Abstract
<title>Abstract</title> <p>Urban floods can increase demand for care while making functioning healthcare facilities harder to reach or creating a need for evacuation. Protecting urban health systems, therefore, depends not only on preventing facilities from flooding, but also on maintaining access to care as roads, services and population needs change throughout an event. Yet flood risk assessments commonly rely on maps of maximum flood extent or depth, which cannot show when healthcare disruption begins, how long it lasts, or which population groups and facilities face the greatest continuity of care challenge. Here we develop a spatio-temporal framework that links hourly flood dynamics, road network passability, healthcare facility exposure, population presence and neighbourhood health vulnerability. We apply it to a one-in-400-year rainfall scenario, resembling a real event that caused the Limburg 2021 flood, to Rotterdam, the Netherlands. We show that healthcare disruption follows a different timeline from flood extent: hospital access loss begins almost a day before maximum inundation, the first facility crosses an operational flood threshold 10 hours before that, and severe road and facility flooding peaks later. Neighbourhoods form a typology of access loss profiles, ranging from early, short, scattered losses to persistent, high losses that continued through the end of the 10-day simulation. At the facility scale, impacts form distinct timing-duration profiles, separating early-hit facilities requiring anticipatory action, a cascade-period group facing simultaneous evacuation and coordination challenges, and later-hit facilities requiring continued support during recovery. Our findings show that urban flood resilience requires maintaining equitable continuity of healthcare services throughout the flood onset, peak, and recovery phases.</p>