Statistical tests for differences in estimated totals?

I've estimated total health care costs for patients by county in the United States by various scenarios (e.g. if they only get to go to specialty doctors vs if they can use nurse practitioners). I used Wilcoxen to test differences in mean costs, but I'm wondering if I should do a statistical test to test differences in total costs across the country by scenario? For example, the weighted mean cost per patient per year for County A is say $5000 in Scenario 1; I compared this mean to $4000 for the same county in Scenario 2. The national total in Scenario 1 is $100 million vs $90 million in Scenario 2. Can I test if these estimated totals are statistically significantly different?