Meta-analyses of short-term effect of air pollution use environmental exposure measurements defined as the concentration average between selected monitors. A simple quality index, Pearson's correlation coefficient between each pair of monitors, has been used in sensitivity analyses and meta-regression. To better characterize the degree of homogeneity in population exposure we propose to use Lin's concordance coefficient and the correlation coefficient between the difference and the average of each pair of values. Using simulated data and real data from the city of Rome (1998-2001) and Milan (1996-2002) we identify three conditions responsible of exposure misclassification when averaged concentrations are used in the analysis.