Institutional transmission of airborne infections such as tuberculosis (TB) is an important public health problem, especially in resource-limited settings where protective measures such as negative-pressure isolation rooms are difficult to implement. Natural ventilation may offer a low-cost alternative. Our objective was to investigate the rates, determinants, and effects of natural ventilation in health care settings.The study was carried out in eight hospitals in Lima, Peru; five were hospitals of "old-fashioned" design built pre-1950, and three of "modern" design, built 1970-1990. In these hospitals 70 naturally ventilated clinical rooms where infectious patients are likely to be encountered were studied. These included respiratory isolation rooms, TB wards, respiratory wards, general medical wards, outpatient consulting rooms, waiting rooms, and emergency departments. These rooms were compared with 12 mechanically ventilated negative-pressure respiratory isolation rooms built post-2000. Ventilation was measured using a carbon dioxide tracer gas technique in 368 experiments. Architectural and environmental variables were measured. For each experiment, infection risk was estimated for TB exposure using the Wells-Riley model of airborne infection. We found that opening windows and doors provided median ventilation of 28 air changes/hour (ACH), more than double that of mechanically ventilated negative-pressure rooms ventilated at the 12 ACH recommended for high-risk areas, and 18 times that with windows and doors closed (p < 0.001). Facilities built more than 50 years ago, characterised by large windows and high ceilings, had greater ventilation than modern naturally ventilated rooms (40 versus 17 ACH; p < 0.001). Even within the lowest quartile of wind speeds, natural ventilation exceeded mechanical (p < 0.001). The Wells-Riley airborne infection model predicted that in mechanically ventilated rooms 39% of susceptible individuals would become infected following 24 h of exposure to untreated TB patients of infectiousness characterised in a well-documented outbreak. This infection rate compared with 33% in modern and 11% in pre-1950 naturally ventilated facilities with windows and doors open.Opening windows and doors maximises natural ventilation so that the risk of airborne contagion is much lower than with costly, maintenance-requiring mechanical ventilation systems. Old-fashioned clinical areas with high ceilings and large windows provide greatest protection. Natural ventilation costs little and is maintenance free, and is particularly suited to limited-resource settings and tropical climates, where the burden of TB and institutional TB transmission is highest. In settings where respiratory isolation is difficult and climate permits, windows and doors should be opened to reduce the risk of airborne contagion.