EMS calls are most likely to be related to socio economic factors than physical geography. Other things being equal the absoute numbers will be proportional to population size.

in terms of regional service provision (call centres, ambulance, paramedics) absolute numbers are relevant.

For the same population, same socio economic level, transport requirements also depend on area to be covered.

Depending on focus of your study, all three measures could be used.

Thank you James Leigh

My purpose is finding high risk areas and also comparing calls distribution in day and night and also different seasons and so on..

As in much social science research, it depends. So I would use all three in the analysis pointing out the similarities and differences. There is often not a single answer, but which is best depends to what purpose it is being put.

Point data such as EMS calls are also amenable to conversion to kernel density showing heat maps of the density of such calls. Kernel density has been used to plot crime statistics as well as health data. Why not use it here? This would enable the viewer to focus in on areas of high call density and implement policies in terms of social or health services in those areas of high need. Kernel density method in ArcGIS fits a quadratic aerial function with a parabolic shape in x- and y- which is centered around a point and decaying quadratically in the x- and y- directions. I have used it successfully in plotting traffic crash data, traffic volume data and even population prediction as in Article Use of Kernel Density and Raster Manipulation in GIS to Pred...