Demographic, Seasonal, and Temporal Determinants of SARS-CoV-2 Infection and Serological Status in Al-Bayda City, Libya: A Cross-Sectional Study Employing Multinomial Logistic Regression

Abstract

This cross-sectional study examined the demographic, seasonal, and temporal determinants of SARS-CoV-2 infection and serological status among individuals tested in Al-Bayda city, eastern Libya, between 2021 and 2025. Despite the global transition toward endemic circulation of SARS-CoV-2, region-specific evidence from North Africa remains limited, particularly with respect to multi-year serological patterns. By applying a multinomial outcome framework, this study aimed to provide a nuanced characterization of population-level exposure beyond binary infection classifications. Laboratory-based serological records were obtained from the Al-Bayda Specialized Diagnostic Laboratory. Individuals were classified into four mutually exclusive serological categories—no exposure, early infection, active infection, and past infection—based on combined IgM and IgG antibody results. Descriptive analyses summarized demographic characteristics and serological distributions, while age differences across outcome categories were evaluated using analysis of variance with post hoc testing. Seasonal and calendar-year patterns were assessed to capture temporal heterogeneity in testing and exposure. To evaluate independent associations of age, gender, season, and calendar year with serological outcomes, multinomial logistic regression with LASSO regularization was employed. Penalized modeling was used to address outcome imbalance, sparse event counts, and potential multicollinearity. Model tuning was conducted using cross-validation, and a conservative regularization parameter was selected to prioritize model stability and generalizability. A total of 309 individuals were included, with a mean age of 32.5 years and an approximately equal gender distribution. Past infection was the most prevalent serological outcome, whereas active and early infections were rare. Unadjusted analyses demonstrated significant age-related differences across serological categories, with older individuals more frequently classified as having active or past infection. However, after adjustment and penalization, none of the examined predictors showed strong independent associations with serological outcomes, as relative risk ratios were approximately unity. These findings suggest that SARS-CoV-2 serological status in this population reflects cumulative and multifactorial exposure processes rather than being driven by isolated demographic, seasonal, or temporal determinants. The use of multinomial ridge regression provides robust region-specific evidence and highlights the importance of advanced analytical approaches in interpreting serological data from resource-limited settings.