Longitudinal dynamics of microbiological persistence and loss of detection in multidrug-resistant organisms: a survival analysis based on individual epidemiological surveillance trajectories.

BACKGROUND: Longitudinal characterisation of microbiological persistence and loss of detection in multidrug-resistant organisms (MDROs) remains limited; most studies report only time to first negative culture, without addressing oscillation or re-positivity. AIM: To reconstruct individual MDRO microbiological trajectories from a hospital surveillance registry, restricting the sustained loss-of-detection event to surveillance samples, and to analyse associated factors using survival models. METHODS: Retrospective cohort study at a public acute-care hospital in the Community of Madrid, Spain (2022-2026). We included 106 MDRO trajectories (at least 2 post-index surveillance controls) from 309 eligible trajectories of methicillin-resistant Staphylococcus aureus (MRSA), carbapenemase-producing Enterobacteriaceae (CPE), vancomycin-resistant Enterococcus (VRE) and multidrug-resistant Pseudomonas aeruginosa (MDRPA). Kaplan-Meier curves were compared using the log-rank test; cumulative incidence functions accounted for death as a competing event (Aalen-Johansen method). Factors associated with loss of detection were analysed using univariable and multivariable Cox regression, with a Fine-Gray subdistribution model as sensitivity analysis. FINDINGS: The median time to sustained loss of detection was 117 days; persistence probability was 51.4%, 44.4% and 38.2% at 90, 180 and 365 days. MRSA lost detection faster than CPE (23 vs 366 days; log-rank p=0.016). ICU exposure was the strongest independent factor associated with loss of detection (adjusted HR 2.69, 95%CI 1.53-4.73; p<0.001), although this association may partly reflect differential surveillance intensity rather than a direct biological effect; nosocomial acquisition was also independently associated (adjusted HR 1.82, 95%CI 1.02-3.25; p=0.043). An oscillating or re-positive pattern was observed in 28.3% of trajectories. CONCLUSION: Microbiological persistence in MDRO patients follows a dynamic, heterogeneous longitudinal pattern. Restricting this event to surveillance samples and classifying trajectory dynamics reveal phenomena undetectable through conventional approaches.