This paper discusses the importance as well as the practical issues that can be expected while working towards the FAIRification of Electronic Health Records (EHRs). Electronic health records are medical records of a patient that document their illness history as well as other relevant personal information. Patients interact with health care professionals in a multitude of capacities and thus, there is a stream of medical data available at any given time. So far, electronic health records have not been utilized to their full potential, which in part is attributable to the many dimensions the health care profession is split into, and the lack of interconnectivity between disciplines. Having patient data either in one source or alternatively interconnected fulfilling the principles of FAIR data, i.e. findable, accessible, interoperable and reusable, will be of immense benefit to both the healthcare system as well as for the patient. Streamlined FAIR compliance electronic health records will mean all the pertinent data should be linked or connected for better findability and accessibility to make it reusable and interoperable. In this position paper we look at the principles of FAIR data, identify priorities in relation to making data reusable, as well as recognizing points of action and complications in delivering FAIRified EHRs.
CEUR ceur-ws.org
An Electronic Health Record (EHR) is a longitudinal collection of electronic health information about a patient. The ubiquity of EHR systems has transformed the health care landscape over several decades. Yet, even as improved patient care and cost savings have begun to emerge, significant usability impediments have been documented. These include the duplication and fragmentation of EHRs that hampers interoperability eforts and impact on patients’ safety. Facilitating better access to and sharing of structured and unstructured health data is crucial to ensure greater accessibility, availability, and afordability of healthcare. It will also stimulate innovation in health and care for better treatment and outcomes and foster innovative solutions that make use of digital technologies.
SeWebMeDA-2022: 5th International Workshop on Semantic Web solutions for large-scale biomedical data analytics, https://www.rcsi.com/people/profile/alihasnain (A. Hasnain) CEUR Workshop Proceedings
© 2022 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
The collection, access, use and re-use of health data in healthcare poses specific challenges in ifnding the right balance between measures that facilitate data sharing while preserving the interests and rights of individuals, including their personal data protection.
It is important to highlight that unstructured data represent the 80% of data contained in
EHRs [
Digital health technology and data pose a critical opportunity and challenge for researching, practicing, and experiencing healthcare, promoting public health policies, and changing the way in which medicine is understood. However, digital health data is currently fragmented and dispersed in diferent and non-homogeneous repositories.
Standardization and common structures for facilitating interoperability and re-use of these
data for clinical practice and for research and innovation is a challenge, especially when it
comes to exploiting unstructured information in a secure way and making this information
interoperable [
Personal data has also emerged in the economic literature as the world economy’s new asset
class. These data ofer a cost-efective and technically feasible pathway to personalized medicine;
create wealth and value, attract private and public investment and provide a return to the whole
society by means of better and more eficient prevention and treatment management protocols.
However, data assetization and exploitation (including its analysis and synthesis) needs further
evidence, developments and research into capitalisation, techno-scientific applications and
socio-cultural and socio-technical identification of resources and possibilities, including legal
and structural facilitators and barriers [
Exchange of medical information has been associated with benefits for patients, health delivery systems and the economy. In fact, real-time health information systems, integrating all relevant information on a patient(s) and their healthcare process, can substantially improve coordinated care, patient safety, quality, and eficiency.
For this reason, it is critical to improve EHRs interoperability to achieve their full potential, that is, the ability of health information systems to work together within and across organizational boundaries to advance the efective delivery of healthcare. While electronic health records (EHRs) are of immense value in healthcare and medicine their potential has not been fully exploited due to fragmentation of the healthcare sector. This is due to the lack of interconnectivity and interoperability between diferent EHR systems.
To this purpose, FAIR data principles [
An argument made is that the research community has not always suficiently acknowledged
the inherent value related to the production of analytical datasets to support research initiatives.
One of the recommendations from the EU commission report (step 3, recommendations 12
and 13) [
In this position paper, we advocate the importance of FAIRification led interconnectivity of EHRs by presenting two motivational scenarios that arise on a daily basis in clinical settings e.g. accidents and emergency units that can be supported using well-connected interoperable EHRs at least at the national level.
Electronic Health Records (EHRs) are primarily known as a digital collection of medical information about an individual patient that includes but is not limited to the information about a patient’s health history, diagnoses, prescription medicines, diagnostic and monitoring tests, known allergies and immunizations. It can also include healthcare encounters and referrals to specialists and ambulatory/outpatient encounters in hospitals.
Exchange of medical information has been shown to be associated with higher quality and safer care for patients. It also has clear benefits for healthcare organizations- providers of care in community, outpatient and hospital settings as well as health insurers and the health sector more broadly. EHRs help healthcare providers better manage care for patients and provide better health care by: • Providing accurate, up-to-date, and complete information about patients at the point of care at specific time • Enabling quick access to patient records for more coordinated and eficient care • Exploiting the potential of sharing electronic information with patients and other clinicians in a secure manner • Helping healthcare providers more efectively diagnose patients, reduce medical or human errors while providing better and safer care • Improving patient and health care provider interaction and communication, as well as health care convenience • Enabling safer, eficient and more appropriate prescribing of drugs
The sensitive nature of healthcare data creates fragmented, siloed and mostly private repositories of data by design. Although online portals have increased patients’ ability to view their records, the structuring of this access is limited by design. A minority of systems allow direct patient access to data and healthcare results. The degree to which data should be shared directly with patients is still being debated within the healthcare community due to ever pressing requirements of safety and security of such sensitive data in the digital space.
On the other hand, holding and oversight of health data has largely been in the hands of multinational technology organizations. This data collection, acquisition, storage and holding of health data has evolved around the concept and motivation of the personal Electronic Health Record data that supports collection of data relating to individuals’ exercise and lifestyle as well as physiological measurements such as pulse rate, rhythm and variability. Initiatives have emerged to develop these person-based EHRs, the most notable examples being Microsoft’s HealthVault1 and Google Health Google Health,2. These technology innovations failed to generate traction because of poor user engagement and concerns about security of personal data. Most critically there were significant limitations in terms of integration and interoperability with EHRs. The overlap between health data relating to diet, exercise and lifestyle activity with more traditional medical data including morbidity conditions, medication codes, and healthcare utilization (diagnostic testing, monitoring of disease, and contacts with health professionals in primary and secondary care) has never been fully resolved. This may well be explained by the confidential nature of medical data and the fact that prescription drugs require input and understanding from health professionals (doctors and pharmacists) who are concerned with prescribing, monitoring and dispensing of medicines needing specialist knowledge and regulatory oversight.
In 2018 the European Commission expert working group on FAIR data produced a detailed
report about how to promote the concept of FAIR data to support more open science in the
European Union [
other research-related materials onto the web and hope that the
motivation and skill of the potential user would be suficient to enable reuse.” [
FAIR is a set of principles that collectively describe requirements to develop a FAIR ecosystem
supporting digital data reuse [
The digital object ensures that the FAIR resource is uniquely identified, semantically described using open technical standards, and appropriately shared according to defined access controls based on licensing information.
An important consideration as highlighted in the EU report is the need for development of
distributed FAIR ecosystems by storing collections of FAIR digital objects across distributed
FAIR repositories that belong to searchable and uniquely defined organizations defined in
FAIR registries [
“Data Federations ofer a means to establish agreements between repositories or registries to
carry out certain tasks collaboratively and therefore will be essential to this distributed system.
Data will increasingly remain at various locations for reasons such as the expense of copying
data or because of legal or ethical restrictions. Distributed queries, managed by brokering
software, will be used to virtually integrate data. The need for such distributed analysis across
multiple data sets is one of the major drivers and use cases for FAIR data: it requires metadata
to find the data resources, protocols to access them, agreed specifications such that the data can
interoperate and rich provenance information so that the data can be reused with confidence”
[
An important distinction to be made is that FAIR data does not necessarily imply ‘Open data’. As part of the metadata associated with FAIR datasets, it is envisioned that licensing will also be described which may allow full access or restricted access to only the metadata descriptions rather than the actual raw datasets themselves where it is considered that the data is sensitive or should be restricted. This also implies that a FAIR infrastructure must allow for authentication mechanisms that define how data can be shared across diferent FAIR data repositories, groups and disciplines while respecting the EU legislation across borders as well (such as General Data Protection Regulations). Whilst open data is considered desirable the EU envisions that data should be ‘as open as possible, as closed as necessary’.
Integration of diferent patient repositories as well as electronic health records add value to the case that such integrations could make healthcare safer for healthcare professionals as well as patients. Consideration of FAIR data principles can be a foundational step for the digitization and integration of EHRs that are more findable, accessible, interoperable and reusable for better accessing clinical scenarios. The following practical real world scenarios show the motivation behind the FAIRification led integration of electronic health records.
Motivating Scenario 1 Susy is a patient who is involved in a road trafic accident and presents to a hospital to be seen by the doctor. Due to the nature of injuries sustained, the hospital orders, among other tests, a CT scan of the brain. The patient has the CT scan but is waiting many hours to be seen by the doctor. She eventually leaves without receiving the results of the tests and presents to another hospital. Since no records are available and the patient is unable to tell the doctor at the second hospital what tests she got in the first place, she has another CT scan at the second hospital. The patient is eventually discharged but has now had two CT scans of the head. The radiation exposure is significant. One in every 1800 CT scans leads to one excess cancer. This patient has now had exposure at two facilities due to data inaccessibility between the two facilities. If data is findable and accessible as proposed by the FAIRify principles, better clinical decisions such as limiting the amount of harmful radiation a patient is exposed to, can be made, and repetitive testing can be avoided. This will lead to better patient outcomes as well as reducing cost of patient care.
In a nutshell the accessibility of patient data could have been achieved through FAIRification led digitization of EHRs
Motivating Scenario 2 In a second scenario, a patient has been discharged from hospital following a prolonged in-patient stay. Upon discharge, a discharge summary has been completed by the hospital many days following discharge. The patient had multiple investigations in the hospital including blood tests, electrocardiogram (ECG) and radiological investigations, but none of the results are available to the GP who is the primary physician of the patient in the community. Some of the doses of medications have been adjusted, and needs referrals to multiple outpatient specialties for continuity of care, but as the GP is unaware due to a delay in the discharge summary and recommendations getting to them, there is a delay ultimately in the patient getting optimal care. If data is findable and accessible as proposed by the FAIRify principles to the corresponding GP (obviously in a secure and closed environment), better clinical decisions as well as timely diagnosis and recommended medication modifications would have been possible by the GP. This will also lead to better patient outcomes as well as reducing cost of patient care.
Motivating scenario 3 FAIRifying EHR will play a significant role in integrating research (interoperable data) along multiple platforms. If an interesting clinical scenario such as a rare cancer or genetic condition is observed in facility A, for instance, with EHR being accessible across a network of healthcare, it will be easy and feasible for another institution to not only gain access to the information due to accessibility of data, but also compare the findings and consolidate their own research with research from other sites. EHR can also then be used to recruit for clinical studies, following trends across institutions.
Accessibility of patient data available in hospital in-patient and its interoperability/ connectivity (in a protected, secured and closed environment) with the EHRs available at GP practices could be supported through FAIRification led digitization of EHRs.
The implementation of FAIR data principles are described in the EC report “Turning FAIR into
reality” [
Based on these priorities, any integrated healthcare data, research outputs, or results can be
subject to a FAIRification process [
This process of FAIRifying the healthcare data, (more precisely Electronic Health Records) and the implementation of technical components needed to support the process of FAIRification are presented as follow but were originally described in the recommendations of the GO FAIR initiative (Figure 1). The seven distinct points to consider as follows: 1. Creation of a FAIR provider registry to uniquely identify digital objects made available by the data provider (e.g., healthcare datasets, publications) 2. Creation of FAIR digital objects (FDOs) from EHR system extracts 3. Creation of local FAIR data repositories (FDRs) for digital objects hosted by each data provider 4. Deployment of FDOs to FDRs. 5. Definition of access policies and licenses to control data sharing and wider access to
FDRs & FDOs 6. Assessment of FDRs based on FAIR maturity models, certification standards and FAIR metrics. 7. Development and deployment of supporting tools and services to allow search, browsing, and controlled access to FDOs between providers and third parties.
It is clear from this description of the FAIRification process that the technical implementation
of the FAIR concept in its entirety and at scale requires several separate software components
that collectively work together to deliver FAIRificiation of data in practice. These important
sets of individual but integrated software components are required to implement FAIR data
objects and FAIR data repositories. We propose three most important set of tools namely 1) FAIR
Data Object Annotation Tool, 2) FAIR Data Object Manager and 3) Local FAIR Data Repository
Manager as described below:
• FDO Annotation Tool: A software that will allow health data providers to describe
their FDOs using metadata tags in a standardized way using accepted ontologies such
as use of appropriate clinical terminologies to provide standardized semantic
interoperability and coding for FAIR data (most notable examples but not limited to these are
ICD [
The pictorial depiction of aforementioned tools, software components and the way these individual components are connected can be seen in Figure 1 below.
It is evident that the delivery of a FAIR data ecosystem that supports sharing of EHR data is therefore not simply a case of adopting health interoperability standards. It must also be supported by a broader ecosystem that uniquely identifies the EHR resources available, makes them searchable, discoverable, and most importantly controls access to only those third parties who are authorized to use those resources.
In the multiple clinical scenarios we have mentioned in this paper, it can be deduced that EHRs, if accessible and interoperable across healthcare modalities, can benefit immensely any given healthcare system in terms of cost reduction by avoiding multiple testing and multiple presentations, as well as by enhancing patient safety. Data collected from example studies of rare genetic disorders or clinical presentations can be consolidated from more than one location and be built upon for clinical studies with more ease.
The original focus for developing FAIR data standards and a broader FAIR ecosystem was
cantered around the sharing and transparency of research-related datasets[
On the contrary, the nature of EHR data is fundamentally diferent. EHR data typically have very wide coverage of collected data. This data may potentially be captured to support decision making across a much wider spectrum of clinical care including phenotypic data, diagnostics, therapeutics, imaging, and genetics. In any discussion about FAIRifying EHR data we therefore need to be clear about whether it is feasible to FAIRify an entire complex EHR in its entirety, or whether several defined subsets of that EHR data are being constructed and FAIRified separately. The data complexity of a typical EHR system suggests that a more manageable approach would be to extract several data subsets from an entire EHR that can be FAIRified to support the defined needs of research or real-world-evidence creation. In the context of FAIR data, we potentially therefore have multiple FAIR datasets created from a single data source that would appear to have some sort of relationship or linkage between them.
EHR data collected in frontline clinical care is also highly dynamic with changes to the underlying data occurring frequently during a single day. The creation of dataset extracts as part of a FAIRification process is therefore only reflecting a point-in-time snapshot of a defined subset of the overall EHR data. If multiple FAIR datasets are required to reflect the data complexity of the EHR as mentioned previously then it only makes sense to create them all at the same point in time to maintain integrity between the diferent related sets of data.
A final distinction needs to be made between research datasets and EHR systems. The data
captured as part of the conduct of research studies is typically subject to ethical approval and
patient consent to gather such data. The ethical constraints under which EHR data that is
captured for the primary purposes of routine clinical care may not be so explicitly defined.
Under General Data Protection Regulation (GPDR) [
These distinctions in complexity, dynamic nature, and ethical approval relating to EHR data
suggest that the FAIR data concept was not originally defined to support the concept of sharing
of data from frontline clinical care systems in its current form. The concept of ‘Findability’ of
data becomes problematic in this context where each data resource should have a uniquely
defined and unchanging identifier that defines it as a uniquely identifiable and shared data
resource. It is not clear if multiple EHR extracts should have a single identifier and whether
that identifier should dynamically change with each snapshot of data that is extracted from the
EHR system [
One potential solution, according to Stein et al, [
The FAIR data concept is still evolving and is now being more universally adopted in the context
of sharing data generated from other areas beyond static research datasets. This has been noted
as being problematic while indicating a gap in the context of applying FAIR data principles to
more dynamic and larger (perhaps linked) datasets that collectively expose EHR data for the
purposes of real world evidence more generally [
Authors also believe that this will emerge and become clearer as expert groups from the EU
have been established to measure adherence to and examine alternative metrics some of which
may be appropriate for measuring compliance with FAIR data principles [
The expert group stated: “A major additional challenge in the data domain is the adoption of
a new set of metrics to assess FAIRness, i.e., compliance with the FAIR principles. We propose
the following as a basic minimum standard: discovery metadata, persistent identifiers and
access to the data or metadata”. It will be important to standardize FAIR metrics globally and to
coordinate initiatives and some are under way to develop FAIR maturity models or assessment
tools [
FAIR data is a set of principles that provides the guidelines and recommendation for data being ifndable, accessible, interoperable, and reusable. Applying FAIR data principles to electronic health records is an evolving process. The complex and dynamic nature of medical data, as well as the ethical complexities that do not critically apply to records in any other field, has demonstrated that more dimensions and associated metrics will be needed to be applied to electronic records if they will be interoperable and accessible. Nevertheless, taking the timely decision the EU has already set up committees to address the feasibility, possibilities and complications of the same which is a judicious step in the right direction.