MODULE ID: SVC-10

Test Ordering Optimization Layer

STATUS: PILOT

Not clinical advice. Protocol-dependent. Must be configured and validated per deployment.

INPUTSPROCESSOPTIMIZEOUTPUTREVIEW
META
LAYER
TEST ORDERING OPTIMIZATION FLOW
01/12MODULE: PROBLEM SPACE

The Challenge

Even with good triage, ordering behavior remains inconsistent. Clinicians often order broad panels because it is faster than reasoning through minimal pathways under time pressure.

→Broad "shotgun" panels
→Duplicate tests across visits
→Protocol drift
OPTIMIZATION IMPACT
Panel Reduction
40-60%
Cost Efficiency
+35%
improvement
02/12MODULE: INTERVENTION

What It Reduces

Broad Panels
BENEFIT 1

Broad Panels

Reduces "shotgun" panel ordering patterns

Duplicate Tests
BENEFIT 2

Duplicate Tests

Eliminates redundant tests across visits

Protocol Drift
BENEFIT 3

Protocol Drift

Maintains consistent ordering standards

03/12MODULE: INPUTS

Input Sources

Engine Outputs
INPUT 1

Engine Outputs

Risk bands and recommendations from SVC-01 to SVC-09

Clinic Protocols
INPUT 2

Clinic Protocols

Local protocol rules (configurable)

Constraints
INPUT 3

Constraints

Patient constraints (cost sensitivity, access)

04/12MODULE: CORE CONCEPT

How It Works

This layer does not detect disease. It converts risk bands into a minimal test plan that is:

✓
Protocol-Aligned
✓
Auditable
✓
Configurable
✓
Reversible
05/12MODULE: OUTPUTS

Output Schema

OUTPUT FIELDS
  • • Minimal next test category
  • • Rationale notes (short, structured)
  • • Confidence and abstain notes
  • • "Stop conditions" (when to escalate)
SAFETY NOTES
  • • Not clinical advice
  • • Protocol-dependent
  • • Must be configured per deployment
  • • Clinician override always available
06/12MODULE: SAFETY GATES

Abstain Conditions

CONDITION 1

No protocol configuration present

CONDITION 2

Conflicting engine outputs

CONDITION 3

Missing constraint inputs

07/12MODULE: MODEL REGISTRY

Model Sources

DomainTaskModel / SourceLink
Decision logicRankingscikit-learnSource →
Tabular rankingBaselineLightGBMSource →
Constraint logicOptimization (optional)OR-ToolsSource →
08/12MODULE: DEPLOYMENT

Deployment Configuration

HARDWARE REQUIREMENTS
CPU: ✓ Required
GPU: ✗ Not required
Edge: ✗ Not supported
Offline: ✗ Not supported
DEPLOYMENT NOTES

Runs CPU-only. Sits above all other wings. Exports a printable "minimal plan" for clinician sign-off.

09/12MODULE: LIMITATIONS

Important Limitations

  • ✗Not clinical advice
  • ✗Protocol-dependent
  • ✗Must be configured and validated per deployment
  • ✗Requires local protocol configuration
10/12MODULE: EXAMPLES

Example Outputs

MINIMAL PLAN

"Minimal plan: monitor + targeted check. Broad panel not recommended unless new red flags appear."

ABSTAIN

"Protocol missing. Abstain. Recommend clinician decision."

11/12MODULE: WORKFLOW

Clinical Workflow Integration

Workflow
WORKFLOW DIAGRAM
Engine outputs → Protocol matching → Constraint optimization → Minimal plan → Clinician review
12/12MODULE: INTEGRATION

How This Service Fits the Engine

The Test Ordering Optimization Layer operates as a meta-service that sits above all other wings (SVC-01 to SVC-09). It converts risk bands and recommendations from individual services into a minimal, protocol-aligned test pathway with auditable rationale, reducing broad panel ordering while maintaining clinician control.

PRIMARY FUNCTION

Meta-service for test ordering optimization

OPERATING MODE

MINIMAL PANEL