Application of logistic regression to explain internet use among older adults: a review of the empirical literature

Abstract

AbstractParticipation in online activities offers older adults enhanced opportunities to cope with age-related limitations and live a socially connected life. For investigating the factors explaining whether older adults use the internet or not, logistic regression has been used, which requires considerable care in the computation and reporting of regression models. The objective of this review is to assess the extent to which the empirical literature examining internet use among older adults (aged 55 +) addresses commonly recommended quality criteria for logistic regression analysis. We carried out a systematic search in Scopus and PubMed for English-language peer-reviewed articles reporting the application of logistic regression, published between 2010 and 2020. Thirty-six articles were eligible for inclusion. We evaluated the articles for six analytic and three documentation criteria. Every article described the selection of independent variables and their coding. Most studies had sufficient events per independent variable (83%), reported point and interval estimates of effect sizes (81%), and stated the fitting procedure used (81%). The reporting rates were lower for the assessment of model fit (53%), consideration of interactions (44%), and testing for conformity with linear gradient (42%) and collinearity (31%). This review provides comprehensive insights into the state of adoption of logistic regression analysis. We suggest improvements for reporting on the studies specific to older adults’ internet use, which can promote the accurate interpretation of articles and the integration of evidence from single studies.