Abstract
BACKGROUND: Perceived organisational support (POS) and psychological resilience help protect nurses from turnover intention (TI), but most studies have used only total scores, so it is unclear which concrete support or resilience elements are most influential. OBJECTIVE: To test whether psychological resilience mediates the association between POS and turnover intention in hospital nurses and to identify, through network analysis, the POS and resilience items most closely linked to turnover-related cognitions. METHODS: A cross-sectional survey was conducted among 439 nurses from a Grade A tertiary hospital in Sichuan Province, China. The Survey of Perceived Organisational Support (SPOS), the 10-item Connor–Davidson Resilience Scale (CD-RISC-10) and the Turnover Intention Questionnaire (TIQ) were administered. Spearman correlations and mediation modelling with 2000 bias corrected and accelerated bootstrap resamples were used to examine construct level associations, with a covariate adjusted sensitivity analysis (gender, years of work experience and job satisfaction). Item level associations were estimated using EBICglasso networks based on polychoric correlations. Expected influence (EI), bridge expected influence (bridge EI), bootstrap accuracy, case-dropping stability, and split-sample network comparison tests were performed. RESULTS: POS was positively associated with resilience (ρ = 0.32, p < 0.001), and both were negatively associated with turnover intention (POS–TI: ρ = −0.33; resilience–TI: ρ = −0.36; p < 0.001). Mediation analysis indicated a significant indirect association between POS and lower turnover intention through resilience (standardised β_a×b = −0.104, 95% CI −0.152 to −0.058), which remained after covariate adjustment. In the three community network comprising POS, resilience and turnover intention, the item “My organisation cares about my opinions” emerged as a key bridge and was conditionally associated with “I do not get discouraged by failure.” Nodes with high expected influence included “My organisation cares about my well-being,” “I can achieve my goals even when faced with obstacles,” “The accumulation of experience has made me stronger,” and “Are you considering resigning from your current job?” Bootstrap based stability analyses suggested acceptable robustness, and split sample network comparison tests did not indicate statistically significant differences. CONCLUSION: Higher POS was associated with lower turnover intention, with a statistically significant indirect association through resilience at the construct level. At the item level, a small set of support and resilience contents, particularly opinion-related organisational care and persistence-oriented resilience, showed greater structural prominence and may help inform more targeted retention strategies in nursing management. CLINICAL TRIAL NUMBER: Not applicable. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12912-026-04499-x.