Vol. 135, No 12 Printed in U S A.
Educational Status and Blood Pressure: The Second National Health and Nutrition Examination Survey, 1976-1980, and the Hispanic Health and Nutrition Examination Survey, 1982-1984
An inverse association between socioeconomic status, as measured by years of education, and blood pressure has been reported in a number of studies. However, two secular trends may have changed the nature of this relation: a higher mean level of education in the population and intervention for high Wood pressure in the community. Given that education is the most commonly used indicator of social class and measurement is a critical issue in epidemiotogic research, a study was initiated to examine the validity of education for predicting Wood pressure among 11,554 examined persons aged 25-74 years from the Second National Health and Nutrition Examination Survey (1976-1980) or the Hispanic Health and Nutrition Examination Survey (1982-1984). In univariate analysis, a consistent, inverse association between education and blood pressure was found for whites and Wacks, but not for Mexican Americans. After adjustment for age and body mass, the effect persisted only for systolic Wood pressure in whites. The association of education and blood pressure was positive in MexicanAmerican females. Education was inversely related to hypertensive status in whites and in black females. These findings suggest that information on education may be of little value for identifying populations at risk of high Wood pressure, particularly if age and body mass are known. However, information on education may be of considerable value in the design and implementation of appropriate interventions. Am J Epidemiol 1992;135:1339-48. blood pressure; educational status; socioeconomic status
An inverse association between education and blood pressure has been reported in a number of studies conducted during the 1960s and early 1970s (1 -14). However, two secular trends which occurred during the same time period may have implications for the association between education and blood Received for publication April 12, 1991, and in final form January 6, 1992. Abbreviations- HHANES, Hispanic Health and Nutrition Examination Survey, NHANES II, Second National Hearth and Nutrition Examination Survey 1 Department of Epidemiology, School of Pubic Health, University of California, Berkeley, CA. 2 Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC Reprint requests to Dr Janet E. Sorel, Department of Epidemiology, 140 Warren Hall, School of Public Health, University of California, Berkeley, CA 94720 Supported by the Katherine McCormick Fund for Women, Stanford University School of Medicine, and National Institutes of Health Training Grant T32HL07365.
1339
pressure. The first was an increase in the education level of the population (15). As a consequence, educational achievement may no longer have the same social meaning or relations to other dimensions of social class such as income or occupation. In addition, the distribution of education may not be sufficient to discriminate between diseased and nondiseased. The second trend with implications for the education-blood pressure association was the implementation of pharmacologic intervention on high blood pressure within the community (16). The application of widespread pharmacologic treatment may be differential by social class, with persons at high risk (e.g., those with less education) more likely to receive medication (17). In this case, the use of blood pressure as a continuous measure may underestimate the
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
Janet E. Sorel,1 David R. Ragland,1 S. Leonard Syme,1 and Wayne B. Davis2
1340
Soreletal.
MATERIALS AND METHODS Study population
The study population consisted of examined persons aged 25-74 years who participated in either NHANES II or the MexicanAmerican portion of HHANES. Both surveys were multistage, stratified probability samples of the civilian, noninstitutionalized population, aged 6 months to 74 years. The sampling universe for NHANES II was the entire US population, while that for Mexican Americans in HHANES was the five southwestern states with the largest proportion of Mexican-American residents: Arizona, California, Colorado, New Mexico, and Texas. Details of the design and collection procedures have been published (19, 20). The nonresponse rate for the examination portion of NHANES II was 27 percent, with the highest levels among the unemployed, the elderly, persons who did not have access to an automobile, and those who did not have health problems that they wished to discuss with a physician (21). For Mexican Americans, the nonresponse rate to the examination was 23.9 percent, with the highest rates among persons with less education and those whose utilization of health care was low (22). The current analysis was based on data from examinees aged 25-74 years for whom there were valid observations on the following variables: age, height, weight, education, and diastolic as well as systolic blood pres-
sures. The study population included 5,398 males (3,705 whites, 495 blacks, and 1,198 Mexican Americans) and 6,156 females (4,031 whites, 597 blacks, and 1,528 Mexican Americans). A total of 1,763 persons reported current use of antihypertensive medication: 471 white, 86 black, and 76 Mexican-American males and 774 white, 187 black, and 169 Mexican-American females. Treated hypertensives were excluded from analyses in which blood pressure was studied as a continuous measure. Variables
In NHANES II and HHANES, education was measured by self-report of highest grade attended and ranged from zero to 17, with the latter indicating an education level greater than 16 years. Four categories of education were created, and education was used as a categorical variable throughout the analysis: zero to eight, nine to 11, 12, and greater than 12 years. Since age and body mass may be correlated with blood pressure and education, both were evaluated as potential confounders of the education-blood pressure association. Body mass index was estimated using the formula: weight (kg)/height (cm)2 x 100 (23). Age was defined in years as recorded at the time of examination. Age and body mass index were retained as continuous variables. Diastolic and systolic pressures were recorded at two separate times during the physical examination using a standard mercury sphygmomanometer and in compliance with the American Heart Association guidelines (19, 20). The mean of the two seated blood pressures was computed for continuous measures of diastolic and systolic pressures. Examinees who met one or more of the following criteria were classified as hypertensive: diastolic pressure >90 mmHg, systolic pressure >140 mmHg, or current use of antihypertensive medication. Statistical analysis
Because HHANES and NHANES II are stratified probability samples, all observations were weighted by the sample weights
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
true association of education and blood pressure. Given that education has become the most commonly used measure of social class in the epidemiologic literature (18) and an important means for identifying populations at risk (17), a study was initiated to evaluate the predictive validity of education in relation to blood pressure among participants from two health surveys of the US population: the Second National Health and Nutrition Examination Survey (NHANES II) in 1976-1980 and the Hispanic Health and Nutrition Examination Survey (HHANES) in 1982-1984.
Education and Blood Pressure
for that particular point and the regression line fit at the midpoint of each education category (i.e., at 4, 10, 12, and 15 years of education). Ninety-five percent confidence intervals around the slope coefficient were calculated. A similar procedure was followed in fitting a slope to the estimated odds ratios for hypertension. Several decision criteria were employed to assess the presence or absence of an association. For contrasts of mean blood pressures, the magnitude of mean differences and of the slope coefficient for the trends analysis as well as the location of confidence intervals around both values were used in determining whether observed differences were meaningful. The direction of the estimated risk of hypertension, the slope coefficient for the trends analysis, and the width of the confidence intervals around both measures of effect were used in evaluating the association between educational status and hypertension. Finally, the persistence of an effect of education across blood pressure outcomes was considered. RESULTS
The sample distribution by sex, ethnicity, and education category is shown in table 1. Whites were more likely to be high school graduates than were either blacks or Mexican Americans. Blacks were somewhat more equally distributed by education category, while Mexican Americans were disproportionately represented among persons with less than 8 years of education. Unadjusted mean blood pressures by sex, ethnicity, and education category are presented in table 2. An inverse association between education and blood pressure was found for all sex-ethnic subgroups with the exception of diastolic pressure in MexicanAmerican males. Blood pressure differences by education category were more pronounced for systolic than for diastolic pressure, in females than in males, and for blacks than for whites or Mexican Americans. Regression coefficients and the corresponding confidence intervals for tests of trends in blood pressure by education category confirmed these findings.
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
which consisted of the reciprocals of the sampling probabilities, adjusted for nonresponse and poststratification by age, sex, and race. The small number of primary sampling units in HHANES and of blacks in NHANES II have been shown to produce unstable estimates of variances using available software programs designed for complex samples. To compensate for this limitation, an alternative method, recommended by the National Center for Health Statistics (24, 25), was applied in the computation of point estimates and variances. First, sample weights were incorporated into procedures from the Statistical Analysis System to produce estimated means, regression coefficients, and standard errors. Second, design effects (a ratio of the variance from a cluster sample to that from a simple random sample of the same size) for diastolic and systolic pressures were generated from SUDAAN: Software for Survey Data Analysis (26) and averaged across sex, ethnic, and 10-year age categories. These average design effects were then used to adjust the standard error estimates obtained from weighted analysis in the Statistical Analysis System. All analyses were ethnic group and sex specific. Potential confounding for age and body mass index was assessed using stratified analysis of mean age and body mass index by education level and by univariate regression models with age and body mass predicting blood pressure. Unadjusted and ageand body mass index-adjusted diastolic and systolic pressure means by education categories were generated using the least squares means procedure within the Statistical Analysis System. Ninety-five percent confidence intervals were computed from the formula fl ± 1.96 (standard error) and used to determine the precision of estimated means and odds ratios. A test for trends in blood pressure means by education category was computed using a technique adapted from Rothman (27). A linear regression model was fit to the four blood pressure means to produce a slope coefficient. Since education categories were not uniform, each point estimate or mean was weighted by the inverse of the variance
1341
1342
Sorel et al.
TABLE 1. Sample distribution by sex, ethnicity, and educational status. NHANES II* and HHANES*: whites, blacks, and Mexican Americans aged 25-74 years Males Whites
Blacks
Females Whites
Blacks
Mexican Americans
1,528
Medication users included No. Education (years) 0-8 9-11 12 >12
3,705
495
1,198
4,031
597
740 535 1,167 1,263
199 105 108 83
552 201 240 205
696 654 1,564 1,117
212 146 154 85
746 262 329 191
1,359
Medication users excluded No. Education (years)
0-8 9-11 12 >12
3,234
409
1,122
3,257
410
616 457 1,014 1,147
149 90 93 77
511 192 227 192
472 513 1,302 970
124 96 120 70
632 239 306 182
• NHANES II, Second National Health and Nutnton Examination Survey; HHANES, Hispanic Health and Nutrition Examination Survey.
Since education may be correlated with age in a cross-sectional survey, we evaluated the role of age in explaining the association between education and blood pressure. Preliminary analyses demonstrated a strong inverse association of age with education. Evidence of an interaction between age and education was present for diastolic pressure only in white males, but the effect was not pronounced, and we controlled for age in the remaining analyses. Age-adjusted blood pressure means by sex, ethnicity, and education category are presented in table 3. Adjustment for age substantially reduced the inverse association between education and blood pressure, which remained significant only for white and for black females and for systolic pressure in white males. There was no consistent association of education and blood pressure in black and in Mexican-American males. Education was positively related to both diastolic and systolic pressures in Mexican-American females, but only the results for diastolic pressure were statistically significant. Because an inverse association between education and body mass index has been reported in a number of studies, we next examined the role of obesity in explaining
the association of education and blood pressure. An inverse association between body mass index and education was found in women from NHANES II and HHANES. Body mass index was added to the models used to compute adjusted mean blood pressures. Age- and body mass index-adjusted mean blood pressures by sex, ethnicity, and education category are shown in table 4. After adjustment for body mass index and age, there was little evidence of a consistent, pronounced, or statistically significant association between education and blood pressure. The exceptions were for systolic pressure in whites and in black females and for diastolic pressure in Mexican-American females. The association was inverse for whites and blacks and positive for Mexican Americans. In females, adjustment for body mass index substantially reduced the educationblood pressure association, particularly in white and in black women, for whom there was a clear inverse relation between body mass index and education. These findings were supported by tests of trends. Given the success of community intervention programs for hypertension treatment and the possible correlation of medication status with education (17), we examined dif-
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
Mexican Amencans
Education and Blood Pressure 1343
TABLE 2. Unadjusted* mean diastolic and systolic blood pressures and standard errors (SE) by sex, ethnicity, and educational status. HHANESt and NHANES lit: whites, blacks, and Mexican Americans aged 25-74 years Mates Whites
Blacks
Mean ± SE
Mean±SE
Mean±SE
Whites
Blacks
Mexican Americans
Mean±SE
Mean±SE
Mean±SE
Diastolic pressure 86.8 ± 1 . 4 82.7 ± 0.9 83.9 ± 0.9 86.1 ± 1 . 4 83.1 ± 1.3 83.3 ± 0.5 82.7 ± 1 . 4 81.5 ± 0 . 5 -0.14 -0.40 Pt 95% Clf -0.25 to -0.04 -0.70 to -0.09
0-8 9-11 12 >12
79.1 ± 0.6 78.0 ± 0.9 79.0 ± 0.8 79.3 ± 0.9 0.003 -0.13 to 0.14
86.5 ± 1.7 84.5 ± 1 . 6 79.5 ± 1 . 3 74.9 ± 1 . 7 -1.03 -0.59 to -0.36 -1.36 to -0.70 81.0± 1.0 78.3 ± 0.8 77.9 ± 0.5 75.8 ± 0.5 -0.47
73.1 ± 0.5 73.6 ± 0.7 73.6 ± 0.6 71.9 ± 0 . 8 -0.04 -0.16 to 0.07
Systolic pressure 0-8 9-11 12 >12
P 95% Cl
133.2 ± 1 . 6 140.2 ± 2 . 7 124.6 ± 1 . 3 128.1 ± 2 . 5 121.5 ± 0 . 8 119.9 ± 2 . 1 116.9 ± 0 . 9 114.9 ± 2 . 8 -0.77 -0.79 -0.38 -1.48 -2.38 -0.94 to -0.60 -1.25 t o - 0 . 3 3 -0.58 t o - 0 . 1 8 -1.68 t o - 1 . 2 9 -2.95 to -1.81 133.5 131.5 129.7 125.6
±1.3 ± 1.3 ±0.8 ±0.7
135.2 132.3 125.4 127.9
±1.9 ±1.9 ±1.8 ±1.9
123.3 120.4 121.2 118.7
±1.0 ±1.6 ± 1.4 ± 1.5
117.2 ± 0 . 8 113.5 ± 1 . 2 114.5 ± 1.0 109.8 ± 1 . 3 -0.55 - 0 . 7 4 to - 0 . 3 7
* Weighted by the sample weights. t HHANES, Hispanic Hearth and Nutrition Examination Survey; NHANES II, Second National Hearth and Nutrition Examination Survey; Cl. confidence interval t Regression coefficients for trends tests represent change In mean blood pressure (mmHg) for each year of change in education.
TABLE 3. Age-adjusted* mean diastolic and systolic blood pressures and standard errors (SE) by sex, ethnicity, and educational status. HHANESt and NHANES lit: whites, blacks, and Mexican Americans aged 25-74 years Males Education (years)
Whites
Blacks
Mean ± SE
Mean ± SE
Females Mexican Americans
Whites
Blacks
Mexican Americans
Mean ± SE
Mean ± SE
Mean± SE
Mean± SE
Diastolic pressure 0-8 9-11 12 >12
fit 95% Clt
82.2 ± 0.9 83.8 ± 0.9 83.9 ± 0.6 82.3 ± 0.5 -0.01 -0.12 to 0.10
85.8 ± 1 4 86.9 ± 1.4 84.5 ± 1.4 84.1 ± 1 . 5 -0.16 -0.48 to 0.16
78.8 ± 0.6 78.5 ± 0.9 79.7 ± 0.9 80.1 ± 0.9 0.11 -0.03 to 0.24
79.9 ± 1.0 78.5 ± 0.8 78.9 ± 0.5 77.4 ± 0.6 -0.22 -0.33 t o - 0 . 1 0
84.2 ± 1 . 7 85.6 ± 1 . 5 82.2 ± 1.4 77.9 ± 1.8 -0.50 -0.84 t o - 0 . 1 6
72.5 ± 0.5 74.4 ± 0.7 74.4 ± 0.6 73.3 ± 0.8 0.14 0.03 to 0.26
133.7 ± 2 . 6 131.3 ± 2 . 3 127.7 ± 2 . 1 123.3 ± 2 . 7 -0.90 -1.45 t o - 0 . 3 5
115.2 ± 0 . 7 116.3 ± 1.1 117.1 ± 0 . 9 114.4 ± 1.2 0.06 -0.11 to 0.22
Systolic pressure 0-8 9-11 12 >12 /3 95% Cl
131.3 ± 1 . 2 131.3 ± 1 . 2 131.9 ± 0 . 8 129.0 ± 0 . 7 -0.23 -0.40 to -0.06
131.6 ± 1.8 135.1 ± 1 . 9 130.6 ± 1.9 133.4 ± 2.0 0.09 -0.37 to 0.55
122.4 ± 1 . 0 122.3 ± 1.5 123.8 ± 1.3 121.8 ± 1.4 0.01 -0.18 to 0.20
129.1 ± 1.4 125.2 ± 1 . 2 125.0 ± 0 7 122.4 ± 0 8 -0.59 -0.77 to -0.41
• Weighted by the sample weights. t HHANES, Hispanic Health and Nutrition Examination Survey; NHANES II, Second National Health and Nutrition Examination Survey; Cl, confidence interval. $ Regression coefficients for trends tests represent change in mean blood pressure (mmHg) for each year of change in education.
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
Education (years)
Females Mexican Americans
1344
Soreletal.
age and body mass index, to estimate the risks of hypertension by education level. The results are shown in table 6. With the highest education level as the reference category, the risk of hypertension was not consistently different by education for any sex-ethnic subgroup. In general, all confidence intervals around the estimated odds ratios included 1.0, with the exception of those for white females. In this subgroup, the risk of hypertension was significantly higher in women
TABLE 4. Age- and body mass index-adjusted* mean diastoJIc and systolic blood pressures and standard errors (SE) by sex, ethnicity, and educational status. HHANESt and NHANES lit: whites, blacks, and Mexican Americans aged 25-74 years Males Education (years)
Females
Whites
Blacks
Mexican Americans
Whites
Blacks
Mexican Americans
Mean±SE
Mean±SE
Mean±SE
Mean±SE
Mean ± SE
Mean±SE
Diastolic pressure
0-8 9-11
82.5 ± 0.8 83.4 ± 0.8 83.4 ± 0.5 82.5 ± 0.5 -0.01 -0.11 to 0.09
12 >12
n 95% Clt
85.7 ±1.3 86.3 ±1.4 84.3 ±1.4 83.7 ±1.4 -0.19 -0.49 to 0.11
79.0 ± 0.6 78.6 ± 0.9 79.6 ± 0.8 79.9 ± 0.9 0.07 -0.06 to 0.20
78.8 ± 0.9
78.1 ± 0.8 78.9 ± 0.5 78.2 ± 0.5 -0.04 -0.15 to 0.07
82.8 ±1.6 85.0 ±1.4 83.0 ±1.3 79.2 ± 1.7 -0.25 -0.56 to 0.07
72.4 ± 0.4
74.2 ± 0.7 74.6 ± 0.6 73.9 ± 0.8 0.20 0.09 to 0.31
Systolic pressure
0-8 9-11
131.7 ±1.2 130.8 ±1.2 131.3 ±0.7 129.2 ±0.7
131.5 ±1.7 134.4 ±1.8 130.3 ±1.8 132.8 ± 1.9
-0.23 -0.39 to -0.07
-0.38 to 0.49
12 >12
P 95% Cl
0.05
122.6 ±0.9 122.4 ±1.5 123.7 ±1.3 121.5 ±1.4 -0.03 -0.21 to 0.16
127.6 ±1.3 124.7 ±1.1 125.0 ±0.7 123.6 ±0.8 -0.35 -0.52 t o - 0 . 1 7
131.7 130.3 128.9 125.3
±2.4 ±2.2 ±2.0 ±2.6
-0.52 -1.04 to 0.001
115.0 ±0.7 116.1 ± 1.0 117.3 ±0.9 115.1 ±1.1 0.12 -0.03 to 0.28
• Weighted by the sample weights. t HHANES, Hispanic Health and Nutrition Examination Survey; NHANES II, Second National Health and Nutrition Examination Survey; Cl, confidence Interval. t Regression coefficients for trends tests represent change in mean Wood pressure (mmHg) for each year of change in education.
TABLE 5. Proportion* of persons on antihypertensJve medication by sex, ethnicity, and educational status. HHANESt and NHANES lit: whites, blacks, and Mexican Americans aged 25-74 years Females
Males Education (years)
0-8
9-11 12
Whites (n - 3,705)
Blacks (n = 495)
14.5 11.5 10.9 6.4
19.4 10.1 9.4
5.9
Mexican Americans (n- 1,198)
Whites (n-4,031)
Blacks (n = 597)
Mexican Americans (n " 1,528)
6.3 3.3 4.1 4.9
28.0 17.1 13.8 8.2
37.1 25.4 19.1 16.9
15.2 7.9 6.4 4.2
• Weighted by the sample weights tHHANES, Hispanic Health and Nutrition Examination Survey; NHANES II, Second National Health and Nutrition Examination Survey.
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
ferences in treatment status by education category in NHANES II and HHANES (table 5). In general, the proportion of hypertensives who reported current use of medication was higher among those with less education. This relation was pronounced for all sex-ethnic groups with the exception of Mexican-American males, for whom the proportion treated was relatively low and did not differ markedly by education status. Next we fit logistic models, adjusting for
Education and Blood Pressure 1345
s
s °P *~.
OJ CM
o
in • * in
^
T-
t
d
T
odd
p d
£• tn
ui
a> p p p p O
^
-i^
i-1
O
*? °? *?
2
LO 00 00
^
m
3
-g I g | I
N S CO O O T^
T^
T^
T^
O
a
I C\irT i-; IV. 6 T^ O T^
T3 C
O
^ O
o
a
g
in p cq p p T^
T^
T^
T^
O
f
i
p o o ir> co sI | o 11 II •o s 8g 5^ « g 11 z _^ I" §
After adjustment for age and body mass, the inverse association between education and blood pressure was weakened or eliminated among participants from NHANES II and HHANES. Only the inverse relation of education and systolic pressure in whites and the direct relation between education and diastolic pressure in Mexican-American females were of sufficient magnitude to be considered important and remained statistically significant. Evidence for an association of education and hypertensive status was not conclusive, although the direction of confidence intervals and tests for trends suggested the presence of an inverse association between education and hypertension in whites and in black females. The absence of a pronounced or consistent association of education and blood pressure in NHANES II and HHANES was unexpected given the weight of evidence from the published literature. The results were, however, similar to those from an earlier study of NHANES II, which reported relatively small differences in age-adjusted blood pressure by education category (28). We explored several possible explanations for disparities between our findings and those reported in other studies. Within the current study, education was categorized into four education strata. Comparability required the use of common cutpoints for three ethnic groups with very dif-
1
fi
g
0 0)0)00 r rdrd
s
5
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
with 0-8 years of education compared with those with 13 or more years of education. However, it should be noted that while the estimated risk of hypertension was not appreciably different by education category for any sex-ethnic group and confidence intervals were wide, the direction of the confidence intervals supported earlier findings of elevated blood pressure in the lowest compared with the highest education categories, particularly for blacks. In addition, regression coefficients and confidence intervals for tests of trends indicated that the risk of hypertension was higher for less-educated whites and black females.
•r
1346
Soreletal. of education as an indicator of social class across population subgroups. As an empirical variable, years of education has become the preferred measure of socioeconomic status in epidemiologic research. However, until recently little attention has been given to an evaluation of the underlying assumptions of the model, to the impact of secular trends on the distribution of education, to the relation between education and other dimensions of social class, or to the meaning of education across population subgroups (18). For Mexican Americans, education may be a more valid measure of cultural orientation, which has been associated with health status (29-36). Second, the documented inverse association between education and blood pressure may be little more than an artifact of study design and methods employed in earlier investigations. Given the association of education with age and body mass index found in the current analyses, failure to examine the role of age and body mass index as potential confounders or effect modifiers and to adjust for these relations in the analysis may misrepresent the association of education and blood pressure. Similarly, given the pronounced association between education and medication status found in the current analyses, one can no longer ignore the issue of treatment which was often disregarded in past studies. If one assumes the presence of an association between social class and blood pressure, the results of our investigation underscore the need to reexamine the underlying assumptions of the model with respect to the validity of education as a measure of socioeconomic status across population subgroups. At the very least, education lacks predictive validity. From a public health perspective, if there is no contribution of education beyond that explained by age and obesity, age and body mass index may be of greater value in the identification of populations at risk. In this case, the primary value of information on educational status may be in the design and implementation of appropriate interventions, particularly those targeted to obesity. If body mass index is the
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
ferent education distributions and resulted in two categories for persons with less than 12 years of formal education. Since a high school education was the reference point in the majority of studies on education and blood pressure, the data were reanalyzed using these education categories. There were no changes in the pattern of results. Often, hypertension is defined using only medication status and diastolic pressure categorized at either 90 or 95 mmHg. We recomputed the risk of hypertension by using several classification schemes and found only minor variations in study findings. All estimated risks from these logistic regressions were within the confidence intervals presented in table 6. Finally, disparities between our findings and those reported in previous studies may be a consequence of selection bias in the design of NHANES II and HHANES. This issue could not be addressed directly by using public data tapes from the National Center for Health Statistics. However, given that both surveys were community samples intended to make inference to the US population, less selection bias would be expected than from previous investigations conducted in occupational settings or restricted geographic areas. Although response rates differed by education in HHANES and by income in NHANES II, nonresponse was highest among the least educated and the poor, who were not likely to have been sampled in existing studies. In summary, selection and information bias do not seem to be plausible explanations for the absence of a pronounced or consistent association between education and blood pressure in NHANES II and HHANES. Nor would these factors explain variations in the effect of education by sex, ethnicity, and blood pressure outcome. Two alternative explanations may be more likely to account for the patterns of association observed in NHANES II and HHANES and for disparities between the findings of our study and those reported in the literature. First, differences in the association of education and blood pressure by sex and ethnicity raise questions regarding the validity
Education and Blood Pressure
primary mechanism by which socioeconomic status is related to high blood pressure, research on the association of social class and obesity is critical for reducing the chronic disease burden of the poor.
14.
16. REFERENCES 1. Dawber TR, Kannel WB, Kagan A, et al. Environmental factors in hypertension. In: Stamler JS, Stamler R, Pullman T, eds. The epidemiology of hypertension. New York: Grune & Stratton, 1967:255-88. 2. National Center for Health Statistics. Hypertension and hypertensive heart disease in adults. Hyattsville, MD: National Center for Health Statistics, 1973. (Vital and health statistics. Series 11: Data from the National Health and Nutrition Examination Survey, no. 13) (DHEW publication no. (PHS) 1000). 3. Syme SL, Oakes TW, Friedman GD, et al. Social class and racial differences in blood pressure. Am J Public Health 1974;64:619-20. 4. Dyer AR, Stamler J, Shekelle RB, et al. The relationship of education to blood pressure: findings on 40,000 employed Chicagoans. Circulation 1976: 54:987-92. 5. Hypertension and Detection Follow-up Program Cooperative Group. Race, education and prevalence of hypertension. Am J Epidemiol 1977; 106:351-61. 6. National Center for Health Statistics. Blood pressure levels in persons 6-74 years. Hyattsville, MD: National Center for Health Statistics, 1977. (Vital and health statistics. Series 11: Data from the National Health and Nutrition Examination Survey, no. 203) (DHEW publication no. (HRA) 78-1648). 7. National Center for Health Statistics. Characteristics of persons with hypertension: United States, 1974. Hyattsville, MD: National Center for Health Statistics, 1978. (Vital and health statistics, Series 10: Data from the National Health Interview Survey, no. 121) (DHEW publication no. (PHS) 791549). 8. Keil JE, Sandifer SH, Loadholt CB, et al. Skin color and education effects on blood pressure. Am J Public Health 1981;71:532-4. 9. Keil JE, Tyroler HA, Sandifer SH, et al. Hypertension: effects of social class and racial admixture. Am J Public Health 1977,67:634-9. 10. Liu K, Cedres LB, Stamler J, et al. Relationship of education to major risk factors and death from coronary heart disease, cardiovascular diseases and all causes: findings of three Chicago epidemiologic studies. Circulation 1982,66:1308-14. 11. Jacobsen BK, Thelle DS. Risk factors for coronary heart disease and level of education: the Tromst) Heart Study. Am J Epidemiol 1988; 127:923-32. 12. Matthews KA, Kelsey SF, Meilahn EN, et al. Education attainment and behavioral and biologic risk factors for coronary heart disease in middle-aged women. Am J Epidemiol 1988; 129:1132-44. 13. Kraus JF, Borhani NO, Franti CE. Socioeconomic
17. 18. 19.
20.
21. 22.
23.
24.
25.
26.
27. 28.
status, ethnicity, and risk of coronary heart disease. Am J Epidemiol I98O;111:407-14. Winkelby MA, Fortmann SP, Barrett DC. Social class disparities in risk factors for disease: eightyear prevalence patterns by level of education. Prev Med 1990;l9:l-12. US Bureau of the Census. Statistical abstract of the United States: 1982-1983. Washington, DC: US GPO, 1983. Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. The 1988 report of the Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. Arch Intern Med 1988; 148: 1023-38. Tyroler HA. Socioeconomic status in the epidemiology and treatment of hypertension. Hypertension 1989; 13(Suppl): 194-7. Liberatos P, Link BG, Kelsey JL. The measurement of social class in epidemiology. Epidemiol Rev 1988,10:87-121. National Center for Health Statistics. Plan and operation of the Second National Health and Nutrition Examination Survey. Hyattsville, MD: National Center for Health Statistics, 1981. (Vital and health statistics. Series 1: Programs and collection procedures, no. 15.) (DHEW publication no. (PHS) 81-1317). National Center for Health Statistics. Plan and operation of the Hispanic Health and Nutrition Examination Survey, 1982-1984. Hyattsville, MD: National Center for Health Statistics, 1985. (Vital and health statistics. Series 1: Programs and collection procedures, no. 19.) (DHEW publication no. (PHS) 85-1321). Forthofer RN. Investigation of nonresponse bias in NHANES II. Am J Epidemiol 1983:117:507-15. Gonzalez JF, McMillen M. Nonresponse and noncoverage analysis in the southwest component of the Hispanic Health and Nutrition Examination Survey. Proceedings of the section on survey research methods. Washington, DC: American Statistical Association, 1986:326-31. National Center for Health Statistics. Obese and overweight adults in the United States. Hyattsville. MD: National Center for Health Statistics, 1983. (Vital and health statistics. Series 11: Data from the National Health and Nutrition Examination Survey, no. 230) (DHEW publication no. (PHS) 83-1680). Division of Health Statistics. National Center for Health Statistics. Analytic guidelines and reliability criteria for analyzing HANES data. (Internal memo). Washington, DC: National Center for Health Statistics, 1989. Delgado JL, Johnson CL, Roy I, et al. Hispanic Health and Nutrition Examination Survey: methodological considerations. Am J Public Health 199O;8O(suppl):6-10. Shah BV, LaVange LM, Barnwell BG, et al. SUDAAN: software for survey data analysis. Research Triangle Park, NC: Research Triangle Institute, 1989. Rothman KJ. Modern epidemiology. Boston: Little, Brown & Co., 1986:336^16. Drizd T, Dannenberg AL, Engel A. Blood pressure levels in persons 18-74 years of age in 1976-80, and trends in blood pressure from 1960 to 1980 in
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
15.
1347
1348
29.
31. 32.
the United Stales. Hyattsville, MD: US GPO. 1986. (Vital and Health Statistics. Series 11, no. 234). (DHHS publication no. (PHS) 86-1684). Ostfeld AM, D'Atri DA. Rapid sociocultural change and high blood pressure. Adv Psychosom Med 1977;9:20-37. Scotch NA. Sociocultural factors in the epidemiology of Zulu hypertension. Am J Public Health 1962;53:1205-13. Alfred BM. Blood pressure changes among male Navajo migrants to an urban environment. Can Rev Soc Anthropol 1970,7:189-200. Prior 1AM, Stanhope JM, Evans JG, et al. The Tokelau Island migrant study. Int J Epidemiol
l974;3:225-32. 33. Beaglehole R, Salmond C, Hooper A, et al. Blood pressure and social interaction in Tokelauan migrants in New Zealand. J Chronic Dis 1977; 30:803-12. 34. Beaglehole R, Eyles E, Salmond C, et al. Blood pressure in Tokelauan children in two contrasting environments. Am J Epidemiol 1978; 108:283-8. 35. Marmot MG, Syme SL. Acculturation and coronary heart disease in Japanese-Americans. Am J Epidemiol 1976;104:225-47. 36. Marikedes KS, Coreil J. The health of Hispanics in the southwestern United States: an epidemiologic paradox. Public Health Rep 1986; 101:253-65.
Downloaded from https://academic.oup.com/aje/article-abstract/135/12/1339/88708 by Tulane University user on 17 January 2019
30.
Soreleta).
