Skip to main content
KBS_Icon_questionmark link-ico
CSMH-SEP-MD-robynne-hu-HOrhCnQsxnQ-unsplash ;

Linking data to understand severe mental illness: what have we learned so far?

Milena Wuerth

Research Assistant

30 November 2022

COVID-19 has proven the importance of gathering reliable and representative health data. Hospitals records, which provide information on every service user, are especially valuable for research in public health. Social and economic factors, like ethnicity and employment status, however, are often not recorded in these records but are collected through the national Census every 10 years. An innovative study based in the Institute for Psychiatry, Psychology & Neuroscience (IoPPN) brings together hospital and Census data to explore relationships between social disadvantage and mental illness.

What is the SEP-MD Study?

An ongoing ESRC-funded project led by Dr Jayati Das-Munshi and a team of researchers at King’s College London and in the NIHR South London & Maudsley (SLaM) Biomedical Research Centre (BRC), called the Social and Economic Predictors of the Severe Mental Disorders: The SEP-MD Data Linkage Study, has linked de-identified hospital records from South London and Maudsley Hospital (SLaM)-- one of Europe’s largest secondary mental health providers - with data from the 2011 Census. The project team aims to use this linked dataset to explore relationships between overlapping forms of social disadvantage, identified through the Census records, to map out the long-term effects of living with severe mental health conditions like schizophrenia. This project is unique because mental health records like these have never been linked to Census in England.

The project began in 2018, after receiving ethical approval. The Office of National Statistics did the work of linking SLaM patient records with Census data and then gave the King’s/ SLaM research team access to the anonymised linked dataset. Members of the team, who have gone through data protection training and clearance, are currently using special secure computers to explore and describe the data.


Eventually, the research team plans to use statistical analyses to see how social disadvantage might predict the onset of severe mental health problems, as well as adverse health outcomes in people with severe mental illness. First, though, they have to address a challenge: a much lower proportion of hospital records in people with severe mental health problems matched with Census records.


A closer look at non-response

This was unexpected as Census tends to be completed by most of the population (usually more than 90%). So, the researchers asked: could this be because of the study design or some error in the matching algorithm? After doing some digging, they have good reason to believe that this is not the primary cause. The researchers think that the records couldn’t be matched because many of the patients included in the study may not have answered the Census in the first place.* This may be because the sample comprised people mostly in contact with mental health services, and potentially living with more severe mental health problems.

Previous studies have found associations between mental distress or illness and non-response to surveys. The SEP-MD study, however, is the first to show non-response to Census – a national mandatory data collection exercise – may also be lower in people living with mental health problems. Because Census data is used to inform policy making and commissioning, high rates of non-response could contribute to the underrepresentation of people living with severe mental illness in service planning.

While important to consider this possibility, non-response is a challenge any researcher working with survey data must grapple with, and it doesn’t mean the data they collect is useless. Researchers address non-response by applying ‘weights’ to the data. These are calculated by taking into account who was less likely to respond and adjusting the data to minimise the effect of losing these records.

Moving forward

As a social researcher by background, this project has given me a window on the daily processes of problem-solving in quantitative data collection and statistical analysis. Since the start of the SEP-MD project, the research team has navigated different hurdles to secure safe access to this valuable population-level data. Now, as they look closely at the linked data set, new questions continue to emerge.

The team is answering these questions – including ones about non-response – by applying statistical methods and by carefully considering the effects of social disadvantage on the data we collect and analyse. This methodological work sets the foundation for our upcoming collaborations with service user groups and for the publication of our findings – stay tuned!

For more information about the project or enquiries about collaboration please email me at

Follow the Centre on Twitter (@kcsamh) or visit our webpage for updates on our progress.

*a paper exploring this in more detail is forthcoming.


In this story

Milena Wuerth

Milena Wuerth

Research Assistant

Latest news