Healthcare and Medicine Reference
In-Depth Information
plays a role. Furthermore, in just about any operational
setting it is difficult to establish the temporal consistency
of (or correct for the temporal variation in) other factors
that contribute to accident risk.
One way to circumvent the difficulties of disentangling
the various factors that contribute to incidents and
accidents and that are or are not related to fatigue is to
predict accident risk on the basis of descriptive modeling
of published incident data. This idea has led to a risk index
for assessing work schedules. 3,21,68 Because this approach is
purely descriptive, however, its generalizability across dif-
ferent shift systems is uncertain. Prediction strategies
based on the neurobiology of sleep-wake and fatigue give
greater confidence that findings for one shift system will
generalize to another, because the underlying mechanisms
involved do not change.
For the purpose of predicting accidents with a biomath-
ematical model of fatigue and performance based on sleep-
wake, accident risk may be defined as an odds ratio. 69 This
odds ratio is expressed as the percentage of accidents co-
occurring with a given range of predicted fatigue (inci-
dence level) divided by the percentage of time spent
working in that range of predicted fatigue (exposure level).
For example, if 20% of accidents occur in a particular
range of predicted fatigue while 10% of work time is spent
in this range of predicted fatigue, then that specific range
of fatigue is associated with a doubling (odds ratio 2) of
accident risk.
A study published by the Federal Railroad Administra-
tion (FRA) applied this technique to validate accident risk
predictions made with a biomathematical model of fatigue
and performance. 4 The data set at hand concerned 400
human-factors accidents and 1000 non-human-factors
accidents in railroad operations. The SAFTE fatigue
model 70 was used to predict operator performance, on an
effectiveness scale from 0 (worst) to 100 (best), at the time
of the accidents. The model predictions were based solely
on the work histories and estimated sleep opportunities of
the locomotive crew in the 30 days leading up to each of
the accidents.
The results of this analysis indicated that there was a
significant, high correlation between reduced predicted
crew effectiveness (i.e., increased fatigue) and the risk of a
human-factors accident, as displayed in Figure 67-4 . No
significant relationship was expected, and none was found,
for non-human-factors accidents. At predicted effective-
ness scores below 70, the risk of human-factors accidents
was elevated above chance level and was greater than the
mean risk of non-human-factors accidents. 71 At such low
levels of predicted effectiveness, accident cause codes
(defined by the FRA to indicate the factors that caused the
accident, such as passing a stop signal or exceeding autho-
rized speed) were of the sort expected to be related to
fatigue, which confirmed that the detected relationship
between accident risk and predicted effectiveness was
meaningful.
A significant relationship was also found between
reduced predicted crew effectiveness (fatigue) and increased
accident damage costs. Human-factors accident costs were
2.5 times greater when predicted effectiveness was below
about 77.5 as compared to above 90. Figure 67-5 shows
the relationship between predicted effectiveness and
80%
65%
60%
40%
22%
20%
6%
10%
7%
0%
-16%
-20%
-40%
100-90 90-80
80-70
70-60 60-50
Below
50
Predicted crew effectiveness
Figure 67-4 Human-factors accident risk, from aggregated
data of five railroads, at specific predicted ranges of decreasing
effectiveness (increasing fatigue). The diagonal line indicates
the linear relationship between predicted crew effectiveness
and relative risk ( r = 0.93). A relative risk of 0% corresponds
to an odds ratio of 1, which is the chance level of risk. Less
than 0% is reduced risk; greater than 0% is elevated risk. A rela-
tive risk of 65% means that accidents were 65% more frequent
than chance.
100%
80%
60%
40%
20%
0%
-20%
-40%
-60%
-80%
86%
$
29%
#
6%
#
$
#
-10%
-25%
$
-57%
Above 90
77.5 to 90
Below 77.5
Predicted crew effectiveness
Figure 67-5 Human-factors accident risk and damage risk as
a function of predicted ranges of decreasing effectiveness
(increasing fatigue). The zero line represents a risk level equal
to the overall (average) risk. #, frequency of accidents; $,
damage costs.
damage risk (i.e., the combination of the risk of human-
factors accidents and their damage costs).
The significant relationships revealed in this FRA study
confirm that accident risk can be predicted, at least to some
extent, on the basis of fatigue as predicted from sleep-
wake-work schedules. The specific risks associated with
given levels of predicted fatigue are likely to be occupa-
tion-specific and related to the demands of the job, and
data from rail operations should not be used to predict
 
Search Pocayo ::




Custom Search