The Thread of Ariadne
Sovereignty, Connection, and the Labyrinth of European Healthcare Data
Abstract
In Crete, there was a labyrinth. No one who entered it ever returned — not because the Minotaur was invincible, but because the corridors were. The beast could be slain with courage; the labyrinth could only be defeated with a thread. Today, European healthcare is that labyrinth. Twenty-seven Member States, each with its own corridors of data, its own walls of regulation, its own dead ends of incompatible systems. The patient walks through them alone, unseen, their clinical history scattered across chambers that do not connect. Meanwhile, two civilizations — one in Washington, one in Beijing — are building something else entirely: not labyrinths, but highways. Centralized, vast, and fast. China deploys 42 AI doctors across a virtual hospital incubated at Tsinghua. The United States trains models on datasets that European regulations have made inaccessible to Europe itself. And Europe? Europe debates. Europe fragments. Europe pilots. This essay is not a lament. It is a thread. It follows the path from the European Health Data Space to the GLP-1 revolution, from anonymized computer vision in home care to the digital biomarkers of neurodegeneration, from Real-World Evidence to the concept of the adaptive patient state — a continuous, multidimensional representation of health that could transform how chronic diseases are managed across the continent. The argument is simple: in a connected world, sovereignty is not the act of building higher walls. It is the act of laying down the thread that lets your citizens find their way through.
In the myth, Ariadne gave Theseus a thread — not a sword, not a map, but a simple, continuous line to follow through the labyrinth. The Minotaur could be defeated, but only if one could find the way back. European healthcare today is the labyrinth: a structure of extraordinary complexity, built over decades by twenty-seven Member States, each with its own corridors, its own architecture, its own dead ends. The clinical data of a heart failure patient in Milan does not communicate with the records of their pulmonologist in Lyon. The electronic health record, where it exists, remains a local chamber — incapable of connecting to the whole.
And yet, in 2025–2026, an unprecedented window of opportunity has opened: the entry into force of Regulation (EU) 2025/327 — the European Health Data Space — the rapid advancement of artificial intelligence applied to medicine, and a new generation of molecules that redefines the very concept of chronic therapy. This essay follows the thread. It traces a path through the labyrinth of fragmented data, sovereign platforms, chronic disease management, and the geopolitics of health AI — from neurodegenerative diseases to the GLP-1 revolution — to argue that the connection itself is the act of sovereignty.
1. The European Health Data Space: A Framework, Not Yet a System
The EHDS Regulation, published in the EU Official Journal on March 5, 2025, and entered into force on March 26, 2025, represents the first sector-specific data space of the Union. Its objective is dual: to enable citizens to access and control their electronic health data across any Member State (primary use), and to allow the reuse of such data — anonymized or pseudonymized — for research, innovation, and health policy-making (secondary use). By June 2025, each Member State was required to designate a National Digital Health Authority. By January 2026, all EHR providers must certify their systems for interoperability. Full applicability is expected by March 2029, with additional data categories operational by 2031.
However, the Regulation is a normative framework, not an operational infrastructure. Fragmentation remains the central problem. Across Europe, healthcare systems operate with radically different IT architectures — from the centrally digitized NHS to the mosaic of Italian ASL systems running on non-communicating legacy platforms. As I analyzed in The Elephants and the Monkeys, Italy's FSE is fragmented across twenty regions using different providers, standards, and data architectures. A physician in Lombardy cannot seamlessly access the health record of a patient treated in Puglia. This is not an Italian problem — it is a European one.
The European Health Data & Evidence Network (EHDEN), the largest federated health data network in Europe, has standardized over one hundred databases using the OMOP Common Data Model. But coverage remains partial. The challenge for the next three years is to build the technical infrastructure — the national data hubs — and integrate it with pan-European systems.
2. The Sovereignty Question: Who Holds the Data?
In 2025, 56.9% of the European digital health market is cloud-based, predominantly running on American hyperscalers — AWS, Azure, Google Cloud. This creates a structural dependency that the GDPR, the AI Act, and the EHDS itself attempt to mitigate but do not resolve. A sovereign platform — understood as proprietary European infrastructure, with guaranteed data residency within Union borders, transparent governance, and native interoperability with EHDS standards — is not an ideological luxury. It is an operational necessity.
The reasons are threefold. First, compliance: the EHDS provisions for high-risk data explicitly allow for sovereign data center storage. Second, trust: the TEHDAS2 consultation demonstrated that transparency about data localization is a prerequisite for citizen consent. Third, geopolitical resilience: the health data of 450 million people is a strategic asset that cannot be governed by foreign jurisdictions.
3. The Asymmetry with China: Tairex and the Centralized Investment Model
To understand the scale of Europe's missed opportunity, it is instructive to observe what is happening in China. The Chinese digital health market reached $94.9 billion in 2025, with projections of $359.9 billion by 2034 at a CAGR of 15.5%. The AI healthcare market specifically grew from $1.59 billion in 2023 to a projected $16 billion by 2028, at a CAGR of 42.5%. In January 2025, the Chinese government created the National AI Industry Investment Fund with initial capital of 60 billion yuan (approximately €8.2 billion). The State Council's development plan envisions a core AI industry exceeding one trillion yuan, with related industries surpassing ten trillion.
The emblematic case is the Tairex Agent Hospital platform, developed by the Institute for AI Industry Research at Tsinghua University. Founded in September 2024 in Wuxi, Jiangsu, Tairex has built a virtual hospital featuring 42 AI doctors across 21 clinical departments, capable of supporting the diagnosis of over 1,000 diseases. In pilot testing during Q1 2025, the platform's AI doctors demonstrated diagnostic efficiency superior to human physicians in standardized contexts. By April 2026, eight real hospitals had integrated Tairex's "real consultation room" module into their existing Hospital Information Systems, with functional trials across more than a dozen departments, primarily in internal medicine. The platform additionally enables the creation of personalized doctor-avatar agents for clinical simulation and training, generating a digital diagnostic record that evolves in parallel with the real clinician's trajectory.
The contrast with Europe is stark. European digital health funding, while growing — $3.4 billion across 182 deals in H1 2025, up 52% year-on-year — remains fragmented across Horizon Europe (€13 billion earmarked for digital activities 2025–2027, of which only €67.5 million for AI-assisted healthcare), uncoordinated national funds, and venture capital with average deal sizes of $20.5 million. Europe produces excellent research but struggles to translate it into operational infrastructure. China invests as a state; Europe invests as an archipelago.
4. Chronic Diseases: The Decisive Field of Application
Chronic diseases absorb over 70% of European healthcare spending, in a context where a deficit of 1.8 million clinicians is projected by 2030. It is in this field that an integrated digital platform generates maximum impact. The idea is not to replace the physician but to create a continuous informational layer — an adaptive digital twin of the patient's health status — that aligns in real time exercise, clinical status, ongoing therapy, and Real-World Data (RWD), thereby generating Real-World Evidence (RWE) usable by both the clinician and the regulatory system.
The EMA published its fourth report in 2026 on the integration of RWE in regulatory decision-making, documenting a progressive shift from the exclusive paradigm of the randomized controlled trial toward a hybrid model that includes structured observational data. The EHDEN network, with its federated infrastructure based on OMOP CDM, has demonstrated that multi-country analyses on harmonized data can be conducted in full GDPR compliance, without centralizing raw data. This is precisely the model that a sovereign platform should scale.
5. Neurodegenerative Diseases and Digital Biomarkers
In neurodegenerative diseases — Parkinson's, Alzheimer's, multiple sclerosis — clinical changes are often subtle, multidimensional (motor, cognitive, sleep, speech), and unfold over years. Scheduled clinical visits, months apart, capture only intermittent snapshots. Digital biomarkers — signals derived from wearable sensors, smartphones, voice analysis, and gait tracking — offer a non-invasive alternative for continuous monitoring.
The NeuroPredict platform, published in December 2025, demonstrated an IoMT (Internet of Medical Things) edge-cloud architecture that integrates commercial wearables, proprietary sensors, and cognitive evaluations, with encrypted transport, role-based access control, and identifier separation compliant with the GDPR. In the WATCH-PD study, body-worn sensors used in-clinic and a mobile application used at-home allowed monitoring of wearing-off fluctuations and dyskinetic movements in Parkinsonian patients, combining activity index and pulse rate as predictive indicators. A framework published in Nature Reviews Bioengineering in April 2026 systematized the field, identifying the need for validated composite biomarkers that connect neuroimaging data, clinical assessments, and patient-reported experiences.
A critical gap remains: despite over 90% of US-based Parkinson's patients expressing interest in new technologies, only 24% use consumer devices for disease management and just 8% use medical-grade wearables. This gap between technological capability and real-world adoption is precisely where a patient-centered platform like HugBrAIn can intervene — not by adding another device, but by creating an integrated layer that reduces friction and makes monitoring invisible to the patient while valuable to the care team.
6. Respiratory Diseases and Anonymized Computer Vision
Chronic respiratory diseases represent an area where remote monitoring and computer vision can generate a qualitative leap. COPD affects nearly 600 million people worldwide according to the GOLD Report 2024. Home management is critical: exacerbations, if intercepted early, can be treated without hospitalization, drastically reducing costs.
Integrated flexible sensor systems with AI — described in a 2026 review in Cell Reports Physical Science — can combine humidity sensors for respiratory rhythm, strain sensors to assess respiratory effort, temperature sensors for environmental compensation, and even biochemical sensors for metabolic anomaly screening through exhaled air analysis. AI compensates for the absence of operational standards in home settings, providing realistic tools for early warning, exacerbation prediction, and personalized management.
Anonymized computer vision adds a further layer. Remote photoplethysmography (rPPG) and remote ballistocardiography (rBSG) techniques allow measurement of subtle variations in skin color and chest movement through simple RGB cameras, without physical contact. Multimodal systems integrating motion tracking and rPPG signals have achieved a mean absolute error of 1.33 breaths per minute in respiratory rate estimation. For therapeutic adherence, object detection models for pill recognition — based on networks such as EfficientNetV2 and YOLO — can verify, through anonymized video, that the patient correctly administers inhaled or oral therapy, recording the administration sequence without acquiring identifiable biometric data. The key is that the system monitors the action — the correct inhaler technique, the pill intake sequence — not the person.
7. Metabolic Diseases and the GLP-1 Revolution
Obesity affects over 890 million adults globally and was associated with 3.7 million deaths in 2024. GLP-1 receptor agonists have redefined the therapeutic paradigm. Injectable semaglutide (Wegovy) demonstrated weight reductions of 15% in clinical trials; tirzepatide (Zepbound), a dual GIP/GLP-1 agonist, achieved reductions up to 22.5% in the SURMOUNT-1 trial. In September 2025, the WHO added GLP-1 therapies to its Essential Medicines List for type 2 diabetes in high-risk groups. In December 2025, the WHO issued its first global guideline for the use of GLP-1 medicines in treating obesity.
The transition from needle to pill represents a crucial step for adherence and scale. Two pivotal phase-3 trials, published in September 2025 in the New England Journal of Medicine, marked this inflection point. The ATTAIN-1 study evaluated oral orforglipron (Eli Lilly) in 3,127 adults with obesity over 72 weeks, demonstrating a mean weight loss of 11.2%, with 54.6% of patients achieving ≥10% body weight reduction. Orforglipron, approved by the FDA in April 2026 under the trade name Foundayo, requires no food restrictions and can be taken at any time of day. The OASIS-4 study tested oral semaglutide at 25 mg in 307 adults over 64 weeks, showing a mean weight loss of 13.6% — though oral semaglutide requires fasting-state ingestion with ≤120 ml of water, followed by a 30-minute wait before eating, which may challenge real-world adherence.
Real-world data reveals a more complex picture. A 2025 study published in Obesity found that real-world weight loss with semaglutide was 7.7% and with tirzepatide 12.4% at one year — roughly half the results observed in randomized trials. Persistence remains problematic: only 63% of patients initiating Wegovy or Zepbound in early 2024 remained on therapy at one year, and only 14% remained on Wegovy after three years. This gap between experimental efficacy and real-world effectiveness is precisely the space where an adaptive digital monitoring platform can intervene: tracking adherence, correlating weight loss with exercise and metabolic parameters, and adjusting multidisciplinary support in real time.
Simultaneously, RWE is revealing unexpected benefits of GLP-1s beyond metabolism. A University of Pennsylvania study (2025), based on target trial emulation using TriNetX electronic health records with 140,169 matched pairs, demonstrated reductions in major adverse cardiovascular events (MACE), major adverse kidney events (MAKE), and all-cause mortality. Emerging evidence also suggests neuroprotective potential and benefits in addiction and neurodegenerative disorders — a field that the NIDDK identified as a research priority in 2025.
8. Cardiovascular Disease and Continuous Monitoring
Cardiovascular disease remains the leading cause of death in Europe. The LEADER, SUSTAIN-6, and REWIND cardiovascular outcome trials have demonstrated the favorable cardiovascular safety profile of GLP-1RAs. But continuous monitoring changes the perspective. Wearable ECG sensors — such as those integrated in the Ametris-AliveCor partnership announced in February 2026 — enable continuous remote cardiac monitoring, detection of asymptomatic atrial fibrillation, and creation of digital endpoints: heart rate variability, gait instability, and activity patterns for heart failure patients. These data, when flowing into a platform that integrates metabolic, respiratory, and adherence parameters, construct a multidimensional patient profile that no quarterly outpatient visit could capture.
9. The Adaptive Patient State: A Unifying Concept
The connecting thread is the concept of the adaptive patient state: a dynamic, continuously updated representation of overall health status that integrates data from four axes — exercise (activity, gait, effort), clinical status (digital biomarkers, physiological parameters), therapy (adherence, dosing, administration routes), and RWE (comparative outcomes at population level). This state is not static: it updates continuously and allows the clinician to rapidly align the therapeutic intervention with the patient's actual condition.
In a patient with COPD and obesity — a frequent comorbidity — the system could integrate: respiratory rate and effort measured by chest sensors, inhaler adherence verified through computer vision, weight loss trajectory under oral GLP-1 therapy, activity level from the wearable, and comparative outcomes with similar patients in the EHDEN network. The physician, instead of navigating four disconnected clinical records, accesses a unified dashboard that flags deviations from the expected trajectory and suggests evidence-based interventions.
This is the multidisciplinary model of next-generation home care: not the patient going to the doctor, but the data going from the patient to the care team — pulmonologist, cardiologist, diabetologist, physiotherapist, nurse — and generating knowledge that improves the system for everyone.
10. From Individual to System: RWD, RWE, and European De-Fragmentation
Every patient monitored in this way generates Real-World Data. Aggregated, anonymized, and standardized in OMOP CDM, these data become the basis for Real-World Evidence that complements randomized trials and informs regulatory decisions, health technology assessments, and health policy. The EMA is already integrating RWE into drug evaluations; the EHDS Regulation explicitly enables the secondary use of health data for research and policy.
The promise is to de-fragment the European system not from the top — through harmonization decrees that take decades — but from the bottom: building a federated infrastructure where every datum collected in home care in Puglia can contribute, anonymously, to a pan-European study on the outcomes of tirzepatide therapy in patients with COPD and heart failure. This is not a theoretical exercise: it is what EHDEN has begun to do, and what a sovereign platform could scale.
11. The Stakes: Healthy Aging and Sustainability
Europe is aging. Chronic diseases multiply with age. The current model — hospital-centric, reactive, fragmented — is unsustainable. The stakes are not technological: they are the capacity of the European healthcare system to offer continuous, multidisciplinary, and personalized care to an aging population at containable costs.
China understood this and invests as a state, with a thirty-year vision — the "Healthy China 2030" plan as the frame, AI as the accelerator, and 38,000 digitized hospitals generating centralized clinical data at scale. Europe has the research, the regulatory framework, and the values — privacy, transparency, patient-centrality — to build something better. But it must decide whether it wants to be an archipelago of pilot projects or a connected continent.
This essay is a call to build. The platform is not a product: it is an infrastructure. The data is not a commodity: it is a European public good. And the chronic patient is not a cost: it is the reason the system exists.
Back to homepage