Underreporting of cancer in medical surveys: a source of systematic error in cancer research.

Systematic errors occur in the reports of disease frequency derived from health surveys based on questionnaire interviews. Five hundred and thirty-three persons with clinically and histologically confirmed disease in a case-control study of cancer were interviewed in their homes by carefully trained interviewers using a standardized questionnaire interview schedule. Comparisons of the information obtained by interview about past history of cancer with cancer registry and hospital medical record information about the same people revealed serious underestimates of correct frequency and wide variation in the rates of correct reporting. The findings reported here and elsewhere add support to an essential requirement in medical survey research: the completeness and accuracy of responses in health interview surveys must be verified and the methods of verification must be reported before the results can be interpreted with confidence.