According to this CDC report, the median number of reported sexual partners for females aged 15-44 is 3.2, and for males 5.1. Tables on pages 19 and 20 report these statistics for a variety of subgroups, and all 40 of the reported subgroups have medians like this.
I am aware of the classic case (usually occurring only in grade-school textbooks) where you have an even number of observations, and the median is defined as the arithmetic mean of the two "middle" observations. However, that does not explain how nearly all of these subgroups could end up with non-whole-numbers as medians. Can anyone explain what is going on?
If all your reported statistics have a small number of distinct reported values, and you have many samples, then according to some conventions all the median statistics could come out to be fractions if there is a duplicate value surrounding a common value in the middle of the data after being sorted. Some people take the convention that if the exact middle value is 3 after sorting, and e.g. there are 9 values equal to 3 on the right of the middle value 3 after sorting, and none on the left, then the median e.g. is 3.1 if the prior value before 3 is 2 and the next value after 3 is 4. This is not universally standard convention for reporting medians, however.