Statistical Modeling of Humoral Immune Response Dynamics to mRNA COVID-19 Vaccines in Nursing Home Residents and Healthcare Workers from Southern Italy

Vaccination has been a cornerstone of the public health response to the COVID-19 pandemic, particularly in protecting older and frail populations. A detailed characterization of antibody titer dynamics and their determinants represents a crucial step toward optimizing vaccination strategies. However, antibody titers are bounded within assay-specific limited intervals and often display skewness and intra-subject correlation, which limit the suitability of conventional modeling approaches. We analyzed longitudinal antibody titer data from 608 residents and staff members of five nursing homes in Calabria (southern Italy) using beta-generalized linear mixed models (β-GLMMs). This framework enabled simultaneous modeling of the mean humoral response (μ), precision parameter (ϕ), and probability of achieving the maximum immune response (α), thereby providing a comprehensive assessment of factors influencing immune dynamics. Two distinct patterns of antibody titer evolution were identified. Among nursing home residents, stroke was associated with higher antibody concentrations, whereas atrial fibrillation, lower body mass index, non-Alzheimer's dementia, and chronic obstructive pulmonary disease were linked to reduced responses. The β-GLMM approach allowed for a more accurate identification of demographic and clinical determinants compared with traditional methods. These findings underscore the utility of β-GLMMs for analyzing bounded longitudinal immunological data and highlight key factors shaping vaccine-induced immunity. Such insights may lead to more tailored immunization strategies in vulnerable older populations.