Loading documentation…
Loading documentation…
PearMedica's AI is powered by a curated, peer-reviewed knowledge base specialised for African diseases. This page explains what's in the knowledge base, how it's built, and how it stays current.
| Category | Examples | Count |
|---|---|---|
| Vector-Borne | Malaria, Yellow Fever, Dengue, Chikungunya | 8+ |
| Gastrointestinal | Typhoid, Cholera, Amoebic Dysentery | 6+ |
| Respiratory | Tuberculosis, Pneumonia, COVID-19 | 5+ |
| Viral Haemorrhagic | Lassa Fever, Ebola, Marburg | 3+ |
| Bacterial | Meningitis, Tetanus, Brucellosis | 4+ |
| Parasitic | Schistosomiasis, Onchocerciasis, Trypanosomiasis | 4+ |
The knowledge base includes location-aware disease prevalence weighting for African regions. When a patient's location is provided, the API adjusts condition probabilities based on regional disease patterns.
Nigeria, Ghana, Sierra Leone, Liberia
Focus: Malaria, Lassa Fever, Yellow Fever
Kenya, Tanzania, Uganda, Ethiopia
Focus: Malaria, Cholera, TB
DRC, Cameroon, Congo
Focus: Ebola, Trypanosomiasis, Malaria
South Africa, Mozambique, Zimbabwe
Focus: HIV/TB co-infection, Schistosomiasis
Every piece of medical content goes through a rigorous 10-step process before reaching production. This process is modelled after peer-reviewed journal standards and clinical guideline development.
Identify target conditions based on African disease burden data, WHO prevalence reports, and pilot feedback.
Medical experts document diagnostic criteria, symptom profiles, risk factors, and regional prevalence patterns.
Content reviewed by independent clinicians for accuracy, completeness, and alignment with current guidelines.
Real-world case studies are built to test diagnostic reasoning and edge cases across demographics.
Chief Medical Officer validates all clinical content against WHO, FMOH, and regional guidelines.
Engineering validates data schema, symptom IDs, and integration with the assessment engine.
Automated test suite runs against all conditions to ensure no accuracy regressions.
Clinical team manually tests edge cases, atypical presentations, and demographic-specific patterns.
Symptom aliases and region-specific terminology are updated (e.g., "body dey pain" → generalised aches).
A/B tested on 50% of traffic for 7 days, then deployed to production after accuracy metrics are confirmed.
The knowledge base is built from authoritative, publicly available medical sources: