BMC Psychol. 2026 Jul 3. doi: 10.1186/s40359-026-04885-9. Online ahead of print.
ABSTRACT
BACKGROUND: Bullying and cyberbullying are well recognized mental health risk factors and are of major public health concern during adolescence, yet inconsistencies in assessment methods and longitudinal design hinder accurate prevalence estimation and understanding of developmental trajectories. This longitudinal study aimed to: (a) compare prevalence estimates of bullying and cyberbullying using single-item versus multi-item assessments; (b) assess developmental changes in victimization and perpetration from 6th to 8th grade (ages 10-13).
METHODS: A cohort of 788 Italian middle school students (50.5% male) completed assessments in sixth (2015-2016) and eighth grade (2017-2018). Bullying and cyberbullying were measured with single-item and multi-item self-report instruments. Analyses included descriptive statistics, Pearson’s correlation alongside point-biserial correlations, McNemar’s tests, paired t-tests, generalized estimating equation models, mixed repeated-measures ANOVAs, Stuart-Maxwell tests, and intraclass correlation coefficients.
RESULTS: Single- and multi-item measures showed moderate convergence (r = .36-0.51). Single-item measures primarily capture global self-identification with victim or perpetrator roles, whereas multi-item measures provide greater specificity regarding the frequency, context, and modality of bullying. Measurement method influenced developmental trends: bullying victimization decreased as children got older using single-item tools (χ² = 24.59, p < .001) but remained stable with multi-item, while perpetration increased only via multi-item (d = -0.28). Cyberbullying findings were more tentative: single-item victimization decreased, multi-item victimization remained stable, and multi-item perpetration showed only a small increase. Supplementary analyses showed no significant Time × Sex interactions, and school-level clustering was negligible.
CONCLUSIONS: Findings highlight that bullying assessment methodological choices strongly influence prevalence and developmental estimates. Multi-method longitudinal designs are crucial for valid assessment and effective prevention.
PMID:42400099 | DOI:10.1186/s40359-026-04885-9
