Thèse Standardisation Pivot des Système d'Information Hospitaliers et Modèles Organisationnels pour la Réplication de l'Innovation H/F Doctorat.Gouv.Fr

Établissement : Institut National Polytechnique de Toulouse École doctorale : EDMITT - Ecole Doctorale Mathématiques, Informatique et Télécommunications de Toulouse Laboratoire de recherche : IRIT : Institut de Recherche en Informatique de Toulouse Direction de la thèse : Imen MEGDICHE ORCID 0000000213318662 Début de la thèse : 2027-09-01 Date limite de candidature : 2026-11-23T23:59:59 Dans les systèmes d'information hospitaliers actuels, une part importante de l'effort d'innovation est consacrée à l'intégration des données : accès aux sources, gestion de formats locaux et non standardisés, développement d'interfaces logicielles spécifiques, respect des exigences de sécurité et mise en oeuvre de mécanismes de traçabilité. Par ailleurs, des innovations de processus souvent demandées par les directions hospitalières et les cliniciens, et nécessaires à une prise en charge optimale des patients, sont freinées ou fortement limitées faute de flux de données adéquats et par manque d'une réelle interopérabilité. La standardisation est généralement perçue comme un moyen d'assurer l'interopérabilité des systèmes et la conformité réglementaire. Cependant, avec l'entrée en vigueur récente du cadre réglementaire régissant l'Espace européen des données de santé et le déploiement du Ségur du numérique en santé, les contraintes et opportunités liées à la standardisation sont devenues des enjeux prioritaires pour les établissements de santé, tant en France qu'à l'échelle européenne. Le principal obstacle à la standardisation réside dans la recherche d'un équilibre entre stabilité et ouverture, ainsi que dans la manière dont la standardisation des données et des systèmes s'articule avec la transformation des processus, et donc avec le management de l'innovation. Un système d'information insuffisamment standardisé nécessite de reconstruire les mécanismes d'intégration à chaque nouvelle innovation organisationnelle de service. À l'inverse, un cadre trop prescriptif risque de rigidifier les pratiques et de limiter la capacité d'adaptation et d'expérimentation.

C'est dans ce contexte que s'inscrit cette thèse, qui vise à traiter deux axes de recherche majeurs : (axe 1) proposer un cadre conceptuel et méthodologique pour l'automatisation de la standardisation des données issues de systèmes d'information hospitaliers hétérogènes ; et (axe 2) évaluer la réplicabilité d'une innovation afin d'en réduire le coût marginal et de faciliter son déploiement au sein de différents établissements de santé. The FHIR (Fast Healthcare Interoperability Resources) standard has progressively matured and is now one of the main interoperability standards in the healthcare sector. However, despite this growing adoption, its implementations remain heterogeneous across healthcare institutions, national systems, and organizational contexts. Variations in implementation profiles, terminologies, and local requirements continue to limit effective interoperability [1, 2]. More recently, the work presented in [3], which focuses on data sharing through FHIR, confirms that the technical, organizational, and institutional determinants of its adoption remain insufficiently documented.
This issue has gained particular importance since the entry into force, on March 26, 2025, of the European Health Data Space (EHDS) Regulation. The regulation establishes a common legal framework for the primary use of electronic health data, particularly for healthcare delivery, as well as for its secondary use for research, innovation, and public policy development [9]. However, its entry into force marks the beginning of a transitional period, with its main obligations being progressively implemented from 2029 onward. Its implementation therefore raises major challenges concerning the harmonization of national governance frameworks, coordination with existing regulations, and the operational integration of standards such as FHIR [10].
The recent review presented in [11] provides an overview of advances in FHIR-based semantic interoperability. It classifies existing approaches into four categories of contributions: data collection; data extraction and annotation; the modeling of patient record data originating from legacy systems; and its transformation into FHIR models and terminologies [11]. The review emphasizes that terminology services remain necessary to accelerate the adoption of these techniques. It also indicates that ontology annotation and matching methods should be further automated to reduce human intervention.
In hospital settings, raw data are generally not produced in a standardized form. Transforming such data into standardized formats is therefore particularly time-consuming and complex. In France, in particular, data are described using other standards and models. Several approaches have been proposed to transform these models into FHIR. These include the FHIR Mapping Language (FML), reusable visual components for transforming CDA documents into FHIR resources [12], and toolchains such as toFHIR, which automate the migration of existing datasets, including MIMIC-IV, into compliant FHIR resources [13]. However, these studies address only part of the problem of automating the transition to a pivot model.
Furthermore, Rogers' Diffusion of Innovations theory is the framework most frequently used to analyze the adoption of digital health innovations [4]. Recent literature nevertheless emphasizes the need to adapt it to the specific characteristics of the healthcare sector, particularly clinical hierarchies, regulatory complexity, and institutional inertia. Recent empirical studies have applied this framework to the adoption of patient portals in Germany [5] and to the assessment of digital health adoption readiness in various national contexts [6]. These studies highlight the central role of certain adopter categories, particularly physicians and nursing managers, in interorganizational diffusion.
The work presented in [7] maps the categories and diffusion pathways relevant to digital health technologies. It also highlights the persistent lack of standardized measures for quantifying the adoption and sustained use of deployed innovations. The replication of an innovation from one healthcare institution to another remains insufficiently supported by methods and tools in the academic literature on health information systems, which tends to focus on single-case studies. Finally, research on the scaling up of digital health technologies explicitly calls for a better characterization of diffusion pathways across different innovation ecosystems [8].
These findings reveal two main research gaps. The first concerns the absence of a framework for determining the required level of standardization and automating [14-19] the transition toward such a framework. The second lies in the lack of methods and indicators for measuring transferability and assessing its technical, economic, and organizational effects.
(axis 1) proposing a conceptual and methodological framework for automating the standardization of data from heterogeneous hospital information systems.

(axis 2) assessing the replicability of an innovation in order to reduce its marginal cost and facilitate its deployment across different healthcare institutions. Research AXIS 1: developing a conceptual and methodological framework to automate the standardization of data originating from heterogeneous hospital information systems. The underlying research questions are as follows:
- What is the minimal set of concepts and relationships needed to represent a hospital information system independently of the specific products and technologies in use, and how can the clinical, administrative, logistical, research, security, and governance dimensions be articulated within a single pivot model without recreating conceptual silos?
- How can the transformation of heterogeneous raw data into this pivot model be automated, and how can metadata extraction from raw data be guided by, and aligned with, the pivot model?
- How can the alignment between source data models (metadata) and the pivot model itself be automated?
Research AXIS 2: assessing the replicability of an innovation in order to reduce its marginal cost and facilitate its deployment across different healthcare institutions. The underlying research questions are as follows:
- How do standardization, governance, and openness shape a hospital ecosystem's capacity to produce, experiment with, adopt, and diffuse innovations?
- Is there an optimal level of standardization - high enough to reduce integration costs and enable reuse, yet not so constraining that it stifles exploration and solution diversity?
- How can the replicability of an innovation across institutions be measured - for instance, through integration lead time, the extent of custom development required, dependence on local teams, component reuse, and the transition from prototype to real-world use?