Kairon Health / Impact Analytics Beta
data as-of 2026-06-30 engine v2.3.1

Impact Analytics

Causal inference over your own claims and clinical data. Build a study from modular patient flags, let the engine enforce statistical validity, and get an effect estimate you can defend in a board meeting.

{{ tab.label }}
Studies + New
{{ s.name }} {{ s.pop }} {{ s.statusLabel }}
Method library
Propensity score matching Inverse probability weighting Difference-in-differences Panel fixed effects Instrumental variables
{{ cfg.name }}
{{ cfg.question }}
Every term below is a versioned definition from the attribute library — click one to open it.
{{ sec.n }} {{ sec.title }} {{ sec.summary }}
{{ sec.statusLabel }} {{ sec.chevron }}
{{ f.label }}
{{ f.value }} {{ f.meta }}
{{ c.label }} {{ c.meta }}
{{ f.hint }}
{{ badCovLabel }} Try it — the engine will refuse the study.
{{ sec.cite }}
Validity checks {{ vHeadline }}
{{ ck.mark }}
{{ ck.title }} {{ ck.detail }}
{{ runNote }}
{{ runLabel }}
Executing pipeline
{{ st.mark }} {{ st.label }} {{ st.t }}
Results {{ res.runStamp }}
{{ m.label }} {{ m.value }} {{ m.sub }}
{{ rt.label }}
Standardized mean differences before and after matching. The conventional balance threshold is |SMD| < 0.10, marked by the vertical rule; 0.05 is the stricter target most terms clear here.
Covariate Pre Post |SMD|
{{ b.name }} {{ b.pre }} {{ b.post }}
before matching after matching vertical rule = 0.10 threshold
{{ bn.label }} {{ bn.value }} {{ bn.note }}
Propensity score overlap
0.0 estimated propensity of treatment 1.0
treated (above axis) comparison pool, pre-match (below)
{{ res.psNote }}
Balance thresholds per Austin (2011), Multivariate Behavioral Research 46(3); matching implementation follows Stuart (2010), Statistical Science 25(1).
Primary outcome {{ res.primary.outcome }} {{ res.primary.contrast }}
Effect {{ res.primary.effect }} 95% CI {{ res.primary.ci }}
Ratio {{ res.primary.ratio }} {{ res.primary.ratioCi }}
Inference {{ res.primary.p }} {{ res.primary.pNote }}
Effect estimates & confidence intervals
{{ fr.name }} {{ fr.tag }}
{{ fr.effect }} {{ fr.detail }}
{{ res.multiplicity }}
{{ pw.label }} {{ pw.value }} {{ pw.note }}
Matching removes imbalance on what was measured. These tests ask how much unmeasured confounding it would take to overturn the finding.
{{ sv.name }} {{ sv.value }} {{ sv.note }}
E-value per VanderWeele & Ding (2017), Annals of Internal Medicine 167(4); Γ sensitivity per Rosenbaum (2002), Observational Studies.
Every run writes an immutable record: the analysis plan as it stood before execution, the exact data snapshot, and the resolved definition of every attribute used.
{{ au.label }} {{ au.value }}
Pre-registration
{{ pr.when }} {{ pr.what }}
Estimand and analysis-plan discipline follow ICH E9(R1) addendum on estimands and sensitivity analysis in clinical trials.
Population bases
{{ p.name }} Unit of analysis — {{ p.unit }} {{ p.rows }} rows
The unit of analysis is declared once, at the base. Each base publishes a spine — UNIT_ID, BENE_MBI_ID, INDEX_DT — that every attribute definition binds to, and everything downstream — filters, matching, clustering of standard errors — inherits it. A per-member study can never silently become a per-claim study.
{{ pop.name }}
{{ pop.description }}
{{ ps.label }} {{ ps.value }}
Cohort attrition
{{ at.label }}
{{ at.n }} {{ at.delta }}
{{ pop.note }}
Attribute library {{ attrCountLabel }}
Attribute Type Category Coverage Status
{{ a.name }} {{ a.type }} {{ a.category }} {{ a.coverage }} {{ a.status }}
Attributes are the modular unit of the engine. Each one is a single versioned definition — a flag, a measure, a risk term — that any study can reference as treatment, outcome, covariate or filter. Definitions never hard-code a cohort: they bind to {population}, the study's spine (UNIT_ID, BENE_MBI_ID, INDEX_DT), so the same attribute runs unchanged on a discharge base, a member-year base, or a monthly panel. Change the definition once and every study that depends on it is marked for re-run.
{{ attr.name }}
{{ attr.description }}
{{ am.label }} {{ am.value }}
Distribution
{{ d.label }}
{{ d.value }}
Definition validated · {{ attr.rowsLabel }} 
{{ attr.sql }}
Referenced by
{{ u.name }} {{ u.role }}