Work

Longitudinal immune trajectories following SARS-CoV-2 vaccination in long-term care facility residents: an unsupervised analysis approach

Year: 2026

Type: article

Abstract: Background: Residents of Long-Term Care Facilities (LTCFs) represent a highly vulnerable population characterized by immunosenescence, multimorbidity, and heterogeneous immune responses to repeated SARS-CoV-2 vaccination and infection. Conventional analyses based on single clinical variables often fail to capture this complexity. We evaluated longitudinal humoral immune responses in institutionalized older adults to identify data-driven antibody profiles and characterize cluster-specific waning kinetics. Methods: In a prospective multicenter cohort of 388 Italian LTCF residents followed across four time points (T0-T3 over 12 months), longitudinal anti-Spike (anti-S) and anti-Nucleocapsid (anti-N) IgG levels were measured alongside demographic and vaccination data. Principal Component Analysis (PCA) was used for dimensionality reduction, followed by K-means clustering. Optimal cluster number was determined using internal validation metrics (Silhouette, Calinski-Harabasz, and Davies-Bouldin indices) and bootstrap stability analysis. Linear mixed-effects models were applied to describe cluster-specific antibody waning trajectories. Results: A longitudinal analysis was conducted on 388 LTCF residents. Three antibody trajectory profiles emerged: (i) a Vaccine-Driven profile (n = 130), characterized by moderate anti-S titers and progressive antibody decline; (ii) a Hybrid Immunity profile (n = 125), characterized by evidence of prior infection and remarkably stable antibody concentrations throughout follow-up despite limited uptake of the 2023/2024 booster; and (iii) a High-Peak Hybrid Responder profile (n = 133), exhibiting the highest baseline anti-S titers followed by the steepest rate of antibody waning. Across the cohort, 12 of 388 residents (3.1%) developed documented symptomatic SARS-CoV-2 infection during follow-up, with no COVID-19-related deaths and no significant difference in infection incidence among profiles. Bootstrap stability analysis showed a mean Adjusted Rand Index (ARI) of 0.74 (SD 0.20). Conclusion: Unsupervised machine learning effectively resolves the complex heterogeneity of SARS-CoV-2 humoral responses in LTCF residents into data-driven trajectory patterns. These findings illustrate the utility of data-driven approach for describing humoral dynamics while distinguishing observational antibody kinetics from direct estimates of clinical protection.

Source: Frontiers in Aging

DOI: 10.3389/fragi.2026.1963384