Using the exposure-response functions, expressed as
relative risk (RR) per 10 g/m3
, and the health frequency
per 1000 000 inhabitants, we calculated for each health
outcome the attributable number of cases (D10) for an
increase of 10 g/m3 PM10 (figure 1), as: D10=(RR-1)*P0
where P0 is the health frequency, given an exposure E0
and RR is the mean exposure-response function acrossthe studies used (table 1). The exposure-response
functions are usually log-linear. For small risks and
across limited ranges of exposure log-linear and linear
functions would provide very similar results. However, if
one may apply the method to populations with very large
exposure ranges, the impact may be seriously
overestimated on the log-linear scale. Thus, we derived
the attributable number of cases (D in figure 1) on an
additive scale. The calculation of D10 requires, however,
prior definition of P0. We defined P0 as the health
outcome frequency that one may expect, given the
reference level of exposure, E0 (7·5 g/m3 annual mean
PM10). We derived P0 from PE, applying the risk function
backwards from PE (the current population mean
exposure), to P0 (chosen to be 7·5 g/m3
). This
procedure is in line with the prudent “at least” approach.
With D10 and the number of people living in each
category of exposure, the total number of cases
attributable to air pollution can be calculated.