To determine the internal relationships pf perceived discrimination and rejection sensitivity with social adjustment among college students with left-behind experiences, so as to provide a reference for targeted prevention and intervention strategies for social adjustment issues in this population.

A cluster random sampling method was used to select 2 999 college students from 3 universities in Qiqihar and Daqing, Heilongjiang Province. In October, 2024 (T1) and April 2025 (T2), the Discrimination Perception Questionnaire of college students, Rejection Sensitivity Questionnaire, and the Chinese College Student Adjustment Scale (CCSAS) were used to assess students' perceptions of discrimination, rejection sensitivity, and social adjustment levels. Independent sample t-test was used for group comparisons. Traditional cross-sectional comparisons and cross-sectional network analysis were performed using T1 data for the core sample of 1 080 students with left-behind experiences (T1). Longitudinal network analysis was performed using paired data from T1 and T2 (n=1 024) and a preliminary cross-lagged model was used to determine the relationships between variables. Sensitivity analysis was performed by including follow-up data from students without left-behind experiences to test the robustness of the results.

After gender adjustment, students with left-behind experiences had higher scores on discrimination perception (13.14±0.16) and rejection sensitivity (92.82±0.73) compared to the non-left-behind group (10.25±0.12) (89.12±0.55) while the social adjustment score (194.94±1.05) was lower than the non-left-behind group (202.82±0.79). The differences were statistically significant (F=212.03, 16.52, 36.02, all P < 0.01). Cross-sectional network analysis showed that the "anticipated rejection likelihood" was the core node (EI=-2.27, BEI=-2.37), playing a key role in inter-network connections and overall dynamic regulation. Longitudinal network comparison revealed statistically significant changes in the intensity centrality of the nodes "individual discrimination perception" and "anticipated rejection likelihood" (both P < 0.05), while other nodes remained relatively stable. Cross-lagged network analysis showed that the predictive effects of T1 on the variables at T2 were not statistically significant (maximum |r|=0.05, FDR-corrected P>0.50), but the binomial test indicated a systematic positive correlation trend between the variables (66.7% positive correlation, P=0.04). Sensitivity analysis showed that the bridge effect of the "anticipated rejection likelihood" and "learning adjustment" nodes at T2 was more significant in the left-behind group compared to the non-left-behind group.

The social adjustment issues of college students with left-behind experiences require focused attention on the negative regulatory effect of "rejection cognition" and the positive regulatory effect of "learning adaptation".