For each region you can fit a beta model to the set of proportions.
Parameterize the beta model so that one of the parameters becomes
logit(mean) and then calculate the fisher information and the score.
Alternatively, if you are familiar with WinBugs and you are happy to report
bayesian credibility intervals insteads of CIs then I think that would be a
lot easier to do.
You may leave the list at any time by sending the command
to [log in to unmask], leaving the subject line blank.