Interoperability is “the ability of two or more systems or components to exchange information and to use the information that has been exchanged”.1 When it comes to interoperability in healthcare, the process grows more complex and multi-layered.
The foundational layer is technical interoperability, which is the ability for two systems to do basic data exchange using communication channels and network protocols.2
The next layer is syntactic interoperability, which is concerned with the format and structure of the data. Standard development organizations like HL7 and W3C have published standards like HL7 v2 and XML to achieve syntactic interoperability. An emerging standard from HL7 is Fast Healthcare Interoperability Resources, or FHIR, which defines 140 common healthcare concepts that can be accessed via state-of-the-art web technologies. FHIR is gaining traction and is the latest attempt to address syntactic interoperability and some basic semantic needs of health data.2
This brings us to the third part of the interoperability puzzle – semantic interoperability – which can be simply defined as a common language for unambiguously transmitting the meaning of medical concepts across a multiplicity of systems. The need for semantic interoperability is why medical terminologies, nomenclatures, and ontologies exist. Several general-purpose terminologies – like SNOMED CT®, domain specific terminologies like LOINC®, HPO (Human Phenotype Ontology), and function specific (e.g., billing) terminologies like ICD-10-CM – are available and widely used today.2
While the health care industry has made progress on the technical and syntactic interoperability fronts – largely due to the need for compliance with government regulations like Meaningful Use – it has lagged in the realm of semantic interoperability. The reasons for this are complex and are explored below.
1. The competing needs of clinician flexibility and semantic uniformity at the point of care
Healthcare professionals don’t use standard medical terminologies to communicate with their patients, nor do they use it to document information about them. For example, no physician speaks to a patient in the language of ICD-10-CM or SNOMED CT. They talk to their patients in a clinically friendly, “natural” language that is conducive to patient interaction. While this preserves clinician flexibility, it also introduces variation in how information is documented in a patient’s chart. Which begs the question: how does one balance the need for individual clinician flexibility with the industry’s need for semantic uniformity?
2. The widespread use of clinical narratives in the patient record
Patient information is documented in the EHR both in structured and unstructured ways. The structured parts of the medical record consist of specific fields in the EHR where information is documented. The unstructured parts are largely clinical narratives like the history and physical (H&P) or progress note. These clinical narratives can have a great deal of variability in how they are written and can differ from clinician to clinician. The lack of uniformity in these clinical narratives makes it hard to communicate the meaning of information from one system to another.
Fortunately, healthcare NLP (natural language processing) – a set of techniques and algorithms to automate the reading and understanding of medical text – can help with this problem. But healthcare NLP algorithms are only as good as the data they are trained on. IMO’s clinical interface terminology can help with this need. IMO’s CIT encompasses a large collection of labels – synonyms for a particular medical concept – that can be invaluable in training NLP algorithms and identifying entities of interest in clinical narratives. Additionally, once the entities are recognized, the same CIT can help automate the mapping of those entities to standard medical terminologies. This ensures semantic uniformity and high data quality in the unstructured part of the patient record.
3. Information loss as data moves out of the EHR and into other systems
Since patients seek care in multiple facilities, their data needs to travel with them. Additionally, de-identified patient data is increasingly being employed for secondary uses like life science and clinical research, which involves the patient record moving into large data warehouses. Both scenarios involve patient records leaving the EHR where the patient was originally treated.
As the data leaves the EHR boundary, information may be lost in transit, and its original clinical intent and meaning needs to be restored in the receiving system. IMO has built algorithms that leverage its CIT to restore and maintain semantic uniformity in the receiving system, thus ensuring high data healthcare quality.
The key to semantic interoperability is maintaining high data quality standards by ensuring semantic uniformity in EHRs at the point of care for both structured and unstructured information. It also involves preventing the loss of information and clinical intent as the data travels between health IT systems. While general-purpose and domain-specific standard medical terminologies exist to aid semantic interoperability, mandating their use needs to be balanced with the equally important needs of preserving clinician flexibility at the point of care; ensuring semantic uniformity in highly variable free-text clinical narratives; and addressing information loss as data is exchanged between health IT systems.
SNOMED and SNOMED CT® are registered trademarks of SNOMED International.
1 IEEE. Institute of Electrical and Electronics Engineers: IEEE standard computer dictionary: a compilation of IEEE standard computer glossaries. (1990).
2 Lehne, Moritz et al. Why digital medicine depends on interoperability. NPJ digital medicine vol. 2 79. 20 Aug. 2019, doi:10.1038/s41746-019-0158-1