LAUSR

Abstract Background The association between loneliness and problematic media use has been evaluated in longitudinal studies and meta-analyses. However, previous meta-analyses have relied heavily on Pearson correlation coefficients, which may not fully account for the complexities of this association. Therefore, an updated meta-analysis incorporating more robust statistical models is needed. Objective This study aimed to examine the longitudinal relationship between loneliness and problematic media use using various statistical models and to explore potential moderators that might influence the strength of this relationship. Methods A systematic search was conducted using Scopus, APA PsycArticles, and PubMed to identify eligible studies up to January 24, 2024. Inclusion criteria included studies written in English, published in a peer-reviewed journal, reporting estimates of the longitudinal relationship between loneliness and problematic media use, and a longitudinal study design. Estimates of the longitudinal relationship were synthesized using a random-effects model. The Joanna Briggs Institute Critical Appraisal Checklist was used to assess the risk of bias. Results A total of 26 studies involving 24,798 individuals were included in the meta-analysis. Random-effects models revealed bidirectional relationships between loneliness and problematic media use. The longitudinal relationships were weaker when examined using estimated beta coefficients (rLPMU=0.10; rPMUL=0.09), followed by other statistical models (rLPMU=0.10; rPMUL=0.10) and the Pearson correlation coefficient (rLPMU=0.28; rPMUL= 0.29). Subgroup analyses demonstrated stable results for beta coefficients across various study-level characteristics (including country, lag length, measure of problematic media use, and measure of loneliness), except for type of problematic media use (Q=16.58; P<.001). Conclusions This meta-analysis identified smaller effect sizes for the longitudinal relationships between loneliness and problematic media use compared with the previous meta-analysis. The weaker longitudinal relationship observed when using estimated beta coefficients highlights important methodological considerations for future meta-analyses. However, the contour-enhanced funnel plot and Egger regression test revealed an asymmetrical pattern, emphasizing the need for more longitudinal studies in this area.