Misclassification of breast cancer as cause of death in a service screening area.

OBJECTIVE: The aim of this study was to assess the misclassification of cause of death for breast cancercases, and to evaluate the differential misclassificationbetween cases detected in an organized screeningprogram and cases found in current clinical practice. METHODS: All deaths occurring between 1999 and 2002 within breast cancercases were linked to hospital discharge records. Deathcertificates and latest available hospital discharge notes were classified into various categories. We created a classification algorithm defining which combinations of categories (of deathcertificates and hospital discharge notes) suggested the probability of misclassificationand the need for an in-depth diagnostic review. Questionable cases were reviewed by a team of experts in order to reach a consensus on cause of death. Based on our algorithmic classification and diagnostic review results, the agreement between original cause of deathand that resulting from the assessment process was analyzed stratifying for every variable of interest. RESULTS: According to death certificates, breast cancerwas the cause of deathin 66.9% of subjects, and after assessment this figure changed to 65.7%. The misclassification rate was 4.3% and did not differ significantly between screen-detected (4.7%) and non-screen-detected (4.3%) cases. Higher misclassificationrates in favor of false positivity (cause of death wrongly attributed to breast cancer in deathcertificates) was observed for subjects with multiple cancers (6.5% vs. 1.9%), with no admission in the year before death (4.6% vs. 2.4%) and with an unknown cancer stage (4.9% vs 2.4% or 2.3%). CONCLUSIONS: The cause of death misclassificationrate is modest, causing a slight overestimate of deaths attributed to breast cancer, and is not affected by modality of diagnosis. The study confirmed the validity of using cause-specific mortality for service screening evaluation.