Under-estimation and over-estimation in gastric cancer incidence registry, Bayesian analysis for selected provinces

Abstract

ncidence is an important index for measuring the burden of cancer in a population. But in medical studies, a difficulty in drawing inference from categorical data is the existence of misclassification error. Misclassification error is the disagreement between the observed and the true value and occurs when new cancer cases diagnosed and registered in neighborhood provinces instead of their hometown due to low facility in their own provinces and difference of quality and quantity of registration system in different provinces. The aim of this study is to use a Bayesian statistical method to assess and correct this misclassification in Gastric cancer incidence, for some selected provinces in Iran. Data extracted from Iranian annual of national cancer registration report in 2008. The Age Standardized Rate (ASR) due to gastric cancer were expressed as rate per/100,000 population for male and female of North, South and Razavi Khorasan provinces. A Bayesian approach was used with Poisson count regression and an informative beta prior distribution assumed for the misclassified parameter. Analyses were carried out using R software version 3.2.0. The Bayesian analysis showed that, there is 34% misclassification in gastric cancer incidence registry from North and South Khorasans in Razavi Khorasan. After the correction, the rate of ASR decreased for Razavi Khorasan provinces, while, these rate increased for the rest of its neighborhoods. This study indicated that there is a major misclassification among provinces in Iran, in which, from neighborhood provinces and cities, the statistics of gastric cancer incidence misclassified on each other’s. In the absence of valid data, Bayesian approach would be a good and flexible alternative to eliminate the effects of Misclassification in incidence registry data for neighboring provinces.