Spatial Analysis Strategies to Study Cancer Incidence Distribution in Córdoba, Argentina

Florianopolis

Congreso; XXV Internacional Biometrics Conference.; 2010

Internacional Biometrics Society

This work proposes a strategy of spatial analysis to study the cancer incidence series in Cordoba province, Argentina. Cancer is the second leading cause of death by disease in Argentina. Its occurrence is typical of each geographic region, cultural patterns and lifestyles. However, there are few studies in our country describing the spatial distribution of the cancer incidence. In this line, Diaz MP et al (2009) and Diaz M et al (2009) show not random patterns for the most of cancers and some social factors associated to their distributions. However, it becomes necessary to explore which spatial pattern could be attributed to each cancer using more specific techniques. The central concept to study was the spatial autocorrelation, i.e. the correlation of a single variable (incidence of cancer) between pairs of neighboring observations. In this study, it is assumed that the spatial data structure has two components of variation, a large-scale, resulting from the global trend, and a small scale, in the residual source after removing the first. From that, it is suitable asking about patterns related to geographical proximity or neighborhood. We used spatial series of incidence rates calculated for the 131 districts of the Cordoba province. First, Moran Index was calculated for each type of tumor and sex using the Euclidean distance as the neighborhood criterion. This measure of spatial autocorrelation was considered as a first indication of the nature of the phenomenon studied. After that, in order to capture the large-scale component, regression model was fitted including latitude and longitude of each district as single covariates. Finally, the Moran index of the residuals series was again calculated to probe about small-scale correlation pattern. According to this strategy, all cancers studied in women showed large-scale correlation and only colon cancer in small scale. In men, it was identified large-scale association in prostate cancer and both associations in bladder cancer. In this study, it is assumed that the spatial data structure has two components of variation, a large-scale, resulting from the global trend, and a small scale, in the residual source after removing the first. From that, it is suitable asking about patterns related to geographical proximity or neighborhood. We used spatial series of incidence rates calculated for the 131 districts of the Cordoba province. First, Moran Index was calculated for each type of tumor and sex using the Euclidean distance as the neighborhood criterion. This measure of spatial autocorrelation was considered as a first indication of the spatialized nature of the phenomenon studied. After that, in order to capture the large-scale component, regression model was fitted including latitude and longitude of each district as single covariates. Finally, the Moran index of the residuals series was again calculated to probe about small-scale correlation pattern. According to this strategy, all cancers studied in women showed large-scale correlation and only colon cancer in small scale. In men, it was identified large-scale association in prostate cancer and both associations in bladder cancer. In this study, it is assumed that the spatial data structure has two components of variation, a large-scale, resulting from the global trend, and a small scale, in the residual source after removing the first. From that, it is suitable asking about patterns related to geographical proximity or neighborhood. We used spatial series of incidence rates calculated for the 131 districts of the Cordoba province. First, Moran Index was calculated for each type of tumor and sex using the Euclidean distance as the neighborhood criterion. This measure of spatial autocorrelation was considered as a first indication of the spatialized nature of the phenomenon studied. After that, in order to capture the large-scale component, regression model was fitted including latitude and longitude of each district as single covariates. Finally, the Moran index of the residuals series was again calculated to probe about small-scale correlation pattern. According to this strategy, all cancers studied in women showed large-scale correlation and only colon cancer in small scale. In men, it was identified large-scale association in prostate cancer and both associations in bladder cancer.