SOCIO-ECONOMIC INDICES AS DETERMINANTS OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE PREVALENCE AND ADMISSIONS

 

Bryden C1, Bastian A2, Laing-Morton T1

1Met Office, FitzRoy Road, Exeter  EX1 3PB

2University of Southampton, University Road, Southampton  SO17 1BJ

Introduction

 

This England-based study aims to create a deprivation index to explain COPD prevalence by postcode district.  This is part of a COPD health forecasting project predicting risk of admission, prompting anticipatory care to reduce this risk.

 

The Met Office has developed a forecast model of risk of COPD admission in 35 Primary Care Trusts, based on seasonality, weather, infectious diseases and health surveillance.  The Met Office is looking into re-basing the model on postcode areas, in which case the COPD indices could be used to gauge different postcode districts’ response to environmental factors, and in targeting anticipatory care.

 

Methods

 

COPD prevalence was calculated as hospital admissions per population.  COPD admissions (ICD10 J40-44) were extracted for 1997-2003 for England, where the postcode district of patient residence was valid (population size = 538,339), and joined to 2001 Census data at a postcode district level for England.  Each postcode district was classified as rural or urban based on the Countryside Agency definition.

 

Four COPD-specific indices were constructed from 16 selected Census variables and the urban-rural classification: a research-based index (RBIC), combining variables reflecting the socio-economic risk factors identified in the literature; a correlations-based index (CBIC), choosing the variables most highly correlated with COPD; a single-variable index (SVIC), using the variable with the highest correlation (permanently sick); and a factor analysis index (FAIC), using principal component analysis to group highly-correlated variables into factors and weighting these in an index.  Two standard deprivation indices based only on Census variables were also constructed: the Townsend index (TDI) and the Carstairs and Morris index (CMDI).

                                                                                                                       

The indices were evaluated as predictors of high and low COPD prevalence, using Pearson’s correlation coefficients and Spearman’s rank correlation analysis, and the FAIC as a predictor of admissions.

 

Results

 

Districts in southern England tend to have lower COPD admission rates than the north, and rural districts lower than urban. Districts along the southern coast with higher proportions of older people have higher prevalence.

 

The TDI and CMDI performed the worst as predictors of prevalence, with Pearson coefficients of 0.378 and 0.550.  The coefficients for the COPD-specific indices were between 0.764 and 0.791.  The FAIC performed the best and the RBIC the worst of the four, but all picked up the patterns in prevalence.  The FAIC explained 55.5% of variation in admissions.

 

Conclusions

 

The standard deprivation indices performed poorly as predictors of COPD prevalence and therefore should not be used in spatial epidemiological analysis of COPD.  The FAIC performed well, but was the most complex index to construct and understand.  Any one of the COPD-specific indices could be used to characterise COPD prevalence in England.

 

 

clare.bryden@metoffice.gov.uk

 

Back to Abstracts Menu