Jcg. Hyndman et al., MISCLASSIFICATION OF SOCIAL DISADVANTAGE BASED ON GEOGRAPHICAL AREAS - COMPARISON OF POSTCODE AND COLLECTORS DISTRICT ANALYSES, International journal of epidemiology, 24(1), 1995, pp. 165-176
Background. Analysis of socioeconomic status (SES), defined on the bas
is of geographical area, will always be subject to misclassification o
f individuals. We studied the relationship between SES and selected he
alth-related measures when SES was defined firstly on the basis of pos
tcode and secondly on the basis of the smaller spatial area of collect
or's district (CD). Method. A Census population of 1.4 million was use
d to investigate the misclassification of individuals to SES group usi
ng postcode as opposed to CD. A field survey of 1000 respondents and a
mortality register of 1756 deaths were used to compare the relationsh
ip between SES and certain outcome variables, when SES group was assig
ned using postcode and CD. Misclassification probability matrices were
used to try to adjust the postcode-based results to approximate CD-ba
sed results. Results. The Census data showed that nearly 50% of reside
nts were misclassified into SES groups by the postcode-based system co
mpared with a CD-based system. In comparing the most socially disadvan
taged group with the least disadvantaged group, the postcode analysis
underestimated the absolute effects of SES by 58% for the increased pr
evalence of smoking, by 19% for the reduced prevalence of participatio
n in junior sporting clubs and by 13% for the increased mortality rate
at ages 0-64 years. Adjustment of postcode-based results using miscla
ssification matrices proved fruitless due to differential misclassific
ation and technical difficulties. Conclusions. Misclassification of in
dividuals to SES groups on the basis of postcode has caused an underes
timation of the true relationship between SES and health-related measu
res. A reduction of this misclassification by using smaller spatial ar
eas, such as CD or census enumeration districts, will provide improved
validity in estimating the true relationship.