Health interoperability has been a stated priority for decades, backed by standards, regulation and substantial spending, and systems still fail to exchange usable information.
The persistent explanation is technical. It is mostly not.
The standards are not the obstacle
Standards for exchanging health data exist, are mature, and are widely implemented. Two systems can generally be made to pass structured messages to each other.
What arrives at the other end is the problem.
Two systems can exchange a message perfectly and still fail, because the receiving system does not mean the same thing by the field it just received.
The gap sits between technical exchange and semantic interoperability: whether the meaning survives the journey. That requires someone who understands what the clinical concept is, how each system represents it, and what is lost in translation.
That person needs clinical knowledge, terminology knowledge and technical knowledge simultaneously, which is the same two-careers problem described in the bioinformatics article and it has the same solution.
What the direction covers
The scope: electronic records, telehealth, interoperability, medical data and the digital transformation of care.
Four capabilities.
Clinical information systems. What an electronic record actually is, how it is structured, and how it is configured for a given organisation.
Terminology and coding. The controlled vocabularies used to represent clinical concepts, and the mapping between them.
Interoperability standards and architecture. How systems exchange data and what the architectural choices imply.
Clinical workflow and implementation. How care actually happens, and how a system either supports it or obstructs it.
Terminology, which is where it actually breaks
The least glamorous and most decisive part of this direction.
Clinical concepts have to be represented in a controlled vocabulary for a computer to do anything reliable with them. Several such vocabularies exist, built for different purposes: clinical recording, billing, laboratory results, medicines, mortality statistics.
Four problems follow.
The same concept is coded differently in different systems, so combining data requires mapping, and mapping is a judgement.
Mappings are lossy. A term in one vocabulary may have no exact equivalent in another, only something broader or narrower. Every map introduces a small distortion, and distortions accumulate across a chain of systems.
Coding practice varies. Two clinicians documenting the same situation may code it differently, both defensibly. Analyses that assume coding is objective will find differences between organisations that are documentation differences rather than clinical ones.
Billing codes are not clinical codes. Where coding drives payment, the codes encode financial incentive as well as clinical fact. Treating billing data as a clinical record is one of the most common errors in health analytics, and it is not correctable by better statistics.
Nobody plans a workforce around terminology specialists, and the organisations that have them find their data is usable.
Where this sits in the domain
Health informatics and digital health is the sixth of ten directions in Astra Trainer's medicine and healthtech domain. It sits underneath medical AI and imaging and precision medicine, because neither works on data whose meaning has not survived, and alongside healthcare systems, regulation and operations.
For partners staffing this from the technical side, the AI, data and computing domain runs eight directions including software engineering, data science and analytics, cloud computing and DevOps, and IT systems and networks. The pairing of clinical informatics with those is the shape most implementation teams actually need. You can see the ten directions here.
Why clinical systems get worked around
Every health organisation has them: the spreadsheet beside the record system, the paper list on the ward, the group chat carrying handover, the field used for something other than its label.
These are usually treated as compliance failures. They are better read as diagnostic information about the system.
The workaround exists because the system does not support the work. Clinicians are not avoiding the system for fun, they are getting a job done that the system makes slow or impossible.
Workarounds carry clinical risk. Information held outside the record is invisible to everyone else, missing from audit, and lost at handover.
Banning them without fixing the cause moves them somewhere less visible.
Understanding this requires people who can watch a clinical process, see why the system fails it, and translate that into a configuration or design change. That is the core informatics skill, and it is not a technical skill or a clinical skill on its own.
The same applies to documentation burden. Time spent on documentation is time not spent on patients, and it is a recognised contributor to clinician burnout. Systems designed without workflow understanding add to it, usually by asking for data at the moment that is most disruptive to collect it.
The roles, named
Clinical informatics specialists. The bridge role. Chronically short everywhere.
Clinical systems analysts and configuration staff. Building and maintaining how the record works in a specific organisation.
Terminology and data standards specialists. Small, decisive, almost never planned for.
Integration engineers. Building and maintaining interfaces between systems.
Health data engineers. Making clinical data usable downstream.
Digital health product managers. Making design decisions with clinical consequences.
Telehealth operations staff. A function that expanded rapidly and is frequently staffed without structured training.
Clinical safety officers for digital systems, a role that carries formal responsibility in several jurisdictions.
Who can be trained into it
Clinicians moving into informatics. The highest-value conversion available and usually self-taught, which means it happens slowly and inconsistently. They hold the clinical and workflow understanding, which is the part that cannot be acquired from documentation, and need the technical and terminology layers.
Organisations that build a deliberate route for this get the bridge role they cannot hire. Those that leave it to chance get a small number of enthusiasts and a persistent gap.
IT staff already working in healthcare. Know the systems and the organisation and frequently lack the clinical grounding to make good design judgements.
Clinical coders. Already terminology specialists, and an underused route into wider informatics.
Software engineers entering healthtech. Need the clinical layer, which is the anatomy and physiology and pathophysiology directions in this domain.
Health service managers. Understand process and governance and need the technical vocabulary.
Medical records and administration staff. Know how data actually gets recorded, which is different from how it is supposed to be recorded.
Clinical safety and data protection. Digital clinical systems carry clinical risk, and in several jurisdictions their development and deployment require formal clinical safety assessment by a suitably qualified clinician. Health data is subject to strict protection law with legal consequences for breach. Training builds capability and awareness of where these obligations apply. It does not constitute clinical safety qualification, information governance certification or authority to approve a system for clinical use.
What to take from this
The standards are mature. The scarcity is people who hold clinical meaning, terminology and technology at once.
Terminology is where meaning is lost, mappings are lossy, coding practice varies, and billing data is not a clinical record.
Workarounds are diagnostic information about a system rather than a discipline problem, and banning them without fixing the cause makes them less visible.
Clinicians moving into informatics are the highest-value conversion and the one most often left to happen by accident.
And clinical coders are already terminology specialists, which makes them an unusually short route into the role nobody can fill.
Why does interoperability keep failing?
Not because standards are missing. Because preserving clinical meaning across systems requires people who understand the concept, each system's representation of it, and what is lost in translation.
What makes terminology so important?
Mappings between vocabularies are lossy, coding practice varies between clinicians and organisations, and billing codes encode financial incentive as well as clinical fact. None of that is fixable downstream with better statistics.
Why do clinicians work around systems?
Because the system does not support the work. Workarounds are diagnostic information, and they carry real clinical risk because the information they hold is invisible to everyone else.
Who makes the best clinical informatics staff?
Clinicians who move into it, because workflow understanding is the part that cannot be learned from documentation. Clinical coders are a close second, since they already hold the terminology layer.
Where does this fit in the domain?
Sixth of ten directions in Astra Trainer's medicine and healthtech domain, sitting underneath medical AI and precision medicine. You can see them here.