Causal Simulator · Clinical Edition 1.0 · in silico counterfactual engine
The engine contrasts Ydo(protocol)(u) against
Ydo(reference)(u) on the same exogenous draw
u per patient — abduction → action → prediction, not two
independent samples.
mean 12-month ΔCAL — negative = attachment gain
per-patient τi = ΔCALprotocol − ΔCALcomparator
edges into an intervened node are severed by the do-operator
| Patient | Phenotype | Defect | Sys load | HbA1c | SPT | CAL V1 | CAL comparator | CAL protocol | τi | Prognosis |
|---|
Tick every option worth considering. The engine evaluates — combinations on this patient's own exogenous draw and ranks them.
Each ticked adjunct is evaluated both on and off, doubling the matrix.
current charted status · no intervention applied
which mechanism is actually driving the attachment loss
Max CAL over 24 months · click any row in the matrix below to plot it
| # | Product | Sequence | Closure | SPT | Adjuncts | 12-mo CAL | ΔCAL | vs SoC | Fill |
|---|
contribution of each decision, in millimetres of attachment
Bars left of zero preserve attachment; bars right of zero lose it. This is the structural decomposition of the selected row — the same arithmetic the engine used, not a post-hoc attribution.