Agreement between aggregate and individual-level measures of income and education: a comparison across three patient groups.

Abstract

The association between lower socioeconomic status and poorer health outcomes has been observed using both individual-level and aggregate-level measures of income and education. While both are predictive of health outcomes, previous research indicates poor agreement between individual-level and aggregate-level measures. The purpose of this study was to determine the level of agreement between aggregate-level and individual-level measures of income and education among three distinct patient groups, specifically asthma, diabetes, and rheumatoid patients.
Individual-level measures of annual household income and education were derived from three separate surveys conducted among patients with asthma (n = 359), diabetes (n = 281) and rheumatoid arthritis (n = 275). Aggregate-level measures of income and education were derived from the 2001 Canadian census, including both census tract-and dissemination area-level measures. Cross-tabulations of individual-level income by aggregate-level income were used to determine the percentage of income classifications in agreement. The kappa statistic (simple and weighted), Spearman's rank correlations, and intra-class correlation coefficient (ICC) were also calculated. Individual-level and aggregate-level education was compared using Chi-Square tests within patient groups. Point biserial correlation coefficients between individual-level and aggregate-level education were computed.
Individual-level income was poorly correlated with aggregate-level measures, which provided the worst estimations of income among patients in the lowest income category at the individual-level. Both aggregate-level measures were best at approximating individual-level income in patients with diabetes, in whom aggregate-level estimates were only significantly different from individual-level measures for patients in the lowest income category. Among asthma patients, the proportion of patients classified by aggregate-level measures as having a university degree was significantly lower than that classified by individual-level measures. Among diabetes and rheumatoid arthritis patients, differences between aggregate and individual-level measures of education were not significant.
Agreement between individual-level and aggregate-level measures of socioeconomic status may depend on the patient group as well as patient income. Research is needed to characterize differences between patient groups and help guide the choice of measures of socioeconomic status.

Full text

Page 1

RESEARCH ARTICLEOpen Access
Agreement between aggregate and individual-
level measures of income and education: a
comparison across three patient groups
Carlo A Marra1,2*, Larry D Lynd1,2, Stephanie S Harvard3and Maja Grubisic2
Abstract
Background: The association between lower socioeconomic status and poorer health outcomes has been
observed using both individual-level and aggregate-level measures of income and education. While both are
predictive of health outcomes, previous research indicates poor agreement between individual-level and
aggregate-level measures. The purpose of this study was to determine the level of agreement between aggregate-
level and individual-level measures of income and education among three distinct patient groups, specifically
asthma, diabetes, and rheumatoid patients.
Methods: Individual-level measures of annual household income and education were derived from three separate
surveys conducted among patients with asthma (n = 359), diabetes (n = 281) and rheumatoid arthritis (n = 275).
Aggregate-level measures of income and education were derived from the 2001 Canadian census, including both
census tract-and dissemination area-level measures. Cross-tabulations of individual-level income by aggregate-level
income were used to determine the percentage of income classifications in agreement. The kappa statistic (simple
and weighted), Spearman’s rank correlations, and intra-class correlation coefficient (ICC) were also calculated.
Individual-level and aggregate-level education was compared using Chi-Square tests within patient groups. Point
biserial correlation coefficients between individual-level and aggregate-level education were computed.
Results: Individual-level income was poorly correlated with aggregate-level measures, which provided the worst
estimations of income among patients in the lowest income category at the individual-level. Both aggregate-level
measures were best at approximating individual-level income in patients with diabetes, in whom aggregate-level
estimates were only significantly different from individual-level measures for patients in the lowest income
category. Among asthma patients, the proportion of patients classified by aggregate-level measures as having a
university degree was significantly lower than that classified by individual-level measures. Among diabetes and
rheumatoid arthritis patients, differences between aggregate and individual-level measures of education were not
significant.
Conclusions: Agreement between individual-level and aggregate-level measures of socioeconomic status may
depend on the patient group as well as patient income. Research is needed to characterize differences between
patient groups and help guide the choice of measures of socioeconomic status.
Keywords: Socio-economic status income, education, aggregate-level, individual-level asthma, diabetes, rheuma-
toid arthritis
* Correspondence: cmarra@exchange.ubc.ca
1Faculty of Pharmaceutical Sciences, University of British Columbia,
Vancouver, BC, Canada
Full list of author information is available at the end of the article
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
© 2011 Marra et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

Page 2

Background
Socioeconomic status (SES) has been shown to be
associated with health outcomes among the general
population [1-3] as well among various patient groups,
including asthma [4-10], rheumatoid arthritis (RA)
[11,12], and diabetes mellitus (DM) patients [13].
Health outcomes associated with SES include level of
asthma control [4], hospital admissions [5,8], emer-
gency department and physician visits [6,7,9], and
asthma-related mortality among asthmatics, disease
activity, physical and mental health, and quality of life
among individuals with RA [11,14], and hospitaliza-
tions among diabetics [13].
The association between lower SES and poorer health
outcomes has been observed using individual-level mea-
sures of SES as well aggregate-level measures, such as
those available from census data. While SES measures at
both of these levels may be predictive of health out-
comes, the validity of using aggregate-level measures as
a proxy for individual-level measures is debatable. In
Canada, studies aimed at quantifying the relationship
between individual-level SES measurements and aggre-
gates derived from Canadian census data have generally
indicated poor agreement. This finding is consistent
across studies using individual-level measures derived
from self-report [15], from structured interviews [16],
and from public health insurance data [17,18]. These
studies indicate that aggregate measures from the
Canadian census function to mask variation in individual-
level measures, the latter being more sensitive to poverty
and poor health outcomes. Studies using US census data
further suggest that aggregate-level SES measures reflect a
construct distinct from individual-level ones [19,20].
Despite their limitations, aggregate-level measures are
considered appropriate for use when individual-level
data are lacking [21]. Particularly in research using
administrative data, aggregate measures are often the
only available means to adjust for SES. In this context,
the question remains whether aggregate-level measures
perform equally well as individual-level proxies across
different patient groups or whether the discrepancy
between aggregate and individual-level measures is exag-
gerated in some populations. The question also remains
whether there are differences across patient groups
according to the aggregate-level measure used.
In Canada, the smallest geographical area for which all
census variables are available is the dissemination area
(DA), which typically contains 400 to 700 residents [22].
Studies may also utilize data from larger census units,
such as the census tract (CT), containing 2500 to 8000
residents [23]. While all of Canada is divided into DAs,
only regions with a population of 50,000 or more are
divided into CTs, which may lead to differences between
DA-and CT-level data.
The purpose of this study was to determine the level
of agreement between aggregate-level and individual-
level measures of income and education among three
distinct patient groups, specifically asthma, DM, and RA
patients.
Methods
Data
Individual-level measures of annual household income
and education were derived from three separate self-
report surveys conducted among patients with asthma
(n = 359), DM (n = 281) and RA (n = 275), respectively.
The methods for these surveys have been published pre-
viously [14,24,25]. All patients were recruited from Brit-
ish Columbia, Canada, and the samples are considered
to be representative of the English-speaking, adult mem-
bers of these patient groups in this region. Ethical
approval for all three studies was obtained from the
University of British Columbia.
Patients with asthma completed the surveys in years
2000 and 2005, patients with RA in 2002, and patients
with DM in 2008. All patients completed the survey
independently. All surveys included the same items on
income and education, which pertained to annual
household income prior to any deductions and to certi-
ficates, degrees, or diplomas obtained. All surveys col-
lected patients’ age, sex, and residential postal codes.
Patients whose residential postal codes were missing
were excluded from the study.
Aggregate-level measures of income and education
were derived from the 2001 Canadian census data,
available online from the University of British Columbia.
Aggregate-level income was based on the median house-
hold income for the census level, which is the
self-reported annual household income prior to any
deductions [26]. Education was evaluated by highest
level of schooling, which in the Canadian census refers
to the self-reported “highest grade or year of elementary
or secondary (high) school attended, or to the highest
year of university or college education completed”.
Patients with the census classification ‘university, with
university degree’ were categorized as having a univer-
sity degree and all other patients were categorized as
not having a university degree [26].
Statistical Methods
Demographic variables between the three patient groups
were compared using ANOVA for continuous variables
and Pearson’s Chi-Square and Fisher’s exact tests for
categorical variables, where appropriate. Statistics
Canada’s Postal Code Conversion File was used to link
patients’ postal codes to their corresponding CT and
DA [27]. The inflation factor from the Canadian Consu-
mer Price Index was used to adjust individual-level
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 2 of 7

Page 3

incomes from across survey years to their 2001 income
equivalents [28]. Both individual-level and aggregate-
level income were categorized as less than $20,000,
between $20,000 and $50,000, and greater than $50,000
[24]. To determine the agreement between aggregate-
level and individual-level income, 3 × 3 cross-tabulations
of individual-level income category by CT-level income
category and of individual-level income category by
DA-level income category were produced. This output
what used to determine the percentage of cases for
which individual-level and DA/CT-level income cate-
gories were in agreement. The kappa statistic (simple
and weighted) and Spearman’s rank correlations were
calculated to determine the degree of nonrandom agree-
ment between individual-level and DA/CT-level income.
The intra-class correlation coefficient (ICC) was also
calculated using the 2-way mixed model for absolute
agreement [29]. For the ICCs, levels of agreement were
adopted as proposed by Fleiss, i.e., <0.40 poor, 0.40-0.75
fair to good, and ≥0.75 excellent [30].
Both individual-level and aggregate-level education
were categorized as at least a university degree or less
than university degree. In the analysis of aggregate-level
education, only the census population in age groups
20-24 and higher was included. Individual-level and aggre-
gate-level education was compared using Chi-Square tests
within patient groups. Point biserial correlation coeffi-
cients between individual-level and DA/CT level education
were also computed.
Results
Patient groups and Number of Corresponding CTs and
DAs
Patients in the asthma sample (n = 359) resided in 198
discrete CTs and 321 discrete DAs, representing a total
of 1,552,655 and 188,235 census respondents, respec-
tively. Patients in the DM sample (n = 281) resided in
169 discrete CTs and 261 discrete DAs, representing a
total of 1,166,395 and 163,095 census respondents,
respectively. Patients in the RA sample (n = 276)
belonged to 144 discrete CTs and 226 discrete DAs,
representing a total of 833,735 and 157,500 census
respondents, respectively.
Individual and Aggregate-level Patient
Sociodemographics
Table 1 shows the sociodemographic characteristics of
each of the three patient groups. Patients in the asthma
sample were significantly younger than those in the RA
and DM samples (p < 0.0001) and there was a signifi-
cantly greater proportion of females in the RA sample
(p < 0.0001). Individual-level household income was
highest among patients with DM, followed by patients
with asthma and RA, respectively (p < 0.0001, with all
pairwise comparisons p < 0.0001). There were no signif-
icant differences in the proportions of patients classed
by CT or DA as having a university degree between the
patient groups. The proportion of RA patients reporting
a university degree was significantly lower than both the
Table 1 Characteristics of the study participants
Asthma
n = 359
Rheumatoid Arthritis
n = 276
Diabetes
n = 281
p-value*
CharacteristicMean (sd†) or n (%) Mean (sd) or n (%)
61.2 (13.8)a
58 (21.2)b
Mean (sd) or n (%)
Age (years)36.8 (8.4)
56.9 (13.1)<0.0001
Sex (males) 128 (35.7)
147 (52.3) <0.0001
Self-reported Income
<20,000
20,000-50,000
>50,000
90 (29.8)
84 (27.8)
128 (42.4)
44 (19.5)
99 (43.8)
83 (36.7)
34 (14.6)
76 (32.6)
123 (52.8)<0.0001
DA‡-Household Income
<20,000
20,000-50,000
>50,000
24 (6.7)
180 (50.4)
153 (42.9)
8 (3.0)
128 (47.8)
132 (49.3)
7 (2.5)
112 (40.1)
160 (57.4)0.001
CT§- Household Income
<20,000
20,000-50,000
>50,000
10 (2.8)
192 (53.8)
155 (43.4)
04 (1.5)
86 (32.2)
177 (66.3)
82 (29.8)d
76 (38.0)
124 (62.0)
45 (17.3)c
<0.0001
Individual-level University degree 124 (34.5)
<0.0001
Aggregate-level University Degree (expected)-DA91 (25.3) 50 (18.1)71 (25.2) 0.06
Aggregate-level University Degree (expected)- CT93 (25.9) 59 (21.3)74 (26.4)0.39
a2 missing;b16 missing;c6 missing;d6 missing;*all p-values were obtained using Chi-square test except for age where ANOVA was used. †Standard deviation;
‡ Dissemination Area; § Census Tract.
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 3 of 7

Page 4

proportions of asthma patients (p < 0.0001) and DM
patients (p < 0.0001). The difference between the
asthma and diabetes samples in individual-level univer-
sity education was not significant (p = 0.21).
Agreement: Individual-level and DA/CT Income Measures
In all patient groups, the proportion of patients who
reported incomes under $20,000 per year was signifi-
cantly higher than the proportion of patients classed in
this income category by DA or CT (all p-values
<0.0001). In the asthma group, the proportion of
patients who reported an income between $20,000 and
$50,000 per year (27.8%) was also significantly lower
than the proportions of patients classed in this income
category by DA (50.4%; p < 0.0001) or by CT (53.8%; p
< 0.0001); the proportions in the highest income cate-
gory (>$50,000 per year) were similar. In the RA group,
the proportion of patients reporting an income over
$50,000 per year was significantly lower than the pro-
portions of patients classed in this income category by
DA (p = 0.005) or by CT (p < 0.0001). Among DM
patients, there were no significant differences between
the proportions of patients reporting incomes between
$20,000 and $50,000 or >$50,000 per year and the pro-
portions classed in these categories by DA and CT,
respectively.
The Spearman’s rank correlations, weighted kappa
coefficients, and intra-class correlations indicating the
association between individual-level and CT-level income
measures, and individual-level and DA-level income mea-
sures, are shown in Table 2. Following the designations
proposed by Fleiss (<0.40 poor, 0.40-0.75 fair to good,
and ≥0.75 excellent), the ICCs generally indicated poor
agreement between individual-level and aggregate-level
income measures among all patient groups.
The extent of perfect agreement between individual-
level and DA-level and CT-level groupings of income is
illustrated in Tables 4 and 5, respectively. Among all
patient groups, both for CT-level and DA-level income
measures, agreement with individual-level measures was
less frequent within the lowest income grouping. Of
note, of the nearly 20% of RA patients who reported an
income under $20,000 in the survey, none were cor-
rectly classed in this income grouping by DA or CT
(Tables 4 and 5). Overall, across all patient groups there
were no significant differences between DA-level and
CT-level income groupings in the proportion of cases
for which there was perfect agreement with individual-
level measures (Table 3).
Agreement: Individual and Aggregate-level Education
Measures
Table 6 shows the comparison of individual-level
university education to DA-and CT-level measures of
university education. In patients with asthma, CT-and
DA-level census data indicated nearly equal proportions
of patients with a university degree. However, this pro-
portion was significantly lower than the proportion of
asthma patients who reported having a university degree
in the survey (35%) (p = 0.01). For the RA and DM
patient groups, differences between individual-level and
aggregate-level measures of education, respectively, were
not significant.
Point biserial correlations between individual-level
university degree and CT-level measures of education
(i.e., proportion of the population with a university
degree) were weak within all patient groups (asthma =
0.31; DM = 0.18; RA = 0.28). Compared to CT-level
measures, DA-level measures of education were not
more highly correlated with individual-level measures
(asthma = 0.28; DM = 0.12; RA = 0.25).
Discussion
This study is the first to compare the agreement
between individual-level and aggregate-level measures of
income and education among three distinct patient
groups. The results suggest that the ability of aggregate-
level measures to approximate individual-level measures
of SES may vary by the patient group as well as patient
income.
In this study, individual-level income was poorly cor-
related with CT-and DA-level measures, which is con-
sistent with several other reports using Canadian census
data [16,17,31,32]. Our findings are also similar to those
Table 2 Spearman’s rank correlation, intra-class correlation and weighted kappa coefficients for the association of
area-based and self-reported household incomes
Census tractDissemination area
rs
0.28
0.23
0.35
ICC (95%CI)k (95%CI)rs
0.35
0.29
0.33
ICC (95%CI)k (95%CI)
Asthma
Rheumatoid Arthritis
Diabetes
0.25 (0.13,0.35)
0.13(-0.01,0.27)
0.26 (0.13,0.38)
0.20 (0.13,0.27)
0.13(0.02,0.23)
0.27 (0.17,0.37)
0.29 (0.17,0.39)
0.15 (0.03,0.28)
0.27 (0.15,0.38)
0.24 (0.16,0.32)
0.16(0.06,0.25)
0.23 (0.13,0.33)
rs= Spearman’s rank correlation.
ICC(3,1) = Intra-class correlation coefficient (2-way mixed model, absolute agreement).
k = Weighted kappa coefficient.
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 4 of 7

Page 5

of Southern and colleagues [31] who observed that cen-
sus-level measures provided the worst estimations of
income among lower-income households. In all three
patient groups, significantly more patients reported
being in the lowest income category than were classed
as such by either aggregate-level measure. Among
asthma patients, this discrepancy reflects CT-and DA-
level measures having classed lowest-income patients in
the middle income category, while among RA patients it
points to aggregate-level measures having classed
middle-income patients in the highest income category.
Thus, aggregate measures of income tended to classify
patients in higher income categories relative to indivi-
dual-level measures. Notably, among RA patients who in
the survey reported being in the lowest income category,
individual-level incomes were never in agreement with
the corresponding DA-or CT-level measures.
Both CT-and DA-level measures were best at
approximating individual-level income in patients with
DM, in whom aggregate-level values were only signifi-
cantly different from individual-level values for patients
in the lowest income category. The frequency of per-
fect agreement between aggregate-level and individual-
level measures of income was also highest overall
among DM patients, the only patient group in which
both CT-and DA-level values agreed with Individual-
level values in more than fifty percent of cases. As well
as performing best among DM patients, both CT and
DA-level measures best approximated individual-level
measures of income among all patients in the highest
income categories. Accordingly, across patient groups
and income categories, the greatest proportion of cases
in perfect agreement with aggregate-level measures
was observed among DM patients of the highest
income category.
With respect to education, all point biserial correla-
tions between individual-level and aggregate-level mea-
sures were weak, with no differences between CT-and
DA-level measures. Despite the weak correlations across
all patient groups, among RA and DM patients there
were no significant differences between individual-level
and aggregate-level measures in the proportions of
patients classified as having a university degree. The
only difference in these proportions was observed within
the asthma group, where the proportion of patients who
reported having a university degree in the survey was
significantly higher than the proportions in the corre-
sponding CT-and DA-level populations.
These findings should be taken in context with the
limitations of the study. First, individual-level income
and education among our three patient groups was self-
reported and could not be verified, and thus reporting
bias may have affected the SES measures that were
derived from surveys. However, the same can be said of
census measures, which are also self-reported; in
Canada, census measures are the most accessible popu-
lation-based data and no ‘objective’ measures of income
and education are available for the Canadian population.
It should also be noted that, despite the risk of bias,
self-reported measures of SES remain powerful predic-
tors of health outcomes [33]. Ultimately, the absolute
accuracy of the individual-level measures does not affect
the conclusions regarding their agreement with aggre-
gate-level measures.
In this study, individual-level measures are assumed to
be better than aggregate-level measures, an assumption
that follows from evidence that individual-level mea-
sures are more strongly associated with health outcomes
[19]. However, this assumption could be inaccurate
under some circumstances. It is possible that among
some patients, income reported in a cross-sectional sur-
vey is not representative of prior income, e.g., income
before retirement among older patients or income prior
to disease onset among patients with work disability.
This could explain the pattern of non-agreement
observed here among RA patients with the lowest indi-
vidual-level incomes, as RA patients are known to have
a high burden of work disability [34,35]. In these cases,
aggregate-level measures could reflect a prior income
Table 4 Percentage of cases in perfect agreement
between individual-level and DA income groupings
Asthma
n(%) n(%)
DiabetesRheumatoid Arthritis
n(%)
p-value*
<20,000
20,000-50,000
>50,000
12 (13)
52 (63)
77(60)
2 (6)
37 (49)
89 (72)
0 (0)
56 (58)
53 (64)
0.03**
0.23
0.12
*Chi-square test; **Fisher’s exact test.
Table 5 Percentage of cases in perfect agreement
between individual-level and CT income groupings
Asthma
n(%)n(%)
Diabetes Rheumatoid Arthritis
n(%)
p-value*
<20,000
20,000-50,000
>50,000
5 (6)
54 (64)
76 (60)
2 (7)
36 (49)
99 (83)
0 (0)
31 (44)
49 (74)
0.44**
0.03
0.0002
*Chi-square test; **Fisher’s exact test.
Table 3 Percentage of cases in perfect agreement
between individual-level and DA/CT income groupings
Asthma
n(%) n(%)
Diabetes Rheumatoid Arthritis
n(%)
p-value*
DA
CT
p-value*
141 (47)
135 (45)
0.62
128(55)
137 (62)
0.14
109 (49)
80 (49)
0.99
0.17
0.0005
*Chi-square test.
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 5 of 7

Page 6

sustained over a longer period and therefore be more
representative of SES. In addition, individual-level edu-
cation may not be a good measure of SES among some
patients, such as women among the oldest old, whose
husband’s educational attainment may be a better
measure of SES than their own [36]. In this context,
aggregate-level measures of education, which reflect
contextual effects, may be more representative of an
individual’s SES. Finally, the aggregate-level data used
here is from 2001, while individual-level data was col-
lected in 2000, 2002, 2005 and 2008. Although methods
were employed to correct income for inflation, no
adjustment could be made to address the time lapse
between the collection of individual-level measures and
this could be a source of bias. However, given the mean
age of the survey participants in the three patient
groups, income and education status may be expected
to have been relatively stable across the study periods
and comparable to 2001 census measures.
Conclusions
This study shows that the agreement between indivi-
dual-level and aggregate-level measures of SES may
depend on the patient group as well as patient income.
While research is needed to characterize patterns of dif-
ferences between patient groups to help guide the
choice of SES indicators, the use of both individual-level
and aggregate-level measures is advised in studies of
health outcomes.
Acknowledgements
Data for this study was derived from projects funded by the Canadian
Arthritis Network and the B.C. and Yukon Lung Association. These funding
bodies played no role in the study design, collection, analysis and
interpretation of data, nor in the writing of the manuscript or decision to
submit the manuscript for publication.
Author details
1Faculty of Pharmaceutical Sciences, University of British Columbia,
Vancouver, BC, Canada.2Centre for Health Evaluation and Outcome Sciences,
Providence Health Research Institute, Vancouver, BC, Canada.3School of
Population and Public Health, University of British Columbia, Vancouver, BC,
Canada.
Authors’ contributions
CAM was involved in the conception of the study, participated in the study
design and data interpretation and was involved in revising the manuscript
critically for important intellectual content and approving the final version.
LDL was involved in the conception of the study, participated in the study
design and data interpretation and was involved in revising the manuscript
critically for important intellectual content and approving the final version.
SSH was involved in data interpretation, drafted the manuscript and gave
final approval for the version to be published. MG performed the statistical
analysis, interpreted the data and revised the manuscript critically for
important intellectual content.
Authors’ information
CAM is a Government of Canada Research Chair in Pharmaceutical
Outcomes and a Michael Smith Foundation for Health Research Scholar in
Health Services Research. LDL is a Canadian Institutes of Health Research
New Investigator and a Michael Smith Foundation for Health Research
Scholar in Population Health. Both CAM and LDL are Associate Professors in
the Faculty of Pharmaceutical Sciences at the University of British Columbia.
SSH is a researcher in the Faculty of Medicine at the University of British
Columbia. MG is a biostatistician in the Faculty of Pharmaceutical Sciences at
the University of British Columbia.
Competing interests
The authors declare that they have no competing interests.
Received: 26 May 2010 Accepted: 31 March 2011
Published: 31 March 2011
References
1.Lee DS, Chiu M, Manuel DG, Tu K, Wang X, Austin PC, Mattern MY,
Mitiku TF, Svenson LW, Putnam W, Flanagan WM, Tu JV: Canadian
Cardiovascular Outcomes Research Team: Trends in risk factors for
cardiovascular disease in Canada: temporal, socio-demographic and
geographic factors. CMAJ 2009, 181(3-4):E55-66.
2. Kozyrskyj AL, Kendall GE, Jacoby P, Sly PD, Zubrick SR: Association Between
Socioeconomic Status and the Development of Asthma: Analyses of
Income Trajectories. Am J Public Health 2009.
3. Franks P, Gold MR, Fiscella K: Sociodemographics, self-rated health, and
mortality in the US. Soc Sci Med 2003, 56(12):2505-2514.
4.Lynd LD, Sandford AJ, Kelly EM, Pare PD, Bai TR, Fitzgerald JM, Anis AH:
Reconcilable differences: a cross-sectional study of the relationship
between socioeconomic status and the magnitude of short-acting beta-
agonist use in asthma. Chest 2004, 126(4):1161-1168.
5.Watson JP, Cowen P, Lewis RA: The relationship between asthma
admission rates, routes of admission, and socioeconomic deprivation.
Eur Respir J 1996, 9(10):2087-2093.
6.Boulet LP, Belanger M, Lajoie P: Characteristics of subjects with a high
frequency of emergency visits for asthma. Am J Emerg Med 1996,
14(7):623-628.
7.Erzen D, Carriere KC, Dik N, Mustard C, Roos LL, Manfreda J, Anthonisen NR:
Income level and asthma prevalence and care patterns. Am J Respir Crit
Care Med 1997, 155(3):1060-1065.
8.Chen Y, Dales R, Krewski D: Asthma and the risk of hospitalization in
Canada: the role of socioeconomic and demographic factors. Chest 2001,
119(3):708-713.
9.Friday GA Jr, Khine H, Lin MS, Caliguiri LA: Profile of children requiring
emergency treatment for asthma. Ann Allergy Asthma Immunol 1997,
78(2):221-224.
10.Cesaroni G, Farchi S, Davoli M, Forastiere F, Perucci CA: Individual and
area-based indicators of socioeconomic status and childhood asthma.
Eur Respir J 2003, 22(4):619-624.
11.Jacobi CE, Mol GD, Boshuizen HC, Rupp I, Dinant HJ, Van Den Bos GA:
Impact of socioeconomic status on the course of rheumatoid arthritis
Table 6 Percentage of patients with a university degree by individual- and aggregate-level measures
Individual-level
University Degree
n(%)
CT-level University
Degree (expected)
n(%)
p-value*DA-level University
Degree (expected)
n(%)
p-value*
Asthma
Diabetes
Rheumatoid Arthritis
124 (34.5)
82 (29.8)
45 (17.3)
93 (25.9)
74 (26.4)
59 (21.3)
0.01
0.40
0.26
91 (25.3)
71 (25.2)
50 (18.1)
0.01
0.25
0.85
*Chi-square test.
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 6 of 7

Page 7

and on related use of health care services. Arthritis Rheum 2003,
49(4):567-573.
Marra CA, Lynd LD, Esdaile JM, Kopec J, Anis AH: The impact of low family
income on self-reported health outcomes in patients with rheumatoid
arthritis within a publicly funded health-care environment. Rheumatology
(Oxford) 2004, 43(11):1390-1397.
Booth GL, Hux JE: Relationship between avoidable hospitalizations for
diabetes mellitus and income level. Arch Intern Med 2003, 163(1):101-106.
Marra CA, Lynd LD, Esdaile JM, Kopec J, Anis AH: The impact of low family
income on self-reported health outcomes in patients with rheumatoid
arthritis within a publicly funded health-care environment. Rheumatology
(Oxford) 2004, 43(11):1390-7.
Southern DA, McLaren L, Hawe P, Knudtson ML, Ghali WA, Investigators
APPROACH: Individual-level and neighborhood-level income measures:
agreement and association with outcomes in a cardiac disease cohort.
Med Care 2005, 43(11):1116-1122.
Demissie K, Hanley JA, Menzies D, Joseph L, Ernst P: Agreement in
measuring socio-economic status: area-based versus individual
measures. Chronic Dis Can 2000, 21(1):1-7.
Sin DD, Svenson LW, Man SF: Do area-based markers of poverty
accurately measure personal poverty? Can J Public Health 2001,
92(3):184-187.
Hanley GE, Morgan S: On the validity of area-based income measures to
proxy household income. BMC Health Serv Res 2008, 8:79.
Geronimus AT, Bound J: Use of census-based aggregate variables to
proxy for socioeconomic group: evidence from national samples. Am J
Epidemiol 1998, 148(5):475-486.
Diez-Roux AV, Kiefe CI, Jacobs DR Jr, Haan M, Jackson SA, Nieto FJ,
Paton CC, Schulz R: Area characteristics and individual-level
socioeconomic position indicators in three population-based
epidemiologic studies. Ann Epidemiol 2001, 11(6):395-405.
Dominguez-Berjon F, Borrell C, Rodriguez-Sanz M, Pastor V: The usefulness
of area-based socioeconomic measures to monitor social inequalities in
health in Southern Europe. Eur J Public Health 2006, 16(1):54-61.
Centre for Health Services and Policy Research. Census Geography.
[http://www.chspr.ubc.ca/resources/census/geographies/census].
Statistics Canada. Appendix 6: Geographical Areas. [http://www12.statcan.
gc.ca/census-recensement/2011/consultation/DissDiffGuide/Appendix-
Annexe-eng.cfm].
Lynd LD, Sanford AJ, Kelly EM, Paré PD, Bai TR, Fitzgerald JM, Anis AH:
Reconcilable differences: a cross-sectional study of the relationship
between socioeconomic status and the magnitude of short-acting beta-
agonist use in asthma. Chest 2004, 126(4):1161-8.
Guimaraes C, Marra CA, Colley L, Gill S, Simpson S, Meneilly G, Queiroz RH,
Lynd LD: Socioeconomic differences in preferences and willingness-to-
pay for insulin delivery systems in type 1 and type 2 diabetes. Diabetes
Technol Ther 2009, 11(9):567-573.
[http://www12.statcan.ca/english/census01/Products/Reference/dict/index.
htm].
Statistics Canada: Census of Canada: Postal Code Conversion File, PCCF+
Version 4J. 2006, Postal Codes, 2001. 2007.
[http://www.bank-banque-canada.ca/en/cpi.html].
Shrout PE, Fleiss JL: Intraclass correlations: Uses in assessing rater
reliability. Psychol Bull 1979, 86(2):420-428.
Fleiss JL: Statistical methods for rates and proportions Oxford England: John
Wiley & Sons; 1973.
Southern DA, McLaren L, Hawe P, Knudtson ML, Ghali WA, Investigators
APPROACH: Individual-level and neighborhood-level income measures:
agreement and association with outcomes in a cardiac disease cohort.
Med Care 2005, 43(11):1116-1122.
Locker D, Ford J: Using area-based measures of socioeconomic status in
dental health services research. J Public Health Dent 1996, 56(2):69-75.
Lantz PM, Golberstein E, House JS, Morenoff J: Socioeconomic and
behavioral risk factors for mortality in a national 19-year prospective
study of U.S. adults. Soc Sci Med 2010, 70(10):1558-1566.
Sokka T, Kautiainen H, Pincus T, Verstappen SM, Aggarwal A, Alten R,
Andersone D, Badsha H, Baecklund E, Belmonte M, Craig-Muller J, da
Mota LM, Dimic A, Fathi NA, Ferraccioli G, Fukuda W, Geher P, Gogus F,
Hajjaj-Hassouni N, Hamoud H, Haugeberg G, Henrohn D, Horslev-
Petersen K, Ionescu R, Karateew D, Kuuse R, Laurindo IM, Lazovskis J,
Luukkainen R, Mofti A, Murphy E, Nakajima A, Oyoo O, Pandya SC, Pohl C,
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
Predeteanu D, Rexhepi M, Rexhepi S, Sharma B, Shono E, Sibilia J,
Sierakowski S, Skopouli FN, Stropuviene S, Toloza S, Valter I, Woolf A,
Yamanaka H, the QUEST-RA study group: Work disability remains a major
problem in rheumatoid arthritis in the 2000s: data from 32 countries in
the QUEST-RA Study. Arthritis Res Ther 2010, 12(2):R42.
Verstappen SM, Watson KD, Lunt M, McGrother K, Symmons DP, Hyrich KL,
the BSR Biologics Register: Working status in patients with rheumatoid
arthritis, ankylosing spondylitis and psoriatic arthritis: results from the
British Society for Rheumatology Biologics Register. Rheumatology
(Oxford) 2010.
Guilley E, Bopp M, Fah D, Paccaud F: Socioeconomic gradients in
mortality in the oldest old: A review. Arch Gerontol Geriatr 2010.
35.
36.
Pre-publication history
The pre-publication history for this paper can be accessed here:
http://www.biomedcentral.com/1472-6963/11/69/prepub
doi:10.1186/1472-6963-11-69
Cite this article as: Marra et al.: Agreement between aggregate and
individual-level measures of income and education: a comparison
across three patient groups. BMC Health Services Research 2011 11:69.
Submit your next manuscript to BioMed Central
and take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit
Marra et al. BMC Health Services Research 2011, 11:69
http://www.biomedcentral.com/1472-6963/11/69
Page 7 of 7