Correcting bias in the estimation of exposure-disease outcome associations

Drs. Wang and Spiegelman have recently begun research that corrects for bias in outcome misclassification as would be pervasively the case when claims and administrative data are used to evaluate interventions. The methods will be illustrated in an analysis of the impact of diabetes and diabetes medications on colorectal cancer incidence in Medicare claims data. By linking a subset of these data to the Nurses’ Health study, the outcome measures from the claims data can be validated.