Private Hospital Group
Seven isolated software environments. One unified definitional layer. No system replaced.
“We had one system for clinical documentation, another for imaging, one for payroll, one for appointments, and yet another for billing. Each one worked. None of them could see the others. People spent their entire day copying data from one screen to another — and everyone accepted it as normal, because it had never been any different.”
Chief Operating Officer, Private Hospital Group, Budapest
Client identity anonymized at the institution's request. Operational metrics reflect the actual deployment environment.
One of Hungary's largest private hospital groups. Budapest headquarters with multiple regional units. Seven separate software environments with no real-time data sharing between units. Stathon deployed at Forge Tier from September 2025 to March 2026 — no system replaced. Lab result processing reduced from 52–90 minutes to 4–9 minutes. Manual billing entries reduced from 15–25 to 2–4 daily. Monthly financial close shortened from weeks to one business day. 11,000–15,000 duplicate patient records identified across units.
The institution.
One of Hungary's largest private hospital groups. Budapest headquarters with multiple regional units, operating in the upper segment of private healthcare in Hungary.
Clinical scope covers internal medicine, surgery, orthopedics, gynecology, cardiology, oncological diagnostics, and rehabilitation. The group serves both self-paying patients and those covered by private health insurance arrangements.
At engagement start, the group operated seven separate software environments with no real-time data sharing between units, and no unified patient or financial record across the network.
Seven separate software environments. medMatrix runs separate databases per unit. No real-time data sharing between units.
Seven islands.
2.1 Isolated Operations
Each unit ran its own isolated medMatrix installation. The same patient visiting two units produced two separate records with no cross-unit linkage. Patient identity duplication was structural, not incidental.
PACS access was local to each unit. Cross-unit imaging requests were handled by phone call, USB transfer, or printed paper. No physician could access another unit's imaging from their workstation.
Lab results arrived as PDFs via the SYNLAB Online portal. Staff manually downloaded each PDF and uploaded it to the corresponding patient record in medMatrix. Processing times ranged from 52 to 90 minutes per result batch.
Billing required full manual re-entry from medMatrix episodes into Szamlazz.hu. This was a daily routine affecting all units. Health fund invoice errors and corrections were a direct consequence of manual transcription.
NEXONber was only partially deployed. The remainder of payroll management ran on Excel. No automated data flow existed between medMatrix attendance data and NEXONber payroll preparation.
Kulcs-Soft was operational only at the Budapest headquarters. Regional units had no unified controlling view. Monthly financial close required manual aggregation across units and took several weeks.
2.2 Workload Assessment
Lab PDFs to medMatrix. medMatrix episodes to Szamlazz.hu. Booking system entries to medMatrix. medMatrix attendance data to NEXONber. In each case, a human being was the integration layer between two systems that had no other connection.
The same patient visiting two different units was registered as two separate persons. This created clinical risk (incomplete care history), billing errors (duplicate or missing charges), and irreconcilable patient counts in financial reporting.
Monthly financial close required manual aggregation across units and consistently took several weeks. Health fund invoice errors generated from manual transcription required regular correction cycles. Discrepancies between medMatrix, Szamlazz.hu, and Kulcs-Soft were structurally irreconcilable.
DIVAS-PRO access was confined to the unit where the imaging was performed. Cross-unit imaging requests required a physician to call a colleague in another unit, who would then retrieve the file and transfer it via USB or printed paper. This process introduced delays and created chain-of-custody gaps for diagnostic-grade imaging data.
Multiple vendors stated that integration with their system was technically impossible. This position reflected commercial interest in maintaining system lock-in, not technical reality. The actual integration surface of each system was assessed independently.
2.3 The Actual Integration Landscape
Independent technical assessment of each system revealed a complete and buildable integration surface. No system required replacement.
Three phases.
Forge Tier deployment. All seven client systems unchanged throughout. No data migration. No replacement. No downtime window.
Arché: Integration Channels & Definitional Model
Sep – Nov 2025Core & Athena: Automation & Intelligence
Dec 2025 – Feb 2026SYNLAB results automatically matched to the correct patient record across all units. Manual PDF download and upload process eliminated. Processing time: 52–90 min → 4–9 min.
medMatrix episodes automatically prepared as Szamlazz.hu invoice entries awaiting human approval. Daily manual data entries reduced from 15–25 to 2–4. Health fund invoice errors dramatically reduced.
For the first time in the group’s history, a physician could see all care events for a patient across all units in a single view. No phone calls. No asking colleagues in other units.
Auto-generated NEXONber-compatible data packages from medMatrix attendance and scheduling data. Excel-based payroll preparation discontinued for covered units.
First unified real-time controlling view across all units. Revenue, capacity, and billing status visible at the network level for the first time.
11,000–15,000 duplicate patient records identified across the medMatrix unit installations. Reconciliation process initiated under clinical governance oversight.
Aegis: Governance & Data Sovereignty
Feb 2026 – ongoingHungarian healthcare data law (Euak.), GDPR Article 9 (special category health data), and EESZT data model compliance. All three regulatory layers addressed in the governance architecture.
Eight defined access roles across clinical, administrative, billing, payroll, and controlling functions. Role definitions based on actual workflow patterns, not system defaults.
Pseudonymization at the point of ingestion. Separate key vault. Data Protection Impact Assessment (DPIA) in progress under Hungarian DPA framework.
What changed.
Measured impact from the first six months across all seven units. Full production since March 2026.
Seven isolated systems now operate beneath a single definitional layer. The change is not integration in the traditional sense — it is the introduction of structural coherence where none previously existed. Each system retained its function. What changed is their relationship to each other.
The administrative hours freed from manual data entry, PDF processing, and cross-system reconciliation were not eliminated — they were redirected. Staff now spend that time on patient service, proactive scheduling, and clinical coordination rather than copying data between screens.
What this case revealed.
"No API" as vendor position, not technical reality
medMatrix runs on MS SQL Server. The database is client-owned and fully accessible for read operations without vendor cooperation. DIVAS-PRO implements DICOM 3.0, a published international standard — header data is extractable by any party with network access. In both cases, the vendor’s claim that integration was impossible described their commercial interest, not the technical state of their software. Vendors protect their own position. Client data sovereignty requires independent technical assessment.
Isolation was incremental, not deliberate
Each of the seven systems was, at the time of its selection, a reasonable and often best-available choice for its function. medMatrix was the correct HIS. DIVAS-PRO was the appropriate PACS. SYNLAB was the contracted lab partner. The problem was never the quality of any individual system — it was the absence of a unifying layer that could relate them to each other. Isolation accumulated gradually, one procurement decision at a time. The result was seven islands connected only by human effort.
Human approval as intentional design
The deployment retained human approval at every high-stakes decision point: billing validation, payroll confirmation, financial close sign-off, duplicate patient reconciliation. This was not a limitation of the architecture — it was a design requirement. In healthcare and finance, institutional control is not a friction to be eliminated. Automation improves decision quality by removing data burden. It does not and should not remove the decision itself.
Seven systems. Seven islands. One definitional layer made them a single institution for the first time.
Stathon deployment conclusion
Forward roadmap.
Athena clinical quality intelligence
Pattern recognition across care episodes, outcome modelling by specialty, and structured decision support for treatment protocol selection.
Cross-unit capacity optimization
Network-wide scheduling intelligence using the unified capacity state model established in Phase 1. Reduces idle slots and cross-unit transfer friction.
Vault tier assessment
Evaluation of sovereign data posture requirements: access governance hardening, audit-grade integrity layer, and EESZT compliance depth review.
Deployment record.
Stathon · Definitional Infrastructure Company. Client identity anonymized at the institution's request. Operational metrics and system references reflect the actual deployment environment.