sample size for dichotomous outcome in retrospective observational study

#1
I really need some help here seeing if I can get this right. I seem to come up with something different every time. I am testing the hypothesis that there is no difference in the timing of removing chest tubes to the development of significant pleural effusions after surgery requiring drainage which is a simple yes/no incidence. I am using a retrospective observational study using chart review. There are two "standard" protocols used. One more restrictive believing leaving tubes in for longer reduces incidence. There might be difficulty in having equal numbers in the two groups based on protocols being assigned due to surgical preference. Most retrospective studies I have seen have imbalanced groups but also do not show any sample size calculations. Historical incidence is around 9% but SD from the limited number of studies I can find is +/- 4.5. (incidence is recorded as anywhere from 4-10%) I would consider a 10% reduction of incidence clinically significant but I cannot seem to find the right calculator to help me figure out the sample size I need for basic 80% power and two tailed test p<0.05.
I have tried openepi amongst others but every one I try comes up with numbers ranging from 278 to over 1100. Annual population having surgery is just 1000
Please help!