Paired Samples
Solution There is a tendency for users to want to generate average change from the Average Patient Profile report. This is an incorrect way to do this. The number should be taken from the Program Change Summary.

Let's run through a simple example to explain why. Its rooted in the theory of Paired Samples.

Bob Smith
---------
Pre Program: 50
Post Program: 70
% Change: +40% ((Post-Pre)/Pre)

Jane Doe
---------
Pre Program: 67
Post Program: 85
% Change: +26.8% ((Post-Pre)/Pre)

Make sense? So if these were your 2 patients, what's your average patient change for your program? Then you average 40% and 26.8%, which is 33.4%.

That's how the program change summary report works.

On the Average Patient Profile report, it gives an average pre and average post value. That's OK. Those values would be:

Pre Program: 58.5 ((50+67)/2)
Post Program: 77.5 ((70+85)/2)

Those numbers represent the "average patient" that walked in the door.
Now everything is correct to this point. The leap that you made was that you then took 58.5 and 77.5 and calculated the % change. Our report does not do that calculation because it violates paired samples. Ignoring that, if you do it, you get 32.47%. That is less than the 33.4% we calculated above.

These numbers were just contrived by me but you can come up with situations where those numbers get much further apart (or closer together). But the key point is that their is a right way and a wrong way to calculate the average change for your program.

Article Details
 Article ID: 43
 Created On: 01 Apr 2011 12:53 PM MDT