Observer bias in blood pressure studies.