Aplicación de un modelo politómico de TRI al test ASSIST para el estudio de sus propiedades métricas

Abstract

Cocaine is, second after cannabis, the most used illegal drug in Europe and Spain. Its use is primarily linked to recreational nightlife settings and young adults who are reluctant to contact with health services and draw on these when they develop serious drug problems. Therefore, screening tools are needed to help early detection of problems related to cocaine use. Aim. The aim of this paper is to analyse the psychometric properties of an online self??applied version of ASSIST (Alcohol, Smoking and Substance Involvement Screening Test). Method. The sample consisted of 1176 cocaine users who completed the ASSIST online. The psychometric properties are studied using the Partial Credit Model, within the family Item Response Theory (IRT). Results. In general terms, the degree to which the test response data are as expected from the model is good. However, ASSIST test shows a poor fit of the response categories. Similarly, the relative positions of the items on the continuum imply poor measurement of people with low levels of risk associated with drug use. Conclusion. ASSIST analysis from the perspective of IRT has showed some shortcomings in this test when applied to this type of sample of cocaine users. Despite good item calibration, several improvements to reduce the measurement error in people with low levels of risk associated with cocaine use (construct) and changes in responses categories are proposed