Silje Ljosland Bakke^a,b, Heather Lesliea^c ,Ian McNicoll^d
a Co-lead, Clinical Modelling Program, openEHR Foundation,
^b Information Architect, Nasjonal IKT HF, Norway
^cb CCIO, Ocean Health Systems, Australia

^d Co-Chair, openEHR, UK

Abstract

Standardisation of clinical models during the last 20 years has mostly been focused on a small set of concepts central to reporting or communication between healthcare organisations, like lab results, diagnoses and procedures. While these concepts are themselves complex, there are a huge number of additional clinical models required to represent the content required to cover the clinical scope of comprehensive electronic health records. Standardisation of clinical models for use within clinical systems have received little attention to date but are crucial to deliver clinical care and underpin decision support. Examples of critical patterns include: the complex, fractal nature of recording examination findings; the scope and detail of therapeutic precautions; managing negation/exclusion: and the varying requirements for social history recording. This workshop will discuss and explore the complexity and patterns that have been developed and published within the openEHR community, with broader applicability to other clinical modelling paradigms.

Keywords:

Informatics, Common Data Elements,

Introduction

Clinical information modelling has historically been a fragmented activity. In recent years standardisation has started to occur but focused on a small number of core concepts that are primarily used to support communication between providers and health sectors. The work required to enable this level of interoperability of clinical data has been huge. The effort required to standardise the full scope of clinical recording requirements for a complete electronic health record, reporting, querying, decision support etc had not been well explored to any significant degree – each vendor usually working from their own unique data models. Breaking down the resulting silos of data created in isolation in each vendor's applications is our next interoperability challenge.
In 2008 the openEHR community commenced collaborative development, peer review and governance of clinical information models intended to meet the clinical recording requirements for comprehensive electronic health records – to cover the complete scope of the clinical record. Clearly the scope is huge and the amount of detail required is enormous, yet until now the complexity and size of work has been largely underestimated. Eight years later there is a library of implementable openEHR clinical models known as archetypes, growing in numbers and complexity. The openEHR specifications mandated a dual-level modelling approach which provides for the ability to create a standardised archetype and constrain it for user/scenario requirements in a secondary artefact known as a template – in this way both standardisation of data patterns as well as the flexibility required for clinical variation are enabled.
The resulting openEHR archetypes reflect a large amount of work by grassroots clinicians and health informaticians. There has already been some direct collaboration with other standards development organisations and it is the desire and intent of the openEHR community that these models are not just implemented by openEHR vendors but can be shared and cross-pollinate other implementation paradigms.
In that spirit, this workshop will identify, explore and discuss some of the learnings about the patterns and complexity identified in building openEHR archetypes for some key areas of clinical information modelling.

Workshop

Workshop speakers

• Silje Ljosland Bakke RN BSN Silje is an informatician and a registered nurse, with a clinical background in surgical nursing as well as clinical research from the University Hospital of Northern Norway. She has worked in health IT in the Norwegian hospital sector since 2009, since 2015 as an information architect in the Nasjonal IKT health trust for strategic IT cooperation within the Norwegian public hospital system. She has been a leading figure in Norway's openEHR modelling, governance, and training effort since 2013, and joined Heather Leslie as Clinical Program Co-Lead at the openEHR Foundation in 2015.
• Heather Leslie MB BS, Dip.Obs (RACOG), FRACGP, FACHI Dr Heather Leslie is Chief Clinical Informatician at Ocean Health Systems and Co-lead for the openEHR Foundation's Clinical Program. Since 2004 she has guided the evolution of 'the openEHR approach' to creation of clinical content for electronic health records using archetypes, including driving development of the online Clinical Knowledge Manager (CKM) tool. She has also provided clinical modelling/clinical knowledge governance consulting services and training to many international eHealth programs & organisations - including Norway's Nasjonal IKT, NHS England, Australian Digital Health Agency, Canada's Alberta Health Services and the Ministry of Health in Brazil.

• Ian McNicoll MBChB, MSc is a former Scottish GP,  and has been involved in healthcare informatics for nearly 30 years, working with openEHR technologies for the last 12 years, most recently as founding Director of freshEHR Clinical Informatics and as Co-Chair of the openEHR Foundation Management Board. He has been involved in a number of openEHR modelling projects both at international, national and vendor-level, is an Honorary Senior Research Associate at CHIME, UCL.

• Each speaker will contribute their knowledge and experience from overseeing the development, quality assurance and governance of clinical archetypes through the openEHR online community of clinicians and other domain experts from over 80 countries.

Workshop topics

The workshop will cover a series of topics the authors have identified as particularly challenging during their own experience with modelling clinical information, including the following six topics:

Examination findings

Clinical examinations is a very large and complex area, covering anything from the visual inspection of the skin for cyanosis, via palpation or percussion of the abdomen, to the gastroscopic inspection of a duodenal gastric ulcer. In many cases a whole stack of examinations are done consecutively, by starting on the level of the whole body and "zooming" down into smaller and smaller areas in a fractal way, while in other cases the same detailed examinations are done directly and without any higher level examinations. This puts a high strain on the flexibility of information models to be used for recording clinical examination findings, and the workshop will present and discuss a proposed pattern for modelling such examination findings.

Therapeutic precautions

Certain information types are critical for clinicians to be aware of while planning a patient's care. Some examples are adverse reactions (to drugs, food, or other substances used in care or present in the environment), contraindications (to specific procedures or interventions), or precautions (about specific conditions that may impact care choices in a multitude of ways). These concepts have some common traits, but also differ in significant ways which impact their respective information models. The workshop will discuss the different concepts and their commonalities and differences.

Social history

Social history is a mixed bag of different concepts, ranging from information about social dependency and dependants, marital status and housing, through religion and education, to occupation, income and drivers licenses. These different concepts will be needed in different combinations and to different levels of detail in each clinical setting, and will therefore have to be modeled as separate consepts to be combined as required under a common heading. The workshop will discuss the different concepts, how to identify, separate and model useful concepts, and how to combine them.

Addiction

Addiction is an entire medical speciality in itself, relating to addictions ranging from nicotine and alcohol through gambling to narcotics. Many tangential concepts such as social history are important in addiction medicine, while the core concepts will in most cases be related to the history of use and consumption of addictive substances. Modelling this area comes with a series of complications, including whether or how to separate substances from methods of administration, and whether to make generic models for several substances, or specialise them for each substance or group of substances. The workshop will explore this area and discuss possible modelling patterns.

Maximum data sets vs. name/value pairs

One tried and tested way of doing information modelling is making name/value pairs, each consisting of a single data element with a corresponding name or definition. While this works well for some data, it comes in short when applied to more complex data where it's important to maintain the relationship between several data elements in a single instance. A second way of modelling which solves the problem with name/value pairs is using minimum data sets to specify the exact required data of a specific use case. This however has the drawback that the same concept invariably will be modeled in different ways and with varying levels of detail across several use cases. A third way which improves on this issue is splitting the model into discrete, reusable concepts, with the intention that the models in time should grow to be maximum datasets for their concepts. This means that the exact same model can be used across use cases even with different levels of detail, but defining the concepts and making a growable pattern places a heavier load on initial modelling efforts. The workshop will discuss the relative merits and drawbacks of each modelling pattern, and how to find a sweet spot.

Exclusion/negation of clinical concepts

When planning care, many concepts need to be excluded due to recorded findings, such as problems/diagnoses, adverse reactions, medications or family history. Exclusions/negations is a semantically difficult area, riddled with questions about such things as double negation, differentiation between the excluded concepts, differentiation between the "flavours" of exclusion, the relevance of exclusions over time, and generic vs. specific exclusions. The workshop will discuss the different problem areas and proposed solutions.

Target audience

This session is directed at students, clinicians, clinical informaticians and system implementers who are interested in collaborating across standards development organisations on standardising clinical content, in the clinical modelling problems and patterns identified by the openEHR community, and how these can be reflected in actual clinical models.

Educational goals

Participants will gain a better understanding of the modelling approach and patterns developed by the openEHR community, and how this has been carried through to relate to real-world clinical modelling challenges and experience, with learning applicable to other related projects beyond those interested in a pure-openEHR approach.

Acknowledgements

The openEHR international community, including associated national and regional communities, of clinicians, health informaticians, implementers and others have made invaluable contributions to the identification and refinement of the concepts and modelling patterns referred in this workshop.
Address for correspondence
Silje Ljosland Bakke – silje.ljosland.bakke@nasjonalikt.no