Digital Transformation

The following questions were asked during the webinar, with responses from Professor Cathie Sudlow below:
 

Question 1:

I work in the Rural and Island Hub within PSD Scotland, and we find it difficult to get data that's usefully disaggregated to rural and island areas. Quite often the best we can do is 'Highland', but that covers an immense area and isn't useful when we're trying to get insights about specific communities. Is there any way this could be addressed?

That’s a great challenge. I don’t have an instant answer for you but it does depend on where/who you are going to for the data and what the purpose you need it for is. Many datasets e.g. those held / available via Public Health Scotland include postcode information which should mean it’s a least in theory possible to pull out the areas and people/patients of relevance. I’d need to understand the specifics more, though, to be able to comment (or even advise) further. Please contact me on cathie.sudlow@ed.ac.uk if you’d like to follow up.

Question 2:

How can we maintain public trust while enabling greater access to and use of health data for research, service improvement, etc.?

A big question! The high-level answer is easy to say and harder to do.

It’s about transparency from the leaders of large organisations (mainly public ones) who generate and use health data – being clear with people who use health and other services across the country that their data, combined with the data of many others, is of great value in better understanding and planning how the health and care service works and can be improved for them and others. All of us in public life need to be on side with this and can help with building and maintaining public trust through being trustworthy and credible professionals. This includes – among others - health and care professionals, researchers and data analysts in universities and NHS, and politicians at local and national govt level. It’s also about illustrating the benefits of using data for these purposes through great, compelling examples of benefits through data use and harms through lack of access and use….like if it’s impossible for the Medicines and Healthcare Product Regulatory Agency (MHRA) to access medicines prescribing and dispensing data linked to data from primary care and secondary care on the  health conditions of people who are or are not taking those medicines / using those products, then the MHRA can’t monitor whether new drugs or devices are working as they should and/or potentially causing unexpected, unpleasant or devastating adverse effects in some people. 

Active safety monitoring systems at population level are an essential part of responsible healthcare and yet no country in the UK has an adequate system for this because of data access and linkage issues. Many people across the country would probably be pretty shocked to hear that and would be keen to see our data linkage and access systems better able to improve the health safety of themselves, their friends and family and the wider population. It’s about demonstrating and describing how data are used responsibly to address important questions for patient and public benefit within secure environments that protect people’s privacy and confidentiality. And it’s about involving patients and the public in the discussions and decisions about how and by whom data from different sources can be used to benefit individuals and society.

Question 3.1:

Given the vast amount of health data already held within healthcare systems, why can't anonymised historical data be analysed retrospectively using modern AI techniques to identify population-level health patterns and outcomes?

It can and it should – but the systems to enable this are not as streamlined as they could or should be and there are often data gaps - certain sources of data that exist may be missing because of regulatory, legal, cultural, technical, human behavioural and/or financial factors.

Question 3.2:

For example, could AI be used to compare long-term trends between vaccinated and unvaccinated populations, as well as uncover broader associations across diseases, treatments, demographics and health outcomes, without relying solely on prospective research cohorts?

Yes and there are examples of this sort of thing happening but there are and have been challenges because it can be harder to build and maintain trust to use AI analysis methods on large scale linked data. This is both because AI seems unfamiliar and scary to many people and organisations, and because additional mechanisms and technical know-how are needed to deploy AI models within secure environments and to address and manage risks of using AI models – for example see https://arxiv.org/abs/2608.16273.

Question 4:

Using modern data analysis tools on historic open data could definitely be really impactful! I wonder if Public Health Scotland already has such projects underway.

Depends what you mean by historic. Public Health Scotland certainly stewards and enables access to a range of health-relevant datasets that go back for up to a few decades. Although PHS’s main focus (as with many healthcare and public health organisations) is on data relevant to the here and now. The general point that there is a huge amount we can and should learn from accessing, linking and analysing historic data that gives us longitudinality through to the present day, whether open or accessible through other means, is a really great one. It’s through longitudinal studies based on such data that we can better understand both the antecedent causes, consequences and potential new approaches to diagnose, prevent and treat a wide range of health conditions.