Assessing Persistence in Spatial Clustering of Disease, with an Application to Drug Related Deaths in Scottish Neighbourhoods
DOI:
https://doi.org/10.2427/13225Abstract
Background: The upward trend in drug related deaths in some countries is a major public health concern. Regarding geographic location within countries, many studies report spatial clustering in drug related deaths. We consider drug related deaths in Scottish small areas, and investigate probabilities that clusters of adjacent neighbourhoods have elevated risk. We focus especially on assessing persistence in spatial clustering, relevant to prioritising area based interventions. We assess impacts of area risk factors on drug deaths, finding a strong link to poverty, and a clear overlap between drug death clustering and spatial poverty clustering.
Methods: We analyse drug related deaths in 1279 Scotland neighbourhoods over two periods, 2009-13 and 2014- 18, during which drug related mortality in Scotland has more than doubled. A fully Bayesian approach is used to identify zones with high mortality risk in both a neighbourhood and its spatial lag (“high-high” clusters), and extended to identify recurring high risk clustering over more than one period. Estimation of mortality risks, and of cluster probabilities through periods, is developed in conjunction with a regression model including area risk factors such as deprivation.
Results: Persistent clustering is concentrated in major urban centres, for example, Glasgow and Dundee. Deprivation is the paramount observed risk factor underlying elevated mortality risk, and persistent clustering in drug related mortality shows strong overlaps with poverty clustering. Social fragmentation modifies the paramount influence of poverty on drug mortality risk.
Conclusion: Cluster persistence is a central feature in small area variability in drug related death risk in Scotland intermediate zones, especially in some urban areas.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2022 Peter Douglas Congdon
This work is licensed under a Creative Commons Attribution 4.0 International License.