gms | German Medical Science

Kongress Medizin und Gesellschaft 2007

17. bis 21.09.2007, Augsburg

Disease mapping: A statistical model for small area variation depending on hospital effects

Meeting Abstract

Suche in Medline nach

  • Verena Barbieri - Medizinische Universität Innsbruck, Innsbruck
  • Alexander Ostermann - Universität Innsbruck, Innsbruck
  • Karl Peter Pfeiffer - Medizinische Universität Innsbruck, Innsbruck

Kongress Medizin und Gesellschaft 2007. Augsburg, 17.-21.09.2007. Düsseldorf: German Medical Science GMS Publishing House; 2007. Doc07gmds020

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter:

Veröffentlicht: 6. September 2007

© 2007 Barbieri 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.



Permanent critical evaluation of a health care system is necessary for efficient health care planning. One important point in this field is to examine how good different regions are supplied with hospitals. Thus, the development of a model that explains spatial variability by the availability of hospitals is needed. The health care situation of Austria is of particular interest because of the special geographical situation of this country. To account for small area variation a Bayesian smoothing model is used. Then, based on the BYM model a GLM model is developed accounting for patient flows between districts, for interactions between hospitals and for population sizes, capacities and distances. From the mathematical point of view a model term for patient flows is developed modeled as a CAR prior with a positive definite variance covariance matrix. The model is tested using several simulation data sets and two selected real data sets from routine Austrian hospital documentation (cataract surgery and bypass surgery). The results are found to be clearly better those from traditional models, as clusters of elevated risk have a higher detection rate and a better model fit is found. If no hospital effects exist, this fact is detected clearly. Investigations on cataract and bypass intervention show a hospital effect on cataract surgery but not on bypass intervention. The model is general and can be applied to several diseases and interventions. It can be extended to any spatial problem and, in a next step, its performance compared to the BYM model can be tested.