What this does: loads the same
cohort_graph.json used by Steps 2–5 and assembles three panels that Section 2
asked the output view to surface together, at a glance, rather than requiring someone to run Steps 3–5 separately to piece them together:
Evidence (what actually happened, per diagnosis group — presentations, interventions, complication rate),
Variation (how outcomes differ by facility tier and resource status, the two structural variables tagged throughout this corpus),
and Confounders (every approximation or data gap introduced in Steps 4–5, collected in one place instead of scattered across pages).
Nothing here is derived beyond what Steps 3–5 already computed — this view re-presents it, it does not re-analyze it.
No file loaded yet — use the cohort_graph.json exported from Step 2 (same file works for Steps 3–6).