When importing from CSV, allow the user to select a field in the incoming data as the 'unique key' and match this to the same named field in the existing collection. Allow the user to choose the standard options (like you currently offer for the _id field) for importing based on the presence of duplicates.