Electronic health records (EHR) interoperability has attracted sustained academic and clinical interest, yet a broadly usable form remains elusive. A scoping review of 24 publications from 2014 to 2023, recently published in Health Informatics Journal, mapped how expectations, approaches and persistent problems shape progress, drawing on the Levels of Conceptual Interoperability Model (LCIM). The evidence points to four visible levels in HER: technical or foundational, syntactic or structural, semantic and process or organizational, rather than the LCIM’s seven. Across literature, semantic interoperability is shown as the central challenge because it must preserve meaning across disparate systems while supporting human and machine interpretation. Standards, ontologies and newer techniques such as natural language processing (NLP) appear promising, but lack of consensus and heterogeneous data continue to slow implementation.
Refining Interoperability Levels
The review consolidates the landscape into four practical levels aligned with, but narrower than, the LCIM. At the technical or foundational level, secure machine-to-machine exchange and connectivity dominate, supported by Internet infrastructure, cloud services and web protocols. Privacy and confidentiality are core concerns, with proposals such as blockchain aiming to protect data sharing across providers while preserving patient privacy. The syntactic or structural level adds common data formats and transfer protocols so that systems can accept and organise each other’s data, although content may still be opaque. Here, approaches reference information and document standards and interface models that define types, formats and messaging.
Must Read: Enriched Validation Boosts EHR Allostatic Load
Semantic interoperability is the focal point for usable exchange because it requires shared meaning that is both human readable and machine interpretable. Expectations at this level encompass correct, consistent and unambiguous interpretation that enables effective use, irrespective of differing local implementations. Definitions vary, and some authors distinguish between partial and full semantic interoperability, adding to conceptual complexity. Above semantics, a process or organisational level addresses coherence of workflows and cross-institutional alignment where ongoing sharing, use and communication occur within and between organisations.
Standards, Ontologies and Emerging Tools
Information models and transfer standards provide the structural backbone for interoperability. Health Level Seven (HL7) artefacts, openEHR, ISO 13606 and related models define how clinical information is organised and exchanged, while standards such as Clinical Document Architecture (CDA) and the Digital Imaging and Communications in Medicine (DICOM) specify documents and imaging exchange. Initiatives like Integrating Healthcare Enterprise promote coordinated use of these established standards. Clinical Information Models underpin modular building blocks, which in some ecosystems are called resources or archetypes, that can be combined with controlled terminologies to support semantic consistency.
Terminology and vocabulary standards including SNOMED-CT, Logical Observation Identifiers, Names and Codes (LOINC), the International Classification of Diseases (ICD) and other domain dictionaries label clinical content in a controlled manner. Coordination between such terminologies seeks to reduce duplication and improve joint use. Ontologies then map concepts across differing information models to form a common, reusable view of the medical domain, aligning structures, terminologies and rules. Rule-based and fuzzy ontologies have been proposed to accommodate imprecision and heterogeneity while preserving semantics. Beyond ontologies, knowledge graphs are suggested as a virtual layer to facilitate semantic interoperability across standards.
Given the prevalence of semi-structured and unstructured data in EHRs, artificial intelligence and NLP are positioned as complementary tools to identify semantic differences, extract meaning from free text and support logical alignment. Some authors also explore bidirectional transformation (BX) to synchronise data between specific systems where one-to-one workflows, such as e-referrals, benefit from targeted mappings rather than broad many-to-many integration.
Persistent Barriers and Practical Focus
Despite the breadth of standards, literature highlights sustained obstacles. Heterogeneity is marked across systems, data sources and local configurations, leading to partial mappings, semantic differences and ambiguity. Legacy content and unstructured notes compound normalisation challenges. Even when standards are adopted, real-world implementations may display inconsistencies that undermine semantic correctness, while governance and operational practices can limit reliability. Ontology projects face contradictions and unsatisfiable results, reflecting the difficulty of maintaining coherency across evolving models.
The review proposes a pragmatic reframing that prioritises the state of the health record over full system integration. At the foundational level, the priority is shareability, enabling exporting, packaging and forwarding regardless of format. At the structural level, the emphasis is integrability at the receiver, independent of whether structures match or exchanges are uni- or bidirectional. At the semantic level, the target is interpretability and usability, which may be achieved through rules, terminology syndication or NLP applied to structured or unstructured data. At the organisational level, maintainability of the record across processes and institutions requires alignment over time, potentially supported by regulation and policy.
Interoperability that preserves meaning is essential for effective use of EHR data, yet consensus on how to achieve it remains limited. The four-level framing clarifies expectations and directs effort toward shareability, integrability, interpretability and maintainability of records rather than comprehensive system fusion. Standards, terminologies and ontologies provide necessary foundations, while AI, NLP, knowledge graphs and targeted bidirectional mappings offer practical routes through heterogeneous, semi-structured and unstructured data. Progress will depend on aligning technical choices with workflow needs and organisational realities so that exchanged records remain usable, consistent and durable across settings.
Source: Health Informatics Journal
Image Credit: iStock
