Floor and ceiling effects when measuring change

I am using a psychological measure with possible scores of 1 to 20. This is measured twice in the same participants (with 5 years between T1 and T2). At both points the score is skewed to the left, but there are participants who score all the way up to 20.

I want to measure the change in this variable as my outcome variable, but that means I have to deal with floor and ceiling effects. Obviously, it is impossible to detect a decrease in a person that has a score of 1 at T1, and to detect an increase in persons who score 20 at T1.

Is there any way I can make sense of the change variable? Would ranking the participants be an option? Or what about using the difference between T1 and T2 as my DV in a linear regression where I control for T1 score? I am very confused right now and don't know if these alternatives make sense at all.

My sample size is approx 2000.