Developed a Medical Records App for a Jeddah Hospital Group to Reduce Chart Retrieval Time by 55%

An interface engine receives HL7 v2 messages from the hospital information system and lab system and converts them into HL7 FHIR R4 resources: Patient, Encounter, AllergyIntolerance, MedicationRequest, Observation, DiagnosticReport and DocumentReference. Radiology reports arrive through the archive's API with a link to the image viewer. A clean internal model keeps source field names and quirks out of the app services.

• Deterministic match first on National ID or Iqama number
• Arabic name normalization: removed diacritics, and a transliteration table for common English spellings
• Matches above a high threshold link automatically, and matches in a middle band go to a records officer for review
• Linked charts keep their original hospital file numbers, so nothing is overwritten

• OpenSearch index with Arabic and English analyzers, n-gram matching for partial names and phonetic matching for spelling variants
• Search by name, National ID, Iqama, file number, bed number or phone
• Results returned in two stages: a short list in under a second, then the full chart in the background
• Index updates triggered by interface layer events, with a nightly full rebuild as a safety net

• A composite chart endpoint returns allergies, active medications, last 5 lab panels and latest imaging report in one response
• Redis stores assembled summaries with a time to live of 60 seconds
• Lab results and allergies are re-read from the source after any new message for that patient, so a cached screen never shows a value that was already corrected
• Critical lab values carry a visible timestamp and the source system name

• Role and attribute-based rules: role, hospital, ward, shift and treating relationship decide what a user can open
• Emergency override with a mandatory reason, an instant alert to the privacy officer and a review task within 24 hours
• Sensitive record categories hidden by default and shown only to named roles
• Patient consent flags checked on every chart request

• Ward census and assigned patient summaries cached on the device in an encrypted SQLCipher database
• Cache wiped on logout, on 8 hours of inactivity or after a remote wipe command
• Biometric unlock plus a short PIN, and certificate pinning on all API calls
• Sync checks the version number of each chart on reconnect and refreshes any that changed

• Correlation identifiers carried from the app through the gateway, services and interface layer
• OpenTelemetry tracing and structured JSON logs with masked patient identifiers
• Dashboards for search p50, p95 and p99 latency, chart load time per source system, interface message lag and failed searches
• Append-only audit records for every chart view, search and override, kept for the retention period the hospital group's policy sets