Page 23 - AIH-1-3

Artificial Intelligence in Health Optimizing EHRs to support AI

relevant regulatory frameworks. New systems are now capabilities to support life-long and person-centered
adopting advanced technologies including cloud-based care, ecosystem-wide safe data sharing through semantic
open (non-proprietary) ecosystem-wide platforms and interoperability, extensive automation of routine reporting,
openEHR’s modeling approach to improve health data secondary data use, and advanced analytics. Widespread
management. They are designed to enable plug-and-play of adoption of data standards and data governance protocols
any number of new devices and niche applications, through is expected to substantially reduce the need for data
architectural standards and frameworks like SMART-on- cleansing. Greater availability of timely, complete quality
FHIR, without losing the ability to share data. data is expected to reduce the costs of routine reporting,
27
Underpinned by open standards-based federated medical research, and other secondary data use including
CDRs, an effective separation of data and application the development, training, and use of AI. The optimization
62
becomes possible. The adoption of open standards enables of EHR data is expected to transform our AI capacity and
secure access to vendor-neutral data by compliant third- the health-care industry generally.
party applications across the whole ecosystem. Fully A set of compatible standards enabling the establishment
standardized health data can be aggregated and utilized for and maintenance of a well-connected national digital
many authorized purposes, including AI. health ecosystem needs to be mandated. The adoption
Ecosystem-wide architectural design is paramount to of data-driven digital health implementation strategies
maximize these potential benefits. Semantic interoperability is now happening in some jurisdictions, such as the UK
71
70
72
requires extensive use of ontologically structured knowledge National Health Service (NHS), Spain, Netherlands,
73
domains and ontology-driven architectures as explained Scandinavian countries, United Arab Emirates, Kingdom
64
63
74,75
in details by Rector et al. Its structure is based on the of Saudi Arabia, and Jamaica, as well as some health-
65
relationships between three resources: (1) Information care facilities. In 2019, the European Union published its
76
models representing, for example, clinical concepts; Common Semantic Strategy for Health.
(2) inference models; and (3) concept system models required Our work in the digital health space has identified an
to reliably undertake data abstraction – a process adopted to urgent need for new knowledge to be acquired regarding
reduce a concept to a set of essential elements. 64,66-68 the scientific underpinnings of health data science among
77
Changing over from the legacy systems’ data/information the health workforce. Such knowledge and skills need
exchange paradigm to knowledge sharing at decreasing to be applied to foster a data use culture to enable greater
levels of abstraction requires the adoption of a reference innovation and transformation. Only then are we in a
architecture that starts at the IT concept level (semantic position to improve the overall health system’s performance.
coordination), through the business domain concept level This review found that essentially there are three
(agreed service function level cooperation), domain level relevant technical standards that need to be considered:
62
78
(cross-domain cooperation), and up to individual context openEHR, HL7 FHIR, and the ISO 13606:2019.
27
(skills-based end-user collaboration). This architectural openEHR provides open standards for the structure,
69
79
model supports ontology/knowledge harmonization storage, and exchange of health-care information. Core
to enable interoperability between, and integration of, openEHR specifications have been adopted by ISO,
80
78
systems, standards, and solutions at any level of complexity making it a full international standard which underpins
without the demand for continuous adaptation or revisions many national programs and vendor implementations
81
of those specifications. worldwide. The ISO 13606 standard consists of five parts
Those marketing the next-generation systems have and was based on the openEHR specification, making
some difficulty gaining a foothold in this market as large these two standards highly compatible.
vendors continue to protect their lucrative business 4.3. openEHR
models unless governments intervene. Most countries
have established a national digital health framework, but openEHR represents the evolutionary result of more than
these tend not to include the establishment of a suitable 20 years of research, innovative development, testing,
national supportive infrastructure designed to optimize implementation, and evaluation undertaken by a growing
data sharing and data quality. international community.
The openEHR archetypes represent health-care
4.2. Knowledge-driven architectural models and concepts (such as blood pressure measurement, laboratory
standards results, and diagnoses) captured in clinical records and
The adoption of a knowledge-driven architectural model messages based on stable technical building blocks. Its
82
means that new-generation systems will have far greater reference model defines generic but healthcare-specific

Volume 1 Issue 3 (2024) 17 doi: 10.36922/aih.3056