Turing / turing.care

Quiet-mode launch page for Epic-native predictive risk governance

Turing is a safety layer around predictive risk workflows. It helps clinical operations teams decide when model output can run, when it should pause, and how to document every decision path with audit-ready evidence.

What Turing is

Evidence-first safety operations for predictive risk programs.

Turing does not replace clinicians or care workflows.

It enforces operational guardrails around model output and keeps a full audit trail of why outcomes changed.

The current MVP is focused on controlled pilot operations and reproducible demo workflows.

Who it is for

Clinical operations, quality/risk, and data platform teams.

Clinical operations leaders who need predictable guardrails before scaling model usage.

Quality and risk teams who need transparent, reviewable evidence across every policy decision.

Technical teams integrating Epic SMART/FHIR read paths into safe operational workflows.

Three core safety scenarios

HOLD

Hard stop when data integrity or safety boundary is clearly violated.

Example: unit mismatch (`Expected C`, `Observed F`) blocks operational use until corrected.

ABSTAIN

Do not render model guidance when current context is out of intended-use scope.

Example: pediatric context outside approved model population.

CAUTION

Continue operation with explicit caution when drift pressure is rising.

Example: cohort mix shift and alert burden growth require monitored mitigation plan.

How the trust layer works

Deterministic, policy-driven flow from event to action.

  1. 1.Read event and context (prediction, encounter attributes, lineage metadata).
  2. 2.Run explicit safety gates (data quality, intended use, drift pressure).
  3. 3.Map gate evidence to outcome semantics (ALLOW, CAUTION, ABSTAIN, HOLD).
  4. 4.Open audit trace and investigation path with deterministic links.
  5. 5.Escalate governed change proposal when thresholds or workflows need adjustment.

What a clinic gets after a 6-week pilot

Operational deliverables, not marketing promises.

  • Week 1-2: environment alignment, data contract check, deterministic demo baseline.
  • Week 3-4: operational safety review with HOLD/ABSTAIN/CAUTION investigations.
  • Week 5-6: governed change loop and handoff pack for clinical + IT stakeholders.

Scope note: current pilot scope is read-only Epic integration with governed safety workflows and audit traceability.

Demo and launch materials

Use the demo kit to run HOLD, ABSTAIN, and CAUTION scenarios in consistent order and connect them to audit, drift, and change-control evidence.

Canonical sequence source: `docs/DEMO_RUNBOOK.md` (HOLD to ABSTAIN to CAUTION to SMART).