Geography Reference
In-Depth Information
Bayesian Mapping of Medical Data
Luk
´
ˇ Marek, V´tP
´
szto, and Pavel Tu
ˇ
ek
Introduction
One of the main aims of the spatial analysis of health and medical datasets is to
provide additional information to specialized medical research. These analyses can
be used for disease mapping; searching for places with a higher intensity and
probability of disease occurrence; or an influence assessment of selected natural
or artificial phenomena. While the location and space are obviously crucial com-
ponents within spatial analysis, they represent also a very problematic part in the
geospatial research of health data. Medical or health data are usually provided as
point data sets—with a direct location, an indirect location, or aggregated frequency
data that can be based on administrative units or grids. The latter case prevails and
immediately evokes visualization in the form of choropleth maps. However, med-
ical data, as well as other demographical data, require careful manipulation,
mapping and subsequent depiction. Suitably selected methods enable proper anal-
ysis of these data and the identification of irregularities and deviations of the
phenomena in the area of interest. The structure of medical data usually needs to
be standardized before comparing different regions. The standardization is based on
the population of the administrative unit and/or the age structure of the population.
Another procedure is the comparison of cases recorded in the area over time, with
the expected number of cases predicted from the known structure of patients in
another region or known probability distribution of the disease. The description of
the latter is well established in Bayesian statistics, which derives posterior proba-
bility as a consequence of prior probability and a probability model for the data to
be observed. Geosciences and geomedicine use Bayesian theory for the smoothing
of data—to help depict the real spatial pattern and its changeability. Bayesian
principles together with the spatial neighbourhood and statistical models are also
