gms | German Medical Science

MAINZ//2011: 56. GMDS-Jahrestagung und 6. DGEpi-Jahrestagung

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie e. V.
Deutsche Gesellschaft für Epidemiologie e. V.

26. - 29.09.2011 in Mainz

Detecting spatial cancer patterns using the adaptive kernel density – A feasibility study

Meeting Abstract

Suche in Medline nach

  • Dorothea Lemke - Westfälische Wilhelms-Universität Münster, Fachbereich Geowissenschaften, Münster
  • Volkmar Mattauch - Epidemiologische Krebsregister NRW, Münster
  • HW Hense - Institut für Epidemiologie und Sozialmedizin/Epidemiologisches Krebsregister NRW, Münster

Mainz//2011. 56. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 6. Jahrestagung der Deutschen Gesellschaft für Epidemiologie (DGEpi). Mainz, 26.-29.09.2011. Düsseldorf: German Medical Science GMS Publishing House; 2011. Doc11gmds142

doi: 10.3205/11gmds142, urn:nbn:de:0183-11gmds1421

Veröffentlicht: 20. September 2011

© 2011 Lemke et al.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen ( Er darf vervielfältigt, verbreitet und öffentlich zugänglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.



Background/Objectives: In spatial cancer epidemiology cancer cluster methods are an important tool to explore the spatial variation of cancer incidences in defined regions. There are many cluster detection methods used in spatial epidemiology. The point pattern analysis (PPA) is one of them, but it is less widely used within routine health services in Germany than might be expected given its potential usefulness demonstrated in research studies. The present study uses the adaptive kernel density to estimate a spatial risk surface based on non-aggregated data for identifying areas of elevated cancer incidences. We tested the applicability of this new method for use as an exploratory screening procedure for elevated cancer risks in a population-based cancer registry database.

Methods: Cases occurring between 1994 and 1998 in the age group 40 to 79 years in the Regierungsbezirk Münster (Germany) were identified from the local epidemiological cancer registry and these address records were geo-coded. Regions of elevated cancer incidence are assessed by using the adaptive kernel density estimation method which adjusts for the underlying population at risk. For the modelling of the population at risk the advantage of the detailed and open access CORINE land cover data was used in order to model a precise representation of the population at risk. We searched for potential clusters of incident carcinoma of the urinary bladder.

Results: The used kernel density method with an adaptive bandwidth shows an increased statistical robustness over an inhomogeneous background and a higher capacity to pick up finer details in densely populated areas than the kernel density estimation with a fixed bandwidth. Detailed geographic population layers permitted the combination of different age-sex categories over different time periods. Geo-coded records of incident cancer cases (observed counts) and average incidence rates in the background population (expected counts) were obtained to estimate the standardized incidence ratios by means of the adaptive kernel density. In a first exploratory study, small areas of elevated risk values were identified in the northern and southern part of the study region.

Conclusion: Our study shows that it is feasible and practicable to link routinely collected data from the health system with geo-codes to produce analytical procedures for the identification of regions with potentially increased risk of disease. We show here that the interplay between different databases and methods can be effectively implemented and that it produces results of potential relevance which may open perspectives for future exploratory screening procedures.