Stats for comparing rates....

[busterboyuk]

I would be very grateful for some pointers re: the following.

I have summary data for 2 years of information regarding the number of blood transfusions given to a group of preterm babies

Year Blood transfusions Pt days Transfusions/1000 pt days
1 293 6032 39.11
2 251 4987 33.52


Some babies received several transfusions whilst others received none. I have calculated there ratios and 95% CI for these ratios BUT is there a more robust way of determing if the rates (transfusion/1000 pt days) is different?

Many Thanks!!!!

[busterboyuk]

Can anybody please assist?

Year 1 39.11 Transfusions/1000 pt days (293 transfusions in 6032 days)
Year 2 33.52 Transfusions/1000 pt days (251 transfusions in 4987 days)

Thanks
C

[TheEcologist]

Can anybody please assist?

Year 1 39.11 Transfusions/1000 pt days (293 transfusions in 6032 days)
Year 2 33.52 Transfusions/1000 pt days (251 transfusions in 4987 days)

Thanks
C

Try an exact binomial test or an equality of proportions test using the number of transfusions has success and the number of days as trials. You are however now working with the daily probability of a transfusion taking place and testing for differences in probabilities.

You will however be assuming that every day is an independent "Bernoulli trial". I don’t know if this is applicable in your situation as your data explanation is sparse at best. You will have to be the judge of the logic.

Alternatively you can use a G-test or Chi-squared test but again you'll work with differences in probabilities/proportions.

If you use R, the functions prop.test() , binom.test() or chisq.test()

In SPSS, use the help functions :P