Learning Graph for Modeling Healthcare Data with Graphs

This section contains the learning graph for this intelligent textbook. A learning graph is a graph of the concepts used in this textbook. Each concept is represented by a node in a network graph. Concepts are connected by directed edges that indicate which concepts each node depends on before that concept can be understood by the student.

A learning graph is the foundational data structure for intelligent textbooks that can recommend learning paths. It is like a roadmap of concepts that helps students arrive at their learning goals.

At the left of the learning graph are the prerequisite or foundational concepts. They have no outbound edges — they only have inbound edges from other concepts that depend on understanding these foundational prerequisites. At the far right are the most advanced concepts in the course. To master these concepts you must understand all of the concepts that they point to.

Learning Graph Statistics

Total Concepts: 513
Total Dependencies: 552
Taxonomy Categories: 14
Foundational Concepts: 2
Average Dependencies per Concept: 1.08
Maximum Dependency Chain: 31 levels
Quality Score: 92/100 (Excellent)

Source Documents and Data Files

Course Description

We use the Course Description as the source document for the concepts that are included in this course. The course description uses the 2001 Bloom taxonomy to order its learning objectives, and now includes a dedicated unit on modeling clinical decision support systems with HL7 FHIR and CQL.

List of Concepts

We use generative AI to convert the course description into a Concept List. Each concept is a short Title Case label, with all labels under 32 characters long. The 513 concepts span graph theory, healthcare domain knowledge, graph analytics, AI/ML, clinical decision support (FHIR, CQL, and CMS authoring/testing tooling), security, and practical applications.

Concept Dependency List

We next use generative AI to create a Directed Acyclic Graph (DAG). DAGs do not have cycles in which concepts depend on themselves. We provide the DAG in two formats: a CSV file and a JSON file that uses the vis-network JavaScript library format. The vis-network format uses nodes, edges, and metadata elements, with edges containing from and to properties. This makes it easy to view and edit the learning graph using an editor built with the vis-network tools.

Analysis & Documentation

Course Description Quality Assessment

This report rates the overall quality of the course description for the purpose of generating a learning graph.

Course description fields and content depth analysis
Validates that the course description has sufficient depth to generate 500+ concepts
Compares the course description against similar courses
Identifies content gaps and strengths
Suggests areas of improvement

The course description scored 100/100, confirming readiness for learning graph generation.

View the Course Description Quality Assessment

Learning Graph Quality Validation

This report gives an overall assessment of the quality of the learning graph. It uses graph algorithms to look for specific quality patterns in the graph.

Graph structure validation — all 513 concepts are connected in a single component
DAG validation (no cycles detected)
No self-dependencies and no orphaned nodes detected
Foundational concepts: 2 entry points (Graph Theory Basics, Healthcare System)
Indegree distribution analysis (most depended-upon concept: Clinical Quality Measure)
Terminal node percentage: 37.6% (within the healthy 5-40% range)
Longest dependency chains (maximum chain length: 31)

The learning graph scored 92/100 (Excellent).

View the Learning Graph Quality Validation

Concept Taxonomy

In order to see patterns in the learning graph, it is useful to assign colors to each concept based on the concept type. We use generative AI to create about a dozen categories for our concepts and then place each concept into a single primary classifier.

A concept classifier taxonomy with 14 categories
Category organization — foundational elements first, capstone project ideas last
Balanced categories, all well under the 30% threshold
Clear 3-5 letter abbreviations for use in the CSV file

The 14 categories are: Foundation Concepts (FOUND), Graph Technologies (GTECH), Graph Analytics and Algorithms (ANAL), Healthcare Domain Fundamentals (HCARE), Patient Data and Clinical Concepts (PAT), Provider Operations (PROV), Payer and Insurance (PAYER), Financial and Business Operations (FIN), Fraud, Waste, and Abuse (FRAUD), AI and Machine Learning (AI), Clinical Decision Support, FHIR and CQL (CDS), Security and Privacy (SEC), Data Governance (GOV), and Capstone and Career (CAP).

View the Concept Taxonomy

Taxonomy Distribution

This report shows how many concepts fit into each category of the taxonomy. Our goal is a somewhat balanced taxonomy where each category holds an equal number of concepts. We also don't want any category to contain over 30% of our concepts.

Statistical breakdown
Detailed concept listing by category
Visual distribution table
Balance verification

The largest category (Clinical Decision Support, FHIR and CQL) holds 59 concepts (11.6%), well within the acceptable range.

View the Taxonomy Distribution Report

Key Features

Multi-Perspective Learning

The learning graph incorporates three critical healthcare perspectives:

Patient Perspective: Clinical data, diagnoses, treatments, care plans
Provider Perspective: Hospitals, clinics, schedules, referrals, performance
Payer Perspective: Claims, policies, coverage, reimbursement

Clinical Decision Support with FHIR and CQL

A dedicated taxonomy category (CDS) covers modeling healthcare decision support systems:

The Four Levels of Knowledge Representation in FHIR (Narrative, Semi-Structured, Structured, Executable)
The three Tiers of Functionality (Data, Logic, Forms/UI)
Clinical Quality Language (CQL) authoring and the Expression Logical Model (ELM)
Clinical Quality Measures (CQMs), CDS Hooks, and guideline-based care pathways
CMS-sponsored authoring, testing, and certification tools: MADiE, CQL Runner, Bonnie, and Cypress

Comprehensive Coverage

Topics span from foundational concepts to advanced applications:

Graph database fundamentals
Healthcare domain knowledge
Query languages (Cypher, GQL, GSQL)
Graph analytics, embeddings, and graph neural networks
Clinical decision support with FHIR and CQL
Fraud detection and compliance
AI/ML integration, LLMs, agentic workflows, token efficiency, context graphs, enterprise knowledge graphs, and the enterprise nervous system concept
Security (HIPAA, RBAC, encryption)
Data governance and explainability

Pedagogically Sound Structure

The learning graph follows established educational principles:

Clear prerequisite relationships
Progressive complexity
Multiple learning pathways
Foundation → Application → Synthesis progression
Capstone projects for knowledge integration

Using the Learning Graph

For Students

Use the concept list to understand the full scope of the course
Follow dependency paths to identify prerequisite knowledge
Track your progress through the 513 concepts
Navigate personalized learning pathways based on your goals

For Instructors

Understand the logical structure of the course content
Identify critical foundational concepts
Design lesson plans that respect dependency relationships
Create assessments aligned with the concept taxonomy

For Developers

Import the learning-graph.json into visualization tools
Build adaptive learning systems using the dependency data
Create personalized recommendations based on concept relationships
Integrate with learning management systems

Technical Details

Format: Learning Graph JSON v1.0
Schema: learning-graph-schema.json
Creator: Dan McCreary
License: CC BY-NC-SA 4.0 DEED
Version: 2.0
Generated: November 6, 2025
Last Updated: September 7, 2026 (regenerated from scratch, 200 → 513 concepts, adding the Clinical Decision Support, FHIR and CQL category)

For questions or feedback about the learning graph, please refer to the course description or contact the course creator.