Formal education level as a significant marker of clinical status in rheumatoid arthritis.

1346
FORMAL EDUCATION LEVEL AS A
SIGNIFICANT MARKER OF CLINICAL STATUS
IN RHEUMATOID ARTHRITIS
LEIGH F. CALLAHAN and THEODORE PINCUS
Clinical status was assessed in 385 patients with
rheumatoid arthritis, according to erythrocyte sedimentation rate, joint count, grip strength, walking time, and
other quantitative measures. All measures indicated
substantially poorer clinical status in patients who did
not complete high school, compared with those who had
completed high school. In general, the poorest results
were seen in patients with only a grade school education.
Progressively better results were seen in patients with
some high school education, high school graduates, and
patients with some college education. No differences in
clinical status were seen among patients who had attended college, graduated from college, or had postgraduate education. Although patients seen at the Veterans Administration Medical Center had lower levels of
formal education than those seen at a university clinic
and private practices, trends in clinical status according
to formal education level were similar in all three
Presented in part at the Southeastern Regional Meeting of
the American Rheumatism Association and the Arthritis Health
Professions Association, Savannah, GA, December 1985, and at the
50th Annual Meeting of the American Rheumatism Association and
the 21st Annual Meeting of the Arthriti? Health Professions Association, New Orleans, LA, June 1986.
From the Division of Rheumatology and Immunology,
Department of Medicine, Vanderbilt University School of Medicine,
Nashville, Tennessee.
Supported in part by the Arthritis Foundation, the Jack C.
Massey Foundation, the Robert Wood Johnson Foundation, NIH
grant AM-21393 to the American Rheumatism Association Medical
Information System, and NIH General Clinical Research Centers
grant 5M-01-RR-00095.
Leigh F. Callahan, BS; Theodore Pincus, MD.
Address reprint requests to Leigh F. Callahan, BS, Division
of Rheumatology and Immunology, B-32 19 Medical Center North,
Department of Medicine, Vanderbilt University, Nashville, TN
37232.
Submitted for publication January 18. 1988; accepted in
revised form June 6, 1988.
Arthritis and Rheumatism, Vol. 31, No. 11 (November 1988)
clinical settings. Differences in clinical status according
to formal education level are not explained by age, sex,
duration of disease, clinical setting, or multiple comparisons. Formal education level may identify an important
marker of clinical status in rheumatoid arthritis.
Morbidity and mortality rates in patients with
rheumatoid arthritis (RA) have been reported to vary
significantly according to level of formal education (1).
These findings are consistent with reports indicating
that lower formal education levels are associated with
increased mortality in cardiovascular disease ( 2 4 ,
and that lower socioeconomic status is associated with
higher disease prevalence and increased mortality in
patients with RA (5-9), cardiovascular disease (10-201,
cancer (21), pulmonary disease (22), and in the general
population (23-37). Formal education level appears to
be a composite or surrogate variable for overall socioeconomic status, although many studies indicate associations between education and health that are independent of other socioeconomic variables (2,9,26,
28,34).
The magnitude of associations between poor
health and lower socioeconomic status suggests that
further studies may lead to useful insights into mechanisms of pathogenesis, morbidity, and mortality of
specific chronic diseases. However, few reports in
medical journals have included markers of socioeconomic status in relation to clinical status (see ref. I),
and most studies of socioeconomic status and health
are based on self-report rather than physical or laboratory measures. Our own previous studies (1) were
limited by the absence of laboratory data, as well as by
a relatively small number of patients. The findings did
suggest that there might be value in studies of larger
EDUCATION LEVEL AND RA
numbers of patients which would include both socioeconomic status a n d clinical status variables.
T h e s e considerations led us to analyze laboratory, physical, a n d functional measures of clinical
status in 385 RA patients whose socioeconomic status
was known. Formal education level w a s chosen a s a
study marker for socioeconomic status, both because
it is more easily and accurately measured than other
markers such a s occupation a n d income, and because
it is much less likely to be affected by disease beginning in adult life than are these other markers. We
report evidence of m o r e severe clinical status in RA
patients with lower educational attainment, according
to all available quantitative measures.
PATIENTS AND METHODS
Patients. The study population included 385 patients
seen at the Division of Rheumatology and Immunology at
Vanderbilt University (VUC), the Nashville Veterans Administration Medical Center Clinic (VAMC), or in the private practices of four rheumatologists in Nashville (Drs.
Joseph W. Huston, J. Thomas John, John S. Johnson, and
John S. Sergent). An effort was made to obtain data on
consecutive patients in the course of a routine visit to a
rheumatologist, without designation of a “study” situation;
it was estimated that more than 95% of consecutive patients
seen during the data collection periods were included. All
201 patients seen at the VUC and VAMC met American
Rheumatism Association (ARA) criteria for definite RA (38).
The other 184 patients were designated by their rheumatologists as having RA. Data were reviewed on a sample of 30
of these patients, and all met ARA criteria for definite RA.
The mean age of the patients was 56 years. Eightyseven percent of the population were white, 63% were
female, and 71% were married. The mean duration of disease
was 1 1 .O years. On the basis of these measures, the patients
appear to be a representative group of RA patients (9).
Measures of clinical status. In the absence of agreement regarding optimal measures by which to assess clinical
status in RA, many types of measures were determined for
these patients. Erythrocyte sedimentation rate (ESR) was
determined using the Westergren method (39). Joint counts
were assessed according to a modified Ritchie index, recording tenderness or pain on motion and swelling in 66 joints
(40,41), i.e., the temporomandibular (2 joints), sternoclavicular (2 joints), acromioclavicular (2 joints), shoulder (2
joints), elbow (2 joints), wrist (2 joints), metacarpophalangeal(l0 joints), proximal interphalangeal (PIP) of the hands
(10 joints), distal interphalangeal (8 joints), knee (2 joints),
ankle (2 joints), tarsus (2 joints), metatarsophalangeal (10
joints), and PIP of the feet (10 joints).
Three physical measures of functional capacity were
obtained using standard techniques (42), as follows: 1) Grip
strength. A blood pressure cuff was inflated to 20 mm Hg,
and the patient was asked to squeeze as hard as he or she
could (43). The score recorded was the mean mm Hg of three
1347
measures of each hand, i.e., six measures. 2) Walk time, The
patient was asked to walk at a normal pace for 25 feet. The
score was recorded in seconds (44). 3) Button test. A
standard board (JA Preston, Clifton, NJ) was used. The
patient was asked to unbutton 5 buttons and then button
them as quickly as he or she could accomplish this task. The
score was recorded in seconds, and the mean for both hands
was recorded (45).
Self-report questionnaires for assessment of difficulty, dissatisfaction, and pain in activities of daily living
(ADL) were administered as previously described (46,47).
Four responses are possible on these ADL scales, analogous
to ARA functional class (48), with a higher score indicating
poorer clinical status (46,47). A standard 10-cm visual analog
scale was used to quantitate pain (49,50). Global self-assessment of functional status was determined using a 1 4 scale (1
= no limitations, 2 = some limitations, 3 = many limitations,
4 = totally disabled), again analogous to ARA functional
class (48).
Analyses according to formal education level and other
demographic variables. The patients were initially classified
into six groups according to formal education level (Table I),
including grade school (24%), some high school (19%), high
school graduate (29%), some college (20%), college graduate
(5%), and postgraduate education (3%). The proportion of
patients with 5 1 1 years of education (43%) is higher than the
25% of individuals in the general population with this level of
education. These data reflect evidence that RA occurs more
frequently in individuals with lower educational attainment
(5-9), as is the case with most chronic diseases (8,37). The
proportions of RA patients according to level of formal
education are similar to those derived from a survey designed to be representative of the United States population,
in which, among all individuals with symmetric polyarthritis,
31% had 8 or fewer years of formal education, 20% had e l l
years, 31% had 12 years, and 18% had more than 12 years of
formal education (9).
The patients seen in private practice settings and at
the VUC were similar in educational attainment, although
both groups showed a higher level of formal education than
patients seen at the VAMC (Table 1). Therefore, analyses
were performed to control and adjust for clinical setting.
Initial analyses according to six educational categories indicated differences among patients in the lowest four
categories, i.e., grade school, some high school, high school
graduate, and some college, but no differences among patients with some college, college graduates, and those with
postgraduate education (Table 2). Therefore, the three categories beyond high school graduate (some college, college
graduate, and postgraduate education) were collapsed into
one group, and the next set of analyses was performed with
the patients grouped into four categories, i.e., grade school,
some high school, high school graduate, and at least one year
of college. This classification appeared advantageous also in
that only 8% of the study population had graduated from
college, which was again consistent with other published
studies (5,9).
Patients in the four education categories did not
differ in terms of race, marital status, or duration of disease,
but did differ significantly in age, age at disease onset, and
sex. The mean ages of the subjects by education category
CALLAHAN AND PINCUS
1348
Table 1. Distribution of 385 rheumatoid arthritis patients studied, according to clinical setting and
level of formal education*
Grade
school
Some
high
school
High
school
graduate
Some
college
College
graduate
Postgraduate
University clinic
Veterans Administration
hospital clinic
Private practice
Total
21
43
20
18
27
14
22
21
8
2
2
2
47
13
22
24
18
19
37
29
17
20
4
5
3
3
40
100
Distribution of individuals with
symmetric polyarthritis in a
population-based survey?
Distribution of all individuals
in US population§
31
20
31
18$
NA
NA
100
10
15
38
37$
NA
NA
100
Total
* Values are percentages. NA
= not available.
9.
$ Includes some college, college graduate, and postgraduate education.
0 Data from ref. 37.
t Data from ref.
high school, 73% in high school graduates, and 67% in those
with more than 12 years of formal education (P< 0,001).
As a result of these findings, all analyses included
statistical adjustments to control for age and sex. Furthermore, analyses to determine clinical status by the study
measures according to formal education level were conducted with patients grouped into four categories according
to sex and age, i.e., males 5 6 0 years, males >60 years,
females 5 6 0 years, and females >60 years. In these analyses, because there already existed four categories according
to age and sex, patients were classified into two groups
according to formal education level, i.e., individuals with
were 62.2, 53.4, 53.6, and 54.7 in those who had been to
grade school only, those with some high school education,
high school graduates, and those with more than 12 years of
formal education, respectively ( P < 0.001, those with grade
school only versus the other three groups). The reported age
at onset of disease was 49.7, 42.2, 43.2, and 42.6 in the four
groups, respectively; this was consistent with the data
showing a similar duration of disease in all four groups. The
age at disease onset in the group with only a grade school
education was significantly (P < 0.001) higher than in the
other groups. The percentage of females was 41% in the
group with a grade school education, 59% in those with some
Table 2. Mean values for laboratory, physical, and self-report measures of disease status in 385 rheumatoid arthritis patients, classified
according to level of formal education
Patients classified according to formal education level
~
Disease status measure*
Laboratory measures
ESR (mmlhour)
Painful joint count
Physical measures
Grip strength (mm Hg)
Walk time (seconds)
Button test (seconds)
Self-report measures
ADL difficulty scale ( 1 4 )
ADL pain scale (11))
ADL dissatisfaction scale ( 1 4 )
Visual analog pain scale (0-10)
Global self-assessment ( 1 - 4 )
All
patients
Grade
school
Some
high
school
40.1
12.1
48.3
16.3
49.4
15.1
98.8
10.3
62.5
93.7
11.2
80.5
92.1
10.0
61.3
1.97
2.37
2.26
5.12
2.68
2.26
2.62
2.54
5.75
3.09
2.04
2.56
2.41
5.85
2.70
High
school
graduate
Some
college
College
graduate
Postgraduate
34.7
9.1
29.3
10.2
41.8
9.3
26.6
8.5
0.002$
O.Ool§
97.9
10.6
60.8
111.6
9.6
46.9
101.6
10.2
58.1
110.9
6.5
54.8
0.079
0.424
0.003f
1.86
2.26
2.12
4.89
2.55
1.73
2.06
2.03
4.26
2.43
* ESR = erythrocyte sedimentation rate; ADL = Activities of Daily Living.
t By analysis of covariance, after controlling for age, sex, clinical setting, and disease duration.
$
P < 0.05 after adjustment for multiple comparisons.
P P < 0.01 after adjustment for multiple comparisons.
1.99
2.46
2.29
4.94
2.61
1.70
2.06
1.81
3.86
2.25
Pi
O.OOO§
0.0019
0.006
0.074
o.oOo(i
1349
EDUCATION LEVEL AND RA
Table 3. Odds ratios of more severe findings on laboratory, physical, and self-report measures of
clinical status in 385 rheumatoid arthritis patients, classified according to level of formal education*
Years of formal education
Grade
school
Unadjusted odds ratios for
clinical status measures
ESR
Joint count
Grip strength
Walk time
Button test
ADL difficulty scale
Visual analog pain scale
Global health self-assessment
Age (>60 years vs. 5 6 0 years)
Sex (male vs. female)
Disease duration (>11 years
vs. < 11 years)
Clinical setting (VAMC vs.
non-VAMC)
Odds ratios for clinical status
measures adjusted for age,
sex, clinical setting, and
duration of disease
ESR
Joint count
Grip strength
Walk time
Button test
ADL difficulty scale
Visual analog pain scale
Global health self-assessment
1.Y1:
2.11
I .8$
4.41:
10.7$
3.11:
2.0$
4.7$
3.2$
2.8t
I .41
Some
high
school
2.31:
1.41
1.71:
2.41:
3.11:
High
school
graduate
1.1
1.41:
1.61:
2.7$
2.01:
Some
college
1 .o
1 .o
1.o
1.o
1
.o
1.o
2.11:
2.01:
I .o
1.41:
1.2
1.41:
1.4$
1.1
1.31:
0.7
I .2$
2.31:
0.7
0.5
1 .o
1.4
1.61:
1.8$
2.71:
4.01:
2.41:
I S1:
2.8$
1.6$
1.1
0.9
0.8
I .3
2.11
1 .5
1.2
1 .o
1 .o
1.o
1.81:
I .61:
2.01:
2.11:
1.4
f .4
I .6$
1.1
1 .o
1.o
1.o
1.o
1.o
Trendt
2.4P
2.3T
2.08
4.06
6.41
3.75
2.49
4.89:
3.48
3.69
1 .o
3.09
1 .o
1.o
I .o
I .o
I .o
1 .o
* ESR = erythrocyte sedimentation rate; ADL = Activities of Daily Living; VAMC = Veterans
Administration Medical Center.
t AnalyLed by the S* statistic.
1: Ninety-five percent confidence limits for these point estimates are > I . I , indicating a 95% probability
that the comparison group differs from the referent group. An odds ratio of 1.3 may show a confidence
limit of > 1.1 if there is little variance and a large number of patients, while in another analysis, a similar
odds ratio may include a confidence limit of 1.0 if there is greater variance or fewer patients.
d P < 0.01.
n P < 0.05.
5 11 years of formal education versus those with 2 12 years
of formal education. This classification was based on the
finding that the most substantial differences in clinical status
according to years of formal education were seen between
those who had and those who had not completed high school
(Tables 2 and 3).
An additional analysis was done with patients classified according to the clinical setting in which they were
seen. All patients were grouped according to one of three
clinical settings, i.e., the VUC, private practice, and the
VAMC. The clinical status of patients within each of these
three groups was analyzed according to the two categories of
formal education used in the studies of age and sex.
Statistical methods. The data were analyzed using the
Statistical Package for the Social Sciences (SPSSx) software
package (5 1). Pearson product moment correlations were
computed initially between all available measures. All measures generally correlated with one another, indicating that
poorer status according to one measure is usually associated
with poorer status according to other measures. However,
other than high correlations among the A D L scores, the
correlations (r) of 0.2-0.5 d o not explain more than 25% of
the variation between two measures. Therefore, it appeared
reasonable t o analyze possible associations between formal
education level and clinical status according to many different types of measures.
Associations between formal education level and
various clinical status measures were analyzed by comparing
mean values of each measure for patients classified into six
groups: grade school, some high school, high school graduate, some college, college graduate, and postgraduate education (Table I). The statistical significance of differences
1350
among all six categories was computed using one-way analysis of variance, as well as analysis of covariance controlling for clinical setting, age, sex, and duration of disease.
The significance of differences between any two individual
groups among the six education categories was determined
in post hoc analyses according to Duncan’s multiple range
test (52). Analyses were controlled for multiple comparisons
by multiplication of the observed P value by the number of
observations, i.e., Bonferroni adjustment (53).
The likelihood that an individual would have more
severe or less severe clinical status according to his or her
formal education level was analyzed by classification of
patients into four education groups, i.e., grade school, some
high school, high school graduate, and some college education, based on analysis of variance (Table 2 ) . The odds that
individuals in the three categories of lower levels of formal
education would have poorer clinical status compared with
those with some college education were then computed as
odds ratios (54). The odds ratio represents the level of risk
for an event associated with the presence or absence of a
study variable. The variables in these analyses are different
measures of clinical status.
For each quantitative measure, more severe clinical
status was defined as a value greater than the median value
for that measure (other than grip strength, for which the
definition was reversed because a high value indicates better
clinical status). When the odds ratio is 1.0, the risk is the
same in a study group as in a referent group, and when it is
> 1.O, an estimate of the odds for more severe clinical status
according to that variable is given. Confidence limits for
odds ratios are computed as 95% Taylor series precisionbased confidence limits for each point estimate (55). If the
confidence limits include 1.0, a difference in an odds ratio
may be secondary to chance; if the lower confidence limit is
2 1.1, a higher odds ratio is significant in 95% of instances.
The confidence limit is a function of the point estimate,
variance, and the number of subjects, so that one odds ratio
may be higher than another but may not exclude a confidence limit of 1.0 if there is greater variance or a smaller
number of patients.
Odds ratios were computed for three sets of analyses
of all the patients. In the first set of analyses, all patients
were classified into four educational groups, i.e., grade
school, some high school, high school graduate, and some
college education, based on analysis of variance (Table 2 ) .
Odds ratios were computed for the grade school, some high
school, and high school graduate categories compared with
the some college education category as the referent group
(odds ratio = 1) each time. The statistical significance of
trends according to the education strata was tested using the
S* statistic described by Brown and Hollander (56).
In the second set of analyses, odds ratios were
computed for all patients categorized by clinical setting, and
in the third set patients were categorized by age and sex,
according to formal education level. As noted above, only
two education level categories were included in the second
and third sets: 5 1 1 years and 2 1 2 years, based on the
findings shown in Tables 2 and 3. This approach was used to
identify six subsets in analyses by clinical setting (VUC,
VAMC, and private practice, each divided into two groups
according to education level), and eight subsets in analyses
CALLAHAN AND PINCUS
by age and sex (males age 5 6 0 , males age >60, females age
560, and females age >60, each divided into two groups
according to formal education level).
In analyses of associations between two variables,
the possibility of confounding by a third variable must be
considered. For example, if an association between lower
formal education level and higher ESR is observed, it is
possible that the associations between formal education
level and ESR are simply explained by associations of both
variables with age, since both higher ESR and lower formal
education levels are seen in individuals who are older. There
are two approaches to analyses of possible confounding of
associations between two variables by a third variable:
stratification and regression modeling. Stratified analyses
were performed by computing odds ratios according to
formal education level, adjusted for age, using the MantelHaenszel chi-square statistic (57), as presented in Table 3.
In regression analyses, possible confounding of associations between two variables by associations between
both variables and a third variable is approached by analysis
of all three (or more) variables simultaneously. Regression
analyses allow identification of the variable that provides the
most explanation of variation in a given variable, known as
a dependent variable, among a group of other variables,
known as independent variables. Furthermore, in these
analyses, the degree to which variation in the dependent
variable is explained by all of the independent variables can
be computed.
Multiple logistic regression analyses were performed
in which formal education level served as the dependent
variable and various clinical status measures as the independent variables, using the maximum likelihood ratio method
(54). A logistic regression allows use of a dichotomous (yes/
no) dependent variable (rather than a continuous variable),
and dichotomous as well as continuous independent variables. Formal education level, the dependent variable, was
dichotomized to 1 1 1 years and 2 1 2 years. Eight separate
logistic regressions were computed, with each containing
age, sex, clinical setting, and a single clinical status variable
as independent variables. A partial odds ratio depicting the
probability of having 511 years of formal education if a
measure indicated poorer clinical status was calculated for
the dependent variable in each regression, controlling for
age, sex, and clinical setting.
To evaluate possible multicollinearity among the
various types of clinical status measures, a logistic regression was computed which included single laboratory, physical, and questionnaire measures of disease status (ESR,
button test, and ADL difficulty scale) as independent variables and formal education level as the dependent variable.
Partial odds ratios are reported for the single laboratory,
physical, and questionnaire measures, after controlling for
the other measures in the equation.
To evaluate associations of three invariant measures
(formal education level, age, and duration of disease) with
clinical status measures, eight stepwise multiple linear regressions were computed. The dependent variables in each
of the eight regressions were the eight clinical status measures: ESR, joint count, grip strength, walk time, button
test, ADL difficulty scale, visual analog scale, and global
health self-assessment scale. The independent variables
EDUCATION LEVEL AND RA
were the three invariant measures. A P \.ralue less than 0.05
was considered significant for a variable to be entered into
the regression model.
RESULTS
Clinical status measures according to formal
education level. All measures analyzed indicated
poorer clinical status in patients with fewer years of
formal education (Table 2). Among traditional disease
activity measures, the mean ESR was 48.3 mm/hour in
patients with a grade school education, 49.4 in those
with some high school, 34.7 in high school graduates,
29.3 in those with some college, 41.8 in college graduates, and 26.6 in those who had attended graduate
school. These differences were statistically significant
after controlling for age, sex, duration of disease,
clinical setting, and multiple comparisons. Mean values for the number ofjoints with tenderness or pain on
motion were 16.3, 15.1,9.1, 10.2,9.3, and 8.5 in the six
groups, respectively. Patterns of statistical significance were similar to those seen for the ESR, and
these patterns also remained significant after controlling for age, sex, duration of disease, clinical setting,
and multiple comparisons.
Among physical measures of functional capacity, mean values for walk time and grip strength
indicated poorer clinical status in patients with fewer
years of formal education, although these differences
were not statistically significant. Results of the button
test, i.e., the time required to button and unbutton 5
buttons on a standard board, indicated values of 80.5
seconds in patients with a grade school education, 61.3
in those with some high school, 60.8 in high school
graduates, 46.9 in those with some college, 58.1 in
college graduates, and 54.8 in those with postgraduate
education. These differences were statistically significant after controlling for age, sex, duration of disease,
clinical setting, and multiple comparisons.
All self-report questionnaire measures also indicated significantly poorer clinical status in patients
with lower formal education levels. The general pattern was similar to that for the physical measures. The
highest scores, indicating greater difficulty, dissatisfaction, and pain in activities of daily living, were seen
in patients with a grade school education. Scores were
progressively lower in the next three groups, i.e.,
those with some high school, high school graduates,
and those with some college education. Again, no
meaningful differences were seen between those with
some college, college graduates, and those with grad-
1351
uate education. The similarity of patterns between
observer-reported and self-reported measures reflects
the general validity of functional status questionnaires
in depicting clinical status in RA (46,50,58).
Relative frequencies (odds ratios) of more severe
disease status according to formal education level. Relative frequencies for age, sex, duration of disease,
clinical setting, and more severe disease determined
by laboratory, physical, and self-report measures of
clinical status were calculated as odds ratios according
to level of formal education (Table 3). The three
groups with formal education levels of some college.
college graduate, or postgraduate education were collapsed into a referent group having >12 years of
formal education, on the basis of findings shown in
Table 2.
The highest odds ratios, indicating the most
severe clinical status, were generally seen in patients
with grade school education, compared with the referent group having >12 years of formal education
(Table 3). Progressively lower odds ratios were seen in
patients with some high school education and high
school graduates. Tests for trend indicated statistical
significance at the P < 0.01 level for six of the eight
comparisons (all except grip strength and joint count,
which were significant at P < 0.05).
Odds ratios for age, sex, duration of disease,
and clinical setting indicated that patients with only a
grade school education were more likely to be older,
male, VA patients, and have a longer duration of
disease (Table 3). Therefore, the odds ratios for the
disease status measures were calculated controlling
for age, sex, clinical setting, and disease duration.
Adjusted odds ratios were lower than unadjusted
results (Table 3), as expected, but remained significantly different according to formal education level.
Further analyses to control for age, sex, and
clinical setting were conducted by classifying all patients into subsets for analyses according to formal
education level. On the basis of the results reported
above, patients were grouped according to two education levels, 5 1 I years of formal education and 2 12
years of formal education, to determine the probability
of poorer status according to formal education.
In the first set of analyses, patients were classified into three subsets on the basis of clinical setting:
VUC, VAMC, or private practice (Table 4). Almost all
odds ratios were >1.5 for the patients with 5 1 1 years
of formal education, indicating that the odds of poor
clinical status were substantially greater for patients
with lower education within all three clinical settings.
CALLAHAN AND PINCUS
1352
Table 4. Odds ratios of having more severe disease status in patients with 5 I I years of formal education versus patients with 2 1 2 years
of formal education, in subsets of rheumatoid arthritis patients grouped by clinical setting"
Private practice
VAMC
vuc
2.63
3.4f
l.9$
1.0
1.0
2.3f
1.0
2.31
1.3
1.0
2.0f
1.0
1.0
2.0f
1.0
1.0
1.0
1.0
2.61;
2.81:
l.3$
1.0
6.5f
1.0
1.0
16.01
1.0
1.0
l.Yf
1.0
1.Yf
3.1$
2.0f
1.0
1.0
1.0
2.5f
2.5f
l.4$
1.0
2.5$
1.0
1.0
8.1$
2.3f
1.0
1.0
1.0
* ESR = erythrocyte sedimentation rate; ADL = Activities of Daily Living; VAMC = Veterans Administration Medical Center: VUC
Vanderbilt University Clinic.
i 5 1 1 and 2 1 2 refer to years of formal education.
1 Ninety-five percent confidence limits for these point estimates are > I . I (see Table 3 for explanation).
=
including the 5 for which partial odds ratios were >2.0
(i.e., ESR,joint count, grip strength, button test, and
global self-assessment). These regression models complemented the stratified analyses reported above, further indicating that associations between lower formal
education level and poorer clinical status are not
explained by age, sex, or clinical setting.
Another consideration was the possible confounding effect of multicollinearity in associations
between lower formal education level and poor clinical
status, i.e., that the eight measures of clinical status
studied might be so highly correlated with one another
that individual analyses of each measure would be
merely depicting the same phenomenon, This consideration was addressed in a logistic regression model
which included three different types of clinical status
measures, i.e., laboratory, physical, and questionnaire
measures (Table 6). The results indicated that clinical
status ratings according to ESR,button test, and ADL
difficulty scale score all vary significantly and independently with formal education level, i.e., associations
In a second set of analyses, patients were
classified into four subsets: females and males, older
and younger than 60 years of age (Table 5). Odds ratios
in all subsets again indicated higher probabilities of
poor status among individuals with 5 1 1 years versus
those with 2 1 2 years of formal education. Odds ratios
according to formal education level were generally
higher for patients in the subsets younger than 60 years
of age, indicating higher relath,e associations of formal
education level and clinical status in younger patients.
Only the ESR, which is known to increase with aging
(39), showed meaningfully higher odds ratios in older
patients.
Regression analysis for multiple variables according to formal education level. Additional studies to
adjust for possible confounding variables were conducted using multiple logistic regression analysis. Initial regression models were computed for each measure of clinical status together with age, sex, and
clinical setting (Table 6 ) . Odds ratios were higher for
all measures adjusted for age, sex, and clinical setting,
Table 5. Odds ratios of having more severe disease in patients with 5 1 1 years of formal education versus patients with 2 1 2 years of
formal education. in subsets of rheumatoid arthritis patients grouped by sex and age*
Males age 560
Males age >60
Females age 560
Females age >60
All patients
1.8t
1.0
2.8f
1.0
2.8f
1.0
1.0
2.4f
1.5
l.3f
3.21;
1.4$
1.0
1.0
1.0
1.0
1.0
1.0
1.0
1.0
3.51
1.5
1.0
1.0
1.0
5.6f
1.0
7.2f
1.1
1.0
1.0
1.0
1.2
1.0
1.0
1.0
5.0f
2.6f
1.6
1.0
1.0
4.6f
0.9
1.0
1.0
2.0$
1.0
1.4$
1.0
1.2
2.9f
1.8$
l.9$
1.0
3.8f
1.0
2.4f
1.0
3.2$
2.1$
* ESR = erythrocyte sedimentation rate; ADL = Activities of Daily Living.
t 5 1 1 and 2 1 2 refer to years of formal education.
f Ninety-five percent confidence limits for these point estimates are >1.1
(see Table 3 for explanation).
3.91
2.61;
2.0
1.1
2.0*
1.0
1.0
1.0
1.0
1.0
7.3$
1.91
1.0
2.71
1.0
1.0
1.0
l.6f
2.95:
1.0
EDUCATION LEVEL AND RA
1353
Table 6. Logistic regression analyses comparing patients with 5 1 1 years of formal education
versus patients with 212 years of formal education*
Results of 8 models including
each disease status measure, as
well as age, sex, and
clinical setting
Clinical status measure
Coefficient
Standard
error
ESR
Joint count
Grip strength
Walk time
Button test
ADL difficulty scale
Visual analog pain scale
Global self-assessment
0.835
0.789
0.709
0.501
1.27
0.667
0.530
0.742
0.258
0.250
0.236
0.240
0.254
0.229
0.270
0.241
Single model including 1
laboratory, 1 physical, and I
questionnaire measure of
disease status
Partial
odds
ratiot
Coefficient
2.318
2.200
2.036
1.657
3.57#
1.958
1.707
2.57#
0.655
I .04
0.740
-
Standard
ertor
0.344
0.351
0.349
-
Partial
odds
ratiof
1.9311
-
2.820
2.107
-
-
-
* In all regressions, formal education level served as the dependent variable. ESR = erythrocyte
sedimentation rate; ADL = Activities of Daily Living.
t Odds ratio for each disease status measure after controlling for other independent variables in the
model (age, sex, and clinical setting).
t Odds ratio for the laboratory, physical, and questionnaire measure of disease status after controlling
for the other measures of disease status in the model.
0 P < 0.01.
n P < 0.05.
# P< 0.001.
between disease status and formal education level
according to the three different types of measures are
not explained by correlations between the measures
themselves.
A final consideration involved a comparison of
three invariate measures, age, duration of disease, and
formal education level, to explain variation in the
various clinical status measures studied (Table 7).
These analyses were conducted because age and duration of disease are generally included in all clinical
Table 7. Analysis of age, duration of disease, and formal education level as explanations of
variation in clinical status measures*
Clinical status
measure (dependent
variable)
ESR
Joint count
Grip strength
Walk time
Button test
ADL difficulty scale
Visual analog pain
scale
Global self-assessment
Independent variable
First step
Second step
Formal education
level
Formal education
level
Disease duration
Disease duration
Age
-t
Formal education
level
Disease duration
Formal education
level
Third step
Multiple
r
R2
0.23
0.05
0.20
0.04
0.21
0.22
0.37
0.04
0.05
0.14
0.30
0.09
Formal education
level
Disease duration
Disease duration
Formal education
level
Age
Age
0.25
0.06
Disease duration
0.30
0.09
* Eight stepwise multiple regressions were computed with each clinical status variable serving as the
dependent variable. The independent variables were age, disease duration, and formal education level.
If the variables met the criterion of P 5 0.05, they were included in the model in the order depicted.
ESR = erythrocyte sedimentation rate; ADL = Activities of Daily Living.
t - indicates no other variables met the level of statistical significance to enter the model.
1354
CALLAHAN AND PINCUS
studies of RA, while formal education level generally
is not included. None of the invariant measures explained more than 14% of the variation in any of the
clinical status measures. However, formal education
level was more explanatory than age or duration of
disease for four clinical status measures, including the
two traditional laboratory and physical measures, ESR
and joint count. These data suggested that studies of
clinical status measures in RA should include, among
the invariant measures considered, formal education
level or some measure of socioeconomic status.
DISCUSSION
Our findings in this cross-sectional study indicated more severe clinical status in RA patients with
lower levels of formal education, according to all
measures studied, including laboratory, physical.
functional, and questionnaire measures. These associations are not explained by associations of formal
education level or clinical status measures with clinical
setting, age, sex, or duration of disease, or by correlations among various types of clinical status measures. Although individuals with lower education levels tended to be older, the odds ratios for poorer
clinical status in individuals with lower education
levels were generally higher in younger patients (for
measures other than ESR, which is known to increase
with aging 1391).
The data reported herein may, in part, explain
the observation that lower formal education level is
associated with substantially higher mortality and
morbidity in RA (1). In our earlier study, patients were
found to not differ significantly at baseline according to
formal education level, although many measures indicated somewhat poorer status in individuals with <8
years of education (1). Those differences were not
statistically significant, in part because the number of
patients was small. The most significant associations
between clinical status and formal education level
were seen in the diflerences between baseline and
9-year review, rather than in baseline and 9-year
review values. These significant differences with small
numbers emphasize the magnitude of changes in clinical status according to formal education level, and
suggest that patients in our previous study may have
been evaluated when their disease was relatively mild.
One important consideration may be that patients with lower levels of formal education may delay
seeing a physician because of financial considerations,
ignorance, limited access, or other variables. How-
ever, differences in status among patients according to
education levels remained significant after controlling
for duration of disease, and no significant differences
were found among the four educational strata according to duration of disease. Furthermore, most measures of clinical status in RA (other than radiographic
measures) are correlated only weakly or not at all with
duration of disease (58,59). Therefore, associations
between formal education levels and disease severity
cannot be explained by differences in duration of
disease, although the possibility cannot be excluded in
these studies that there may be differences in percepfiori of onset of disease according to education level.
Another consideration in explaining our findings is the fact that the patients were seen at referral
settings, rather than a primary care setting. Although
patients seen in the VAMC had less formal education
than patients seen in the VUC or private practice
settings, the trends of more severe clinical status in
patients with fewer years of formal education remained when clinical setting was controlled for using
stratification and modeling analyses. In the patient
populations from all three settings, the proportion of
individuals with <12 years of education was higher
than that in the general population, confirming reports
that RA is more frequent in individuals of lower
socioeconomic strata (5-9). It is possible that differences according to formal education level might not be
seen in primary care locales, although other studies
indicate higher frequencies of symmetric polyarthritis
in a population-based data set (8,9,37). The present
study indicates that among individuals seen in referral
center settings, those with lower educational attainment have more severe clinical status regardless of
age, sex, or clinical setting. However, the possibility
cannot be excluded that patients seen in referral
centers and those seen in primary care locales might
differ according to clinical status or formal education
level.
Associations between greater disease severity
and formal education level do not appear to be unique
to RA: Social class may be an important predictor of
the occurrence and course of disease in general (1-37).
It has been documented that mortality from coronary
artery disease over 7 years in 17,530 civil service
workers in London was highly associated with employment grade, and the relative risk of death was
explained more by employment grade than by the sum
of recognized risk factors including cholesterol level,
smoking, and hypertension (16,17). Furthermore, evidence has been presented that the prevalence of all
EDUCATION LEVEL AND RA
common chronic diseases is higher in individuals with
lower levels of formal education (37).
The mechanisms whereby health and formal
education level are associated remain poorly understood. Formal education level appears to be a surrogate or composite for many variables including income, diet, compliance with medical care, general
health habits, efficiency in consuming medical services, and overall lifestyle (1,23). Associations between
formal education level and health may be viewed as
resulting from biologic, social, or methodologic
sources. The finding of associations between all types
of measures and formal education level in this study,
as well as the associations between education and
health documented in numerous other studies (1-37),
would appear to exclude a methodologic explanation.
While social class is highly correlated with formal
education level, the number and strength of the associations also suggest biologic, as well as social, explanations. Our finding that ESR and formal education
level are associated may provide initial clues to underlying biologic mechanisms to explain the observed
association between education and health.
Our findings do provide strong evidence that, if
the number of years of education is known, variation
in clinical status is not likely to be random in a
population of RA patients, even after adjusting for age,
sex, clinical setting, and duration of disease. Indeed,
formal education level appears to explain variation in
clinical status measures as effectively as do age and
duration of disease, and might be routinely included in
clinical studies as a significant demographic variable
associated with clinical status. Formal education level
is an easily measured socioeconomic variable which is
less likely to be affected by chronic disease than are
variables such as income and occupation.
The findings may have pragmatic implications
for clinical studies and formulation of health policies.
For example, the use of Medicaid data to analyze drug
toxicities or other phenomena might yield results that
are not necessarily applicable to the entire population.
In contrast, the unexpectedly good outcomes in
chronic diseases reported by several individuals (6&
62) may be related to the high education level of the
patient-authors. Finally, most health policies are
based on the premise that risk of disease and responses to therapy are likely to be equal in individuals
of a given age and sex. A need for some reassessment
may be indicated on the basis of associations between
formal education level and health status.
1355
ACKNOWLEDGMENTS
We thank Raye H. Brooks for careful evaluation of
patients and collection of data, Drs. Alan N. Baer, Howard
A. Fuchs, Joseph W. Huston, J. Thomas John, John S.
Johnson, Nancy J. Olsen, and John S . Sergent for encouraging their patients to participate in the study, Carolyn
Burnette and Ginnie Farley for assistance in data processing,
and Marilyn Welch-Fava for expert editorial assistance.
REFERENCES
1. Pincus T, Callahan LF: Formal education as a marker
for increased mortality and morbidity in rheumatoid
arthritis. J Chronic Dis 38:973-984, 1985
2. Weinblatt E, Ruberman W, Goldberg JD, Frank CW,
Shapiro S, Chaudhary BS: Relation of education to
sudden death after myocardial infarction. N Engl J Med
299:60-65, 1978
3. Ruberman W, Weinblatt E, Goldberg JD, Chaudhary
BS: Education, psychosocial stress and sudden cardiac
death. J Chronic Dis 36:151-160, 1983
4. Ruberman W, Weinblatt E , Goldberg JD, Chaudhary
BS: Psychosocial influences on mortality after myocardial infarction. N Engl J Med 311552-559, 1984
5. Cobb S, Kasl SV, Chen E, Christenfeld R: Some psychological and social characteristics of patients hospitalized for rheumatoid arthritis, hypertension, and
duodenal ulcer. J Chronic Dis 1831259-1278, 1965
6. Engel A: National health examination survey: rheumatoid arthritis: comments, Population Studies of the
Rheumatic Diseases. Edited by PH Bennett, PHN
Wood. Amsterdam, Excerpta Medica, 1968
7. Wolfe AM: The epidemiology of rheumatoid arthritis: a
review. I. Surveys. Bull Rheum Dis 19518-523, 1968
8. Pincus T, Callahan LF: Taking mortality in rheumatoid
arthritis seriously: predictive markers, socioeconomic
status and comorbidity. J Rheumatol 132341-845, 1986
9. Mitchell JM, Burkhauser RV, Pincus T: The importance
of age, education, and comorbidity in the substantial
earnings losses of individuals with symmetric polyarthritis. Arthritis Rheum 31:348-357, 1988
10. Syme SL, Hyman MM, Enterline PE: Some social and
cultural factors associated with the occurrence of coronary heart disease. J Chronic Dis 17:277-289, 1964
11. Lehman EW: Social class and coronary heart disease: a
sociological assessment of the medical literature. J
Chronic Dis 20:381-391, 1967
12. Hinkle LE, Whitney LH, Lehman EW, Dunn J, Benjamin B, King R, Plakun A, Flehinger B: Occupation,
education, and coronary heart disease. Science 161:238246, 1968
13. Antonovsky A: Social class and the major cardiovascular diseases. J Chronic Dis 21:65-106, 1968
14. Jenkins CD: Recent evidence supporting psychologic
1356
and social risk factors for coronary disease. N Engl J
Med 294:987-994, 1976
15. Holme 1, Helgeland A, Hjermann I, Leren P, LundLarsen PG: Coronary risk factors in various occupational groups: the Oslo study. Br J Prev SOCMed 3 1 :96100, 1977
16. Marmot MG, Rose G , Shipley M, Hamilton PJS: Employment grade and coronary heart disease in British
civil servants. J Epidemiol Community Health 32:244249, 1978
17. Rose G, Marmot MG: Social class and coronary heart
disease. Br Heart J 45:13-19, 1981
18. Keil JE, Sandifer SH, Loadholt CB, Boyle E Jr: Skin
color and education effects on blood pressure. Am J
Public Health 71:532-534, 1981
19. Jenkins CD: Psychosocial risk factors for coronary heart
disease. Acta Med Scand [Suppl] 660: 123-136, 1982
20. Dobson AJ, Gibberd RW, Leeder SR, O’Connell DL:
Occupational differences in ischemic heart disease mortality and risk factors in Australia. Am J Epidemiol 122:
283-290, 1985
21. Jenkins CD: Social environment and cancer mortality in
men. N Engl J Med 308:395-398. 1983
22 Lebowitz MD: The relationship of socio-environmental
factors to the prevalence of obstructive lung diseases
and other chronic conditions. J Chronic Dis 30599-61 1 ,
I977
23, Stockwell EG: A critical examination of the relationship
between socioeconomic status and mortality. Am J
Public Health 53:95&964, 1963
24. Fisher S: Relationship of mortality to socioeconomic
status and some other factors in Sydney in 1971. J
Epidemiol Community Health 32:4146, 1978
25. Auster RD, Leveson I, Sarachek D: The production of
health: an exploratory 5tudy. J Human Resources 4:411436. 1969
26. Kitagawa E, Hauser P: Differential mortality in the U.S.
Cambridge. Harvard University Press, 1973
27. Nagi MH, Stockwell EG: Socioeconomic differential in
mortality by cause of death. Health Serv Rep 88:449465, 1973
28. Grossman M: The correlation between health and
schooling, Production and Consumption. Edited by NE
Terleckyj. New York, National Bureau of Economic
Research, 1975
29. Syme SL, Berkman LF: Social class susceptibility and
sickness. Am J Epidemiol 104: 1-8, 1976
30. Rosen S, Taubman P: Changes in the impact of education and income on mortality in the U.S., Statistical
Uses of Administrative Records with Emphasis on Mortality and Disability Research. Edited by L DelBene, F
Scheuren. Social Security Administration, 1979
31. Wiley JA, Comacho TC: Life-style and future health:
evidence from the Alameda County study. Prev Med 9:
1-21. 1980
I
CALLAHAN AND PINCUS
32. Fuchs VR: Time preference and health: an exploratory
study, Economic Aspects of Health. Edited by VR
Fuchs. Chicago, University of Chicago Press, 1982
33. Wingard DL, Berkman LF, Brand RJ: A multivariate
analysis of health-related practices: a nine-year rnortality follow-up of the Alameda County study. Am J
Epidemiol 116:765-775, 1982
34. Leigh JP: Direct and indirect effects of education on
health. SOCSci Med 17:227-234, 1983
35. Slater C, Carlton B: Behavior, lifestyle, and socioeconomic variables as determinants of health status: implications for health policy development. Am J Prev Med
1:25-33. 1985
36. Kleinman JC, Madans JH: The effects of maternal
smoking, physical stature, and educational attainment
on the incidence of low birth weight. Am J Epidemiol
121:843-855, 1985
37. Pincus T, Callahan LF, Burkhauser RV: Most chronic
diseases are reported more frequently by individuals
with fewer than 12 years of formal education in the age
18-64 United States population. J Chronic Dis 402365874, 1987
38. Ropes MW, Bennett GA, Cobb S , Jacox R, Jessar RA:
1958 revision of diagnostic criteria for rheumatoid arthritis. Bull Rheum Dis 9:175-176, 1958
39. Baum J, Ziff M: Laboratory findings in rheumatoid
arthritis, Arthritis and Allied Conditions. Tenth edition.
Edited by DJ McCarty. Philadelphia, Lea & Febiger,
1985
40. Ritchie DM, Boyle JA. Mclnnes JM, Jasani MK, Dalakos TG, Grieveson P, Buchanan WW: Clinical studies
with an articular index for the assessment of joint
tenderness in patients with rheumatoid arthritis. Q J
Med 147:393406, 1968
41. Camp AV: An articular index for the assessment of
rheumatoid arthritis. Orthopedics 4:3945, 1971
42. Pincus T, Callahan LF. Brooks RH: Quantitative nonlaboratory measures to monitor and predict the course
of rheumatoid arthritis, Rehabilitation Management of
Rheumatic Conditions. Second edition. Edited by GE
Ehrlich. Baltimore, Williams & Wilkins, 1986
43. Lee P, Baxter A, Dick WC, Webb J: Assessment of grip
strength measurement in rheumatoid arthritis. Scand J
Rheumatol 3:17-23, 1973
44. DeCeulaer K, Dick WC: The clinical evaluation of
antirheumatic drugs, Textbook of Rheumatology. Edited by WN Kelley, ED Harris Jr, S Ruddy, CB Sledge.
Philadelphia, WB Saunders, 1981
45. Clawson DK, Souter WA, Carthum CJ, Hymen ML:
Functional assessment of the rheumatoid hand. Clin
Orthop 77203-210, 1971
46. Pincus T, Summey JA. Soraci SA Jr, Wallston KA.
Hummon NP: Assessment of patient satisfaction in
activities of daily living using a modified Stanford Health
EDUCATION LEVEL AND RA
47.
48.
49.
SO.
51.
52.
53.
54.
Assessment Questionnaire. Arthritis Rheum 26: 13461353, 1983
Callahan LF, Brooks RH, Summey JA, Pincus T: Quantitative pain assessment for routine care of rheumatoid
arthritis patients, using a pain scale based on activities of
daily living and a visual analog pain scale. Arthritis
Rheum 30:630-636, 1987
Steinbrocker 0, Traeger CH, Batterman RC: Therapeutic criteria in rheumatoid arthritis. JAMA 140:659-662,
1949
Scott J , Huskisson EC: Graphic representation of pain.
Pain 2: 175-184. 1976
Fries JF, Spitz P, Kraines RG, Holman HR: Measurement of patient outcome in arthritis. Arthritis Rheum 23:
137-145, 1980
SPSSx User’s Guide. Second edition. New York, McGraw Hill, 1986
Godfrey K: Comparing the means of several groups. N
Engl J Med 313:145&1456, 1985
Cupples LA, Heeren T, Schatzkin A, Colton T: Multiple
testing of hypotheses in comparing two groups. Ann
Intern Med 100:122-129, 1984
Kleinbaum DG, Kupper LL, Morgenstern H, editors:
Epidemiologic Research, Principles and Quantitative
1357
55.
56.
57.
58.
59.
60.
61.
62.
Methods. Belmont, CA, Lifetime Learning Publications,
1982
Greenland S: Interpretation and estimation of summary
ratios under heterogeneity. Stat Med 1:2 17-227. 1982
Brown BW Jr, Hollander M: Statistics: A Biomedical
Introduction. New York, John Wiley & Sons, 1977
Mantel N. Haenszel W: Statistical aspects of the analysis of data from retrospective studies of disease. JNCI
22:719-748, 1959
Pincus T, Callahan LF, Sale WG, Brooks AL, Payne
LE, Vaughn WK: Severe functional declines, work
disability, and increased mortality in seventy-five rheumatoid arthritis patients studied over nine years. Arthritis Rheum 27:864-872, 1984
Pincus T, Callahan LF, Vaughn WK: Questionnaire,
walking time and button test measures of functional
capacity as predictive markers for mortality in rheumatoid arthritis. J Rheumatol 14:240-251, 1987
Alsop S: Day of Execution. New York. Harper & Row,
1973
Cousins N: Anatomy of an Illness as Perceived by the
Patient. New York, Bantam, 1979
Abram M: The Day is Short: An Autobiography. New
York, Harcourt Brace Jovanovich, 1982