Sample size or CI help

#1
Trying to solve a work problem on Sample Size and confidence intervals

Hello! Maybe it’s because I’m fried by now, but I’m trying to help some physicians at my job with a problem. Say we have information that says our entire population’s prevalence of diabetes is 20%. And the population is 50,000 patients (so 10,000 diabetics). We need to chart audit to make sure that 20% is correct. Reading the charts on 10,000 would take way too long. How many supposed diabetic charts would need to be reviewed to be 90% confident the accuracy?
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
So if the truth is 20%, how many charts to review to have a CI that includes 20%, 90% of the time?
 

katxt

Active Member
#3
I haven't quite got my head around the problem. It sounds like you have a collection of charts from supposedly diabetic patients. You want to review these charts and from them determine the proportion of diabetics in the wider population. This doesn't seem like the right data to me. What am I missing?
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
Agreed. The typical process is that you don't know the truth and a person wants to know how many charts need to be collected to have an estimate with blank amount of precision. Think voting polls. They randomly select so many poll exit'ers to get an estimate and the number of people sampled can be based on how precise they want the estimate's bounds to be.

There is a push to do sample size calculation this way - based on precision.
 
#5
Trying to solve a work problem on Sample Size and confidence intervals

Hello! Maybe it’s because I’m fried by now, but I’m trying to help some physicians at my job with a problem. Say we have information that says our entire population’s prevalence of diabetes is 20%. And the population is 50,000 patients (so 10,000 diabetics). We need to chart audit to make sure that 20% is correct. Reading the charts on 10,000 would take way too long. How many supposed diabetic charts would need to be reviewed to be 90% confident the accuracy?
Video on calculating sample size.