Skip to content

Medical Ontologies & Terminology Standards

Overview

Entheory.AI integrates with major medical ontologies and coding systems to ensure semantic accuracy, interoperability, and regulatory compliance. This document outlines how we leverage these standards for clinical data normalization, automated billing, and ABDM alignment in the Indian healthcare context.

1. UMLS — Unified Medical Language System

Purpose

UMLS serves as the terminology integration backbone, enabling cross-mapping between disparate coding systems (ICD-10, SNOMED CT, CPT, RxNorm) for semantic normalization.

Entheory.AI Integration

Aspect Implementation
Primary Use Concept normalization, synonym resolution, cross-ontology mapping
Data Source UMLS Metathesaurus (NLM license required)
Key APIs UMLS REST API, Local MRCONSO/MRREL tables
Update Cycle Biannual (aligned with NLM releases)

Key Capabilities

  1. Terminology Integration
  2. Map free-text clinical terms to standard concepts (CUI mappings)
  3. Resolve synonyms: "Type 2 Diabetes" → "Diabetes Mellitus Type 2" → CUI: C0011860

  4. Link ICD-10, SNOMED CT, and RxNorm via shared Concept Unique Identifiers (CUIs)

  5. Semantic Accuracy

  6. Validate clinician-entered diagnoses against standardized terminology
  7. Detect ambiguous or deprecated terms during data ingestion
  8. Support multilingual lookups (English, Hindi transliterations)

India-Specific Adaptations

[!IMPORTANT] UMLS requires an NLM Metathesaurus license. For ABDM compliance, we prioritize mappings to SNOMED CT (ABDM's recommended terminology).

  • Regional Term Mapping: Indian clinical abbreviations (e.g., "T2DM", "HTN") mapped to UMLS concepts
  • Ayushman Bharat Alignment: Cross-reference with Health Benefit Packages (HBP) coding

2. SNOMED CT — Clinical Terminology

Purpose

SNOMED CT provides the clinical terminology foundation for precise documentation, problem lists, and automated billing workflows.

Entheory.AI Integration

Aspect Implementation
Primary Use Clinical concept coding, problem list standardization
Edition SNOMED CT International + India Extension (when available)
FHIR Binding http://snomed.info/sct
ABDM Profile SNOMED CT required for Condition, Procedure resources

Key Capabilities

  1. Clinical Terminology
  2. Encode diagnoses, procedures, findings, and anatomical sites
  3. Hierarchical relationships enable clinical decision support

  4. Example: "Invasive Ductal Carcinoma of Breast" → 254837009

  5. Automated Billing Support

  6. Map clinical concepts to billable codes via SNOMED-to-ICD-10 crosswalks
  7. Reduce manual coding effort for insurance claims

  8. Support audit trails for billing compliance

  9. EHR Interoperability

  10. FHIR R4 resources use SNOMED CT as primary clinical coding
  11. Enables seamless data exchange with ABDM-compliant systems

ABDM Compliance

[!NOTE] ABDM mandates SNOMED CT for clinical terminologies in FHIR bundles. Entheory.AI validates all clinical concepts against SNOMED CT before FHIR export.

FHIR Resource SNOMED CT Usage
Condition Diagnosis codes (e.g., 254837009 for breast carcinoma)
Procedure Procedure codes (e.g., 387713003 for mastectomy)
Observation Clinical findings and measurements
AllergyIntolerance Allergen and reaction coding

India-Specific Adaptations

  • India Extension: Track availability of SNOMED India module from MoHFW
  • Regional Disease Patterns: Ensure coverage for high-prevalence conditions (TB, dengue, malaria)
  • TPA Integration: Map SNOMED concepts to Third Party Administrator (TPA) required codes

3. ICD-10 — Disease Classification

Purpose

ICD-10 provides the disease classification standard for epidemiological reporting, insurance claims, and EHR interoperability.

Entheory.AI Integration

Aspect Implementation
Primary Use Diagnosis classification, morbidity/mortality reporting
Version ICD-10-CM (Clinical Modification) for diagnoses
FHIR Binding http://hl7.org/fhir/sid/icd-10-cm
Update Cycle Annual (October releases)

Key Capabilities

  1. Disease Classification
  2. Systematic classification of diseases, injuries, and health conditions
  3. Hierarchical structure: Chapter → Block → Category → Subcategory

  4. Example: "Type 2 Diabetes with Nephropathy" → E11.65

  5. EHR Interoperability

  6. Standard coding for diagnosis exchange between systems
  7. Required for hospital discharge summaries (ICD-10 mandatory in India since 2017)

  8. FHIR Condition resources include ICD-10 as secondary coding

  9. Insurance & Claims

  10. Mandatory for CGHS, ECHS, ESI, and most private insurance claims
  11. Map clinical narratives to appropriate ICD-10 codes via NLP
  12. Reduce claim denials through accurate coding

India-Specific Adaptations

Requirement Implementation
MoHFW Mandate ICD-10 mandatory for HMIS reporting since 2017
Ayushman Bharat PM-JAY packages mapped to ICD-10 codes
State Registries Support state-specific disease surveillance (IDSP)
NCRP Reporting Oncology cases aligned with National Cancer Registry Programme

NLP-Assisted Coding

Entheory.AI provides AI-assisted ICD-10 coding:

flowchart LR
    A["Clinical Text"] --> B["NLP Entity Extraction"]
    B --> C["UMLS Concept Mapping"]
    C --> D["ICD-10 Code Suggestion"]
    D --> E["Physician Review"]
    E --> F["Confirmed Code"]

4. CPT — Procedure Codes

Purpose

CPT (Current Procedural Terminology) provides procedure coding for billing, reimbursement, and explainable AI audit trails.

Entheory.AI Integration

Aspect Implementation
Primary Use Procedure documentation, billing automation
License AMA CPT license required for production use
Alternative HCPCS for public domain procedure codes
India Context Limited direct use; mapped to NABH/TPA procedure lists

Key Capabilities

  1. Procedure Codes
  2. Standardized codes for medical, surgical, and diagnostic services
  3. Categories: Evaluation & Management, Surgery, Radiology, Pathology, Medicine

  4. Example: "CT Chest with Contrast" → 71260

  5. Explainable AI

  6. AI-generated recommendations linked to specific CPT procedures
  7. Audit trail: "Recommended based on CPT 71260 (CT Chest) findings"

  8. Transparency for clinical decision support explanations

  9. Billing Integration

  10. Automate charge capture from clinical documentation
  11. Map procedures to institutional fee schedules
  12. Support DRG calculation for inpatient services

India-Specific Adaptations

[!WARNING] CPT is not directly mandated in India. Local procedure coding follows NABH, CGHS rates, and TPA-specific procedure lists.

India Standard CPT Relationship
NABH Procedure Codes Crosswalk mapping for accredited hospitals
CGHS Rate List Map CPT equivalents for government beneficiary billing
PM-JAY Packages Health Benefit Packages with procedure components
TPA Procedure Lists Insurer-specific codes mapped via crosswalk tables

5. Cross-Ontology Architecture

Mapping Flow

flowchart TB
    subgraph "Clinical Input"
        A["Free Text / Voice"]
        B["Structured Forms"]
        C["HL7 / FHIR"]
    end

    subgraph "Normalization Layer"
        D["NLP Entity Extraction"]
        E["UMLS Concept Lookup"]
        F["CUI Resolution"]
    end

    subgraph "Target Ontologies"
        G["SNOMED CT"]
        H["ICD-10"]
        I["CPT / NABH"]
        J["RxNorm"]
    end

    subgraph "Outputs"
        K["FHIR Bundle"]
        L["Insurance Claim"]
        M["Registry Report"]
    end

    A --> D
    B --> D
    C --> E
    D --> E
    E --> F
    F --> G
    F --> H
    F --> I
    F --> J
    G --> K
    H --> K
    H --> L
    I --> L
    H --> M

Mapping Tables

Source System Primary Ontology Secondary Ontology Use Case
Discharge Summary OCR SNOMED CT ICD-10 Clinical coding + billing
Lab Results (ORU) LOINC SNOMED CT Lab observation coding
Pathology Reports SNOMED CT ICD-O-3 Oncology histology
Prescriptions RxNorm — Medication normalization
Procedures SNOMED CT CPT/NABH Clinical + billing

6. ABDM Terminology Compliance

Required Standards

The Ayushman Bharat Digital Mission mandates specific terminology usage for FHIR interoperability:

FHIR Resource Required Ontology Entheory.AI Status
Patient — (demographics) ✅ Compliant
Condition SNOMED CT ✅ Compliant
Observation LOINC / SNOMED CT ✅ Compliant
Procedure SNOMED CT ✅ Compliant
MedicationStatement RxNorm (preferred) ✅ Compliant
DiagnosticReport SNOMED CT / LOINC ✅ Compliant
AllergyIntolerance SNOMED CT ✅ Compliant

Validation Pipeline

All outbound FHIR bundles pass through terminology validation:

  1. Code Verification: Validate concept codes exist in target ontology
  2. System URI Check: Ensure correct FHIR system URIs
  3. Display Consistency: Verify display text matches preferred term
  4. Fallback Handling: Flag unmappable concepts for manual review

7. Implementation Notes

Licensing Requirements

Ontology License Cost Notes
UMLS NLM UMLS License Free Requires license agreement
SNOMED CT SNOMED International Free (India member) India is SNOMED member country
ICD-10 WHO Free Public domain, no restrictions
CPT AMA License Paid Required for commercial use
LOINC Regenstrief Institute Free Open license

Update Strategy

  • Quarterly Reviews: Check for ontology updates and new mappings
  • Version Tracking: Maintain ontology version metadata in provenance records
  • Deprecation Handling: Monitor for deprecated concepts and migration paths

Performance Considerations

  • Local Caching: Frequently used concepts cached in-memory
  • Batch Mapping: Bulk terminology mapping during ingestion pipelines
  • Lazy Loading: Full ontology not loaded; API-based lookups as needed

Document Owner: Data Architect / Interoperability Lead
Last Updated: 2024-12-08
Related: Data Model | APIs & Interoperability | NLP Use Cases