Civil Engineering Reference
In-Depth Information
Because incidence is a rate calculated by following people over time, and many studies are cross-sec-
tional or retrospective (case-control), other measures, such as the prevalence ratio and the odds ratio, have
been developed to summarize the association between exposure and outcomes for these other study
designs. Our analysis focused on associations expressed by such risk estimates as the odds ratio and
the relative risk. These estimates were retrieved from the original article or calculated when sufficient
raw data were presented.
2.1.3.2 Definition of Measure: Attributable Risk
The attributable risk is another measure used to help generate inferences. In its simplest form, it is the
difference between the incidence in those exposed and those unexposed — a risk difference. This risk
difference is thought of as attributable risk in that, in theory, removing this exposure entirely would
reduce the frequency of the outcome to the level of those who are unexposed. Rotham and Greenland
(1998a, b) discuss some of the limitations of this simple assumption. Attributable risk is often calculated
as a ratio rather than a difference: risk in the exposed is divided by risk in the unexposed, producing an
attributable fraction. The attributable fraction is the proportion by which the rate of the outcome among
the exposed would be reduced if the exposure were eliminated. This fraction is calculated as the ratio of
(RR
RR, where RR is the relative risk or the prevalence ratio of risk in the exposed compared with
the unexposed:
1)
2
/
AF
e
¼
(RR
1)
=
RR
(2
:
1)
The attributable fraction helps scientists and policy makers recognize that in many cases a variety of
factors contribute to the total incidence of a disease or other outcome, so that removal of an exposure
typically does not reduce the outcome rate to zero. However, in its simplest form, the attributable risk
is a measure that suggests that if the offending exposure were removed (by intervention or regulation),
then the amount of disease outcomes would be estimated to be reduced by the calculated amount. As is
noted in the following para, this simple summary is enmeshed in caveats.
It is important to recognize in this calculation that the result depends on what is included. This is, if
one considers a calculation of one factor as it relates to an outcome and then performs a separate calcu-
lation for another factor for the same outcome, there is overlapping (correlation) between factors that
could make the sum of the two separate factors sum to more than 100%. Attributable fraction, then,
represents a crude but important estimation of the impact of control of risk factors. An estimate of
the attributable fraction for a multifactorial disease such as a musculoskeletal disorder provides only
an estimate of the relative importance of the various factors studied. It is not, and cannot be, considered
a direct estimate of the proportion of the disease in the population that would be eliminated if only this
single factor was removed (Rotham and Greenland, 1998a). Rather it provides guidance to the relative
importance of exposure reduction in those settings in which the exposure under study is prevalent. Con-
sequently, we have not attempted to rank or further interpret the findings for attributable fractions and
have chosen only to report them as a rough guide to the relative importance of the factors in the study
settings in which they have been examined.
In this review, the relative risk in longitudinal studies and the prevalence or odds ratio in cross-
sectional surveys were used to calculate the attributable fraction for the risk factors studied. For
example, if workers exposed to frequent bending and twisting have a prevalence of low back pain that
is three times that of those not exposed, then among the exposed the attributable fraction will be:
AF
e
¼
(3
1)
=
3
¼
0
:
67
(2
:
2)
By this hypothetical calculation, 67% of low back pain in the exposed group could be prevented by
eliminating work that requires bending and twisting.
Search WWH ::
Custom Search