Discrimination of psychotropic drugs over‐consumers using a threshold exceedance based approach. (5th November 2012)

• Record Type: Journal Article
• Authors: Bellanger, Lise; Vigneau, Caroline; Pivette, Jacques; Jolliet, Pascale; Sébille, Véronique
• Is Part Of: Statistical analysis and data mining. Volume 6:Number 2(2013)
• Page Start: 91
• Page End: 101
• Publication Date: 2012-11-05
• Subjects: extreme value theory -- cluster analysis -- discrimination -- logistic regression -- ROC curve -- misuse -- evaluation; Data mining -- Statistical methods -- Periodicals; 006.312
• DOI: 10.1002/sam.11165
• Languages: English
• ISSNs: 1932-1864

Abstract

Abstract: Use of some medication, such as tranquilizers or hypnotics may carry important risks for patients including the emergence of abuse and/or dependence. The problem we tackle consists of identifying and discriminating the group most "at risk" of abuse and/or dependence for a given drug, in order to provide an estimation of its prevalence and to develop preventive measures targeted toward the corresponding drug prescription. A criterion, currently employed to characterize patients' consumption of a drug, is the ratio between their daily average consumption and the maximum recommended daily dose as specified in the drug monograph, called the F factor. In theory, any patient having an F factor greater than 1 should be classified as an over‐consumer for the corresponding drug, but in practice this threshold might not be very relevant for all drugs. The proposed approach, combining different statistical methods (extreme value theory with the Peaks Over Threshold Model, logistic regression, ROC curve), is an innovative way to study consumption behaviors of psychotropic drugs. Two drugs are studied: an antidepressant, tianeptine and a hypnotic, zolpidem. From one drug to another, different thresholds for the F factor and patient's characteristics associated with the risk of extreme consumption are found, revealing different consumption behaviors. © 2012 Wiley Periodicals, Inc. Statistical Analysis and Data Mining 6: 91–101, 2013