About the project
Consolidate and extend the reach of the CPFT Research Database
The CPFT Research Database is a database of anonymised medical records. The Cambridge and Peterborough NHS Foundation Trust's electronic clinical records are put through a software tool called CRATE (Clinical Records Anonymisation and Text Extraction), which removes information that might identify a patient. The database will be invaluable as researchers carry out their work into what causes conditions such as dementia, and which treatments are best for dementia and other mental health conditions.
The Cardinal team developed CRATE to de-identify arbitrary relational databases. CRATE extracts identifying information from a source database and uses this to “scrub” sensitive free text from the source database creating a de-identified research database (the CPFT Research Database).
Further improvements to CRATE’s linkage and de-identification performance will support linkage of CPFT data (on mental health and community services) to data from acute medical services and to other national and local data sets. Completion of this step will provide cryptographically robust access to integrated healthcare data.
Further development and application of Natural Langage Processing (NLP) tools, including the prototyping of a national secure cloud-based NLP service and application programming interface (API), will support partnership working on NLP tool development and their application to federated datasets. Completion of this step will support ongoing collaboration with partners in a national mental health platform.
Team members working on this: Rudolf Cardinal, Tim Dalgelish, Kristel Klaus, Francesca Spivack, John Suckling
Improve the range and depth of structured clinical and cognitive data
We will build on CamCOPS to deliver an extensible neuropsychiatric assessment system suitable for use at the bedside, in the clinic, or in the community, delivering research-grade data and saving clinicians’ time.
This system will be free and open source, will integrate via open standards to Electronic Patient Records (EPRs) and other research data capture systems, may be extensible to very large scale population testing and to medical device approval, and will improve the harmonization of mental health measures between research cohorts and NHS populations.
For example, we will use CamCOPS’s psychiatric phenotyping tools for assessment, including of risk and resilience factors, in large research cohorts in Cambridge (MRC Cognition and Brain Sciences Unit; CBU) from childhood through to old age that can be linked securely to NHS EPRs; we will apply them to studies examining the relationship between immune dysfunction and psychiatric illness; and we will validate CamCOPS tools for use in routine clinical practice.
Team members working on this: Cardinal, Burchell, Khandaker, Parkinson
Tackle the mortality gap in serious mental illness
We have already used the CPFT Research Database to confirm what others have reported: life expectancy is reduced by >15 years in CPFT service users with serious mental illness. We now need to understand the causes of this in greater detail and to predict mortality so as to provide early warning and be able to intervene better.
We will use the data provided by our linkage and de-identification tools and sophisticated machine learning algorithms to develop new predictive models of outcomes in serious mental illness and to develop new ways of working between mental health services and industry whilst maintaining confidentiality of NHS records.
Team members working on this: Cardinal, Spivack, Banerjee, Jones, Lio
Democratising Mental Health Research
We already have excellent arrangements for service user engagement in the CPFT Research Database but we want to build on this. We consider it very important that research-grade data on mental health are accessible to clinicians and service users, as well as professional researchers, and that all service users are empowered to participate in research as promised by the NHS Constitution.
We will extend our existing information-sharing and consent-for-contact systems and work towards a national sign-up register for mental health research, engaging with patients and the public as to the best ways to develop simple tiered consent models for information sharing and participation applicable across the NHS.
We will develop data visualization tools so that clinicians and service users can more easily ask research questions of de-identified NHS data, and bring clinical value by making research database search tools directly accessible to clinicians for their own patients.