Limited discriminant validity of health belief and 3C models in high HPV vaccine coverage settings: the moderating role of hukou and structural barriers

Abstract

Background Whether traditional health belief models (HBM) and 3C models can distinguish residual non-vaccinators in settings with >70% HPV vaccine coverage is unknown. The moderating role of China’s household registration (hukou) on psychological factors and behavior has rarely been quantified. Methods From April 2025 to February 2026, we surveyed 1,257 women aged 18–65 from six Shenzhen communities. We collected demographics, HBM (15 items), 3C (12 items), social support, HPV vaccination intention (measured by a single item: willingness to receive the HPV vaccine within the next 6 months), self-reported vaccination history, reasons for non-vaccination, and information channel preferences. Analyses included t-tests (Cohen’s d), observed-variable multi-group path analysis, and cross-sectional comparisons of vaccination proportions among women with high intention. Results HPV vaccination rate was 70.0% (880/1,257); cervical cancer screening rate 58.3%. Vaccinated and unvaccinated groups differed on only two of ten psychological constructs (trivial effect sizes: perceived susceptibility d = 0.13, self-efficacy d = 0.13). HPV vaccination intention did not differ significantly between vaccinated (M = 3.72, SD = 0.83) and unvaccinated women (M = 3.67, SD = 0.80), t = 0.81, p = 0.417. Main reasons for non-vaccination were high cost (42.3%) and safety concerns (33.8%). Observed-variable multi-group path analysis showed good fit (CFI = 0.968, RMSEA = 0.051). Perceived barriers were significantly negatively associated with HPV vaccination intention among non-Shenzhen hukou women (β = −0.247, p < 0.001), but not among Shenzhen hukou women (β = −0.102, p = 0.128). The formal test of hukou moderation was not statistically significant (Δχ2 = 2.358, Δdf = 8, p = 0.968). Most preferred information channels: WeChat official accounts (91.5%), face-to-face doctor consultations (85.1%), short video platforms (76.5%). Conclusion At 70% coverage, HBM and 3C scales show limited discriminant validity between vaccinated and unvaccinated individuals. Residual non-vaccination is associated with cost and safety barriers. Hukou showed a suggestive but not statistically significant moderating effect on the perceived barriers-intention association. Interventions may need to target structural barriers and tailor information channels.