Sponsoring Section/Society: ASA Biometrics
Session Slot: 2:00- 3:50 Sunday
Estimated Audience Size: 100-200
AudioVisual Request: overhead and slides projector
Session Title: Spatio-Temporal Modelling of Small Area Health Data
This session focusses on the a developing area within epidemiology which has considerable import for informing policy in health resource allocation and assessment of environmental health risk. The analysis of spatio-temporal patterns of disease incidence has considerable importance in areas of health surveillance(early detection of adverse incidence) and in understanding of epidemiological processes. This session features a number of diverse approaches to this field which is fast evolving, and a spectrum of aopplication areas is also represented,from disease cluster detection to the analysis of hidden structure via mixtures.
Theme Session: Yes
Applied Session: Yes
Session Organizer: Lawson, Andrew University of Abertay Dundee
Address: DR Andrew B. Lawson Mathematical Sciences Division University of Abertay Dundee Bell Street Dundee DD1 1HG UK
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 0 minutes First Speaker - 25 minutes Second Speaker - 25 minutes Third Speaker - 25 minutes Fourth Speaker - 25 minutes Floor Discusion - 10 minutes
Session Chair: D'Agostino, Ralph B. Boston University
Address: Statistics and Consulting Unit, Department of Mathematics Boston University 111 Cummington Street Boston, MA 02215
1. Space-Time Mixture Modelling in Health Data
Boehning, Dankmar, Free University, Berlin
Address: PD Dr Dankmar Bohning Department of Epidemiology Institute of Social Medicine Free University Berlin Haus 562 Fabeckstr 60-62 14195 Berlin Germany
Schlattmann, Peter, Free University Berlin
Abstract: Mixture models for the construction of disease maps have been suggested in parametric(Clayton and Kaldor, Biometrics 1987, 671-687) and nonparametric form (Schlattmann and Boehning, Statistics in Medicine 1993,1943-1950). The discrete nature of the nonparametric mixture model estimator has been exploited to bind each component with a colour (or grey pattern) in the map which leads to a natural way of disease map construction with the program DISMAP=20(Schlattmann, 1996, Statistics in Medicine, 931). However, in these models time effects are not allowed. Here we consider ways to incorporate replications of area counts of disease cases in the mixture model which then would enable the disesase-mapper to model space-time clustered data simultaneously.
2. On Space-Time Cluster Modelling via MCMC Methods
Lawson, Andrew, University of Abertay Dundee, UK
Address: Mathmatical Science Division University of Abertary Dundee Bell Street Dundee DD1 1HG UK
Clark, Allan, University of Abertay
Abstract: The analysis of disease clustering is of considerable importance when time of occurence as well as spatial location are observed. This is especially true when surveillance systems are to be established. In this work we adopt a parametric modelling approach where clusters are regarded as objects to be estimated and where different types of clusters can be defined,depending on their space-time characteristics. The development of MCMC methods based on reversible jump sampling is extended to the case where temporal and spatial and spatio-temporal clusters are allowed.
3. On the Bias of the Knox Method and Other Space-Time Interaction Tests
Kulldorff, Martin, National Cancer Institute
Address: Dr Martin Kulldorff Biometry Branch DCPC National Cancer Institute Bethesda Maryland 20892-7354 USA
Hjalmas, Ulf, Ostersund Hospital, Sweden
Abstract: The Knox method, as well as other tests for space-time interaction, are biased when there are geographical population shifts so that different regions have different percentage population growth. In this talk, the size of the population shift bias is investigated for the Knox test, and it is shown that it can be a considerable problem. A method for constructing unbiased space-time interaction tests is then presented and illustrated. Practical implications are discussed in terms of the interpretation of past results and the design of future studies.
4. Spatio-Temporal Modelling of Mortality Data for Multiple Causes of Death
Leyland, Alistair, University of Glasgow, UK
Address: Dr Alastair Leyland Public Health Research Unit University of Glasgow 1 Lilybank Gardens Glasgow G12 8RZ UK
Langford, Ian, University of East Anglia
Goldstein, Harvey, University of London
Rashbash, J., University of London
Abstract: This paper illustrates how it is possible to develop a spatial model for counts of diseases occurring in discrete geographical areas as a multilevel model with heterogeneity effects and spatial effects considered to be random and correlated across areas. Such an approach has the advantage that it permits the extension of the model to include other levels of nesting. The model can therefore be extended to include temporal effects either by considering the time points to be repeated measures within areas or by modelling random variation as a function of temporal distance across areas in the same way that physical distance is modelled for the spatial effects. Similarly, a further extension means that multiple responses can be modelled for each area within each time point. This approach is illustrated on cause-specific mortality data for small geographical area population approximately 1000) within Scotland.
List of speakers who are nonmembers: None ???