Fewer information-related delays, repeated explanations and administrative journeys. These are goals for a future partner evaluation.
Synaporia Health · Development stage
Help the next step
Help the next step
of care happen.
Synaporia Health explores how care teams could use AI outputs with traceable evidence, professional authority and visible unfinished actions. The first test is referral text and handoff.
Less lost context.
More accountable follow-through.
For receiving clinicians
Relevant information with its sources, missing details and contradictions visible for professional judgement.
For care coordinators
Less reconstruction of records, with a clear owner and next action when a handoff remains unresolved.
Kenya’s Ministry of Health identifies referral fragmentation as a policy concern. That establishes a relevant problem to investigate, not evidence that this prototype improves care. Read the Ministry’s March 2026 announcement.
The wider opportunity
Keep clinical AI accountable.
Where an institution already uses a qualified screening, specialist-reporting or maternal-alert tool, a future Synaporia integration could preserve its source or study version, model identity, permitted reviewer and next-action evidence. Each setting needs its own partner and clinical validation.
The first test
Start where evidence is possible.
The proposed six-week experiment identifies Mistral as a study-specific candidate for the model-neutral platform for source-linked text drafting and a simulated referral handoff. It does not read scans, diagnose disease or validate a partner model.
Illustration · fixed synthetic scenario
A handoff still needs a next step.
A clinician has requested a referral and reviewed the document. Explore what changes when a delivery receipt arrives. No model, patient record or live service is connected.
Clinical decision recorded
The clinician specifies the referral and destination.
Information reviewed
The professional checks source passages and approves the final document.
-
Delivery starts as unknown in this illustration. A receipt establishes document delivery only; later care milestones remain unverified.
Owner: referral coordinator
Delivery outcome unknown.
The coordinator checks with the receiving service before any repeat dispatch. The case stays open and assigned.
- Document delivery
- Unknown
- Receiver acceptance
- Not evidenced
- Appointment / attendance
- Not evidenced
- Clinical completion
- Not evidenced
Where AI may help
Bring evidence into view.
The proposed model organises narrative records and drafts a summary linked to source passages. Staff inspect the originals, resolve gaps and review the final document.
Search is restricted to authorised sources for the selected episode. Similarity does not prove factual support. Clinicians retain clinical judgement, the referral decision and release authority.
The value test
AI has to earn its place.
Compare assistance with structured templates and conventional extraction. Measure factual errors, missing information, total review effort and whether unresolved work stays visible.
If a simpler workflow works as well with less burden, prefer it. Transport, staffing and receiving-facility capacity need their own interventions.
What exists, and what comes next
Current components
Coordination contracts, policy checks, workflow components and simulated handoff adapters exist in the repository. A complete institutional workflow, live model qualification and clinical evaluation remain outstanding.
Next evidence
Work with clinicians and coordinators to test the problem, review synthetic cases and measure usefulness and burden. No clinic partner, deployed healthcare service or patient benefit is currently claimed.
This project does not provide diagnosis, triage, treatment recommendations or emergency care. Scope reviewed 19 October 2026.
Does this match a gap in your workflow?
Seeking conversations with clinicians, referral coordinators and institutional workflow owners. Use non-identifying examples; please do not send patient records.