IPUMS USA is currently re-developing the STEPMOM and STEPPOP variables for the 1970-onward samples as part of the update to family interrelationship variables. However, for most modern samples, the primary way stepparents are identified is by improbable age difference (outside the range 15-49 for mothers and 15-64 for fathers), and if the parent was only linked because they were married to a clear parent of the child. So using age-differences and the MOMRULE and POPRULE variables to determine the strength of these parental links should allow you to create a systematic criteria. I recommend also using the attach characteristics tool when creating your extract to attach the ages and relationship codes of parents to children, along with any other variables you think may be helpful in identifying non-biological links.
I hope this helps!