Home landline telephone coverage and potential bias in epidemiological surveys.