Roger Cooley

Referral patterns to renal services: what has changed in the past 4 years?

BACKGROUND Awareness of chronic kidney disease (CKD) has been prompted by the publication of several large epidemiological studies since 2002. This has led to various initiatives for the early identification and management of CKD, including the introduction of automated glomerular filtration rate (GFR) reporting and renal indicators in the primary care quality and outcomes framework (QOF) since April 2006. These initiatives were intended to promote identification of CKD and have had an impact on referral patterns to renal services. The aim of this study was to understand the nature of this impact in a catchment population of 1.2 million people. METHODS Data were collected and recorded from all written referrals from primary care between 1 April 2004 and 31 March 2008. Referral patterns for each postcode sector were mapped using Microsoft MapPoint 2004. The effect of chance on referral patterns was modelled by using small area analysis techniques. The association between the CKD prevalence reported from QOF data and the estimated CKD prevalence was examined at post-code district level. RESULTS There were 1461 referrals in 2 years prior to the introduction of the initiatives and 2890 referrals in the 2 years post-introduction. The main reason for referral in both groups was impaired renal function or previously established renal disease. Reported comorbidity was similar between the groups. Mapping showed that there was wide heterogeneity in referral behaviour in the first 2 years of the study, which was less in the second period. Small area analysis suggested that the variation that led to the extremal quotients observed in both of the study periods was not due to random variation in referral pattern alone. There was no correlation between the reported CKD prevalence and the referral rates. CONCLUSION Referral patterns have changed between 1 April 2004 and 31 March 2008. The main findings were an increase in referral rate and in the age at referral without a significant change in reported comorbidity of the people referred. The main increase in referral rates was seen in more advanced CKD suggesting more targeted referral of patients with CKD to renal services.

Using simulated students to evaluate an adaptive testing system

This paper describes a strategy for evaluating an assessor algorithm through the use of simulated students. The assessor is an adaptive testing strategy elicited from a human expert. The assessor claims an accurate assessment of the students domain knowledge with a minimum number of questions presented to the student. An evaluation of the assessor is carried out to test the claim. Three types of students are simulated-one who knows all the relevant skills in the subject domain, one who knows a subset of the relevant skills, and one who has one or more malrules. The performance of the assessor in predicting the mastery of skills of each student is analysed. Future work involves the evaluation of its accuracy in the assessment of other types of students, such as students with lucky guesses or careless slips.

From decision support to expert systems

This paper discusses an approach to developing an expert system as a combination of a decision support system and several simulations of expert users of that support system. Particular attention is given to the problems of coordinating the simulated experts.

The Use of Contraint Logic Programming in the Development of Adaptive Tests

This research shows how Constraint Logic Programming can form the basis of a knowledge elicitation tool for the development of adaptive tests. A review of literature reveals that adaptive testing has, in recent years, been used as a student modelling tool in intelligent tutoring systems. Efforts in the construction and delivery of such tests have involved the use of state-space type constructs, such as granularity hierarchies and knowledge spaces, to represent a syllabus. The use of Bayesian probability networks has been proposed as a technique for finding an optimal route through these ”spaces” so as to find the shortest sequence of problems to put to the student being evaluated. The research presented here sets out to use ”expert emulation” as means of performing the same tasks. The aim is to construct an adaptive test in order to model a student’s knowledge, skills and tutorial requirements. The context is the development of such a system to support autonomous revision for examination in a particular domain in elementary mathematics.