The use of snapshot personal exposure monitoring in estimating annual exposure to air pollutants for epidemiological studies

Background While advancements in sensor technology have broadened the use of personal monitors to estimate individual exposure to air pollution in recent years, significant challenges remain in extending their use for application in epidemiological studies. By collating measurements from several personal exposure (PE) campaigns in London, we aimed to develop a method for extending snapshot PE monitoring to estimates of annual mean exposure to pollution from outdoor sources at individual level. Methods Our conceptual model for annual extrapolation included three factors: a) variation in ambient concentrations, b) variation in infiltration efficiency of residence and other frequently inhabited buildings, and c) mobility behaviour. We utilised a dataset of 200+ million PE measurements from 427 participants of diverse demographics - school children, COPD patients, professional drivers and healthy adults - to characterise these three factors during snapshot monitoring (varying from four days to six months). Location-adjusted ambient concentrations, coupled with seasonal infiltration factors were used to extend results to annual mean personal exposure estimates for PM2.5, NO2, O3 and BC. Methods described elsewhere were used to censor exposure to indoor sources from the measurements. Results Residential infiltration factors did not show significant variation between seasons, suggesting that ventilation behaviour was not a major determinant. Each demographic exhibited distinct patterns in mobility behaviour, although impact was limited by relatively small proportion of time spent travelling. Behaviour modelling of participants with at least four months' PE measurements showed that some participants had repeated routines, but others were unpredictable. Conclusions The extension of snapshot PE monitoring to annual mean exposure estimates requires assumptions relating to participant behaviour over time, notably mobility and residential ventilation. The impact of these assumptions will vary between individuals and demographics. Extrapolation of snapshot monitoring is more robust for those with predictable routines. Keywords Air pollution, personal exposure, sensors, measurement error