The reliability of retrospective treatment reports