Premium
Internet Gaming Disorder as a formative construct: Implications for conceptualization and measurement
Author(s) -
van Rooij Antonius J,
Van Looy Jan,
Billieux Joël
Publication year - 2017
Publication title -
psychiatry and clinical neurosciences
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.609
H-Index - 74
eISSN - 1440-1819
pISSN - 1323-1316
DOI - 10.1111/pcn.12404
Subject(s) - construct (python library) , conceptualization , formative assessment , scale (ratio) , psychology , reliability (semiconductor) , cronbach's alpha , the internet , computer science , social psychology , psychometrics , artificial intelligence , developmental psychology , mathematics education , world wide web , power (physics) , physics , quantum mechanics , programming language
Some people have serious problems controlling their Internet and video game use. The DSM‐5 now includes a proposal for ‘Internet Gaming Disorder’ (IGD) as a condition in need of further study. Various studies aim to validate the proposed diagnostic criteria for IGD and multiple new scales have been introduced that cover the suggested criteria. Using a structured approach, we demonstrate that IGD might be better interpreted as a formative construct, as opposed to the current practice of conceptualizing it as a reflective construct. Incorrectly approaching a formative construct as a reflective one causes serious problems in scale development, including: (i) incorrect reliance on item‐to‐total scale correlation to exclude items and incorrectly relying on indices of inter‐item reliability that do not fit the measurement model (e.g., Cronbach's α); (ii) incorrect interpretation of composite or mean scores that assume all items are equal in contributing value to a sum score; and (iii) biased estimation of model parameters in statistical models. We show that these issues are impacting current validation efforts through two recent examples. A reinterpretation of IGD as a formative construct has broad consequences for current validation efforts and provides opportunities to reanalyze existing data. We discuss three broad implications for current research: (i) composite latent constructs should be defined and used in models; (ii) item exclusion and selection should not rely on item‐to‐total scale correlations; and (iii) existing definitions of IGD should be enriched further.