This study has provided evidence to indicate that ingestion and inhalation, together with inoculation, are important routes for the development of melioidosis in Thailand. A range of activities were found to be independently associated with melioidosis, including presumed inoculation during unprotected occupational exposure to soil or environmental water, ingestion by eating contaminated food or drinking untreated water, and inhalation by outdoor exposure to rain. We also confirmed the presence of B. pseudomallei in water obtained from wells and boreholes and from piped water supplies, and recorded that a number of cases had consumed untreated water from these sources prior to presentation with melioidosis.
This is the first study to show that ingestion is an important route of human B. pseudomallei
infection. Based on data obtained from The Provincial Waterworks Authority of Ubon Ratchathani province, only people living in the town of Ubon Ratchathani (3% of the provincial population) receive piped chlorinated water 
. Tap water quality control does not include assessment for the presence of B. pseudomallei
, which we found in 12% of tap water samples (32/273) and which a number of cases had consumed without adequate treatment. Unlike observations made in Hong Kong (14), none of the collected rainwater samples tested positive for B. pseudomallei
. It is customary for the drinking rain water to be collected in large earthenware pots situated close to the house, the water in which can reach temperatures in excess of 40°C. This may explain the negative culture results in our setting, although it is also possible that the bacterial count was below the level of detection of our methodology. We recommend that all non-bottled water should be boiled prior to consumption. Although filtration is an alternative method of water purification, we observed that filters were poorly maintained and detected B. pseudomallei
in some filtered water samples. In view of this, we do not recommend the use of filtration.
This is also the first evidence to indicate that exposure to rain is an independent risk factor for melioidosis. Exposure to dust clouds was a significant risk on univariable but not multivariable analysis. This is the first study to identify that smoking may be associated with acquiring melioidosis. Smoking could decrease the effectiveness of the local inflammatory response and increase the risk of infection by inhalation. Although microbiological confirmation of aerosolized B. pseudomallei
has not been published, this could be due to poor sensitivity of the techniques used or a very low bacterial concentration. In experimental mice, inhalation of only 5 CFU can result in death within a few days 
Our study has several limitations. Relatives were asked about activities of daily living when cases or controls were not capable of providing this information, and it is possible that they were not aware of the full spectrum of activities undertaken. It is also possible that cases who were aware of their diagnosis of melioidosis might mention risk factors more readily than controls. This would only be the case if people were knowledgeable about melioidosis, but in a recent survey most Thai people (72%) had not heard of melioidosis, and the remainder had heard of the word but did not know what it meant (unpublished data). The education about melioidosis was given to all participants after the interview. There may be other factors associated with a risk of melioidosis that we failed to examine, and we cannot evaluate the relative risk of a matched variable. The study was not powered to identify risk factors with a relative risk less than 2.0. The criteria specified for matching were stringent and we were unable to find controls for some patients. A diagnosis of diabetes was more common in patients with melioidosis who were excluded because of failure to find a control, compared with those enrolled as cases (72% v.s. 42%). This is because the prevalence of diabetes in patients admitted to the hospital with non-infectious conditions (potential controls) was low, and finding matched controls for diabetic cases was more difficult than that for non-diabetic cases. However, diabetes is the strongest risk factor for melioidosis 
, and matching for diabetes is very important to control the possible confounding effect. We consider it likely that our findings are applicable to similar settings in neighboring Asia but may be less applicable to more distant geographical settings including Australia.
Current efforts are being directed toward increasing public awareness and implementing preventive measures for melioidosis in endemic areas, particularly Thailand. A vaccine that protects against B. pseudomallei
infection is not available and there is no prospect of one being developed and ready for use in the near future 
. There is, therefore, every reason to look for alterative solutions to prevent melioidosis, both in people living in regions of the world that are endemic for melioidosis, and visitors (e.g. travelers and military personnel) to these regions. The Ministry of Public Health in Thailand has included melioidosis on a priority list of emerging diseases in Thailand, and public health campaigns for melioidosis prevention based on the knowledge of this work are being developed. These actions will include the implementation of an education programme based on the recommendations provided here, the improvement of infrastructure relating to effective treatment of public water supplies and access to protective clothing, tractors and other machinery to reduce contact time of farmers with soil and environmental water. Further studies are required to model the cost-benefit of guidelines for the prevention of melioidosis, together with their acceptability, up-take and impact on rates of melioidosis.