Reading
This week, whilst continuing to wait for a computer and access to the organisations internal/HR systems, I did some wider reading around NHS patient data flows and the stakeholders, and about the consumers of NHS data.
I discovered this fantastic paper which details an incredibly comprehensive piece of work by Zhang et al (2023). The research also has an accompanying microsite.
There is a lot more to say about the findings of this study, but one of the aspects it highlighted was the OpenSafely Platform. OpenSafely was also featured in Price et al (2024), which provides an overview of a variety of the current data infrastructure solutions for cancer research data in the Greater Manchester Area in the UK, including SREs (Secure Research Environments). The paper touches briefly on the potential and some of the challenges of Federated Learning approaches in this context, including the challenges for risk assessment and quality control when the analysts are working at level of abstraction completely cut off from the underlying data, or the context from which that data was gathered.
A Three Stage Transition in Data access
What this threw up for me, combined with my previous work on Federated Learning was the idea of a three stage transition in the mental models of our users when we are working with Data Science, but also as we are asking them to move from requesting data extracts, and working directly with data in a local context, to booting up a cloud container - as is the case with SDEs and working with data in more service-oriented ways, to potentially, at some point in the future an evolution to a Federated Learning systems, whereby they send their algorithm to a completely invisible or intangible data backend system and have no direct or indirect access to data at all.
Obviously, this is a very high level of generalisation of the user-mental model, and a higher level of abstraction of thinking about the user-journey, but it struck me that the mental model our users, and backstage stakeholders might have, might be oriented more around an extractive and transactive model of data acquisition and use vs. this new, more emergent model of data provision as a service where they never possess the data at all, but rather gain access to the data through a cloud container. Such a evolution of the mental model of the service points to the need for us to do some careful work both to understand user's existing mental models and expectations, but also content work in terms of signposting users through this more complicated transition and new modes of data access and use.
Doing
I finally got access to my computer on Wednesday of this week, so the subsequent days involved a fair amount of the necessary HR administrative onboarding and mandatory training.
Towards the end of the week I was invited in to a few of the meetings that were taking place in the SDE (Secure Data Environments) Service team that I will be joining, and taking over as Service Designer on... it is very early days for me, and a lot of information to absorb, but it is really exciting to see the work that the team have been doing to date, as the service approaches it's Beta Assessment.
Service Handovers
One thing that struck me, and it ties together with the above reflection about transitioning users mental models of a service, as part of the evolution or upgrade or roll-out of a new digital service, is also the concept of "Service Handovers". Lou and Sarah touched on this recently on the Dead Ends podcast as it relates to working with Legacy IT. But, as it relates to what I observed in the SDE work, I have been reflecting on the process of linking together different silos or backend services, as part of the process of creating a joined up or seamless frontend user experience and how this often involves the handover, or handling callbacks from a variety of backend systems in order for the user to progress through their user journey. It's also often the case that silo-ed working practices and a lack of communication or collaboration between backend systems or teams / projects running different backend systems, can often be the reason why various micro-services or backend systems can't integrate, or don't talk to each other properly, and this can often be the root of user-experience problems such as users being asked to re-enter information that they have already entered at previous stage in a process. Or simply, this lack of integration or collaboration can result in a confusing or disjointed experience for users.
References
Price, G., Peek, N., Eleftheriou, I., Spencer, K., Paley, L., Hogenboom, J., Soest, J.V., Dekker, A., Herk, M.V. and Faivre-finn, C. (2024). An Overview Of Real-world Data Infrastructure For Cancer Research. Clinical Oncology, DOI: 10.1016/j.clon.2024.03.011, Elsevier Bv http://doi.org/10.1016/j.clon.2024.03.011
Zhang, J., Morley, J., Gallifant, J., Oddy, C., Teo, J.T., Ashrafian, H., Delaney, B. and Darzi, A. (2023). Mapping and evaluating national data flows: transparency, privacy, and guiding infrastructural transformation. The Lancet Digital Health, Vol. 5, pp. e737-e748, DOI: 10.1016/s2589-7500(23)00157-7, Elsevier Bv http://doi.org/10.1016/s2589-7500(23)00157-7