Skip to content

Visual Authoring Concepts

This page summarizes the conceptual model from your WHO visual authoring material and aligns it with TRICC implementation.

Why visual CDSS authoring

  • Transforms static, paper-like guidance into dynamic and computable decision support.
  • Supports local customization (protocols, endemicity, drug availability).
  • Clarifies calculations by formalizing inputs and algorithm logic.
  • Supports comorbidities and more integrated patient-centered logic.

Authoring roles

  • Subject matter experts (SME/clinicians): own clinical content and final decisions.
  • End users: provide workflow and usability feedback.
  • IT specialists: ensure implementability and technical consistency.
  • Other stakeholders (MoH, implementers, sponsors): ensure contextual fit.

Core challenge themes

  • Collaboration and shared ownership across stakeholders.
  • Balancing technical precision with readability.
  • Treatment modeling with prioritization and safety.
  • Reusable activity design without duplication.
  • Sequence logic without loops/dead ends.
  • Controlled starts by segment/process.
  • Reliable conversion with low regression risk.

Segmented approach

Authoring is split into care/process segments so teams can:

  • decide when each segment should run,
  • merge several guideline segments coherently,
  • jump between segments when needed.

This aligns with process-based starts and modular activity linking in TRICC.

Layered approach

The material describes layered authoring:

  • Layer 1: Segment overview (WHAT should happen).
  • Layer 2+: Activities (HOW a segment executes).
  • Layer 3+: Nodes/tasks (specific actions and logic).

TRICC implements this through start/process orchestration, activity diagrams, and typed nodes.

How this maps to TRICC

  • Segment-level orchestration maps to process starts (start with process).
  • Activity-level logic maps to activity diagrams (activity_start, activity_end, goto).
  • Task-level logic maps to node types (inputs, messages, calculations, sequence nodes).

Practical recommendations from implementation slides

  • Keep diagrams readable; move dense logic into CQL when appropriate.
  • Maintain a concept/code-driven data dictionary.
  • Work in small iterations: change, convert, test.
  • Keep at least one file per segment for parallel authoring.
  • Treat warnings/errors as first-class signals before debugging assumptions.