Influencing factors of hormone replacement therapy acceptance intention and construction of a nomogram prediction model among women aged 35–60 years with menopausal dry-eye symptoms in China: a cross-sectional study based on the health belief model

Background Perimenopausal women are at high risk of several eye diseases, particularly dry eye disease (DED), yet their acceptance of hormone replacement therapy (HRT) remains low, and no quantitative screening tool is available for ophthalmic settings. This study aimed to explore factors influencing HRT acceptance intention among Chinese women aged 35–60 years with menopausal dry-eye symptoms based on the Health Belief Model (HBM), and to construct the first nomogram prediction model for opportunistic screening through ophthalmology–gynecology collaboration. Methods This cross-sectional survey recruited 902 women aged 35–60 years with dry-eye symptoms by convenience sampling between May and September 2026 (1,161 questionnaires distributed; 902 included in the analysis, 77.7% of the number distributed). Dry-eye status was defined by self-reported OSDI symptoms (score > 0), not objective ophthalmic testing. The outcome was the continuous HRT behavioral intention score (2–14). After univariate screening, multiple linear stepwise regression was fitted to build a nomogram. Because residual diagnostics indicated violations of normality, homoscedasticity and linearity, inference additionally used HC3 robust standard errors and nonparametric bootstrap confidence intervals (5,000 resamples), with prespecified sensitivity analyses (transformation, robust, quantile and ordered logistic regression). Performance was assessed for the primary continuous outcome by R 2 , calibration and bootstrap validation (1,000 resamples); AUC and DCA were secondary analyses on an artificially dichotomized low-intention endpoint. Results Six factors independently predicted HRT acceptance intention (R 2 = 0.716, adjusted R 2 = 0.714): monthly household income, menopausal symptom score, perceived severity, perceived barriers, self-efficacy, and cues to action; self-efficacy was the strongest predictor ( β = 0.615). Coefficient directions were identical under conventional, HC3 and bootstrap inference, and all bootstrap 95% CIs excluded zero; perceived severity was marginal ( P = 0.0027, 0.048 and 0.042). Across 12 sensitivity specifications all six directions were 100% consistent and self-efficacy ranked first in 12/12, whereas the ordering of the remaining five was unstable and is not claimed. After bootstrap optimism correction the adjusted R 2 was 0.706 and the C-index 0.943. For the low-intention endpoint, a native logistic model gave an AUC of 0.945 (95% CI, 0.929–0.962; optimism-corrected 0.942), but three of six predictors lost significance after dichotomization. Eight predictions (0.89%) exceeded the scale maximum; truncation to 2–14 did not worsen performance (MAE 0.908 to 0.899). Conclusion This nomogram shows potential value as a communication-assisting screening tool for identifying low HRT acceptance intention in ophthalmic clinics, but it must undergo external validation before any routine clinical application.