Multiple cause-of-death data as a tool for detecting artificial trends in the underlying cause statistics: a methodological study