There is actually a diagnosis done at the screening visit, which comes before the baseline visit. The Clinical Core recommends the use of the diagnosis at baseline, since more information is available at that time to make a clinical assessment. The way the diagnosis information was captured, however, will reflect a "change" in diagnosis from the screening visit (which as you notice happens very infrequently). If a "change" is noted at the baseline visit, treat the diagnosis as what it changed to (for example, "MCI to Dementia" should be a Dementia diagnosis at baseline).

Hi, thanks for the answer.
So, if the patient changes from state x to state y it gets the label 'x to y' on the visit that the change occurred. However, when I was analysing the table 'UPENNBIOMK_MASTER' I have noticed some patients did have label change between baseline and the m12 visit, but it was not documented in that format:

label_in_the_bl -> label_in_the_m12 Number of occurrences

MCI -> MCI to Dementia 25
MCI -> Dementia 9
MCI -> MCI to NL 3
MCI to Dementia -> Dementia 1
MCI -> NL to MCI 1
Dementia to MCI -> MCI 1
NL -> NL to MCI 1
NL -> MCI 1
Name: DX, dtype: int64

As we can see, 25 patients were described as 'MCI to Dementia' in the m12 visit, but there were other 9 patients who have also transitioned. I thought it could depend on whether it was in ADNI1, GO or ADNI2, but they were all from ADNI1.

Why is that? Shoudn't all labels be documented in the same way when there is a change of diagnosis?

I have not looked closely at the data to determine an exact answer, but it is possible that the differences arise due to the way in which diagnosis information was captured in ADNI-1/GO/2 (even for ADNI-1 participants that continued into ADNI-GO/2, though that would be an unlikely explanation for a m12 visit). If you send an RID that has the MCI to dementia at m12 and another that has just dementia (but starts as MCI), I can look into it a bit more closely. In general, there has been some confusion in the recording of diagnoses, so to identify transitions, it is generally a good idea to compare diagnoses across visits to make sure you capture all of the transitions.