Application of Vaccination Behavioural Models Based on Modified Proxy Measures for C Scales and Characteristics of Their Determinants in Patients Vaccinated Against Infectious Diseases: A Cross-Sectional Survey

Background: C scales are important measures that can support the vaccination process by analysing the possible reasons for vaccination hesitancy. In some studies, it is not possible to utilise typical models based on the traditional 3C–7C scales, but it is possible to adjust them by using modified proxy measures and converting individual determinants into variables of similar significance in the logistic regression analysis of a vaccination behavioural model. Objective: The aim of this study was to compare the relationship between determinants defining a vaccination behaviour model based on the classic 3C, 5C, and 7C scales with modified proxy measures and two variables (low vaccine-related fear instead of complacency and exposure to misinformation-prone digital sources instead of conspiracy) and the dependent variable, namely, the vaccination status of patients of primary care centres in Poland from 2024 to 2025. Methods: A cross-sectional study was conducted using a survey and questionnaire data linked with patient records from three Polish health centres (N = 1206, 46.9% rural residents and 53.1% city residents). Demographic data and determinants of three vaccination behaviour models were compared between vaccinated and non-vaccinated patients using Pearson’s χ2 test, while behavioural models based on the 3C, 5C, and 7C scales describing the relationship between independent variables and vaccination status (dependent variable) were prepared using logistic regression. Two of seven determinants were changed. Results: All determinants used in the three logistic regression behavioural models differed significantly between vaccinated and non-vaccinated patients (p < 0.001). Furthermore, independent variables in each behavioural model were significantly related to the dependent variable (p < 0.001). With an increasing number of model determinants, Nagelkerke’s R2 increased from 0.303 to 0.435 and the ROC-AUC from 0.828 to 0.882. However, the VIF value indicated very low multicollinearity of determinants. The Wald test did not reveal a significant relationship between confidence in public health and the dependent variable (p > 0.05, OR 1.17–0.92 for Models 1 and 3, respectively). Conclusions: In the vaccination behavioural models prepared based on empirical data, all determinants were found to be significantly related to the dependent variable. As the number of independent variables in the behavioural model increased, parameters such as the ROC-AUC and Nagelkerke’s R2 also increased, while the Akaike information criterion, which determines the balance between fitting the model to data and its complexity, decreased. No significant individual relationship was found for the determinant “confidence in public health”, but this result confirms observations from other previous studies describing the impact of pandemic fatigue, misinformation, and conflicting expert information on institutional trust.