Two sides of the same coin: Operational versus research use of administrative data
In this blog, Emily Oliver, Head of Training and Capacity Building at ADR UK, explores the distinctions and convergences between using administrative data for operational and research purposes.
Why is administrative data so amazing? It includes everyone. Anyone who goes to school, pays tax, interacts with the courts, or goes to see their GP is counted. This is because the original purpose of administrative data is to generate information for public services to operate efficiently.
So, we’re automatically included in a dataset when we interact with a service, rather than being asked to take part in the data collection as we would be for a survey, for instance. When administrative data is used for running a service like this, we refer to it as ‘operational’ use of the data - because it’s helping a service to operate.
Sometimes this data is also linked together and made securely available to accredited researchers to analyse trends and patterns in the data. Researchers do this to try to improve people’s lives with insights that lead to better, evidence-based decision-making. This is called ‘research’ use of administrative data.
When administrative data is linked and used for research purposes, the privacy of individuals has to be protected because individuals did not explicitly consent to their data being used for research. The data is carefully de-identified so that any information that could identify an individual person is removed.
While de-identified data can’t be used to manage the issues of individuals, government analysts have found it offers valuable insights into patterns of individual behaviour and needs. It can therefore inform the targeting and delivery of services, meaning the de-identified data can be used operationally to improve services as well as by researchers for research.
Raising awareness and understanding
It’s important that people are aware of how their data is used. As part of an ADR UK-funded project, Dr Dougal Hargreaves and his team of researchers at Imperial College London have been discussing the risks and benefits of linked data with a range of public and professional groups in northwest London.
This includes working with children and young people to explain data use in a way that people can understand and relate to. As a paediatrician (a doctor who specialises in children), Dougal sees the operational versus research use of data as two sides of the same coin: “In explaining the data and how we use it, one approach that often works well is to present both the operational and research advantages of linked data in parallel. For example, we might explain how linked data can help us deliver more joined up services for people living in this area, and how it also allows us to do research on which kinds of services are most effective. This might then help children living in other parts of the country.”
Efforts to improve understanding of linked administrative data shouldn’t be limited to patients and the public. Stakeholders such as government departments or local authorities can also benefit from thinking about how it can be used for operational and research agendas in parallel. This is set out nicely by Professor Sir Chris Whitty in his recent blog on the topic of healthcare data. He reminds us of how important data linkage was during the Covid-19 pandemic for clinical care of individual patients, effective management of NHS services, and research (in this case into new treatment and vaccines for Covid-19).
Linking data for multiple purposes can present a compelling case in terms of logistical efficiency, as well as enriched insights. If we are to maximise the opportunity it affords across services, we must explain this effectively to build trust and understanding.
The safeguards in place
The lack of explicit consent around the collection and use of linked administrative data for research purposes can still worry some people. That’s why it’s important to explain the additional safeguards for research data. It is stored and accessed with the utmost care to protect it from being misused and to ensure that individuals can’t be identified. It’s also worth noting that the process for ensuring the safety of one de-identified dataset is different to when two de-identified datasets are linked together.
There are many measures in place to protect against both data breaches and re-identification: for instance, technological privacy measures are taken and physical checks are made to avoid the data being accessed by unauthorised individuals. Researchers must also go through a rigorous clearance process before they can publish their findings, protecting individuals from being identified in their research.
Hearing from the public
But it isn’t enough to simply put these measures in place; it is crucial to also understand and be informed by public views. Understanding public attitudes to data use, whether operational or research, must be central to how we think about the systems which collect, store, and use data. We know from other studies that people are broadly happy for their de-identified data to be used to improve local services, as well as for research to address real-world needs and benefit others. However, many are not happy for their data to be sold for private profit, or for anything harmful such as perpetuating stereotypes about certain groups of people.
Inevitably, public dialogues and consultations afford the opportunity for the public to consider data in different ways - and in doing so, contextualise it within their own experiences. This directly diverts their attitudes, perceptions, or responses towards something that is much more personal and meaningful. And it is here that the harmony between the use of data for research versus operational purposes becomes apparent. After all, what’s the point of research if it’s not going to inform how something might (or might not) be changed for the better?
If you have suggestions for ways to improve communication and transparency about how data is used for research, please get in touch by emailing firstname.lastname@example.org.