Analytic Mapping of Spatial Patterns of Prostate Cancer Mortality in South Korea, 2008-2011

Abstract

Among the epidemiological studies have been conducted in South Korea, most studies has been focused on the causes of the disease occurrances and their problems. Therefore, the studies have not been done actively on calculating and mapping statistically reliable mortality. This research suggest a guideline for disease mapping by using prostate cancer death data for the expansion of the spatial dimension of epidemiological studies. In addition, spatial pattern of mortality and logistic regression modeling with individual-level variable were conducted as an application study. We need to consider the way to process and offer the data to maintain statistical reliability of data in the procedure of making maps based on the disease data. Also, we need to consider the dimension of users data interpretation to avoid distorted interpretation of mortality and ratio scale data. First of all, we applied age-sex standardization techniques to mortality data. Standardization is a statistical method to remove the effects of age and sex differences between populations for comparison purposes. In addition, we applied Bayesian techniques to standardized death rate. Mortality can be calculated with statistical reliability by reducing the variability caused by small populations. Occurring the distortion of mortality can be prevented through this process.

Spatial pattern analysis was applied as an application section. Cluster analysis is an exploratory method to find significant cluster areas with high or low disease risk. In this context, the study to detect areas with high death rate of a disease can be used to find materials to conduct other epidemiological studies. And the spatial pattern analysis studies can be used to give a solid scientific facts to epidemiological researches. In this study, both global pattern analysis and local cluster analysis is conducted. Global pattern analysis provides an indication of significance of clusters. In addition, local cluster analysis is conducted to identify additional information of the size and location of clusters. Logistic regression modeling was applied to identify the vulnerable sub-class of variables of the individual-level data of death by prostate cancer. The variables included for modeling were restricted to the individual variables that contained cause of death data. It contains information about age, education level, and marriage status of each person who died. We can identify the vulnerable characteristics of variables on the death by prostate cancer within the range of the individual-level variables that were provided by cause of death data. We hope that this research will be useful to publish the disease atlas of a variety of disease in South Korea. In addition, we hope that this research contributes to publish various researches of disease in geographical dimension, and to make a better public health environment.