Trajectory, not snapshot
A single value can miss the direction of travel. MedArchitex considers longitudinal patterns, while preserving the date, source, and context of each input.
Transparent by construction
The number has to be explainable.
The central question is not simply whether a condition is present today. It is whether a person may be moving toward preventable disease, over what horizon, and with which evidence.
Disease-specific models
Cardiovascular disease, heart failure, type 2 diabetes, chronic kidney disease, COPD, liver disease, and stroke require distinct models, inputs, horizons, and limitations.
Deterministic calculations
Validated equations are implemented in versioned code with reproducible test cases. Generative AI can explain a result, but it cannot produce or modify the number.
Context that changes action
Medical history, laboratory trends, medications, family history, behavior, social conditions, and environmental exposure can affect calibration or what intervention is realistic.
Visible uncertainty
A report shows its time horizon, data completeness, missing inputs, provenance, model version, and known limitations. Not available is better than false precision.
Ongoing governance
Model performance is monitored across relevant populations. Sensitive attributes require scientific, clinical, legal, and fairness review before any production use.
The MedArchitex standard