Bayesian Analysis of Breast Cancer Mortality to Reduce the Effects of Misclassification

Breast cancer (BC) is the most common cancer in Iranian women. Studying the mortality statistics is important to monitor the effects of screening programs or the influence of earlier diagnosis on the burden of this chronic disease. Misclassification is still a problem in the Iranian death registry data and about 20% of death statistics are recorded in misclassified categories. The aim of this study is to re-estimate the mortality rate of Breast cancer in the Iran by a Bayesian model in order to exclude the bias due to misclassification. The mortality data were extracted from national death Statistic, which reported or published by the Ministry of Health and Medical Education, from 1995 to 2004. The rate of mortality due to BC [ICD-10; C50] were expressed as the annual rates per/100,000 population in age group (< 15, 15 - 49 and ≥50 years of age) and also, age standardized rate (ASR) calculated in this study. To correct the misclassification effect, a Bayesian approach was used with Poisson count regression and beta prior. The results of the Bayesian analysis indicated that there were between 20 to 30 percent misclassified deaths statistic in mortality records which did not reported and re-estimated by this model for BC. Also the rate of BC mortality has increased in recent years; however it seems that the rate would be leveled off from 2002 to 2004. The findings revealed a substantial misclassified mortality statistics of BC in the Iranian women. Therefore policy makers in health scope should notice to this unreported data.