The idea is to generate the checksum right after encryption has occurred. Then whenever the integrity of the file needs to be checked, the checksum of the encrypted file can be recomputed and compared to the baseline value. This will detect corruption to a high level a confidence, if a suitable checksum function is used. It appears in your case that this was not done, so there is no baseline with which to compare. Detecting if the encrypted files have been corrupted from their original state is now impossible.

Computing the checksum of the files in their current state will enable you to detect any further corruption. If the concern is deteriorating media, this might be useful.

I understand that certain encryption methods leave telltales in their encrypted files, some of which could be detected via a statistical analysis. This would allow you to tell if a certain file still looks "pretty much like most files encrypted using this method." If that would actually provide any level of reassurance to you, go for it.