The HIS data mountain and the health insurance settlement puzzle. The hospital's HIS system is a massive data warehouse with tens of thousands of tables spanning years of operations. Key issues:
Panda connects directly to Oracle using a SELECT-only account — read-only, completely safe for the hospital's operational data. No writes, no modifications, no interference with HIS workflows.
Out of ~17,000 tables across the entire system, Panda maps >2,000 HIS master tables against a normalized business map — every table and column is clearly defined in business terms.
Data from 01/2018 → 07/2026 lives across ~100 monthly schemas. Panda consolidates them into a single unified source, enabling queries across 8+ years of data with no time boundaries.
With a normalized business map, every management, settlement, and reconciliation report flows through the same data source — ending the era of conflicting departmental numbers.
37+ core business tables normalized, including the most critical ones:
The BHYT cost model is fully built: insurance fund share (BHYTTRA), patient out-of-pocket (BNTRA), co-payment — with a 2-line structure per medical order, cleanly separating the fund portion from the patient portion.
Built complete QTCT, QTLL, QTNHOM settlement views — consistent structure across settlement types, ending the era of mismatched view formats.
Join BIEN_LAI → admission/registration to correctly retrieve medical record type (medical record type) — fixing the root cause of mismatches when comparing data across sources.
Accurate bed occupancy from a standardized bed catalog: ~1,000 beds across ~20 departments — a management metric previously impossible to compute correctly.
Advances retrieved via LEFT JOIN HOANTRA — preserving all unmatched records, with full audit trail from HUY tables to trace every data change.
🔴 Project in progress — Pilot 2026 at General Hospital
