Roadmap
From verified surveillance to decision simulation
Global Malaria Watch is built in four sequential phases. A phase is only entered once the layer beneath it is sourced, documented and reproducible — which is why the current release deliberately ships surveillance alone.
- Phase 1 of 4
- Research Preview v0.1
- Research Preview
- FSI-AIBR
Historical surveillance intelligence
✓ Current release
Climate and environmental intelligence
Planned
Predictive risk intelligence
Planned
Digital twin and decision support
Planned
Delivery phases
- Current release
Historical surveillance intelligence
Establish an auditable national record that every later layer can be validated against.
Gate — Complete when every displayed figure is reproducible from the published snapshot.
- WHO GHO burden, intervention, diagnostic, treatment and financing series, 2000–2024
- World Bank population denominators with correctly re-derived incidence and mortality
- Executive, programme and clinical analytical lenses over one dataset
- Year scrubbing, country intelligence briefings and CSV export on every table
- Published methodology, indicator register and source provenance
- Planned
Climate and environmental intelligence
Introduce the environmental covariates that condition transmission, with the alignment method published before any covariate is displayed.
Gate — Gated on a documented spatial and temporal alignment protocol.
- Rainfall and land-surface temperature series aligned to national boundaries
- Vegetation and surface-water indices as proxies for vector habitat
- Lag structure analysis between environmental signal and reported burden
- Uncertainty intervals rendered on burden estimates alongside covariates
- Scheduled refresh against source APIs with snapshot versioning
- Planned
Predictive risk intelligence
Short-horizon national risk forecasting under an explicit validation standard, with explainability as a requirement rather than an addition.
Gate — No forecast is displayed without its validation record and horizon.
- Out-of-sample validated forecasting models with published error metrics
- Prediction intervals displayed on every forecast, without exception
- Per-prediction feature attribution for explainability
- Anomaly detection across national reporting streams
- Sub-national layers where national programmes share data under agreement
- Planned
Digital twin and decision support
Move from description and prediction to decision testing — intervention and financing scenarios evaluated before commitment.
Gate — Gated on calibration transparency and published sensitivity analysis.
- Calibrated national simulation environment for intervention scenarios
- Budget-constrained allocation and cost-effectiveness comparison
- Elimination pathway scenarios with sensitivity reporting
- Briefing-quality export for ministerial and donor decision cycles
- Architecture transfer to tuberculosis, HIV and neglected tropical diseases