Comparing the before and after of filling np.nan values on a dataframe level using pandas describe()

I am trying to compare the difference for the before and after of filling the NA values and then using describe() method.

for example, first dataframe:

idx A   B
1   NA  5
2   NA  4
3   3   3
4   5   NA
5   6   7

after fill na

idx A   B
1   3   5
2   3   4
3   3   3
4   3   3
5   6   7

i wish to compare describe between the difference in data after filling the NA values, with random data combination.
Original dataframe has 80k rows with around 30% na of different columns(total of 30 columns)

ideal result:
results with no change should show 0
results with change will show the difference (ie, mean=2 (from 3 to 5))

attempt 1:
subtract them manually with a method, but it is not as clean as I would like

attempt 2:
create two dataframe, and use compare, and then describe, can this be cleaned up ?

Many thanks.