This study employed two ventilation indexes: local mean age of air and air change rate per hour, to investigate wind-induced natural ventilation of 260 wards of a multi-storey hospital building in suburb of Guangzhou using computational fluid dynamics simulations. Using the surface-grid extrusion technique, high-quality hexahedral grid cells were generated for the coupled outdoor and indoor airflow field. Turbulence was solved by the renormalisation group k-model validated against experimental data with grid independence studies. Homogeneous tracer gas emission was adopted to predict room age of air. The air change rate of cross ventilation and single-sided ventilation can reach 30-160 h-1 and 0.5-7 h-1, respectively. Due to different locations of room openings on the balconies, natural ventilation of a room can be greatly better than its neighbouring room. The wind-induced cross ventilation highly depends on the distance from the room opening to the stagnation point and on the resulting pressure distribution on the target building surface. Furthermore, it is significantly influenced by the upstream buildings, the bent shape of the target building, and the prevailing wind directions. The coupled computational fluid dynamics methodologies with integrated ventilation indexes are useful for assessing the natural ventilation performance in other complex built environments.