Person-Centered Clinical Analytics — trajectories, not averages

synthetic data · educational scaffold

Cohort parameters

Generation runs entirely client-side. Same seed + parameters reproduce the same cohort. Loading a file replaces the in-memory cohort used by all three tabs.

Condition mix — 20 conditions, three families

Deselect a condition to exclude it from generation. Weights are equal among selected conditions; comorbidities are drawn from the same pool minus a patient's own primary condition.

Cohort summary

no cohort generated yet

Once generated, switch to Nested Viewer or Pivot Explorer — both read this same in-memory cohort.

View controls

Checkpoints: Diagnosis → Presentation → Initial plan → Escalation → Advanced/recurrent → Disposition. Node size = number of patients converging on that value. Dashed nodes = value not seen elsewhere in the loaded population. Patient A traces in teal/amber tones, patient B in blue/violet tones — use both to eyeball pairwise divergence.

Vertical 100%
Horizontal 100%
100%
hub node (convergent) outlier / not in population drag to pan · scroll to zoom · click a node for detail

Patient vs. population

Select a patient to overlay to see percentile placement.

Top trajectories

Full diagnosis→disposition chains, ranked by patient count.

Pivot & filters

100%
pivot center matched subgroup divergent path drag to pan · scroll to zoom · click an outcome for detail

Confounders

    Clinical output view

    Age × sex stratified breakdown

    Splits the matched subgroup by age band and sex instead of collapsing them to filters, to surface interaction effects a flat share can hide.

    Pinned pivots

    Pin up to 3 pivots (e.g. Diagnosis=T1D vs Diagnosis=T2D) to compare their disposition breakdowns side by side, synced on the same outcome rows.

    No pivots pinned yet.

    Cohort synthesis

    Age distribution

    Sex mix

    Condition prevalence — all 20 conditions

    Escalation-rate leaderboard

    Comorbidity co-occurrence

    Cell shade = observed co-occurring patients ÷ expected under chance (independence). Coral = pair clusters together more than chance; blue = less. Grey = too few observations to read much into.