Supplementary evidence from Dr Jeni Tennison OBE, CEO at the Open Data Institute following an evidence session on Tuesday 25 February.
I was pleased to join you to provide oral evidence on Tuesday 25th February 2020. I would like to use this opportunity to supply pointers to materials I mentioned during that evidence, and, as requested, provide an answer for Question 4 which was asked by Lord Mair.
The following materials may be of interest to the committee:
● On educating children on health data: Jeni Tennison (2017) “What would ‘data literature’ look like?”, Jeni’s Musings, https://www.jenitennison.com/2017/05/19/data-literature.html
● On the collection of characteristics of the users of digital public services: Edafe Onerhime et al (2019) “Monitoring equality in digital public services”, Open Data Institute, https://theodi.org/article/monitoring-equality-in-digital-public-services-report/
● On barriers to use of NHS data: Seb Bacon & Ben Goldacre (2019) “Barriers to Working With National Health Service England’s Open Data”, Journal of Medical Internet Research, https://www.jmir.org/2020/1/e15603/
● On data trusts and data institutions: Jeni Tennison (2020), “What do we mean by data institutions?”, Open Data Institute, https://theodi.org/article/what-do-we-mean-by-data-institutions/
Question 4 was “What new processing and analytical tools could be used to make better use of health data? In what ways could better and wider use of healthcare data help meet the needs of older people and a growing older population?
● “Is the wider infrastructure in place, including digital skills, to make the best use of the data we already have?
● “Is there a gap between expectations of what data could achieve, and what is possible in reality?”
There is a lot of attention on AI and machine learning, which can be very good at detecting signals in noisy data, such as images such as scans or free text files such as medical records. There is a tendency to be more excited about these new technologies than about existing technologies, when there is a great deal that could be done with fairly straightforward uses of data, if the data were readily available.
The National Audit Office’s 2019 report “Challenges in using data across government” was not specifically about the use of data in the health system. However, it highlighted that across the public sector, data is not always seen as a priority, its quality is not well understood, and that there is a culture in government of tolerating and working around poor-quality data.
At the ODI, we find that thinking of data as a new form of infrastructure (similar to the road network) is a useful way of understanding its importance to decision making and the provision of services. Data needs to be accessible, its quality and authoritativeness understood, and it needs to be reliably maintained. New technologies make real-time, detailed and unstructured data more valuable than previously, but there is still some way to go to make even basic data available, as documented in the 2017 report “Open data in the health sector” by Giuseppe Sollazzo and David Miller.
There is also a need to improve skills in working with data. As well as skills in data science, medical informatics and evidence-based medicine, we would highlight skills in data strategy, data governance and how to manage access to data through procurement. At the ODI, we have developed a Data Skills Framework that describes the skills needed around data within any organisation. We also believe it is important that the NHS adopts more open ways of working, particularly through open analytics and open science, so that people in one organisation are more able to build on the work of others.
Finally, there is a strong requirement for careful roll-out and critical clinical and systemic evaluations of new technologies. As an example, the accuracy claims of Babylon's GP at Hand service have been called into question; it is also prominently used by younger people, which has systemic effects on the NHS (see this Wired article for more details). When technologies are initially rolled out to low-risk patients, which may exclude many elderly patients, evidence needs to be built separately about applicability to those with higher risk or more complex conditions and needs. An infrastructure that supports rigorous, iterative evaluation with a rapid turnaround would increase patient safety, help them be used effectively, and help speed their adoption. The ODI is currently undertaking work funded by the Health Foundation to examine the data infrastructure required for these kinds of evaluations.
6 March 2020