The effects of spatial population dataset choice on estimates of population at risk of disease