Bayesian correction model for over-estimation and under-estimation of liver cancer incidence in Iranian neighboring provinces.

AIM

The aim of this study was to obtain more accurate estimates of the liver cancer incidence rate after correcting for misclassification error in cancer registry across Iranian provinces.

BACKGROUND

Nowadays having a thorough knowledge of geographic distribution of disease incidence has become essential for identifying the influential factors on cancer incidence.

METHODS

Data of liver cancer incidence was extracted from Iranian annual of national cancer registration report 2008. Expected coverage of cancer cases for each province was calculated. Patients of each province that had covered fewer cancer cases than 100% of its expectation, were supposed to be registered at an adjacent province which had observed more cancer cases than 100% of its expected coverage. For estimating the rate of misclassification in registering cancer incidence, a Bayesian method was implemented. Beta distribution was considered for misclassified parameter since its expectation converges to the misclassification rate. Parameters of beta distribution were selected based on the expected coverage of cancer cases in each province. After obtaining the misclassification rate, the incidence rates were re-estimated.

RESULTS

There was misclassification error in registering new cancer cases across the provinces of Iran. Provinces with more medical facilities such as Tehran which is the capital of the country, Mazandaran in north of the Iran, East Azerbaijan in north-west, Razavi Khorasan in north-east, Isfahan in central part, and Fars and Khozestan in south of Iran had significantly higher rates of liver cancer than their neighboring provinces. On the other hand, their neighboring provinces with low medical facilities such as Ardebil, West Azerbaijan, Golestan, South and north Khorasans, Qazvin, Markazi, Arak, Sistan & balouchestan, Kigilouye & boyerahmad, Bushehr, Ilam and Hormozgan, had observed fewer cancer cases than their expectation.

CONCLUSION

Accounting and correcting the regional misclassification are necessary for identifying high risk areas of the country and effective policy making to cope with cancer.