Funnel plot for rates

[sarahcjj]

Has anyone had success adapting Rick Wicklin's funnel plot for proportions code https://blogs.sas.com/content/iml/2011/11/23/funnel-plots-for-proportions.html for rates that can be >1? 

Briefly, I'd like to identify "high-intensity" antibiotic prescribers. My y-axis would be prescriber's antibiotic prescriptions / number of patient visits (which can be >1) in 1 year and my axis-would be number of patient visits in 1 year.

Thanks in advance for any help.

[Rick_SAS]

Yes. See

• A funnel plot for immunization rates
• Use a funnel plot to visualize rates: The case fatality rate for COVID-19 in North Carolina counties

Both articles include links to the SAS programs that generate the funnel plots.

[sarahcjj]

Thank you so much. I read in one of the comments that the funnel plot code is for binomial proportions between 0 and 1. I have observations where the number of events is greater than the number of trials. Would the code still be appropriate for these data? 

[Rick_SAS]

You can try it. The upper/lower curves on the funnel plot are determined by the average proportion, which is the total number of events divided by the total number of trials. If this value is less than 1, you can compute and plot the curves. The actual data values are just displayed in the scatter plot.

I encourage you to think about what statistical model might be appropriate for your data.  If you have rates that are greater than 1 or negative, then there are probably more appropriate models.

[sarahcjj]

Thanks so much for your help