When there isn't much data, you shouldn't be segmenting your population. Instead it is common in data analysis to collapse variables to have meaningful observations. Why would that be any different here?