For 20 years, the Department of Medical Affairs kept its data in dozens of standalone systems that worked like black boxes, with no links between divisions or between systems. When leadership needed a cross-unit analysis, staff spent weeks compiling it by hand. Under the Medical Administration Digital Transformation Program, we built the department a data integration hub.

Medical data integration hub (illustrative)

Illustrative

Dozens of systems

Medical administration

Pediatric physicians

Hospital accreditation

+ Dozens more systems

External data

PopulationHealth insuranceLong-term careDeath statistics

Integration hub

Cleanse & standardizeFHIR

De-identify

Unify keys

Delivered as

API

SFTP

SQL Endpoint

BI analytics

RAG knowledge center

Runs in the agency's dedicated data center

Client: Department of Medical Affairs, Ministry of Health and Welfare
Program: Medical Administration Digital Transformation Program
Contract: Accepted and closed in 2024; renewed for 2026
Environment: Ministry of Health and Welfare dedicated data center

Dozenscore systems connected

FHIRinternational healthcare data standard

3external interfaces: API, SFTP, SQL Endpoint

Renewedaccepted in 2024, renewed for 2026

The challenge

The data was all there, but none of it could be used together.

Core systems for medical administration, pediatric dedicated physicians and hospital accreditation each ran on their own. Any cross-unit analysis meant manual compilation, sometimes even rounds of official letters. This kind of "data dysfunction" made the road to digital transformation far longer.

The goal was more than compiling and querying data. It was to reshape the data itself: turning scattered administrative records into digital assets with real decision-making value, and laying the groundwork for smart health administration within the agency's security and privacy rules.

How we did it

01 · Data cleansing & standardization

ETL cleansing, unified to the FHIR standard

Facing dozens of heterogeneous back-office systems, we carried out deep ETL cleansing: removing redundant and noisy values, unifying key logic, and standardizing formats to the international FHIR healthcare data standard, so data can be processed plug-and-play across divisions, units and even ministries.

A Data Fabric architecture with a NoSQL cloud data lake
Structured and unstructured data held side by side
Linked with external data: Ministry of the Interior population data, national health insurance, long-term care and death statistics

ETL cleansing job (illustrative)

Illustrative

Source records · 3 systems

Medical administration

hosp_code:0101-A

lic_date:113/5/2(ROC year)

dept:Pediatrics

Pediatric physicians

HospID:0101A

start_dt:2024/05/02

dept_name:Peds clinic

Hospital accreditation

org_no:0101 A

date:2024-5-2

spec:PED.

Cleansed · FHIR resourcesFHIR R4

resourceType:"Organization"

identifier:"0101A"

address.city:"Taipei City"

resourceType:"Practitioner"

qualification.code:"PED" · Pediatrics

period.start:"2024-05-02"

managingOrganization:"Organization/0101A"

4 duplicate fields merged

Redundancy removedKeys unifiedFHIR

Data Fabric + NoSQL data lake: structured and unstructured data side by side

02 · Data integration hub

One hub, three external interfaces

The hub offers API, SFTP and SQL Endpoint interfaces, so the Department of Medical Affairs and other divisions can plug BI tools straight in. Every medical administration record can now inform policy decisions.

Unified keys that join across tables
Consistent access control on every interface
AI-ready architecture, the foundation for AI inference

External endpoint console (illustrative)

Role-based access

APIConnected

GET /v1/practitioners?region=north&specialty=PED

200 OK · FHIR JSON

SFTPExported daily at 02:00

practitioner_20261110.csv.gz

organization_20261110.csv.gz

accreditation_20261110.csv.gz

SQL Endpoint

SELECT o.region, COUNT(p.id)
FROM practitioner p
JOIN organization o
  ON p.org_key = o.org_key
GROUP BY o.region;

BI preview: dedicated physicians by region (illustrative)

NorthCentralSouthEastIslands

Dept. of Medical AffairsFullOther divisionsAnalytics viewExternal unitsDe-identified only

03 · RAG knowledge center

AI answers you can check against the data

On top of the hub we built a RAG knowledge center that uses structured health administration data as the knowledge anchor for AI inference. The AI doesn't just talk: it understands the context of health policy, pulls the latest figures in real time and helps staff interpret policy.

Every answer traces back to its data source
Pulls the latest data from the hub in real time
Supports interpretation; staff still make the decisions

RAG knowledge center (illustrative)

Illustrative

How has the number of dedicated physicians changed by city/county over the past three years?

AI answer

Overall, dedicated physician numbers have risen over the past three years, most clearly in northern and central areas. Eastern and outlying-island counties have barely changed and remain relatively short-staffed, so they may merit review in the next phase.

City/county2023→2025Trend

Taipei CityRising

Taichung CityRising

Kaohsiung CityFlat

Hualien CountyFlat

Sources (illustrative)

1Pediatric dedicated physician rosterUpdated 2026-11-01

2Medical personnel practice registryUpdated 2026-11-01

Every answer traces back to its source · staff make the decision

04 · Privacy & security

Privacy built in from the architecture up

Inside the Ministry of Health and Welfare's dedicated data center, we adopted a secure software development lifecycle (SSDLC): every stage, from requirements and design to development and go-live, passes a security review.

All sensitive medical data is de-identified before processing
VM and SQL server isolation
Zero-trust principles to protect every citizen's personal data

De-identification (illustrative)

Illustrative

BeforeAfter

National IDA12•••••89PID-7F3A2CTokenized

NameWang *******Masked

Date of birth1987-03-141985–1989Banded

SpecialtyPediatricsPediatricsKept

Practice locationTaipei CityTaipei CityKept

Raw data is processed only inside the dedicated data center; masked on screen

MOHW dedicated data centerVM isolationSQL server isolation

SSDLC security reviewRequirementsDesignDevelopmentGo-live

Results

Health administration data, viewed across systems for the first time.

Area	Outcome
Data integration	Dozens of core systems connected, including medical administration, pediatric dedicated physicians and hospital accreditation
External links	Population, national health insurance, long-term care and death statistics, viewed alongside health administration data
Analytics	API, SFTP and SQL Endpoint, so every division can run BI analysis directly
Contract	Accepted and closed in 2024, renewed for 2026; also delivering the 2025–2026 Medical Workforce Management Efficiency and Digital Application Program

More public-sector projects