Healthcare Ontology
Comprehensive domain knowledge for healthcare organizations, including hospitals, clinics, health systems, and telehealth providers.
Overviewβ
The Healthcare ontology equips AI agents with deep understanding of healthcare operations:
- Patient Management: Registration, scheduling, care coordination
- Clinical Workflows: Orders, results, documentation
- Revenue Cycle: Billing, claims, collections
- Compliance: HIPAA, regulatory reporting
- Quality & Safety: Metrics, incident reporting
Ontology Layersβ
Role Layerβ
roles:
- name: Patient Services Agent
responsibilities:
- Handle appointment scheduling
- Process registration updates
- Answer patient inquiries
- Manage referrals
permissions:
- read: demographics, appointments, insurance
- execute: schedule, reschedule, send_reminder
phi_access: limited # Only what's needed for task
- name: Clinical Support Agent
responsibilities:
- Process order requests
- Track results delivery
- Manage clinical documentation
permissions:
- read: orders, results, notes
- execute: flag_critical, notify_provider
phi_access: clinical # Full clinical context
- name: Revenue Cycle Agent
responsibilities:
- Process claims
- Handle denials
- Manage patient billing
permissions:
- read: claims, payments, balances
- execute: submit_claim, appeal_denial, payment_plan
phi_access: billing # Financial and demographic only
Domain Layerβ
Key Entitiesβ
| Entity | Description | Key Attributes |
|---|---|---|
| Patient | Healthcare recipient | MRN, Name, DOB, Insurance, PCP |
| Encounter | Care episode | ID, Type, Date, Provider, Facility |
| Provider | Clinician | NPI, Name, Specialty, Credentials |
| Order | Clinical request | ID, Type, Status, Urgency, Results |
| Claim | Billing submission | ID, Amount, Payer, Status, Codes |
| Appointment | Scheduled visit | ID, Time, Type, Provider, Status |
Entity Relationshipsβ
erDiagram
PATIENT ||--o{ ENCOUNTER : has
PATIENT ||--o{ APPOINTMENT : schedules
PATIENT ||--o{ INSURANCE : covered_by
ENCOUNTER ||--o{ ORDER : contains
ENCOUNTER ||--o{ DIAGNOSIS : documents
ENCOUNTER ||--|| PROVIDER : seen_by
ENCOUNTER ||--o{ CLAIM : generates
ORDER ||--o| RESULT : produces
Interaction Layerβ
Common Workflowsβ
Appointment Schedulingβ
sequenceDiagram
participant P as Patient
participant A as Agent
participant S as Scheduling
participant E as EHR
P->>A: Request Appointment
A->>E: Check Patient Status
E-->>A: Patient Record
A->>S: Find Available Slots
S-->>A: Available Times
A->>P: Offer Options
P->>A: Select Time
A->>S: Book Appointment
S-->>A: Confirmation
A->>P: Send Confirmation
Clinical Order Processingβ
graph LR
A[Order Placed] --> B{Urgent?}
B -->|Yes| C[Priority Queue]
B -->|No| D[Standard Queue]
C --> E[Process Order]
D --> E
E --> F[Perform Test]
F --> G[Results Available]
G --> H{Critical?}
H -->|Yes| I[Immediate Alert]
H -->|No| J[Standard Notification]
Regulatory Complianceβ
| Regulation | Scope | Agent Actions |
|---|---|---|
| HIPAA | Protected Health Info | Access logging, minimum necessary, encryption |
| HITECH | EHR meaningful use | Audit trails, breach notification |
| CMS CoP | Hospital conditions | Quality reporting, safety protocols |
| Joint Commission | Accreditation | Documentation standards |
| State Laws | Varies by state | Consent management, reporting |
HIPAA Considerations for AI Agentsβ
const hipaaGuardrails = {
// Minimum Necessary Rule
dataAccess: {
principle: "minimum_necessary",
enforce: true,
auditAll: true
},
// PHI Handling
phi: {
neverLog: ["ssn", "full_dob", "credit_card"],
maskInLogs: ["mrn", "name", "address"],
encryptAtRest: true,
encryptInTransit: true
},
// Access Controls
access: {
requireAuthentication: true,
requireAuthorization: true,
logAllAccess: true,
breakGlassAudit: true
}
};
Prompt Templatesβ
Patient Inquiryβ
You are a patient services agent for {organization_name}.
Patient Context:
- Name: {patient_name}
- MRN: {mrn}
- Upcoming appointments: {appointments}
- Outstanding balance: {balance}
Patient Question: {question}
Guidelines:
1. Be empathetic and professional
2. Only share information the patient is authorized to access
3. For clinical questions, direct to their care team
4. For urgent symptoms, advise ER/urgent care
Respond helpfully while maintaining HIPAA compliance.
Appointment Managementβ
You are a scheduling agent processing an appointment request.
Request Details:
- Patient: {patient_name} (MRN: {mrn})
- Requested service: {service_type}
- Preferred times: {preferences}
- Insurance: {insurance}
Provider Availability:
{available_slots}
Find the best appointment option considering:
1. Patient preferences
2. Clinical urgency
3. Insurance requirements
4. Travel/accessibility needs
Integration Pointsβ
EHR Systemsβ
| System | Capabilities | Agent Actions |
|---|---|---|
| Epic | Full EHR | MyChart messages, scheduling, orders |
| Cerner | Full EHR | Patient portal, results, documentation |
| Athenahealth | Cloud EHR | Scheduling, billing, messaging |
| Meditech | Hospital EHR | ADT, orders, results |
Healthcare APIsβ
| Standard | Purpose | Use Case |
|---|---|---|
| HL7 FHIR | Data exchange | Patient records, clinical data |
| HL7 v2 | Messaging | ADT, orders, results |
| X12 EDI | Claims | 837/835 transactions |
Metrics & KPIsβ
| Metric | Description | Agent Monitoring |
|---|---|---|
| Patient Wait Time | Avg time to appointment | Alert if > 14 days |
| No-Show Rate | Missed appointments | Alert if > 10% |
| Clean Claim Rate | First-pass acceptance | Alert if < 95% |
| Patient Satisfaction | Survey scores | Alert if < 85% |
Sample Agent Configurationβ
const healthcareAgent = {
name: "Patient Services",
ontology: "healthcare-v1",
autonomyLevel: "L2", // Supervised
capabilities: {
scheduling: true,
patientInquiries: true,
insuranceVerification: true,
referralManagement: true
},
guardrails: {
hipaaCompliance: "strict",
phiHandling: "minimum_necessary",
requireApprovalFor: [
"cancel_appointment",
"update_demographics",
"share_records"
],
blockedActions: [
"clinical_advice",
"diagnosis_discussion",
"medication_changes"
]
},
integrations: [
"epic_mychart",
"insurance_eligibility",
"scheduling_system"
]
};
HIPAA Compliance
All healthcare agents must operate under strict HIPAA guidelines. Ensure proper Business Associate Agreements (BAAs) are in place and all PHI access is logged and auditable.
Getting Started
To deploy a Healthcare agent:
- Select the Healthcare ontology during workspace setup
- Configure EHR integration (FHIR preferred)
- Enable HIPAA-compliant audit logging
- Start with L1 (Assisted) autonomy for clinical workflows