Synthetic evidence only

This demonstration explains a modelling workflow with generated observations. It is not a medical device, does not represent any person, and must not guide care.

Loading the synthetic score, timeline, and evidence context.

Synthetic risk review

Inspect the generated score, compare it with an illustrative review threshold, and see how threshold choice changes the number of flagged windows.

Synthetic six-hour probability

-- Loading

Generated from deterministic observations for interface and evaluation testing only.

Illustrative review threshold 60%

Waiting for timeline

For demonstration only. This is not a recommended clinical threshold.

Prediction horizonNext 6 hours

Runtime modeSynthetic demo

Last refreshNot loaded

Heart rate112 bpmAbove reference

Mean arterial pressure64 mmHgBelow reference

Lactate3.4 mmol/LAbove reference

Synthetic risk timeline

Six generated observation windows show non-linear change. Move the threshold above to compare review burden.

Probability Threshold

Largest synthetic contributors

Directional terms describe the generated score only. They are not causal explanations.

Loading contributor details.

Evidence boundary

What this public surface can and cannot establish.

DataGenerated only

PurposeResearch workflow

Clinical validityNot established

Care useNot supported

How to inspect it

Three checks keep the score from becoming a context-free headline.

Start with probabilityRead it as synthetic output, never as a patient estimate.
Move the thresholdSee review count change across the same six windows.
Check the limitsCalibration on generated data does not establish clinical performance.

Technical method

The public demo uses a deterministic synthetic feature vector and a fixed replay timeline. The repository evaluates a named baseline and calibrated challenger on a separate deterministic synthetic cohort. Restricted research artifacts stay local.