Application of the ant colony optimization algorithm for the construction of a short version of the German alcohol decisional balance scale

Abstract

Abstract Self-report questionnaires must be psychometrically sound, but also brief and efficient to avoid participant nonresponse and fatigue, especially in the health and prevention sciences. Meta-heuristics such as the Ant Colony Optimization (ACO) algorithm overcome limitations of the traditional stepwise approach of selecting items based on few or a single statistical criterion. The aim of this paper was to demonstrate the use of the ACO algorithm by constructing a short version of the German Alcohol Decisional Balance Scale (ADBS). Self-report data from three studies (N = 1,834; 19% women; mean age = 31.4 years) was used that proactively recruited alcohol consumers from the general population and general hospitals in Germany. All participants rated the perceived importance of different pros and cons in their decision to drink alcohol (decisional balance) on a 5-point Likert scale. Optimizing different model fit indices and theoretical considerations simultaneously, the ACO algorithm produced a psychometrically valid and reliable 10-item short scale that was superior to the 26-item full ADBS scale and an already established 10-item short version of the ADBS with respect to the a priori defined optimization criteria. The paper provides a customizable R syntax for building reliable, valid, and theoretically well-founded short scales.