Kairon

Care Management

Interactive demo

Day at a glance · Jordan Mercer, RN

Caseload, prioritized by the Care Graph

Ranked by graph-derived acuity and open signals extracted from the record, notes, and call transcripts — no lag, not sorted by hand. Each patient leads with why they surfaced today and the next best action.

#{{ c.rank }} {{ c.trendArrow }}{{ c.risk }} {{ c.acuity }}

Why now {{ c.why }}

Recommended next action

Stage

{{ caseCount }} of 42 patients shown · rest below graph-acuity threshold

Readmission risk, scored at the moment of discharge

Every discharge is scored from the patient's live care graph — ADT feed plus full claims history — in a single model call. No DRG, no billing detail from the index stay. Just what a care manager actually has on discharge day.

Model

Care Graph LLM

AUROC

0.71

LACE AUROC

0.58

30-Day Readmission Risk

597 discharges · 70 readmitted (11.7%)

Patient	Discharge	Risk	LACE	Outcome	Top signals
{{ p.name }} {{ p.mbi }}	{{ p.discharge }}	{{ p.riskPct }}	{{ p.lace }}	{{ p.outcome }}	{{ p.topSignals }}

{{ sel.mbi }}

Risk

{{ sel.confidence }} confidence

LACE

{{ sel.lace }} / 19

Outcome

LACE missed this one. Scored low-risk, flagged high by the graph.

Model reasoning

Risk factors from the graph

Against LACE

{{ pt.demo }} · {{ pt.mbi }}

Program

CHF Transitions · TCM

Risk

78%

Layers

AI-detected episodes click to focus the graph

>50% of signals resolved

≤50% resolved

{{ corr.lift }} {{ corr.liftNote }}

Risk

78%

high confidence

LACE

7 / 19

low risk

Outcome

✓

no 30-day readmission

Model reasoning

Risk factors from the graph

Against LACE

Telehealth session Phone · reached · 12:04 · outcome saved

Call transcriptauto-transcribed

TCM 14-Day Post-Discharge Assessment {{ eng.progLabel }}

"{{ f.source }}"

{{ otherLabel }} lives here in the full product.

This demo focuses on the Care Graph and AI Insights — try those tabs.

Program Effect — TOC High-Acuity Cohort

Retrospective cohort of 770 highest-acuity transition-of-care discharges (LACE ≥ 14, PY2025), drawn from 2,628 discharges and 10,906 graph-extracted signals. Outcome: 30-day all-cause readmission. Exposure: majority (>50%) of graph-extracted signals resolved during the episode. Estimates from multivariable logistic regression adjusted for LACE and RAF, landmarked at day 7 (immortal-time corrected), and analyzed within the reached population. This is an adjusted association, not a randomized effect — the landmark removes immortal time and the model adjusts for LACE and RAF, but patients whose signals resolved may still differ from those whose did not in ways neither captures. The balance table below is what is checkable about that.

Adjusted OR

0.49

0.27–0.87

ARR

9.6 pp

3.4–15.8

NNT

11

7–30

.016

Wald

30-day cumulative readmission incidence

Kaplan–Meier failure estimate, by signal-resolution status

Effect estimates

Measure	Estimate	95% CI
{{ e.measure }}	{{ e.est }}	{{ e.ci }}

Outcome by exposure stratum

Reference: ≤50% of signals resolved

Stratum	N	Readmit n (%)	Median d-to-event	Unadj OR	Adj OR [95% CI]	p
{{ r.stratum }}	{{ r.n }}	{{ r.readmit }}	{{ r.median }}	{{ r.unadj }}	{{ r.adj }}	{{ r.p }}

Adjusted odds ratio by program

>50% of signals resolved vs not · OR < 1 favors resolution

Signal → action → outcome (cohort)

Baseline characteristics by exposure

Standardized mean difference (SMD) < 0.1 indicates balanced covariates

Characteristic	Resolved >50% (n=385)	Resolved ≤50% (n=385)	SMD
{{ b.char }}	{{ b.closed }}	{{ b.open }}	{{ b.smd }}

Time-to-detect program signal: ~3 weeks with the Care Graph (signal/action nodes update in near-real-time) vs ~2 quarters for claims-only analysis (14-day claims lag plus quarterly reconciliation).