Step 4 — Pivot Lens Selector

cohort_graph.json → re-anchor the same graph around Disease / Intervention / Disease+Intervention combo (Section 5) · same file as Step 3, no re-ingestion
Scope note (carried from Section 0): this is an audit/analytics view surfacing evidence, variation and confounders across a synthetic cohort — it is not a decision-support tool and makes no treatment recommendation for any individual patient.
What this does: loads the same cohort_graph.json from Step 2/3 and re-indexes it around a chosen anchor instead of a chosen patient. Three lenses, same underlying graph, same data, no rebuild: Disease-centered (anchor = a Diagnosis node — "given patients diagnosed with X, what walked in and what did we do?"), Intervention-centered (anchor = an Intervention node — "when we deploy Y, who receives it and how do outcomes vary?"), and Disease + Intervention combo (anchor = the composite hyper-node — "when Diagnosis A patients receive Intervention B, why do outcomes diverge?"). Selecting a lens re-clusters the same patients around the new anchor; it never re-derives or special-cases T1D vs. T2D in code.
No file loaded yet — use the cohort_graph.json exported from Step 2 (same file works for Step 3 and Step 4).