Statistical and Knowledge-Based Approaches to Clinical Decision-Support Systems, with an Application in Gastroenterology

Abstract

Many attempts have been made to support clinical decisions by formal statistical reasoning, but the practical impact of these efforts has been limited. Developers of "expert systems", who use the techniques of artificial intelligence to represent clinicians' personal "knowledge", have suggested one reason for this lack of success may be that the probabilistic methodology itself is often inappropriate to the clinical problems or opaque to the user. We contrast the statistical and "knowledge-based" paradigms, with an emphasis on their different approaches to the manipulation and explanation of uncertainty. A statistical application to the diagnosis of "dyspepsia" is described, in which data are obtained by computer interview of the patient, and both diseases and symptoms form a hierarchy. We argue that the flexible use of "weights of evidence" overcomes many of the previous criticisms of statistical systems while retaining a valid probabilistic output. We conclude by discussing the complementary roles of deductive and probabilistic reasoning.

Indicators of Acute Bacterial Meningitis in Children at a Rural Kenyan District Hospital

“…Variables found to have crude likelihood ratios of Ն1.5 or Յ0.67 were adjusted for the potential confounding effects of related variables in a multivariate analysis according to the method of Speigelhalter and Knill-Jones. 8,13,14 Variables with adjusted likelihood ratios (ALRs) of Ն1.5 or Յ0.67 were regarded as potentially useful, independent clinical indicators. Practical screening rules incorporating the indicators identified were evaluated by calculating the area under the receiver operating characteristic (ROC) curve and identifying the point of maximum discriminatory value.…”

The BUGS project: Evolution, critique and future directions

“…For example, extremely influential work from Stanford featured rule-based systems in which uncertainty was handled using a system of 'certainty factors' attached to rules and manipulated according to somewhat arbitrary principles-this was labeled as 'ad hoc quantitative mumbo-jumbo' by a prominent statistician (Smith, discussion of [31]). The challenge was to come up with a more rigorous process based on probability theory but that still retained the attractive aspect of local computation.…”

Mass movement susceptibility mapping - A comparison of logistic regression and Weight of evidence methods in Taounate-Ain Aicha region (Central Rif, Morocco)

“…L'analyse spatiale bivariée par régression logistique est un modèle non-linéaire, fondée sur une fonction (3) de répartition sigmoïdale entre la probabilité de présence d'un glissement de terrain (Vd) et différents facteurs de prédisposition (Vd), [4,8]. Le modèle logistique pour la présence ou l'absence de glissements de terrain est défini par :…”

“…Après une brève présentation du secteur d'étude, la démarche méthodologique ainsi que les résultats seront présentés et discutés. [4], est une version log-linéaire du théorème général de Bayes utilisant les principes de calcul de probabilité fondés sur les notions de probabilité a priori et de probabilité a posteriori. La probabilité a priori, qui est la probabilité qu'une unité de terrain (un pixel) contienne une variable dépendante (Vd, dans notre cas les glissements de terrain), est calculée suivant la densité sur la zone d'étude.…”