Uncalibrated illustrative simulation. Pointwise Monte Carlo uncertainty, not clinical confidence intervals.
Choosing a preset or stratum loads its defaults. Expand comparisons after editing advanced settings to retain your edited base.
Loading the saved question
95% pointwise Monte Carlo intervals are conditional on the model. Empty evidence is not normal oxygenation; inspect the U state below.
Advanced scenario and comparison settings
Fields labeled “varied by comparison” are controlled by the experiment.
Paired transitions
Rows: comparator. Columns: variant. Select a nonempty cell to explain its lowest-ID synthetic patient. Suppressed singletons retain algorithm zero.
All condition estimates and paired contrasts
Deterministic rule explorer
Explain a synthetic encounter
Select a transition cell, or enter a stable patient ID.
Documented FiO2 source records
Measurement and availability times are minutes relative to admission. Source IDs link these records to the scoring table above.
Methods and export
Imports check versions, hashes and patient-row reconciliation. Output remains labeled as imported until a new run completes.
For larger runs:
resp-sofa-experiment run --request request.json --output
output-bundle
Immutable result request
Saved synthetic benchmark
The historical benchmark used 200 legacy replicates. It is not a clinical reference or an accuracy target.
Open legacy reproduction (Scenario / Sweep)
Rule contrasts are not additive attribution. No population-weighted total is inferred. Baseline deltas are score changes, not sepsis incidence.