The interviewers conducting a survey asking about deaths were poorly trained and included deaths which occurred before the time period of interest.