When you count of pregnencies noted in Y(n) and deliveries in Y(n+2) there is something I can't explane using my highly rectricted knowledge in this area, and commn sense.

If you count pregranceis per member in Y1 and deliveries by same mebmers in Y2, numbers are as follows:

Y1: Registerd pregnencies in Y1: 33061, registered deliveries is in Y2: 4342, no deliveris for same members in Y2: 28719.

Y1: Registerd pregnencies in Y2: 33407, registered deliveries is in Y3: 3698, no deliveris for same members in Y3: 29709.

As far as I know, after registerd prregnancies in Y[i] without deliveries in Y[i+1], there should there should be at least 3/4-ths of deliveries in Y[i+1]. Still, this is not the case in actual data.

So, what happeded to wemen with registered pregrancies in Y[i] and did not deliver in Y[i+1] for 3/4-ths of them?

Is there some logical explanation for this, or it can be simply not consistent data?

But, if data is incosnsitent on such kind of level, how we shuld built our models? Based on what intput?

I've gone through the same analysis but found there was a strong ability to predict delivery in advance. The original market makers GBM captures about 2/3 of pregnancy hospitalizations