Self-report versus medical record data in a study of infant leukemia

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): ABSTRACT The case-control approach is the standard in conducting epidemiological studies of childhood cancer. It is unclear, however, how much bias exists in evaluating maternal recall in these studies. Medical records around the time of pregnancy provide an advantage in that they are collected prior to the diagnosis of cancer, thereby eliminating much recall bias. However, medical records can be difficult and expensive to collect. We seek to determine the validity of a mother's self-reported events around pregnancy by comparing that data to her medical record. Self-reported data from a case-control study of infant leukemia will be used. The interview asked mothers to recall events during and after her pregnancies, including medical conditions, menstrual, hormone-use, and contraceptive history, labor and delivery, and infant birth characteristics and hospitalization. Our goal is to calculate sensitivity and specificity values for items that are comparable in the two datasets, using the medical record as the "gold standard." Our purpose is to use these data to adjust the odds ratios estimates from the maternal interview for misclassification using probabilistic uncertainty analyses. Results of this analysis will provide us with valuable information regarding the need to collect medical records in future case-control studies. The analyses will indicate which health conditions mothers are able to recollect more accurately. If the mothers are able to recall events during pregnancy, for instance, then medical records may not be needed for future case-control studies. Further, the findings from our research will provide investigators and other interested parties with specific values for nonrandom error rates for more comprehensive evaluations of associations from the infant leukemia study. Lastly, the current application along with our other published research will provide a broad range of knowledge of the uncertainty associated with nonrandom errors. These published results will help illustrate to epidemiologists how to implement quantitative analysis of nonrandom error into their own research and thereby improve the interpretation of epidemiologic study results. PUBLIC HEALTH RELEVANCE: Projective Narrative (Public Health Sentence): This is an important and under-investigated area of epidemiology that has potential applicability to improving the collection of data for future case-control studies.