Latent class analysis of sexual behavior and HIV testing profiles among sexually experienced college students in China: a cross-sectional online survey

Abstract

Rising human immunodeficiency virus (HIV) infections among young people underscore the need to strengthen campus-based prevention. We used latent class analysis (LCA) to summarize sexual behavior and HIV-testing profiles among sexually experienced college students, examine associated factors, and assess whether the selected solution added information beyond an observed-response typology. This cross-sectional study analyzed 5,218 students reporting lifetime sexual intercourse in an anonymous online survey in Nanning, China. LCA included sexual exposure/protection status, multiple sexual partnerships and same-sex sexual behavior in the past six months, and HIV testing in the past year. Modal LCA membership was compared with a post hoc rule-based typology using agreement measures. Multinomial logistic regression examined associated factors. Among 108,552 valid respondents, 5,218 were analyzed. The three-class model balanced fit, class size, parsimony, and interpretability. Class 1 ( n = 509, 9.76%) combined multiple partnerships and/or recent same-sex sexual behavior with past-year HIV testing; Class 2 ( n = 2,535, 48.58%) comprised participants with 0–1 recent partners, no recent same-sex sexual behavior, and low testing uptake; and Class 3 ( n = 2,174, 41.66%) resembled Class 1 in partner patterns but lacked past-year testing. The rule-based typology exactly reproduced modal LCA membership (agreement = 100%; κ = 1.000; adjusted Rand index = 1.000), indicating limited additional person-level classification. Compared with Class 2, Class 1 membership was positively associated with junior college status, pre-exposure prophylaxis (PrEP) awareness, and ever PrEP use, and inversely associated with town household registration, heterosexual orientation, and an opposite-sex first sexual partner. Class 3 membership was positively associated with ever PrEP use and inversely associated with male sex, town household registration, comprehensive university attendance, an opposite-sex first sexual partner, and adequate HIV-related knowledge. Many participants reporting multiple partnerships and/or recent same-sex sexual behavior had not undergone HIV testing in the past year, indicating a need for confidential and accessible testing. Because LCA mainly summarized strongly separated observed patterns, brief, non-stigmatizing assessment of partner patterns, condom-use consistency, and testing history may support behavior-informed campus prevention and linkage to PrEP/PEP services.