From February 2017, information about the work of the MRC/CSO Social and Public Health Sciences Unit, University of Glasgow is available and updated on the University of Glasgow website.

Modelling the Geography of Health

Inequalities between geographical units can easily be described by the variance of the outcome between units, but it is unclear what effect the way in which a country is broken down into geographical areas for the purpose of administration may have on this variance. We explored the extent to which the variance of the mortality rate for different European countries depends on the mean size and variation in size of the regions in each country. Countries in which regions have a greater variation in population size tend to display greater variation in death rates, suggesting that international comparisons of geographic inequalities may be hampered by such administrative constraints. Research into the modifiable areal unit problem based on different sized areas in Helsinki demonstrated that the choice of area size had a limited impact on the assessment of geographical inequalities in mortality after taking account of individual level characteristics such as occupational social class, economic activity, education, housing tenure and living arrangements.

Current research is exploring the relationship between area size and socioeconomic inequalities. In Scotland, measures of area deprivation tend to be assessed at the postcode sector (PS) level (average population 5,233 in 2011) using Carstairs deprivation scores or at data zone level (average population 815) using the Scottish Index of Multiple Deprivation (SIMD). Output areas (OAs) also exist and are the smallest unit of Census geography available with an average population of 114 people.  For the first time, in 2011, Carstairs deprivation scores have been calculated for all three area levels. This analysis examines inequalities in mortality at each area level to see what happens to health inequalities in Scotland as the average size of the areas decrease.
We have used spatial models to investigate the relationship between air pollution and mortality in a study of 105,359 people aged 50-74 living in 468 neighbourhoods in Oslo, Norway. We showed that some of the excess mortality associated with air pollution could be explained by contextual measures of disadvantage, even after adjustment for individual socio-economic circumstances.
As part of a study led by Basile Chaix from Community Medicine and Public Health at Malmö University Hospital, we investigated the spatial patterning of mental disorders in Malmö, Sweden. After taking individual factors into account, we found mental disorders due to psychoactive substances to be related to both neighbourhood deprivation and the incidence of violent crimes in the neighbourhood, whilst neurotic disorders were related only to neighbourhood deprivation.
We have developed spatial multilevel models and have shown how these can be extended to include multiple outcomes (multivariate spatial models). Applications of such models have included the joint modelling of mortality due to circulatory disease and neoplasms in Greater Glasgow Health Board and of lung cancer mortality for 187 regions in 11 European countries. Other work includes a review of the use of Empirical Bayes methods (both spatial and non-spatial) for disease mapping.



Molaodi OR, Leyland AH, Ellaway A, Kearns A, Harding S. Neighbourhood food and physical activity environments in England, UK: does ethnic density matter?. International Journal of Behavioral Nutrition and Physical Activity 2012;9:75

pubmed  open access  


Jackson A, Davies CA, Leyland AH. Do differences in the administrative structure of populations confound comparisons of geographic health inequalities?. BMC Medical Research Methodology 2010;10:74

pubmed  open access  

Tarkiainen L, Martikainen P, Laaksonen M, Leyland AH. Comparing the effects of neighbourhood characteristics on all-cause mortality using two hierarchical areal units in the capital region of Helsinki. Health & Place 2010;16:409-12



Lumme S, Leyland AH, Keskimäki I. Multilevel modeling of regional variation in equity in health care. Medical Care 2008;46:976-983



Manda SOM, Leyland AH. An empirical comparison of maximum likelihood and Bayesian estimation methods for multivariate disease mapping. South African Statistical Journal 2007;41:1-21

Næss Ø, Piro F N, Nafstad P, Smith DG, Leyland AH. Air pollution, social deprivation, and mortality: a multilevel cohort study. Epidemiology 2007;18:686-694

pubmed  open access  


Chaix B, Leyland AH, Sabel CE, Chauvin P, Rastam L, Kristersson H, Merlo J. Spatial clustering of mental disorders and associated characteristics of the neighbourhood context in Malmo, Sweden, in 2001. Journal of Epidemiology & Community Health 2006;60:427-435



Leyland AH, Davies CA. Empirical Bayes methods for disease mapping. Statistical Methods in Medical Research 2005;14:17-34



Davies CA, Leyland AH. Spatial patterns of cancer mortality in Europe. In: Kirch W, editor Public health in Europe: 10 years EUPHA. Berlin: Springer-Verlag, 2004:227-243.

Leyland AH. Increasing inequalities in premature mortality in Great Britain. Journal of Epidemiology & Community Health 2004;58:296-302

pubmed  open access  


Leyland AH, Langford I H, Rasbash J, Goldstein H. Multivariate spatial models for event data. Statistics in Medicine 2000;19:2469-2478



Langford IH, Leyland AH, Rasbash J, Goldstein H. Multilevel modelling of the geographical distributions of rare diseases. Applied Statistics 1999;48:253-268



  • Geographical unit Area.
  • Neurotic disorders (neuroses) Mental imbalance causing distress but not affecting daily functioning.
  • Psychoactive substance A chemical substance (drug) that alters brain function, possibly affecting mood, perception or behaviour.
  • Variance A statistical measure of the dispersion in a variable.
View all glossary entries