Impact of geographic area level on measuring socioeconomic disparities in cancer survival in New South Wales, Australia

A period analysis

Julia F. Stanbury, Peter D. Baade, Yan Yu, Xue Qin Yu

Research output: Contribution to journalArticleResearchpeer-review

Abstract

Background: Area-based socioeconomic measures are widely used in health research. In theory, the larger the area used the more individual misclassification is introduced, thus biasing the association between such area level measures and health outcomes. In this study, we examined the socioeconomic disparities in cancer survival using two geographic area-based measures to see if the size of the area matters. 

Methods: We used population-based cancer registry data for patients diagnosed with one of 10 major cancers in New South Wales (NSW), Australia during 2004–2008. Patients were assigned index measures of socioeconomic status (SES) based on two area-level units, census Collection District (CD) and Local Government Area (LGA) of their address at diagnosis. Five-year relative survival was estimated using the period approach for patients alive during 2004–2008, for each socioeconomic quintile at each area-level for each cancer. Poisson-regression modelling was used to adjust for socioeconomic quintile, sex, age-group at diagnosis and disease stage at diagnosis. The relative excess risk of death (RER) by socioeconomic quintile derived from this modelling was compared between area-units. 

Results: We found extensive disagreement in SES classification between CD and LGA levels across all socioeconomic quintiles, particularly for more disadvantaged groups. In general, more disadvantaged patients had significantly lower survival than the least disadvantaged group for both CD and LGA classifications. The socioeconomic survival disparities detected by CD classification were larger than those detected by LGA. Adjusted RER estimates by SES were similar for most cancers when measured at both area levels. 

Conclusions: We found that classifying patient SES by the widely used Australian geographic unit LGA results in underestimation of survival disparities for several cancers compared to when SES is classified at the geographically smaller CD level. Despite this, our RER of death estimates derived from these survival estimates were generally similar for both CD and LGA level analyses, suggesting that LGAs remain a valuable spatial unit for use in Australian health and social research, though the potential for misclassification must be considered when interpreting research. While data confidentiality concerns increase with the level of geographical precision, the use of smaller area-level health and census data in the future, with appropriate allowance for confidentiality

Original languageEnglish
Pages (from-to)56-62
Number of pages7
JournalCancer Epidemiology
Volume43
DOIs
Publication statusPublished - Aug 2016
Externally publishedYes

Fingerprint

Local Government
South Australia
New South Wales
Social Class
Demography
Survival
Vulnerable Populations
Neoplasms
Confidentiality
Censuses
Health Status
Research
Health
Registries
Age Groups
Outcome Assessment (Health Care)
Population

Cite this

@article{500d9b64eadc41c08457ac89144a1e4e,
title = "Impact of geographic area level on measuring socioeconomic disparities in cancer survival in New South Wales, Australia: A period analysis",
abstract = "Background: Area-based socioeconomic measures are widely used in health research. In theory, the larger the area used the more individual misclassification is introduced, thus biasing the association between such area level measures and health outcomes. In this study, we examined the socioeconomic disparities in cancer survival using two geographic area-based measures to see if the size of the area matters. Methods: We used population-based cancer registry data for patients diagnosed with one of 10 major cancers in New South Wales (NSW), Australia during 2004–2008. Patients were assigned index measures of socioeconomic status (SES) based on two area-level units, census Collection District (CD) and Local Government Area (LGA) of their address at diagnosis. Five-year relative survival was estimated using the period approach for patients alive during 2004–2008, for each socioeconomic quintile at each area-level for each cancer. Poisson-regression modelling was used to adjust for socioeconomic quintile, sex, age-group at diagnosis and disease stage at diagnosis. The relative excess risk of death (RER) by socioeconomic quintile derived from this modelling was compared between area-units. Results: We found extensive disagreement in SES classification between CD and LGA levels across all socioeconomic quintiles, particularly for more disadvantaged groups. In general, more disadvantaged patients had significantly lower survival than the least disadvantaged group for both CD and LGA classifications. The socioeconomic survival disparities detected by CD classification were larger than those detected by LGA. Adjusted RER estimates by SES were similar for most cancers when measured at both area levels. Conclusions: We found that classifying patient SES by the widely used Australian geographic unit LGA results in underestimation of survival disparities for several cancers compared to when SES is classified at the geographically smaller CD level. Despite this, our RER of death estimates derived from these survival estimates were generally similar for both CD and LGA level analyses, suggesting that LGAs remain a valuable spatial unit for use in Australian health and social research, though the potential for misclassification must be considered when interpreting research. While data confidentiality concerns increase with the level of geographical precision, the use of smaller area-level health and census data in the future, with appropriate allowance for confidentiality",
keywords = "Area level, Cancer, Disparity, Geographic, Socioeconomic variation, Survival analysis",
author = "Stanbury, {Julia F.} and Baade, {Peter D.} and Yan Yu and Yu, {Xue Qin}",
year = "2016",
month = "8",
doi = "10.1016/j.canep.2016.06.001",
language = "English",
volume = "43",
pages = "56--62",
journal = "Cancer Epidemiology",
issn = "0361-090X",
publisher = "Elsevier",

}

Impact of geographic area level on measuring socioeconomic disparities in cancer survival in New South Wales, Australia : A period analysis. / Stanbury, Julia F.; Baade, Peter D.; Yu, Yan; Yu, Xue Qin.

In: Cancer Epidemiology, Vol. 43, 08.2016, p. 56-62.

Research output: Contribution to journalArticleResearchpeer-review

TY - JOUR

T1 - Impact of geographic area level on measuring socioeconomic disparities in cancer survival in New South Wales, Australia

T2 - A period analysis

AU - Stanbury, Julia F.

AU - Baade, Peter D.

AU - Yu, Yan

AU - Yu, Xue Qin

PY - 2016/8

Y1 - 2016/8

N2 - Background: Area-based socioeconomic measures are widely used in health research. In theory, the larger the area used the more individual misclassification is introduced, thus biasing the association between such area level measures and health outcomes. In this study, we examined the socioeconomic disparities in cancer survival using two geographic area-based measures to see if the size of the area matters. Methods: We used population-based cancer registry data for patients diagnosed with one of 10 major cancers in New South Wales (NSW), Australia during 2004–2008. Patients were assigned index measures of socioeconomic status (SES) based on two area-level units, census Collection District (CD) and Local Government Area (LGA) of their address at diagnosis. Five-year relative survival was estimated using the period approach for patients alive during 2004–2008, for each socioeconomic quintile at each area-level for each cancer. Poisson-regression modelling was used to adjust for socioeconomic quintile, sex, age-group at diagnosis and disease stage at diagnosis. The relative excess risk of death (RER) by socioeconomic quintile derived from this modelling was compared between area-units. Results: We found extensive disagreement in SES classification between CD and LGA levels across all socioeconomic quintiles, particularly for more disadvantaged groups. In general, more disadvantaged patients had significantly lower survival than the least disadvantaged group for both CD and LGA classifications. The socioeconomic survival disparities detected by CD classification were larger than those detected by LGA. Adjusted RER estimates by SES were similar for most cancers when measured at both area levels. Conclusions: We found that classifying patient SES by the widely used Australian geographic unit LGA results in underestimation of survival disparities for several cancers compared to when SES is classified at the geographically smaller CD level. Despite this, our RER of death estimates derived from these survival estimates were generally similar for both CD and LGA level analyses, suggesting that LGAs remain a valuable spatial unit for use in Australian health and social research, though the potential for misclassification must be considered when interpreting research. While data confidentiality concerns increase with the level of geographical precision, the use of smaller area-level health and census data in the future, with appropriate allowance for confidentiality

AB - Background: Area-based socioeconomic measures are widely used in health research. In theory, the larger the area used the more individual misclassification is introduced, thus biasing the association between such area level measures and health outcomes. In this study, we examined the socioeconomic disparities in cancer survival using two geographic area-based measures to see if the size of the area matters. Methods: We used population-based cancer registry data for patients diagnosed with one of 10 major cancers in New South Wales (NSW), Australia during 2004–2008. Patients were assigned index measures of socioeconomic status (SES) based on two area-level units, census Collection District (CD) and Local Government Area (LGA) of their address at diagnosis. Five-year relative survival was estimated using the period approach for patients alive during 2004–2008, for each socioeconomic quintile at each area-level for each cancer. Poisson-regression modelling was used to adjust for socioeconomic quintile, sex, age-group at diagnosis and disease stage at diagnosis. The relative excess risk of death (RER) by socioeconomic quintile derived from this modelling was compared between area-units. Results: We found extensive disagreement in SES classification between CD and LGA levels across all socioeconomic quintiles, particularly for more disadvantaged groups. In general, more disadvantaged patients had significantly lower survival than the least disadvantaged group for both CD and LGA classifications. The socioeconomic survival disparities detected by CD classification were larger than those detected by LGA. Adjusted RER estimates by SES were similar for most cancers when measured at both area levels. Conclusions: We found that classifying patient SES by the widely used Australian geographic unit LGA results in underestimation of survival disparities for several cancers compared to when SES is classified at the geographically smaller CD level. Despite this, our RER of death estimates derived from these survival estimates were generally similar for both CD and LGA level analyses, suggesting that LGAs remain a valuable spatial unit for use in Australian health and social research, though the potential for misclassification must be considered when interpreting research. While data confidentiality concerns increase with the level of geographical precision, the use of smaller area-level health and census data in the future, with appropriate allowance for confidentiality

KW - Area level

KW - Cancer

KW - Disparity

KW - Geographic

KW - Socioeconomic variation

KW - Survival analysis

UR - http://www.scopus.com/inward/record.url?scp=84977123742&partnerID=8YFLogxK

U2 - 10.1016/j.canep.2016.06.001

DO - 10.1016/j.canep.2016.06.001

M3 - Article

VL - 43

SP - 56

EP - 62

JO - Cancer Epidemiology

JF - Cancer Epidemiology

SN - 0361-090X

ER -