I needed to use
df_dm.dropna(axis = 1, how="all", inplace = True)
I was only dropping rows with all Nans since:
axis = 0
is the standard.