Is it possible to determine whether households completed the survey more than once if over 5 years of PUMS data are being combined?
From the ACS design and methodology report (page 45 of the PDF):
"One of the ACS design requirements is that no HU address be in sample more than once in any five-year period."
I just reread your question and missed the "over 5 years of data" part.
The way households are sampled, a HU could be sampled again 5 years later (but not 6 or 7 because of rotating sub-frames). But there is no way of identifying if two observations 5 years apart came from the same HU.