RE: Which control chart to monitor insulin protocol in an ICU

["David Maley" <dmaley1@columbus.rr.com>]

Interesting, real-life question.  I suggest you start your thinking by
being clear about the aim or goal of your study.  I'm assuming your
interest is in detecting individual patient condition in order to react
to their situation appropriately and perhaps more quickly.  (Alternative
aims could be comparing one treatment method to another, or how the
population is trending over time, etc. and my subsequent comments would
not apply.)  

If you're interested in monitoring individuals, never mix reading from
different individuals.  Individual readings would give the most
information.  Each individual's glucose checks are likely not happening
often enough to warrant an X-bar chart.  I suggest you use the X-mR
chart (individual and moving range) -- one for each patient.  This would
show the natural variation in each individual's glucose level over time.
There is a common misperception that the underlying distribution needs
to be Normal.  It does not.  The control limits and the other control
signals will indicate when a true change has occurred in the glucose
level, not just due to common cause variation.  

It's OK that the readings do not go negative and I see no reason to
transform the data with natural logs.   Just plot each individual's
readings and see what the data tells you.  

Let me (and the list) know what you find.  It sounds like an interesting
study.

Regards,
David Maley
dmaley1@columbus.rr.com 

From: robert.shaw@med.va.gov [mailto:robert.shaw@med.va.gov]