I am guessing income can be a pretty ugly distribution and they are trying to address this. Think about the 1%'ers. It would be nice if they provided a histogram of income, so we could understand what the raw data looked like. I would imagine it is an acceptable application, but I would recommend that you look at comparable studies and see what is common practice in the field. Lastly, it is interesting that there is not a trend across quintiles, just all others are higher risk than referent group.