It depends on what you are looking for. In case you are looking for the prevalence of affected people with an event, then you might consider the population so as to avoid duplication or over estimation. In case you are looking for a magnitude of an event then you might consider per tooth estimation.

Thank you Sir. Does using Patient level data, instead of Tooth level data, cause possible aggregate bias? Is that what you meant by overestimation or duplications? For example, I am trying to estimate amount of bone loss (in milimeters) that would occur around a dental implant in well-controlled vs poorly-controlled diabetics. So, as you explained, I am looking for the magnitude of bone loss (continuous outcome) - hence i used per-implant estimation. Is this a correct thing to do? I am not looking for the prevalence of bone loss in the population of well or poorly controlled diabetics.

There are two sorts of questions : questions about events and questions about people. As Mohanad Kamaleldin Mahmoud Ibrahim pointed out, you may be interested in the proportion of people who are affected. This tells you the number of people who require services of some sort. Per tooth can tell you how much work is required per person, for example.

Have a look at the issues discussed here : Article Description of comprehensive dental services supported by th...

Thank you for the answers, it is of great help to me Mohanad Kamaleldin Mahmoud Ibrahim and Ronán Michael Conroy .