Impact of the COVID-19 pandemic on emergency department utilization: a time series analysis with compositional covariates

Emergency Department (ED) overcrowding represents a persistent and critical challenge for healthcare systems globally. Understanding the factors driving patient admissions is essential for effective management and resource allocation. The COVID-19 pandemic significantly impacted healthcare utilization patterns, including ED access, introducing complex dynamics related to fear of contagion, altered healthcare-seeking behavior, and evolving access protocols. This study investigates the evolution of daily ED admission in pandemic and post-pandemic phases. We specifically focus on analyzing the influence of various input factors on the number of daily accesses. Crucially, we recognize that some key determinants, such as the daily distribution of patient triage codes (indicating urgency levels) and the daily proportion of admissions by gender and age, are compositional in nature. To appropriately account for the constrained sample space and inherent dependencies of these variables, we utilize Additive Mixed Models incorporating the Isometric Log-Ratio transformation for compositional covariates. The models also include other temporal, environmental, and pandemic-specific factors, alongside an Auto-Regressive Moving Average error structure to capture temporal autocorrelation. Our findings reveal significant shifts in overall admission numbers and, importantly, in the composition of patients accessing the ED during and after the pandemic, highlighting the value of employing compositional data analysis techniques to better understand complex healthcare dynamics. The results provide valuable regional insights applicable to healthcare planning and response strategies in similar contexts.