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

["Prevette, Steven S" <Steven_S_Prevette@rl.gov>]

My first thought was an X-bar and s chart. 

SSP - yes that should be sufficient.  X-bar R would also likely work and
might be easier.

1) My concern is that the data is naturally skewed. You 
can have a blood glucose of 300, but not -200

SSP - True, but this happens a lot in control charts.  You are usually
plotting quantities that only exist as positive numbers.  I'd also want to
look at the distribution of the data for "typical" patienst who have been in
control to see what some of the actual distributions are before becoming
overly concerned.

2) Would a natural log transformation, then create a 
control chart help?

SSP - Likely it won't be necessary, but certainly nothing prevents you from
trying it and seeing if it is more effective than a non-transformed chart.
Control charting has been shown to be "robust" and distribution-independent.

3) Changing the continuous data to attribute data (in 
control / out of control ) does not seem correct as a 
glucose of 111 is very different that 300 although both 
would be out of control.

SSP - Correct.  You would be losing a lot of valuable information if you
converted the continuous data to go - no go.

Steve Prevette
Occupational Safety and Health
Fluor Hanford, A Fluor Government Group Project
ASQ Certified Quality Engineer
steven_s_prevette@rl.gov
509-373-9371