The main project was originally funded by a NHS NEAT grant and began in 2003. Christopher Buckingham, who started at Aston University in 2000, was the Principal Investigator and continues to lead the ongoing research. The project aimed to improve mental-health risk assessments by developing a universally accessible and innovative clinical decision support system (CDSS) called the Galatean Risk Screening Tool, GRiST. GRiST records service-user data (cues) and provides risk estimates for suicide, self-harm, self-neglect, vulnerability, and harm to others.
The key innovation of GRiST is the use of a psychological model of classification at the heart of the CDSS. This Galatean model was developed by Buckingham [4], who also conducted parallel research into how classification is part of a generic model of clinical decision making [1-3].
The Galatean model of psychological classification explains how people use cues to determine class or outcome likelihoods. It proposes that clinicians respond to conditional probabilities of outcomes given cues and that these probabilities compete with each other for influence on classification. The model s validity is demonstrated by explaining people s response patterns in psychological experiments and provides the evidence for using it within CDSSs.
The work on clinical decision making shows how it can be interpreted as three linked, iterative classification tasks: diagnosing, assessing potential outcomes, and making intervention decisions. A coding scheme was developed to encapsulate the underlying elements of these classification tasks and successfully applied to analysing differences between US and UK primary-care practitioners decision-making behaviours [1,2].
Together, these two programmes lie at the core of GRiST by providing the theory and framework for capturing how mental-health clinicians conduct risk assessments. The work was carried out by Buckingham at Aston in collaboration with researchers at Warwick Medical School, who provided the bridge to clinical expertise. The research is important because one of the most critical problems with current risk-assessment tools in mental health is the lack of evidence or formal models for processing the interrelationships of cues. This is why actuarial methods are not sufficient and why the Department of Health (2007), along with the Institute of Psychiatrists (2008), advocate approaches that combine structured clinical judgement with empirical evidence. GRiST is the only approach attempting to do this, using its formal psychological model of structured clinical judgement linked to and validated by sophisticated probabilistic and statistical analyses of the patient database.
The psychological underpinnings of GRiST allowed us to derive knowledge engineering techniques [5] that elicited clinical expertise in a transparent format such that the intuitive nature of the end-product was propagated through the elicitation and implementation cycles. The result was a unique formal model of knowledge structures used in mental-health risk assessment [6]. Ongoing PhD EPSRC CASE and ORSAS funded studentships (2007 2011) investigated methods of learning the multiple galatean parameters from the clinical database so that experts do not need to provide them [8] and also relationships between the fuzzy-set formalisation and alternative probabilistic approaches so that GRiST can exploit their complementary analyses [9].