Healthcare • Data Engineering • AI

[REDACTED] Healthcare Ministry

Sovereign Patient Record Linkage Platform

Client:GCC Government Healthcare Ministry
Timeline:9 months (2023)
Team:Data Engineering, AI, Security & Mobile
Entity ResolutionSnowflakeArabic NLPAirflowSovereign AIBlockchain
System Architecture

The Challenge

A national ministry of health operated 14 separate legacy hospital systems with no unified patient identity. Patients often had up to 7 duplicate records, making cross-hospital history invisible to clinicians and causing dangerous gaps in care. Manual reconciliation was taking an average of 25 days per case.

  • Duplicate patient identities causing medication errors and redundant testing
  • Fragmented data silos across 14 hospital management systems
  • Complexity of Arabic name normalization and demographic variability
  • Requirement for 100% data residency on air-gapped infrastructure

The Solution

Architected a sovereign data lake and an AI-powered entity resolution service to unify patient records with 99.1% precision confirmed by manual audit.

  • AI Record Linkage Engine using custom probabilistic matching models
  • Arabic NLP Normalization for handling varying transliteration and phonetic variants
  • Sovereign Data Lake on air-gapped infrastructure using Snowflake
  • Clinician-Facing Portal (Next.js) for a unified 360-degree patient timeline
  • Blockchain Audit Trail for forensics and regulatory compliance

Lexer System's Approach

1

Probabilistic Matching Logic

Developed a custom ML classifier trained on anonymized demographics, utilizing Jaro-Winkler similarity and phonetic Arabic matching to handle naming complexities.

2

Unified Data Mesh

Deployed ETL pipelines using Apache Airflow to pull data from all 14 legacy systems into a unified Snowflake data lake within the ministry's network.

3

Sovereign Air-Gapped Setup

Configured a fully isolated VPC with zero trust architecture, satisfying HIPAA and PDPL (Personal Data Protection Law) requirements.

4

Clinician Experience Design

Built a high-performance clinical portal and an offline-first mobile app for field workers to access and update patient records in real-time.

Results & Impact

340,000+
Record Merges

Duplicates eliminated in initial phase

25d → 4d
Processing Time

Time to reconcile complex patient cases

99.1%
Linkage Precision

Confirmed accuracy via manual audit sampling

Air-Gapped
Sovereignty

Zero data leakage outside government boundary

Technical Highlights

Arabic Name Normalization

Specialized NLP models to handle the vast variance in Arabic name spelling and family name prefixes.

HL7/FHIR Ingestion

Scalable adapters for multiple legacy hospital formats, converting them to unified FHIR-standard entities.

Forensic Audit Trail

Immutably logging every data access and record merge event using a private blockchain layer.

Lessons Learned

  • Entity resolution in healthcare requires a hybrid approach of ML and phonetic rules for highest safety
  • Sovereignty requirements dictate that the entire AI/ML lifecycle must run within the air-gapped perimeter
  • Data cleaning from legacy systems is 70% of the effort in a record linkage project

Next Steps

  • Integrate AI-powered diagnosis suggestion based on unified history
  • Expand to national pharmacy and insurance system integrations
  • Implement predictive risk modeling for chronic disease management

Have a Similar Challenge?

We specialize in building production-grade systems that solve complex operational problems. Let's discuss how we can help architect your solution.

Let's TalkView More Projects