Biomedical Engineering Reference
In-Depth Information
distribution and is said to show only
common cause
variation
. For an example, an endemic disease
process such as the incidence of influenza might
be considered in control if the charted daily frac-
tion of hospital admissions diagnosed as ILI lie
within control limits, as in Figure 3.2. The process
is held
out of control
and suggestive of
special
cause variation
, such as a rise in disease inci-
dence to epidemic levels, when the confidence
limits are exceeded. Since surveillance applications
are usually concerned with unexpected increases,
interest is often restricted to upper confidence
limits and one-sided tests.
Accounting for common cause variation depends
on the type of quantity chosen for monitoring and
on its underlying distribution. Several common
distributions are associated with canonical chart
types. For example, for continuous quantities (e.g.,
blood pressure and temperature) often considered
having a normal distribution, X-bar and S charts
are used for the mean and standard deviation,
respectively. P-charts, with limits derived from the
binomial distribution, are used for proportions such
as the fraction of total admissions with gastroin-
testinal chief complaints. U-charts derived from
the Poisson distribution are often used for count
data such as the weekly number of admissions for
neurological disorders. Numerous other basic and
hybrid chart types are available.
14
UCL
12
10
8
6
4
LCL
2
0
2
4
6
8
10
12
14
16
18
20
Day
Figure 3.2
Sample control chart.
