The impact of an improved wood burning stove (Patsari) in reducing personal exposures and indoor concentrations of particulate matter (PM(2.5)) and carbon monoxide (CO) was evaluated in 60 homes in a rural community of Michoacan, Mexico. Average PM(2.5) 24-h personal exposure was 0.29 mg/m(3) and mean 48-h kitchen concentration was 1.269 mg/m(3) for participating women using the traditional open fire (fogon). If these concentrations are typical of rural conditions in Mexico, a large fraction of the population is chronically exposed to levels of pollution far higher than ambient concentrations found by the Mexican government to be harmful to human health. Installation of an improved Patsari stove in these homes resulted in 74% reduction in median 48-h PM(2.5) concentrations in kitchens and 35% reduction in median 24-h PM(2.5) personal exposures. Corresponding reductions in CO were 77% and 78% for median 48-h kitchen concentrations and median 24-h personal exposures, respectively. The relationship between reductions in median kitchen concentrations and reductions in median personal exposures not only changed for different pollutants, but also differed between traditional and improved stove type, and by stove adoption category. If these reductions are typical, significant bias in the relationship between reductions in particle concentrations and reductions in health impacts may result, if reductions in kitchen concentrations are used as a proxy for personal exposure reductions when evaluating stove interventions. In addition, personal exposure reductions for CO may not reflect similar reductions for PM(2.5). This implies that PM(2.5) personal exposure measurements should be collected or indoor measurements should be combined with better time-activity estimates, which would more accurately reflect the contributions of indoor concentrations to personal exposures.
Practical implications: Installation of improved cookstoves may result in significant reductions in indoor concentrations of carbon monoxide and fine particulate matter (PM(2.5)), with concurrent but lower reductions in personal exposures. Significant errors may result if reductions in kitchen concentrations are used as a proxy for personal exposure reductions when evaluating stove interventions in epidemiological investigations. Similarly, time microenvironment activity models in these rural homes do not provide robust estimates of individual exposures due to the large spatial heterogeneity in pollutant concentrations and the lack of resolution of time activity diaries to capture movement through these microenvironments.