Generating risk-prediction models from mental-health data

Responsible KEG Investigator:
Dr Christopher Buckingham

Project Collaborator(s):

  • Birmingham & Solihull Mental Health Trust
Type of Project: Case Award

Funder: EPSRC

Date Commissioned: 10/2006
Date Completed: 09/2009

Project Summary
EPSRC CASE PhD Studentship

Background: The CASE PhD studentship will be linked with and use data from an ongoing research project funded by the Department of Health for 275,000 pounds. The goal is to produce a decision support system (DSS) that assesses risks associated with mental-health problems using a psychological model of clinicians' risk-assessment expertise.

Aims of proposed project: The main aim is to enhance the DSS by providing mathematical tools for analysing risk data that complement the risk assessments based on human expertise. A secondary aim is to improve the quality of risk assessments within the collaborating organisation, Birmingham and Solihull Mental Health Trust (BSMHT). Specific objectives are to investigate: how Bayesian belief networks might be both informed by and inform expert knowledge structures; the creation of optimum belief-net structures that complement the mental-health expertise; how changes in patient data over time affect event probabilities; the relationship between the DSS judgements and those given directly by mental-health practitioners; pattern recognition tools that do not rely on domain structure inherent within the patient data set; how the DSS can validate higher-level judgements given by clinicians, with the intention of detecting inconsistent judgements for patients; matching of patient information profiles, both for data validation purposes and to simulate case-based reasoning; current methods of risk assessment within the Trust, including the influence of workplace procedures and the influence of information technology (IT); and how the analyses of risk data should best be presented to clinicians to maximise the chances of the DSS being adopted.

Outcomes: The general outcome will be a web-based risk-assessment DSS that includes both human expertise and sophisticated mathematical prediction tools. This combination will greatly enhance confidence of clinicians in the outputs, encouraging both mental-health Trusts and front-line services to adopt the DSS, and increasing the public's access to mental-health advice and help. The research will help develop innovative ways of processing and disseminating mental-health risk information, making it readily available through IT and easily accessible to both clinicians and those without a mental-health background. The proposed work will inform best practice for IT tools, including empirical evidence for the data that should be collected. It will provide crucial input to the National Programme for Information Technology currently being implemented within the NHS.