PREVENTIVE
MEDICINE
19, 1-12 (1990)
Social Class Disparities in Risk Factors for Disease: Eight-Year Prevalence Patterns by Level of Education’ MARILYN Stanford
A. WINKLEBY,
PH.D.,’ STEPHEN DONALD C. BARRETT,
P. FORTMANN, M.S.
M.D., AND
Center for Research in Disease Prevention, Stanford University School of Medicine, loo0 Welch Road, Palo Alto. Culifornia 94304-1885
This article examines the associations between education, a primary indicator of social class, and six risk factors for disease. Data are presented on a sample of 3,349 individuals ages 25-74 years who participated in one of four cross-sectional surveys conducted by the Stanford Five-City Project between 1979and 1986.The six risk factors examined are knowledge about health, cigarette smoking, hypertension, serum cholesterol, body mass index, and height. A highly significant pattern of associations was found between education level and the six risk factors, in the direction of higher risk among those with lower education (all P values 16 years (postgraduate). Income level is defined as total gross annual family income and is divided into 11 categories, ranging from $5,000 to $75,000 per year. All analyses are adjusted for time of survey as a surrogate adjustment for inflation. Occupation is coded according to the 13 standard census occupational codes, which are collapsed into the following 10 categories and used as indicator variables: executives, managers, administrators, clerical workers, skilled and semiskilled workers, unskilled workers, unemployed, homemakers, students, and retired. Knowledge about health and cigarette smoking status were determined by selfreported questionnaire data. Health knowledge encompassed questions about cardiovascular disease risk factors and is presented as a summary index of 17 items where the highest score reflects the highest level of knowledge. Cigarette smoking, coded as “yes” if the respondent had smoked one or more cigarettes in the last 48 hr and had ever smoked on a regular basis, was confirmed by expired air carbon monoxide and serum thiocyanate (29). Height, weight, and blood pressure were collected according to standardized study protocols. Height was measured without shoes to the nearest 0.6 cm by metal rule and weight was measured to the nearest 0.1 kg without shoes on a beam balance scale. BMI, used as a measure of obesity because of its high correlation with other estimates of body fat (30), is defined as wt(kg)/ht(m*). Serum TC, derived from nonfasting venous samples and analyzed fresh for lipids and lipoproteins by methods established by the Lipid Clinics Research Program (31), is measured in milligrams per deciliter. Blood pressure (BP) was determined by the mean of two manual measurements obtained by standard protocol (32), using a mercury sphygmomanometer. Measurements were taken on the right arm with an appropriate-sized cuff on participants seated at rest for 2 min. The first and fifth Korotkoff’s sounds were recorded as the systolic and diastolic pressures, respectively. Hypertension is defined as (a) systolic BP 3140 mm Hg, or (b) diastolic BP a90 mm Hg, or (c) current use of antihypertensive medication. A secondary measure of BP that used the mean of two readings from an automated BP recorder was also analyzed. Since findings for the manual and automated measurements were similar, results from the manual readings only are reported. RESULTS
Study Population The total age-eligible population residing in the two control cities was 93,200. Approximately 6% of this target population was included in the sampling frame which, after a mean response rate of 64% yielded a study population of approx-
4
WINKLEBY,
FORTMANN,
AND
BARRETT
imately 800 participants per survey and a total of 3,349 participants from all surveys combined (Table 1). Demographic characteristics, including mean years of education, were generally stable across the g-year survey period, allowing for a pooling of data across surveys in several of the analyses described below. Relationships between Education Level and Risk Factors The age-adjusted associations between education level and the six risk factors are presented graphically in Fig. 1. This figure uses data pooled across the four study time periods and presents separate slopes for males and females. In general, there is a clear relationship between years of education and each risk factor, with higher education levels associated with lower risk. As shown, increased educational attainment is related to greater knowledge about health, taller stature, lower smoking and hypertension prevalence, and lower obesity and cholesterol levels. The slopes are remarkably similar for males and females for all risk factors with the exception of body mass index and cholesterol level, which show stronger relationships for females. The magnitude of the differences in risk by education level is generally large. For example, prevalence of smoking for males and females combined was 39% among those with < 12 years of education compared to 13% among those with > 16 years of education. This translates into a threefold difference in risk. For the other five risk factors, the differences between the extreme education levels persist, although most are not as strong as those for smoking. In order to test whether the observed relationships were statistically significant and to explore the consistency of the associations across sex and age groups, separate regression analyses between years of education and each risk factor were conducted using multiple regression models for continuous variables and multiple logistic models for dichotomous variables (Table 2). Each regression coefftcient is adjusted for age, sex, and time of survey, as appropriate. For the total study population, all relationships between years of education and risk factors are statistically significant (P values ranging from
MEDICINE
19, 1-12 (1990)
Social Class Disparities in Risk Factors for Disease: Eight-Year Prevalence Patterns by Level of Education’ MARILYN Stanford
A. WINKLEBY,
PH.D.,’ STEPHEN DONALD C. BARRETT,
P. FORTMANN, M.S.
M.D., AND
Center for Research in Disease Prevention, Stanford University School of Medicine, loo0 Welch Road, Palo Alto. Culifornia 94304-1885
This article examines the associations between education, a primary indicator of social class, and six risk factors for disease. Data are presented on a sample of 3,349 individuals ages 25-74 years who participated in one of four cross-sectional surveys conducted by the Stanford Five-City Project between 1979and 1986.The six risk factors examined are knowledge about health, cigarette smoking, hypertension, serum cholesterol, body mass index, and height. A highly significant pattern of associations was found between education level and the six risk factors, in the direction of higher risk among those with lower education (all P values 16 years (postgraduate). Income level is defined as total gross annual family income and is divided into 11 categories, ranging from $5,000 to $75,000 per year. All analyses are adjusted for time of survey as a surrogate adjustment for inflation. Occupation is coded according to the 13 standard census occupational codes, which are collapsed into the following 10 categories and used as indicator variables: executives, managers, administrators, clerical workers, skilled and semiskilled workers, unskilled workers, unemployed, homemakers, students, and retired. Knowledge about health and cigarette smoking status were determined by selfreported questionnaire data. Health knowledge encompassed questions about cardiovascular disease risk factors and is presented as a summary index of 17 items where the highest score reflects the highest level of knowledge. Cigarette smoking, coded as “yes” if the respondent had smoked one or more cigarettes in the last 48 hr and had ever smoked on a regular basis, was confirmed by expired air carbon monoxide and serum thiocyanate (29). Height, weight, and blood pressure were collected according to standardized study protocols. Height was measured without shoes to the nearest 0.6 cm by metal rule and weight was measured to the nearest 0.1 kg without shoes on a beam balance scale. BMI, used as a measure of obesity because of its high correlation with other estimates of body fat (30), is defined as wt(kg)/ht(m*). Serum TC, derived from nonfasting venous samples and analyzed fresh for lipids and lipoproteins by methods established by the Lipid Clinics Research Program (31), is measured in milligrams per deciliter. Blood pressure (BP) was determined by the mean of two manual measurements obtained by standard protocol (32), using a mercury sphygmomanometer. Measurements were taken on the right arm with an appropriate-sized cuff on participants seated at rest for 2 min. The first and fifth Korotkoff’s sounds were recorded as the systolic and diastolic pressures, respectively. Hypertension is defined as (a) systolic BP 3140 mm Hg, or (b) diastolic BP a90 mm Hg, or (c) current use of antihypertensive medication. A secondary measure of BP that used the mean of two readings from an automated BP recorder was also analyzed. Since findings for the manual and automated measurements were similar, results from the manual readings only are reported. RESULTS
Study Population The total age-eligible population residing in the two control cities was 93,200. Approximately 6% of this target population was included in the sampling frame which, after a mean response rate of 64% yielded a study population of approx-
4
WINKLEBY,
FORTMANN,
AND
BARRETT
imately 800 participants per survey and a total of 3,349 participants from all surveys combined (Table 1). Demographic characteristics, including mean years of education, were generally stable across the g-year survey period, allowing for a pooling of data across surveys in several of the analyses described below. Relationships between Education Level and Risk Factors The age-adjusted associations between education level and the six risk factors are presented graphically in Fig. 1. This figure uses data pooled across the four study time periods and presents separate slopes for males and females. In general, there is a clear relationship between years of education and each risk factor, with higher education levels associated with lower risk. As shown, increased educational attainment is related to greater knowledge about health, taller stature, lower smoking and hypertension prevalence, and lower obesity and cholesterol levels. The slopes are remarkably similar for males and females for all risk factors with the exception of body mass index and cholesterol level, which show stronger relationships for females. The magnitude of the differences in risk by education level is generally large. For example, prevalence of smoking for males and females combined was 39% among those with < 12 years of education compared to 13% among those with > 16 years of education. This translates into a threefold difference in risk. For the other five risk factors, the differences between the extreme education levels persist, although most are not as strong as those for smoking. In order to test whether the observed relationships were statistically significant and to explore the consistency of the associations across sex and age groups, separate regression analyses between years of education and each risk factor were conducted using multiple regression models for continuous variables and multiple logistic models for dichotomous variables (Table 2). Each regression coefftcient is adjusted for age, sex, and time of survey, as appropriate. For the total study population, all relationships between years of education and risk factors are statistically significant (P values ranging from
