Self-assessed water-safety knowledge and high-risk swimming behaviour: a cross-sectional network analysis of 17,623 Chinese children and adolescents
Drowning is a leading cause of injury-related death among children and adolescents, and high-risk swimming behaviour is a potentially modifiable antecedent. Psychological determinants have largely been examined individually. We mapped their conditional-association network and compared children with adolescents. In this cross-sectional study, 17,623 students aged 8–18 years from 54 schools in four Chinese provinces completed measures of self-assessed water-safety knowledge, risk perception, risk attitude, sensation seeking, swimming self-efficacy and high-risk swimming behaviour. Gaussian graphical models were estimated using EBICglasso on Spearman correlations for the total sample and separately for children (8–12 years) and adolescents (13–18 years). School cluster-robust standard errors were used for mean comparisons. Age-group differences and uncertainty were evaluated using a 20,000-permutation school-stratified Network Comparison Test and 1,000 school-cluster bootstrap resamples. Sensitivity analyses assessed regularisation, alternative behaviour nodes, stability under school dropping and province-specific networks. The nodes separated into risk propensity–behaviour and knowledge–perception communities. Aggressive violations and risk attitude were central. Self-assessed water-safety knowledge had negative expected influence but substantial cross-community connectivity; together with self-efficacy, it was a principal bridge node. The school-stratified comparison identified no difference in global strength ( P = 0.8537) or network structure ( P = 0.8023). Risk attitude–sensation seeking was the only edge to survive Holm correction, but its school-cluster bootstrap interval included zero. An exploratory analysis restricted to the 17 mixed-age through-schools, where all 127 adolescents were aged 13 years, yielded a divergent structure difference and is reported as a limiting sensitivity result. Province-specific networks were broadly correlated with the total-sample network, and the negative expected influence of self-assessed knowledge persisted when correlated behaviour nodes were combined. The water-safety network was organised into propensity–behaviour and knowledge–perception communities, with risk attitude and aggressive violations central. Self-assessed water-safety knowledge had negative signed expected influence but substantial cross-community connectivity, and self-assessed knowledge and self-efficacy were principal bridge nodes. This shows how limited signed influence can coexist with connectivity, offering a systems-level account of knowledge–behaviour relations. School-aware analyses detected no robust age-group network difference. Longitudinal and experimental studies should evaluate attitude- and norm-focused components and competence calibration alongside knowledge education, environmental safeguards and supervision.
Authors
- Zhiping Li (ORCID: https://orcid.org/0000-0001-7513-8272)
- Zhe Xiao
- Chao Xie
- Xi Yang
- Changlin Yuan
- Hui Zhang
- Shuai Wang
Institutions
- Hubei University for Nationalities (CN)
Publication Details
- Journal
- BMC Public Health
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1186/s12889-026-30388-8
- Primary Topic
- Injury Epidemiology and Prevention
- Type
- article
- Field-Weighted Citation Impact
- 0.00