Pre/Post Data

#1
Hello.

I have pre/post data that I would like advice on what would be the most appropriate data analysis method. I visited a group of clinics and conducted medical chart audits on patients 65 years and older and collected data on who received a flu shot. Either they did or did not (1/0). I calculated a coverage rate for that clinic (n/N). We then implemented different types of interventions to increase coverage rates, or at least we hope. Either the clinic adopted the intervention or did not (1/0). After a certain period of time, I re-visited the clinics and conducted another series of chart audits to measure coverage rates. The two different audits per clinic were not the same people. So I have baseline data and post-intervention data. First, I want to compare coverage rates for each clinic for baseline and post-intervention to test for signficant differences. I would also like to examine whether adoption of any of the interventions had any impact on coverage rates. Since a clinic could have adopted more than one intervention, I would like to examine the combination of interventions on coverage rates. I also have race and gender data for each clinic as well as the amount of vaccine the clinic had.

Any assistance would be greatly appreciated. I need a starting point.

Thanks
Enrique Ramirez
 

JohnM

TS Contributor
#2
In terms of a statistical test, you can directly compare pre-intervention rates with post-intervention rates for each clinic. Use a two-sample test of proportions.

Since the pre- and post-intervention samples were not the same people, you'll need to do an independent-samples test. This link goes into more details:
http://davidmlane.com/hyperstat/B73479.html

If the difference between the two rates is significant, you can say that there is general evidence supporting the effectiveness of intervention.

In terms of comparing the effectiveness of different intervention measures, all you can do here is make general observations that would require further study to statistically verify, since this was not a "controlled" aspect of the study.

Same idea for the race and gender and vaccine amount data - just general observations that may/may not lead to a more rigorous study...

For instance, a clinic may/may not have adopted a particular intervention to the same degree as another clinic, etc. and there may be other factors involved, such as degree of staff commitment....