A Bayesian belief network for quality assessment: application to employment officer support.

BACKGROUND: 'Supported' employment stipulates that disabled people should have real jobs for real pay. This paper models kinds of supported employment, assesses how the support and placement features affect its outcomes and its quality from the perspective of the employees, and provides a dynamic model to help explore what types of interventions might promote greater social inclusion for people with learning and other disabilities. METHOD: Bayesian belief networks (BBNs) provide the general framework for modelling the relationships between the variables and features of interest. The structure, probabilistic specification and quality indicators were elicited from project advisory groups, including people with learning disability, and took into account a pilot survey of 30 individuals. A subsequent survey of 1,461 supported employees was used to update the model and to provide actual assessments of quality of placement. RESULTS: We present the BBN methodology in some detail, as novel to this discipline. We show how the model was constructed, and its implications for supported employment. We derive indices for quality of placement, taking into account the views of clients. We show how survey and individual results can be used to update the model. Use of the model suggests that quality of placement is, on average, relatively high, with small differences between groups with differing primary disability. CONCLUSIONS: The BBN is the appropriate methodology to model complex relationships and interventions for problems such as these. The model developed in this study can be used to assess and improve the fit between people and jobs, both at the individual level and for groups of employees, and can take into account different kinds of quality for different stakeholders.

How good is zoom endoscopy for assessment of villous atrophy in coeliac disease?

BACKGROUND AND STUDY AIMS: The advent of magnification endoscopy may allow the macroscopic detection of unrecognised villous atrophy in patients with unsuspected coeliac disease. In addition, it may also be possible to use this method to assess the degree of villous atrophy. The aim of this study was to determine the accuracy of zoom endoscopy for the macroscopic evaluation of villous atrophy, in comparison with histological evaluation. PATIENTS AND METHODS: The zoom endoscope provided a magnification capability of x 115. A scoring system (Z score) was devised for grading the appearances of villous atrophy: "Z1" for normal mucosa, "Z2" for stunted villi, "Z3" for markedly stunted villi (with ridges and pits) and "Z4" for a flat mucosa. A total of 53 consecutive patients with treated coeliac disease were followed up over almost 2 years using the Olympus GIF-Q240Z zoom endoscope; a total of 80 procedures were carried out. Four biopsies from the second part of the duodenum were taken from each patient for histological assessment. Histological assessment of villous atrophy was made by a pathologist blinded to the Z score. The correlation between the Z score and the histological score was assessed using the weighted kappa method. RESULTS: The kappa score for the correlation between the macroscopic assessment of villous atrophy and the histology was 0.631, indicating fair to good reproducibility. Agresti's method revealed a very strong baseline association between the two methods ( P < 0.001). Zoom endoscopy had a positive predictive value of 83 % and a negative predictive value of 77 % in detecting villous atrophy. CONCLUSIONS: Our findings suggest that zoom endoscopy may be valuable in assessing the degree of villous atrophy. However, further studies are needed to assess its efficacy in routine practice as a screening or case-finding tool.

A simple modelling approach for water quality: the example of an estuarine impoundment.

With the increasing development of water quality standards for all forms of water bodies and all forms of their uses, there is a greater need for models that can predict water quality in relation to these standards. This paper proposes an empirical method based on logistic regression techniques that deliver probabilities that given quality standards are met (i.e. water above or below the given standard). This technique is applied to data from a saline intrusion barrage (River Wansbeck, Northumberland, England) to predict low water quality events on the basis of UK water standards. The approach accurately classifies the overwhelming majority of pollution events that are caused by processes internal to the system. Misclassification can be ascribed to external processes, such as combined sewer outfalls. The general applicability of this method is compared to physically-based models.