PHCSG Annual Conference 2021: Doing IT better

I attended as a representative of BCS Health & Care, with my own perspective of being responsible for the implementation of a shared care record service for an English ICS. GP data is an important component of this shared care record service, and I was keen to improve my insight into the culture, capabilities and perspectives of primary care. My recent experiences of multiple NHS integration initiatives had already highlighted the need to improve data management practice.

The breaking of bread, a meeting of minds

The conference dinner the previous night had done its job. Old friends reunited, new acquaintances made, batons passed, and prizes awarded. The evening was steeped in fond reminiscences combined of past achievements, challenges and frustrations, as well as a keen anticipation of what was to come.

The delivery of modern healthcare has become increasingly dependent on the use of digital technology. The shift of paradigm from paper to digital records is profound and has far reaching implications, many of which are only recently becoming apparent. The term ‘digital transformation’ can contribute to the confusion since the real transformative effect lies not in the tools that are used (software and storage) but in the outcome of their use (the data) and the way that we choose to use them (the processes).

A thought-provoking keynote speech

The conference programme began with Tim Ballard, who spoke about the role of the CQC and highlighted the creation of the Multi-Agency Advisory Service for AI and Data-driven Technology. This important initiative is intended to provide a one stop shop for guidance on regulation, evaluation, and adoption of AI solutions.

The project will be overseen and funded by NHSx, which sets the strategy for the NHS's digital transformation. NHSx will work with various national bodies including the National Institute for Health and Care Excellence (NICE), the Care Quality Commission (CQC), the Health Research Authority, and the Medicines and Healthcare Products Regulatory Agency (MHRA) to ensure new innovations meet high standards regarding safety, effectiveness and data governance.

A heath service of the future

Artificial intelligence and other computational approaches can inform, characterise, and guide the treatment and management of disease and well-being. The development of analytic capabilities to exploit the vast potential of this new world of big data must, however, be supported by a mature approach to data management. In this context, data management should not simply be understood to be “what a Data Manager does”. Data management is everyone’s responsibility.

Marcus Baw, a locum GP and Emergency Physician from Yorkshire, gave us a practical example of the potential benefits of open API-driven approaches to support delivery of clinical tools. He eloquently demonstrated the importance of close clinical involvement in the design and development of software tools as well as the opportunities offered by modern software development tools and approaches that reduce the technical challenges. His presentation reinforced the message from the previous speaker, demonstrating the immediate need for an advisory service and frameworks for implementation that promote standards and best practice.

The management and use of data

Data is not an asset like a car or even oil, in that it is not depletable. On the contrary, each time data passes through someone's hands it generates more data, otherwise known as metadata, where this metadata provides context, structural and descriptive information about the data and associated processes as well as providing a range of qualitative information. The challenge that we face with data (and metadata) therefore comes with understanding how it can and will be used, along with a multitude of quality expectations.

For you

Be part of something bigger, join BCS, The Chartered Institute for IT.

Philip Scott, a Reader in Health Informatics at Portsmouth University, explained the relevance of Computable Biomedical Knowledge and its role in the establishment of Learning Health Systems. He understandably focused on the opportunities to represent knowledge derived from clinical research and informatics in a form that could provide computable content to inform and drive improvements in patient care.

However, another aspect of this challenge relates to the slightly more prosaic problem of capturing knowledge associated with existing business processes and projects in a standardised and computable form. Anyone who has experience of attempting to bring about change in the NHS will be all too familiar with the opaque and heterogeneous nature of operational and business knowledge.

How can we reuse this data?

This concept of ‘data use’ is fundamental. In fact, more accurately, we really need to consider the distinctions between ‘data use’, ‘data reuse’, and ‘data repurposing’. This highlights the importance of the underlying design of processes that lead to the generation or collection of data, otherwise known as the ‘data lifecycle’.

Different characterisations have been proposed to describe this data life cycle. Its phases would typically include:

• Planning: What data needs to be collected and at what point?
• Design and enable: What data collection instruments will be used and what are the applicable standards and data quality rules?
• Creation: What processes will be in place to support the collection of data? What is the context within which data will be collected and by whom?
• Store and maintain: Where will the data be stored and how will this be managed? What safeguards need to be in place to assure the confidentiality, integrity, and availability of the data?
• Use: What is the intended use of the data?
• Enhancement: What opportunities exist to enhance the data by addition of metadata, computational enhancement, or linkage with other data?
• Disposal: Under what circumstances does the data need to be destroyed, how will this be done and by whom? What records need to be kept relating to the disposal?

Many speakers, including Julia Hippisley-Cox (winners of the John Perry Prize for her work on COVID-19 risk stratification and prioritisation), Robbie Foy (ASPIRE Project) and Julian Brown (who established the Eclipse analytic system) clearly demonstrated the challenges of the current levels of data governance maturity across the NHS.

Consideration of the data life cycle provides some important insights:

1. Data management must be executed with an understanding of how data is produced, or obtained, as well as how data is used. It costs money to produce data. Data is valuable only when it is consumed or applied.
2. Data quality management is central to data management. Low quality data represents cost and risk, rather than value. Data is often created as a by-product of operational processes. Data quality can be affected at any point during the data life cycle and therefore data quality must be actively managed accordingly.
3. Metadata is a form of data, and organisations rely on it to manage other data. Therefore, metadata quality must be managed in the same way as the quality of other data.
4. Data management also includes ensuring that data is secure, and the risks associated with data are mitigated. Data that requires protection must be protected throughout its life cycle, from planning to disposal.
5. Data management should focus on the most important data. Organisations produce a lot of data, a large proportion of which is never actually used. Trying to manage every piece of data is not possible. Life cycle management requires focusing on organisations’ most critical data and minimising data that is redundant, obsolete, or trivial.

One size does not fit all

Both Ian McNichol and David Stables offered an optimistic perspective on the future. They have both spent many years considering the implications of various modelling approaches to health data and have concluded that on size does not fit all. We are inevitably going to have to accept that no single information model can be used for the representation of health data. Our management approach needs to reflect this truth. If we can’t standardise the models that we use for the data, perhaps we can standardise how we describe them; by standardising the metadata.

Public trust is essential. Very real and justifiable concerns about the approach taken by the NHS in handling and sharing health data were raised during the final panel discussion. Historical attempts to enable greater data sharing have often faltered due to inadequate consultation combined with opaque processes. Improved professional standards in data management, underpinned by clear and effective data governance is surely an essential enabler.