Hypertension Among Female Workers in Sao Paulo, Brazil. Predictors and Joint Effects MYRIAM B. DEBERT RIBEIRO, M.D.,

Transcription

1 Hypertension Among Female Workers in Sao Paulo, Brazil Predictors and Joint Effects MYRIAM B. DEBERT RIBEIRO, M.D., DR.P.H. SUMMARY Previous analysis based on data derived from a prevalence study of hypertension among workers in different economic activities in the Greater Metropolitan Area of Sao Paulo, Brazil, have shown separate effects of biologic and social attributes on diastolic blood pressure levels (). The present paper explores joint effects of sociodemographic, anthropometric, and behavioral characteristics on blood pressure levels. For that purpose, the Morgan-Sonquist procedure on interaction effects in the prediction of levels was used. Six independent variables emerged as predictors: age, Quetelet index, occupation, schooling years, subsector of the economy, and smoking. Very young and lean women who were also smokers had the lowest mean levels (66.1 mm Hg). Young, overweight women without college education showed intermediate levels ( mm Hg). Highest levels were found among older women in production-linked occupations engaged in transportation and metallurgy activities (89.5 mm Hg). The study suggests that the potential role of work-related characteristics on blood pressure levels is greatest among older women. (Hypertension 5 (supp V): V-144-V-148, 1983) KEY WORDS workplace sociodemographic factors women age Quetelet index occupation education smoking obesity MULTIFACTORIAL etiology of hypertension has been widely acknowledged. This assumption, however, has raised methodological issues regarding analytical procedures and interpretation of results. Refinement and elaboration of associations between factors not only need to take into account separate or additive effects but also must try to capture potentially fruitful interconnections. A prevalence study of hypertension undertaken among workers in 10 subsectors of the economy in Sao Paulo, Brazil, estimated the prevalence of the condition in a highly urbanized and industrialized metropolitan area of the Third World. The study also provided descriptive data on the epidemiology of hypertension in that population, having shown separate effects of certain attributes on blood pressure levels. In the present paper, the joint effects of biological and social dimensions on blood pressure levels are explored by means of identifying internally homo- genous subgroups of subjects that differ in respect to blood pressure levels. These subgroups represent combinations of subclasses of biological and social characteristics. Methodology Study Population The study population was composed of subjects in 10 subsectors of the economy. In the industrial sector, the following subsectors were selected: textile, metallurgy, housing and automobile construction. In the trade and service sector, the following were selected: trade, insurance, loans and finance, transportation. In the liberal art professions advertising, journalism, and teaching were included. A two-stage sampling procedure was used. The sample size totaled 5500 workers 4013 men and 1457 women aged years from 57 firms. Detailed study design and methodology are described elsewhere. 1 This report presents results for the female group in the sample. From the Department of Preventive Medicine, Escola Paulista de Medicina, Sao Paulo, Brazil. Dr. Mynam B. Debert Ribeiro is an Established Investigator of the National Council for Technological and Scientific Development (NCPq). Financial support was provided by Fundacao de Amparo a Pesiquisa do Estado de Sao Paulo (FAPESP). Address for reprints: Myriam B. Debert Ribeiro, M.D., Department of Preventive Medicine, Escola Paulista de Medicina, Rua Botucatu no. 740, Sao Paulo, Brazil. V-144 Variables Explanatory variables introduced in the analysis included sets of demographic (age, ethnicity, parity, nativity, family), anthropometric (Quetelet index), social class indicators (income, education) as well as work characteristics (subsector, occupation, time in the job, weekly hours of work) and behavioral aspects (smoking and oral contraceptive use).

2 HYPERTENSION IN BRAZILIAN FEMALE WORKERS/Ribeiro V-145 The Quetelet Index (QI) was calculated as weight/ height 2 (kg/m). Three categories were defined using the overall mean for the group of women and its variability. In the first group, (lean) women were included whose QI was below the mean + 1 SD. The second included subjects with QI values lying between the mean + 1 and 2 SD. The third group was composed of women with QI values greater than the mean + 2 SD (overweight). With regard to occupation, a 13-point scale of occupation was constructed taking into account the heterogeneity of the very concept of production in the various subsectors. The scale involved two dimensions. On the one hand, production-linked occupatibns were separated from those in administration and from a third group of workers in charge of command functions linked either to production or administration. Each of the first two groups was further subdivided according to the amount and/or nature of skills required. Thus, regardless of the formal engagement of workers in the industrial or service sectors, their occupation was considered from the point of view of production. Education (schooling) was conceived as an ordinal variable representing number of school years completed. In the first subclass, subjects were included who had up to 4 school years; women in the second subclass had completed elementary school and pursued up to the 8th grade; the third up to the 1 lth grade; subjects in the last subclass had entered college. Procedure The Sonquist-Morgan procedure 2 has been applied. The basic technique is the sequential identification of subgroups, one at a time, nonsymmetrically. The method makes it possible to order a set of independent TABLE 1. Diastolic Blood Pressure () Levels According to Categories of Explanatory Variables Name (scale) Age* Quetelet index (kg/m 2 )* Ethnicityt Parityt Migration statust Familyt Schooling* Income* Cigarette smokingt Oral contraceptive uset Categories =S =s Whites Black and Mulatos Orientals Nuliparae Multiparae Natives Migrants Single no children Married no children Married with children Single with children Elementary School High school (up 8th grade) High school (up 11th grade) College education < 3 (MW) 3-5 (MW) 5-10 (MW) 3= 10 (MW) Yes No Yes No No MW =» minimum wage; 1 MW was equivalent to $70.00 in American money at the time of the study. Ordinal scale. tnominal scale. (mm Hg)

3 V-146 PROCEEDINGS/INTERAMERICAN SOCIETY SUPP V, HYPERTENSION, VOL 5, No 6, NOV/DEC 1983 study variables (predictors) in relation to a quantitative dependent variable. In the first step, it will take into account all feasible divisions of the sample or original group on the basis of each explanatory variable to be included. It will result in the division of the categories of any variable so that the dichotomization of the group into two subgroups provides the largest reduction in the unexplained sum of squares. The method continues for each one of the already formed subgroups. It halts according to a preassigned stopping rule involving the group size and/or the reduction in the unexplained sum of squares. Both nominal and ordinal scale explanatory variables can be handled in the analysis. TABLE 2. Diastolic Blood Pressure Levels According to Categories of Explanatory Variables Name (scale) Categories No. (mm Hg) Subsector of economyt Occupationt Hours of work/weekt Time on job (yrs)* Ordinal scale. tnominal scale. Textile Metallurgy Auto construction Housing Trade Insurance loans & finance Transportation Advertising & journalism Teaching Professionals Production Nonskilled, manual Semiskilled, manual Skilled, manual Technicals Semiskilled, nonmanual Technicians Skilled, nonmanual Administration Nonskilled Semiskilled Technical Skilled Supervision Production Administration < Median > Median < * In the case of ordinal variables, a natural order is preserved. In contrast, when dealing with nominal explanatory variables, categories are ordered according to the means taken by the dependent variable and not their own sequential values. In our analysis, the division process or dichotomization stopped when the group to be split had less than 100 subjects and the reduction in the sum of squares was less than 0.6%. Student's t test was used for assessing differences between means of parental groups. Results Results are reported for the sample of 1411 women for whom information on all variables was available. Categories of explanatory variables, frequency distribution, and associated levels are shown in tables 1 and 2. The sequential division process is illustrated from the first step building up to the complete model. Overall mean for women was 72.3 mm Hg. Age was the first selected predictor, leaving women over 35 years of age with the higher mean (80.8 mm Hg.) (see fig. 1). In the following step (fig. 2), two predictors were entered. For younger women (Group 2), the Quetelet index provided the best split, indicating higher levels for overweight women (77.8 mm Hg, Group 5). Among women over 35 years of age (Group 3), higher diastolic was found among those engaged in production linked occupations (82.9 mm Hg, Group 7). In the next step (fig. 3) for the group of lean young women, age was reintroduced as the best predictor, leaving those in the age group on the higher blood pressure side ( mm Hg, Group 9). For the group of overweight young women, those without college education showed higher levels (80.9 mm Hg, Group 10). For women over 35 years of age and in production-linked occupations, subsectors of the economy provided the best division; transportation and metallurgy workers showed higher levels (89.5 mm Hg, Group 15). In the last division, for the group of very young and lean women, nonsmokers had higher mean (70.3 mm Hg, Group 17). Table 3 describes the characteristics of the eight terminal groups generated in the analysis in regard to combinations of the selected predictors. Differences <35 ^ ) 1 AGE 72.3 (1411) ^ FIGURE 1. Splitting of the sample of women according to predictors of diastolic blood pressure. 3

4 HYPERTENSION IN BRAZILIAN FEMALE WORKERS/Ribeiro V-147 <35 LEAN 69.8 (1066) V (1173) OVERWEIGHT 77.8 (107) ADMINISTRATION OCCUPATION 80.8 PRODUCTION (63) (175) FIGURE 2. Splitting of the sample of women according to predictors of diastolic blood pressure. TABLE 3. Diaslolic Blood Pressure () Levels and Profile of the Terminal Groups Group Description (mm Hg) years old; lean; smokers years old; lean; non smokers years old; lean < 35 years old; overweight; college education < 35 years old; overweight; without college education 3= 35 years old; administration linked occupation ^ 35 years old; production-occupations; textile, teaching, etc. S 35 years old; production-occupations; transportation & metallurgy Analysis by Student's ( test; (comparisons between parental groups). *p < tp < t * between means of terminal parental groups were tested by the t test. The group of lean, young, nonsmokers showed significantly higher levels when compared to smokers. Overweight women under 35 years of age without college educations had higher levels. Older women engaged in metallurgy and transportation activities showed significantly higher levels, when compared to those in others activities. Discussion Of the 14 explanatory variables introduced in the model, six were found to be predictors of leveis in women. Predictors were selected from each of the five sets of independent variables, namely, age (demographic), Quetelet index (anthroporpetric), education (social class indicator), occupation and subsector (work characteristics), and smoking (behavioral). Most predictors defined by the Morgan-Sonquist procedure in the present study have been extensively (1411) < ) AGE ^otjetele OCCUPATION < / 8 SMOKINC LEAN 4 /-\ 6a7 637) 69.6 (1066) OVERWEIGHT (107) -25^ \ «5 NO COLLEGE COLLEGt ) 10 80S (72) II 35) ADMINISTRATION (63) PROOUCTION 7 82S (175) SUBSECTOR TEXT. / TEACH, JOURN ETC J44) TRANSP.METAL YES (294) \NO (343) FIGURE 3. Splitting of the sample of women according to predictors of diastolic blood pressure.

5 V-148 PROCEEDINGS/INTERAMERICAN SOCIETY SUPP V, HYPERTENSION, VOL. 5, No 6, NOV/DEC 1983 reported as risk factors for hypertension. Thus, age was the attribute that accounted for most of the explained variance in pressure, generating internally homogenous subgroups which, nevertheless, differed in respect to diastolic pressure. One finding that has to be pointed out refers to a situation where the same predictor is used again to divide a given group, as in the case of age in the present analysis. The possibilities of interactions among variables in different stages can be handled by reintroducing the second analysis, attributes whose simple effects have already been removed but which may also mediate the effects of factors at one of the later stages. In the present context, it seems reasonable to assume that the reintroduction of age is related to the effects of smoking on blood pressure levels. Smokers have been shown to have slightly lower blood pressures than nonsmokers, the association having been interpreted on the basis of weight differences. Although differences in due to smoking are trivial (3.4 mm Hg) and are restricted to the very young group of women, it cannot be accounted for by weight since it was only detected among lean women. In the present study, the relationship of education to blood pressure levels is clearly shown for the younger age group. The finding of an inverse relationship between blood pressure and education level is consistent with those reported in a number of studies involving both blacks and whites. 3 " 7 It is also important to point out that the cutoff point for lower levels among women is. given by college education. The finding that college-educated overweight women had lower levels may be explained by greater awareness of hypertension as a health hazard among the more educated overweight women. It should also be stressed that, in the older age group, predictors tended to concentrate on work-related characteristics such as occupation and subsector. Women who work in that age group often assume social roles such as head of household, housewife, and wage earner. Moreover, such women more frequently fill lower positions in the occupational, educational, and income scales, and may also tend to engage in more stressful activities, such as those represented by transportation and metallurgy in the present study. The study strengthens the view of interconnections among biological factors, social dimensions, in general, and work characteristics, in particular, regarding to blood pressures levels. It also suggests that the potential influence of work-related characteristics may be of greater magnitude in the older age group. References 1. Debert Ribeiro MB, Ribeiro AB, Stabile Neto C, Chaves CC, Kater CE, Iunes M, Saragoc,a MAS, Zanella MT, Ancjo MS, Marson O. Kohlmann O Jr, Franco RJS, Nunes SF, Ramos OL: Hypertension and economic activities in Sao Paulo, Brazil. Hypertension 3 (suppl II): , Sonquist JA, Baker EL, Morgan JN: Searching for structure: an approach analyses of substantial bodies of microdata and documentation for a complete program. Ann Arbor, Michigan: University of Michigan, Institute for Social Research, Kannel WB, Sorlie P: Hypertension in Framingham. In Epidemiology and Control of Hypertension, Edited by Oglesby P. New York: Stratton, Antonovsky A: Social class and the major cardiovascular diseases. J Chron Dis 21: 65, Syme SL, Oakes TW, Friedman GW: Social class and racial differences in blood pressure. AmJ Public Health 64: 619, Hypertension Detection and Followup Program Cooperative Group. "Race, education and prevalence of hypertension." Am J Epidemiol 106: 351, Roberts J, ed: Blood Pressure Levels of Persons 6-74 Years, United States, Publication No. DHEW (HRA) (series 11, number 203). Washington, DC: US Department of Health, Education, and Welfare, 1978