Skip to content
Field Report · Forge Tier · Healthcare Operations

Iberian Private Healthcare Network

Iberian Peninsula · Two Autonomous CommunitiesNovember 2024 – Active

Six sites. Three HIS environments. One definitional spine for revenue cycle integrity across acquisitions.

“Until this engagement, the network’s revenue performance depended on three or four people connecting dots in their heads. The dots are now connected at the system level. Growth is no longer a question of whether we can find another set of people who can do that.”

CEO, Iberian Private Healthcare Network
Note

Client identity withheld by agreement. Deployment metrics reflect production conditions as of May 2026.

Executive Summary

A six-site Iberian private healthcare network, assembled through three transactions over thirty months, was running on three different hospital information systems and four different financial back-ends. Each site managed its own coding, denial workflow, and contract reconciliation locally. Net collection ratio varied by 5–9 percentage points between sites delivering similar specialty mix. Stathon deployed a Forge Tier definitional spine connecting all three HIS environments — SAP IS-H, Dedalus ORBIS, and Selene — into a unified event stream and a single patient entity model. One capability is in production: cross-site revenue cycle anomaly surfacing, validated against legacy controller workflows over a 90-day cohort 1 window.

Denial overturn+25–35%Cohort 1, 90-day window
Coding recovery~€340KValidated, orthopedic
Decision latency9 daysFrom quarterly cycle
Sites unified63 HIS environments
01 · The Environment

Six sites. Three HIS.

A six-site PE-backed private healthcare network operating across two Spanish autonomous communities. Approximately 1,500 staff, around €140M in revenue, and roughly 140,000 outpatient visits per year. One platform investment, two follow-on add-on transactions in 2023–2024.

Two sites carry SAP IS-H deployed by Common MS during the platform investment phase. Three sites carry a proprietary HIS in active migration toward Dedalus ORBIS — the leading EHR provider in Spain by Black Book Q2 2025 ranking. The sixth site, acquired in early 2024, runs Selene (CGM Clinical España, formerly Cerner).

Insurer EDI flows toward Sanitas, Adeslas, DKV, Mapfre, and the MUFACE mutualidad were configured per site. Each site’s revenue cycle team maintained its own denial workflow, its own coding cross-check process, and its own reconciliation against the convenio terms.

Operational systems at engagement start
HIS — Sites A & B (platform)SAP IS-H (Common MS deployment)
HIS — Sites C, D, EProprietary HIS · Dedalus ORBIS migration in progress
HIS — Site F (acquired Q1 2024)Selene (CGM Clinical España, ex-Cerner)
Imaging (PACS)Agfa Enterprise Imaging (4 sites) · Sectra (2 sites)
Laboratory (LIS)Werfen Modulab
Financial Back-EndSAP S/4HANA · MS Dynamics 365 BC · regional Spanish ERP
Insurer EDISanitas · Adeslas · DKV · Mapfre · MUFACE

Three HIS environments. Four financial back-ends. Five insurer EDI relationships. No shared definitional layer across sites.

02 · The Challenge

The dots in their heads.

2.1 The Visible Problem

Net collection variance5–9 points between similar-mix sites

Two sites delivering the same surgical mix and the same payor distribution posted net collection ratios eight points apart. The Group CFO opened the engagement with a single question: why does this gap exist, and why does it surface only in February for the prior calendar year?

Three controllers, three spreadsheetsThree different cadences

Three site controllers, each working from their own spreadsheet refreshed at their own cadence, were the network’s only mechanism for understanding revenue performance across the group. The CEO had been told this was the operational reality. He was being told it was also the architectural ceiling.

Knowledge in human memoryThree or four senior people

The dots were being connected, when connected at all, in the heads of three or four senior operators carrying the legacy of each acquired entity. Two were over fifty-five. One had given notice during the diligence phase of a planned third acquisition. The growth thesis was structurally dependent on knowledge that lived in no system.

Prior consolidation closedReported as architectural ceiling

A 2023 IT consolidation engagement had concluded that meaningful cross-site harmonization required either replacing SAP IS-H or replacing the Dedalus-migration environment. The platform investor declined for cost; clinical leadership declined for disruption. The engagement closed with a recommendation that the HIS environments could not be reconciled at the data level.

Definitional driftPer-site master data

Each site had its own definitions, its own master data, its own coding traditions, its own conventions for what counted as a billable encounter and what counted as a write-off. There was no shared definition of what a patient was, what an episode of care was, what an insurer obligation was, or what a denial event was — across sites.

Quarterly board cadenceVisibility lag of weeks

Material variances surfaced in quarterly board packs four to twelve weeks after the originating event. Insurer tariff changes, coding inconsistencies, and denial pattern shifts were observable only in arrears. The network had no mechanism for intervening inside the current operational cycle.

2.2 The Structural Gap

01
No shared patient definition across sites

A single patient who attended cardiology consultation at one site, an MRI at a second site, and a surgical procedure at a third — whose claim sequence touched three different insurer relationships — existed in three different identity systems. Inside each site, the information was complete. Across the group, it did not connect.

02
No shared episode definition

What counted as a billable encounter at one site was a write-off at another. Coding traditions had drifted in three independent directions over thirty months of acquisitions. The same orthopedic procedure, billed against the same insurer, was coded inconsistently across three sites for eleven consecutive months before any controller noticed.

03
No shared denial event class

Each site’s revenue cycle team carried its own taxonomy of denial reasons, its own escalation cadence, and its own contract reconciliation logic against the convenio terms. The same insurer behavior across sites was not recognized as the same pattern.

04
Definition gap, not analytics gap

The challenge was not a missing dashboard or a missing report. It was the absence of a definitional layer beneath the analytics — a layer in which a patient was a patient, a denial was a denial, a coded procedure was a coded procedure, regardless of which site, which HIS, or which controller had recorded it.

05
Selene as integration friction

The acquired Selene environment had no live API. Connection required a nightly batch extract in CSV format with a documented 6–8 hour data latency window. This was named explicitly in the Phase 1 closeout review: that site would not see real-time inclusion until the planned Selene-to-Dedalus migration completed in late 2026.

2.3 Integration Surface Per HIS

Independent technical assessment confirmed that no system required replacement. The clinical and operational systems remain in place; the intelligence layer that governs them is what the engagement establishes.

Integration channels established
SAP IS-HHL7 v2 feed + partial OData surface (Common MS deployment, sites A & B).
Dedalus ORBISHL7 v2 via the ORBIS interoperability stack (sites C, D, E — migration in progress, completion Q4 2026).
SeleneNightly CSV batch extract with documented 6–8 hour latency flag (site F). Real-time inclusion follows Selene-to-Dedalus migration in late 2026.
Imaging (Agfa & Sectra)DICOM 3.0 standard interface. Metadata only ingested into entity graph; image payload remains in source PACS.
Werfen ModulabHL7 v2 result feed. Analyzer integration native to source system.
Financial back-endsSAP S/4HANA via IDoc; MS Dynamics 365 BC via OData; regional Spanish ERP via structured file exchange.
Insurer EDIPer-site flows captured (Sanitas, Adeslas, DKV, Mapfre, MUFACE) and reconciled to a unified claim event schema in the entity graph.
03 · The Engagement

Three phases.

Forge Tier deployment. SAP IS-H stays where it is. The Dedalus migration proceeds on its own timeline. The Selene environment is connected, not replaced.

Phase 1

Arché: Definition & Integration Spine

Weeks 1–6
Definitional model built
Entity graph: single patient identity across all six sites, regardless of how many local MRNs
Event schema: encounter, billable episode, claim event, denial event, coding finalization, payment receipt
Compliance model: Ley 41/2002, Cataluña Ley 21/2000, AEPD subprocessor, AI Act deployer-style
Workflow state model: legitimate transitions encounter → claim → denial → overturn → payment
Core: operational systems connected
SAP IS-H: HL7 v2 + partial OData surface
Dedalus ORBIS: HL7 v2 via interoperability stack
Selene: nightly CSV batch with 6–8h latency flag
Source system fingerprinting: provenance per event
Idempotent state transitions: corrections update, do not duplicate
Event log schema: audit-grade record of every transition
Phase 2

Athena & Aegis: First Live Capability

Months 2–4
Cross-site denial surfacing

Rule-based escalation logic for known denial categories (missing prior authorization, coding-payor mismatch, eligibility lapse, contract terms outside agreed convenio bands) layered beneath a probabilistic scoring engine trained on the 18-month historical denial corpus, normalized through the entity graph.

Advisory-mode launch

First three weeks of production showed false positive rate at approximately 1.6× the operational target the controller team had agreed to accept. Two tuning cycles over the first 60 days brought the rate within the operational band.

Parallel-run discipline

First six weeks ran as parallel process: every flag reviewed against the controller team’s existing workflow. Discrepancies in either direction — flag missed by controllers, controller finding missed by flag — logged and analyzed.

Orthopedic coding finding (14 Aug 2025)

Single procedure code mismatch surfaced across three sites: same procedure, same insurer, coded inconsistently for the prior eleven months. Recovery on that single coding alignment, validated against insurer re-submission, returned approximately €340,000 in annualized revenue exposure.

Tariff variance event (Oct 2025)

Insurer convenio silently re-priced a category of outpatient consultations through a channel that did not reliably reach all six sites. Variance surfaced against contract terms within nine days of the first re-priced encounter — versus the prior six-to-ten-week Q4 close lag. ~€110K disputed and recovered against the original tariff.

Aegis: entity resolution policy

Patient identity matching is deterministic-first (DNI/NIE + DOB + structured name). Probabilistic-band cases route to a manual review queue staffed by master data team — no auto-merge below threshold. False merge / false split are first-class event classes in the audit event taxonomy: every entity merge is logged, reversible, and traceable.

Phase 3

Cohort 2 Rollout & Auto-Flag Transition

Q1–Q2 2026
Cohort 1 validation closed

90-day production validation window completed early February 2026. Advisory-mode flagging now moving toward auto-flag with human-in-the-loop on cohort 1 sites, pending AEPD prior consultation closure.

Cohort 2 in advisory mode

Sites 4–6 — including the Selene-environment acquired site — in active rollout. Same definitional spine extended; same scoring engine running advisory-only against cohort 2 data while site-specific tuning cycles complete.

DPIA & AEPD prior consultation

Auto-flag mode requires a Data Protection Impact Assessment and, given the residual risk profile under prevailing AEPD doctrine on AI-on-special-category data, a prior consultation track. Realistic clearance window: 8–14 weeks under GDPR Art. 36.

04 · Outcomes

What changed.

Validated business impact measured over the 90-day production window on cohort 1, November 2025 to February 2026. Sample basis: full cohort 1 denial population in window (n ≈ 2,400 denial events).

Denial overturn rate
Baseline
+25–35%
Cohort 1, 90-day production window
Coding recovery (orthopedic)
~€340K
Validated end-to-end against insurer re-submission
Decision latency
Quarterly
9 days
From source event to controller review
Tariff variance recovery
~€110K
Within contractual contest window
Liberated controller capacity
Manual reconciliation
1,200–1,500 hrs/yr
Redirected per cohort, not eliminated
Cohort 1 coding run-rate
€0.6–0.9M
Annualized estimate, broader specialty validation pending
Nature of the change

Decision latency on cross-site revenue cycle exceptions has compressed from quarterly board reporting to within nine days of the source event. This is a structural shift, not a speed improvement: the network is no longer reactive to last quarter’s performance — it intervenes inside the current operational cycle.

Capacity redirection

Approximately 1,200–1,500 hours per year per cohort have been redirected from manual cross-site reconciliation to contract negotiation support, insurer-specific exception strategy, and specialty-level coding discipline. The shift is not toward eliminating the capacity. It is toward redirecting it.

05 · Observations

What this case revealed.

Definitional gap, not integration gap

The prior IT consolidation engagement had concluded that meaningful cross-site harmonization required replacing one of the two HIS environments. That conclusion was correct under the assumption that integration meant moving data. The first month of this engagement produced no visible operational change — it produced the only thing without which everything else would have been a heavier reporting layer. Connecting systems moves data. Establishing what the data is about is the prerequisite that connection alone cannot supply.

Knowledge in heads as structural risk

Two of the three or four senior operators carrying the institutional integration logic were over fifty-five. One had given notice during the diligence phase of the planned third acquisition. The CFO understood, before the engagement began, that the network’s growth thesis was structurally dependent on knowledge that lived in human memory and was not represented in any system. Each subsequent acquisition will now integrate against the same definitions — not against a fresh round of human reconciliation.

Patient identity as patient safety constraint

The single-patient-entity claim across six sites is governed by a deterministic-first match policy with a manual review queue — not by model confidence alone. Both false merge (two patients incorrectly unified) and false split (one patient incorrectly fragmented) are first-class event classes in the audit event taxonomy. Every entity merge is logged, reversible, and traceable to operator and evidence basis. Patient safety, not analytical convenience, is the operative constraint on entity resolution.

Six sites. Three HIS. The categories that lived in three or four operators’ working memory now live at the system level.

Stathon deployment conclusion
06 · What Is Live, What Is Next

Forward roadmap.

In production (cohort 1)
Cross-site denial pattern detection
DRG/GRD coding inconsistency flagging
Advisory mode with 90-day validation completed
Auto-flag mode pending Q2 2026
In pilot (cohort 2)
Same anomaly surfacing, advisory-only
Site-specific tuning cycles ongoing
Cross-site patient continuity surfacing
Auto-flag follows cohort 1 by ~one quarter
On roadmap
OR utilization optimization
No-show prediction (15–20% benchmark)
Contrast media demand sensing
Network-level capacity planning
Q2 2026

Auto-flag mode (cohort 1)

Transition from advisory mode to auto-flag with human-in-the-loop on cohort 1 sites. Pending closure of AEPD prior consultation under GDPR Art. 36 (estimated 8–14 week clearance window).

Q2 2026

Cohort 2 production validation

Sites 4–6 complete site-specific tuning cycles and 90-day advisory-mode parallel-run. Auto-flag follows cohort 1 by approximately one quarter.

Late 2026

Selene-to-Dedalus migration completion

Acquired Site F transitions from nightly CSV batch (6–8 hour latency flag) to live HL7 v2 stream via the ORBIS interoperability layer. Full real-time inclusion across all six sites.

2026–2027

OR utilization & no-show prediction

Surgical capacity optimization across the four sites with OR perimeters; intent-driven scheduling intervention for outpatient no-show rates currently sitting at the high end of the 15–20% European benchmark band.

07 · Engagement Parameters

Deployment record.

Engagement TypeForge Tier · Multi-site PE-backed roll-up
Engagement StartNovember 2024
Cohort 1 Validation WindowNovember 2025 – February 2026 (90-day production)
Cohort 2 RolloutQ1–Q2 2026 (advisory mode)
Network Footprint6 sites · ~1,500 staff · ~€140M revenue · ~140,000 outpatient visits/year
Client Systems (unchanged)SAP IS-H · Dedalus ORBIS · Selene · Agfa Enterprise Imaging · Sectra · Werfen Modulab · SAP S/4HANA · MS D365 BC

Stathon · Definitional Infrastructure Company. Client identity withheld by agreement. Deployment metrics reflect production conditions as of May 2026.