Complexity of hospital demand during the COVID-19 pandemic in Mexico City

BackgroundThe COVID-19 pandemic posed unprecedented challenges to healthcare systems worldwide. In densely populated urban areas such as Mexico City, the strain on hospitals was amplified due to the high volume of cases and resource limitations. Understanding the spatial and temporal dynamics of hospital demand is crucial for informing effective public health strategies and improving system resilience. MethodsA retrospective analysis of COVID-19 hospitalization data in Mexico City was conducted utilizing a line-list dataset from the SISVER surveillance system. The analysis included the spatial distribution of hospital demand using the weighted centroid of hospitalizations as well as a model system of the interactions between residential areas and hospitals as a bipartite network. The emergence of giant components in the network was used as indicators of system strain and the relationship between network strain and patient outcomes was evaluated. FindingsHospital demand in Mexico City exhibited significant spatial dynamics, with a northward shift in the weighted centroid of hospitalizations as the pandemic progressed. Despite the changing distribution of cases, a small subset of 17 hospitals managed the majority of hospitalizations. During high-demand periods, the network transitioned to a more disordered state, characterized by a giant component encompassing multiple neighborhoods. This disordered strain was associated with higher case fatality rates, particularly in patients over 40 years of age. InterpretationOur findings highlight the complex, adaptive nature of the healthcare system in response to the pandemic. The emergence of giant components in the hospital demand network can serve as an early warning indicator of the health system overload. Adaptive measures, such as the establishment of temporary COVID-19 units, were effective in mitigating strain. These insights can guide future public health strategies for rapid response and resource allocation in similar crises. FundingThis project was partially funded through CONACYT Project 320557 (to GA-J).