Platform Intelligence

HealthCloud Knowledge Graph

A living, connected model of patients, conditions, diagnostics, and AI — enabling context-aware healthcare applications that understand relationships, not just records.

What It Is

The HealthCloud Knowledge Graph connects every clinical entity into a unified, queryable system. Instead of isolated records in separate databases, you get a rich network of relationships spanning the entire care journey.

Patients→ connected in one unified graph
Conditions→ connected in one unified graph
Observations→ connected in one unified graph
Diagnostics→ connected in one unified graph
AI Models→ connected in one unified graph
Workflows→ connected in one unified graph

Core Graph Path

Patient
→ Condition (UTI)
→ Observation (Urinalysis)
→ Diagnostic (UTI Model v2)
→ Recommendation (Antibiotic)

Graph Node Types

Each node type maps to a clinical or AI entity. FHIR resources are native graph nodes.

👤Patient

FHIR Patient resource with demographics, identifiers, and coverage linkages.

FHIR: Patient
🩺Provider

Clinician or care team member with credentials, specialty, and NPI.

FHIR: Practitioner
🏥Condition

ICD-10 and SNOMED CT coded diagnoses and problem list entries.

FHIR: Condition
📊Observation

Lab results, vital signs, and device readings encoded with LOINC codes.

FHIR: Observation
🧪Diagnostic

AI model output including confidence scores, evidence, and recommendations.

🤖AI Model

Clinical ML model with version, training dataset, and performance metrics.

🗄️Dataset

Curated training or evaluation dataset with provenance and lineage metadata.

🔄Workflow

Orchestration definition linking agents, APIs, and data steps.

Relationship Edge Types

Edges express the relationships between entities. Every edge is typed, bidirectional, and queryable.

EdgeFromToDescription
has_conditionPatientConditionPatient is diagnosed with a condition
has_observationPatientObservationPatient has a clinical observation or measurement
evaluated_byConditionAI ModelA condition is evaluated by an AI diagnostic model
trained_onAI ModelDatasetAn AI model was trained on a specific dataset
triggersObservationWorkflowAn observation anomaly triggers a clinical workflow
recommended_byDiagnosticProviderA diagnostic recommendation is surfaced to a provider
ordered_byProviderObservationA provider ordered a lab test or observation
producesWorkflowDiagnosticA workflow execution produces a diagnostic result

Example Graph Query

Query the graph using HealthCloud's graph query language — traverse relationships across patients, conditions, and AI models in a single request.

Query

GraphQL
query {
  patients(
    condition: "diabetes_type2",
    observation: {
      code: "LOINC:2339-0",
      value_gt: 140
    }
  ) {
    id
    name {
      given
      family
    }
    conditions {
      code
      onset_date
    }
    related_models {
      id
      name
      accuracy
    }
  }
}

Response

200 OK
{
  "data": {
    "patients": [
      {
        "id": "pat_123",
        "name": { "given": "Jane", "family": "Doe" },
        "conditions": [
          { "code": "E11.9", "onset_date": "2024-06-15" }
        ],
        "related_models": [
          {
            "id": "glucose_risk_v2",
            "name": "Glucose Risk Predictor",
            "accuracy": 0.94
          }
        ]
      }
    ]
  }
}

AI + Knowledge Graph

The Knowledge Graph is what makes HealthCloud's AI different. Models don't operate on isolated inputs — they reason across connected clinical context.

🧠

Context-Aware AI

Models understand the full patient context — prior diagnoses, observations, and care history — not just the current input.

🔍

Explainability

Trace any AI output back through the graph to the source data and model weights that produced it.

🔄

Continuous Learning

Link clinical outcomes to model predictions to enable feedback loops that improve accuracy over time.

Real-Time Updates

The graph updates automatically when new observations arrive, workflows execute, or diagnostics complete.

Use Cases

📈Clinical

Disease Progression Tracking

Trace how conditions evolve over time by linking observations, encounters, and model outputs across the patient timeline.

🧪Clinical

Diagnostic Decision Support

Surface AI-recommended next steps by querying related models, evidence datasets, and prior outcomes for similar patients.

🤝Operational

Care Coordination

Identify care gaps by traversing patient → condition → workflow edges to surface unaddressed clinical triggers.

🌎Operational

Population Health Analytics

Aggregate cohort-level insights by querying shared condition or observation patterns across thousands of patients.

🗄️AI

Training Dataset Curation

Identify high-quality training examples by traversing model → dataset → patient edges with outcome metadata.

AI

Model Validation

Trace model outputs back to input observations and ground truth labels to audit accuracy and detect drift.

Graph API Examples

The Knowledge Graph is accessible via a REST API. Query patients, models, and relationships programmatically.

POST/v1/graph/queryQuery patients with specific conditions and elevated lab values

Request Body

{
  "query": "patients",
  "filters": {
    "condition": "diabetes_type2",
    "observation": {
      "code": "LOINC:2339-0",
      "value_gt": 140
    }
  },
  "include": ["conditions", "related_models"]
}

Response

{
  "patient_id": "pat_123",
  "conditions": ["diabetes_type2"],
  "related_models": ["glucose_risk_v2"]
}
GET/v1/graph/patients/{id}Retrieve the full knowledge graph subgraph for a single patient

Response

{
  "id": "pat_123",
  "nodes": 14,
  "edges": 22,
  "conditions": ["E11.9"],
  "observations": 47,
  "workflows_triggered": 3
}
GET/v1/graph/models/relatedFind AI models related to a given condition or observation type

Response

{
  "condition": "diabetes_type2",
  "models": [
    { "id": "glucose_risk_v2", "accuracy": 0.94 },
    { "id": "hba1c_predictor_v1", "accuracy": 0.91 }
  ]
}

Most Platforms Store Healthcare Data. HealthCloud Understands It.

Query across patients, conditions, diagnostics, and AI models in a single request. Build AI that reasons about context — not just data.