Sm. Sri. Med. Vol. 32, No. 5, pp. 579-590, Printed in Great Britain. All rights reserved
1991 copyright
0
0277-9536191 $3.00 + 0.00 1991 Pergamon Press plc
GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR: SOME UNANSWERED QUESTIONS* MARY-ANNEKANDRACK Department of Sociology & Anthropology,
Carleton University, Ottawa, Ontario, Canada KlS 5B6
KAREN R. GRANT and ALEXANDERSEGALL Department of Sociology, University of Manitoba, Winnipeg, Manitoba, Canada, R3T 2N2
Abstract-To date, no single explanation has accounted for discrepancies between male and female morbidity rates and health care ultilization patterns. The sociomedical approach to sex/gender differences in health related behaviour has generated a variety of hypotheses. However, despite extensive study, many unanswered questions remain. The findings of this study fall short of offering conclusive evidence as to the causes of variations in morbidity and health services use between women and men. However, an effort is made to identify the salience of social role and related social status characteristics (e.g. labour force participation) in accounting for variation in health, illness and sick role behaviour. This paper utilizes data from the 1983 Winnipeg Area Study. Findings of this study raise questions about the adequacy of current concepts and measures for studying sex/gender differences in health related behaviour. The study concludes with a critical discussion of conceptual, methodological and theoretical issues which must be considered in our efforts to advance our understanding of why women experience greater longevity, but experience greater morbidity and make more extensive use of health services. Key words-gender,
health, behaviour
INTRODUCT’ION In the past decade, the issue of sex/gender differences in health and illness has gained popularity as a research topic among medical sociologists. It has been observed, for example, that women live longer than men and have lower mortality rates for most causes of death [l-8]. Nathanson [9, p.57 states that “the sex differential in mortality statistics has usually been explained by women’s constitutionally greater resistance to both infections and degenerative disease.” However, epidemiological evidence contradicts this explanation. Studies of sex/gender differences in health and illness indicate an excess of female morbidity as compared to males [lO-151. The problem then becomes one of accounting for the discrepant pattern of morbidity data [16]. Clearly, constitutional resistance and/or susceptibility cannot adequately explain all health and illness differences between men and women. The study of sex/gender differences has been approached from a number of perspectives. For example, a biomedical approach essentially tests illness hypotheses, whereas a sociomedical approach tests illness behaviour hypotheses. According to the biomedical approach, it has been hypothesized that observed sex differences in morbidity are the product of biologically based inherited risks. If being female means greater risk of disease, and disease may cause death, then this should be reflected in sex differences in mortality rates. Given the inability of this perspective to account for differences in both morbidity and mortality *Revised version of a paper presented at the Western Association of Sociology and Anthropology Meetings in February, 1985, Winnipeg, Canada.
rates, an alternative approach has been formulated. The sociomedical perspective hypothesizes that these differences can be better explained in terms of social role obligations, acquired risks, health reporting behaviour, and/or the illness and preventive health orientations of women and men. In short, it is argued that differences in both morbidity and mortality are related to sociocultural and social-psychological factors (i.e., gender). The findings of research informed by sociomedical hypotheses are, unfortunately, often confusing and inconsistent. To complicate this area of study even further, it is readily acknowledged that health related behaviour may also be- influenced by the interaction of biological and sociocultural factors. Yet, our ability to separate out the effects of nature vs nurture are extremely limited. How can one show where biology stops and environment starts? Many hypotheses that have been formulated both within the biomedical and sociomedical approaches have considerable descriptive value. This study, informed by a sociomedical perspective and using a Canadian sample, provides additional data on the subject of sex/gender differences in health related behaviour. In particular, the salience of role characteristics is explored in relation to health, illness and sick role behaviour. While circumspect in terms of the scope of the hypotheses tested, this study attempts to move the level of discourse from description toward an explanation of sex/gender differences. This is in recognition of the need for a synthetic theoretical model to integrate the diverse empirical findings which already exist, and to serve as a guide to future research on male/female differences in health and illness behaviour. 579
580
MARY-ANNEKANDRACK~~ AND HEALTH CARE UTILIZATION: DESCRIBING SEX/GENDER DIFFERENCES
MORBIDITY
Before reviewing some of the literature on sex/gender differences in health and illness, it is worthwhile to consider the conundrum which characterizes the study of sex/gender. Writers in gender relations typically differentiate between the concept of sex and the concept of gender, the former referring to a physiological attribute and the latter representing the sociocultural expression of sex [ 171.Although this is an important distinction, its utility is often lost at the empirical level, owing to our inability to capture these differences. Furthermore, as Eichler asserts, “no human activity exists that is not mediated through culture.. . , i.e., there are always aspects of gender in sex, if we use this language. In principle, therefore, the distinction cannot be clear” [18, p. 3601. In light of this point, it is worth remembering that while some writers speak of sex differences and others speak of gender differences, the line between the two is often quite fine. This is illustrated by the body of research evidence showing the greater longevity of women. In most industrialized nations, women outlive men by approximately 7 years. Though there has been a recent stabilization of the life expectancy differential, women’s apparent constitutional resistance continues. In 1976, Waldron posed the question “Why Do Women Live Longer Than Men?“. Waldron and others suggest that sex differences in life expectancy can be attributed to the biophysical differences between males and females; i.e. it is generally thought that males are less durable biologically than females [l-8, 19-221. Sex differences in life expectancy have also been related to occupation [23-281. In this case, the discussion becomes one of gender differences because the intervening variable (occupation) is a social role, which may systematically vary by sex, but is still not a biophysical fact. Similarly sex differences in morbidity may be examined more closely by exploring occupational and familial roles that have been linked to the health of men and women [6, 16,29, 301. Again, as soon as these behavioural dimensions are introduced, the discussion shifts to one of gender, and not simply sex differences in morbidity. In keeping with the task at hand, interpretations and explanations of sex/gender differences in health and illness behaviour must be sensitive to the problem of mistaking these two dimensions for each other. What does the existing empirical evidence reveal about the nature of sex/gender differences in health related behaviour? Health survey data indicate an excess of morbidity among females as compared to males [l, 2,4, 5, 7-9, 15, 16, 3 1, 321. This excess appears for both acute and chronic conditions with the qualification that the latter are more “consequential” and/or “incapacitating” for males [S, 7-91. In a summary of American and British evidence regarding sex differences in morbidity, Nathanson [9, p. 571 reports that “women have higher rates than men for almost all indices of morbidity and utilization of health services.” Clearly et al. [33, p. 1061 cite a variety of sources which suggest that “women consistently have been found to use outpatient medical services more frequently than
al.
men.” According to Verbrugge and Wingard, females utilize more medical and health services for curative as well as preventative, purposes. Thus, studies of utilization typically show an excess in female rates, which in turn “boost their rates of morbidity compared to males” (32, p. 291. While morbidity and mortality rates and health services utilization patterns are apparently interrelated, it is extremely difficult to establish the causal connection between these factors. There tends to be some agreement between the morbidity and mortality data for men and women, however, the linkages between these health status indicators are not always direct. When one considers health care utilization, the picture becomes even more complicated. The basic assumptions of the biomedical perspective suggest that higher morbidity is correlated with higher utilization, and that sicker individuals will more likely die than those who are healthy. Furthermore, it is argued that if individuals avail themselves of medical services, then improvements to health will follow. In other words, it is assumed that women may be healthier (i.e. have a greater life expectancy) in part because they are more active consumers of health care services. This line of reasoning obviously has some validity, although several critics of modern medicine point to the fact that the use of medical services is neither a necessary, nor a sufficient, condition for the production of health [34-371. While the debate regarding the efficacy of medical measures is beyond the scope of this paper, it would be naive to argue that medical services alone are responsible for improvements in health. Lalonde [38], among others, has pointed out that medical services are but one element in the health field. Human biology, the environment and individual lifestyle are also important to understanding the distribution of health, disease and death in any human population. Sociomedical studies of sickness have emphasized the important role of demographic characteristics (e.g. age, education, occupation) and socio-structural conditions (e.g. access to medical services and alternative types of health care) as possible mediating factors in the relationships between morbidity, mortality and utilization. To date, however, no single explanation had been able to account for the discrepancies between male and female rates of utilization (and the bearing of utilization on morbidity and mortality). THE SEARCH FOR AN EXPLANATION: RESEARCH HYPOTHESES ON SEX AND GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR
The selective evidence summarized in the preceding section demonstrates that male/female differences do indeed exist in the distribution of disease and death, and in the utilization of health services. Marcus and Seeman [16] suggest that it must now be asked why do women experience higher rates of milder conditions? In addition, it is important to seek an explanation for women’s greater utilization of health services and experienced disability days, in view of the finding that their excess morbidity is primarily of a mild form. One possible answer to these questions may be found in the nature of male
Gender differences in health related behaviour vs female biological risks of disease and death. Other answers focus on acquired risks, role occupancy and the psychosocial aspects of illness. Once the social psychological and sociocultural dimensions of health and illness are introduced, the central question becomes-why are there gender differences in health related behaviour? An exclusively biomedical approach to this question will not suffice. Although biological differences may explain certain symptoms or conditions, such differences cannot explain the overall pattern. For example, Waldron has noted that genetic and hormonal factors may account for some differences between males and females in their susceptibilities to congenital anomalies, infectious disease, atherosclerotic disease [l, 2,5]. It has been suggested, as well, that sex-specific conditions such as pregnancy account to some extent for females’ greater use of health services. However, it has been reported that “even when conditions associated with reproduction are excluded, women have higher rates than men for all indices of illness experience employed” [9, p. 581. Inherited health risks simply do not fully explain morbidity patterns, or the discrepancy between morbidity and mortality rates. It is clear, therefore, that a systematic explanatory model must incorporate the psychosocial and sociocultural aspects of the experience of illness. To date, the sociomedical approach to gender differences in health and illness has generated a wide variety of hypotheses, which focus on the following: illness behaviours [ 10, 39-421; reporting behaviours [39,43]; acquired risks related to social roles, lifestyles, and stress [16,21,25,20,41,42]; methodological artifacts [44]; and a sex bias in medicine [7,45]. Verbrugge [7], Nathanson [9] and Wingard [3] have each summarized the major hypotheses in their review articles. Since the present study extends some of the work previously done regarding social roles and illness behaviours, our review of the literature is restricted to these areas. Despite extensive study, there is no conclusive support or refutation to be found in the research investigating any of these hypotheses. Indeed, in some instances, conflicting findings have been reported [e.g. 11, 16,4042]. Marcus and Seeman [ 16, p. 18I] found that “fixed role obligations are particularly relevant for the study of sex differences in reduced activity-but may be less relevant for explaining male/female differences in symptom recognition.” Similarly, Verbrugge [7] has aptly noted that the salience of any given hypothesis tends to vary with the outcome measure under consideration. The present study examines self-reported health related behaviours in relation to social roles. The central research question is: how do role attributes (particularly, marital status and social status characteristics) influence individuals’ beliefs and actions related to health? The purpose of this analysis is to extend the empirical basis of the exploration of gender and health. TESTING GENDER DIFFERENCES IN HEALTH AND ILLNESS
The Winnipeg Area Study (WAS) is an annual cross-sectional, city-wide survey of a sample of WM3215-E
581
Winnipeg residents [46]. The 1983 WAS data were not collected for the explicit purpose of studying sex/gender differences in health related behaviour. However, this community survey included questions on gender and health and consequently the data can be used to illustrate some of the conceptual and methodological problems associated with testing hypotheses pertaining to illness behaviour and the acquired risks of social roles. Sample design and data collection
The 1983 WAS population was designated as all dwelling units that were listed in the 1982 assessment file for the City of Winnipeg, Canada. A simple random sample of 701 addresses was selected for personal interviewing from a computerized list of addresses compiled by the City Planning Department. Nursing homes and temporary residences were deleted from the sample and the household was the primary sampling unit. Within the household one eligible person was selected as the respondent for a 1 hr interview. A person was eligible to be the respondent if the dwelling was her/his usual place of residence and s/he was 18 years of age or older. Interviewers were instructed to try to obtain an equal number of male and female respondents within their given allotment of addresses. As adult males are generally more difficult to contact, set procedures were followed by the interviewers to assist them in obtaining their quota. If eligible respondents refused, the household was not replaced. This method for selecting interviewees in households has been used successfully in the WAS over the past decade, without producing evidence of biased samples. A total of 524 respondents were interviewed for a response rate of 75%. The sample was compared to the 1981 Census of Canada for Winnipeg and the results reveal that the sample is representative of the city on a number of key characteristics such as, household size, age and sex distributions. For example, the sample is 56% female compared to 53% of the city population (over 20yr of age). It may be of interest to note that on average the person answering the door in Winnipeg Area Studies was female in 57% of the households. There was no significant difference between the 1981 Census distribution of males and females (53%) in Winnipeg, the sex of the first contact within households (57% female), and the sex of the respondents who participated in this survey (56% female). Operationalization
Studies of sex/gender differences have varied considerably in the specific aspects of health/illness behaviour investigated. This study focused upon the relationship between gender and health status, health behaviour, illness behaviour and sick role behaviour. The major independent variable is, of course, gender. In measuring gender, the first step was to determine the respondents’ sex and then to explore the various roles (e.g. marriage, employment) occupied by each person. The intent was to try to learn about the role commitments of men and women which may be linked to health. The underlying assumption was that a better understanding of the link between male/female social roles and acquired
582
MARY-ANNE KANDRACK et
health risks may provide further insight into gender differences in health related behaviour. A number of social factors were measured including: age (actual age in years); marital status (married/common-law, single/never married, divorced/separated/widowed); level of education completed (elementary, junior high, high school, vocational, university); income (total individual income); and employment status (employed, unemployed, retired, in school, keeping house). Turning to the dependent variables, two indicators of self-reported health status were assessed. Respondents were asked: how would you describe your present health; and how healthy have you felt in the past twelve months. The response categories provided for these questions were: excellent, good, fair, and poor. The health locus of control belief scale was used in this survey to measure beliefs about health behaviour. Each respondent was presented with a list of twenty statements designed to gauge their views about an individual’s ability to control health matters. Responses ranged from strongly agree (5) to strongly disagree (1). An internal preventative subscale that reflects beliefs about one’s ability to avoid illness was derived from a factor analysis of responses to the locus of control items included in the interview schedule [47]. To assess one dimension of actual health behaviour, the respondents were asked about their use of informal social networks (i.e. spouse, friends, parents, children) as health resources. For example, each respondent was asked who s/he speaks to about health problems. A summary measure was examined for gender differences in the number of informal social network members consulted for discussions of health problems. A number of measures of illness behaviour were used to assess the respondents’ illness experience and their symptomatic responses. Measures focused upon the disability days experienced by each respondent during the past year including the number of days that an illness required a cut down in one’s usual activities, kept the individual from work, and/or kept him/her in bed. These were included as indicators of illness behaviour, as many studies have concentrated on sex/gender differences in the level of restricted activity resulting from illness. In addition, illness behaviour was gauged by the type of response (i.e. ignore, self-help, see doctor) made to a variety of conditions (e.g. dizziness, loss of appetite). Respondents were also asked to report the number of different medications presently being taken. Finally, sick role behaviour was assessed in terms of willingness to adopt the sick role and actual utilization of medical services. The interview schedule contained a number of Likert type attitudinal items which focused upon: (1) self-reliance, or one’s willingness to relinquish well-roles and become dependent upon lay others; and (2) medical skepticism, or one’s faith in physicians and willingness to become dependent upon doctors [48]. Included among the measures of actual medical care contact were having a regular doctor; number of different doctors visited in the past twelve months; recency of last general medical checkup and the total number of visits made to all of one’s doctors in the past twelve months.
al.
Some of the measures used here are less than ideal for the purposes of the present analyses. For example, measures of self-reported health may confound aspects of reporting with actual health status. Our measures cannot unambiguously capture the nuances of socially constructed meanings attached to health and illness for women and men. As well, measures of health services use are not sufficiently sensitive to differences in help-seeking related to preventive vs reactive care. Plus, we cannot differentiate between help-seeking related to reproduction and other reasons for frequenting medical practitioners. The latter point is particularly crucial in the study of gender differences in utilization, as has been noted previously. Thus caution is exercised in the interpretations of data offered here. At a later point, we will return to a consideration of these methodological issues, as they bear upon this field of research. Data analysis
The data analysis was done in stages. In the first step, a test of the statistical significance of relationships between sex and the selected set of indicators of health status, health, illness and sick role behaviour was conducted. These bivariate relationships were assessed with measures of association and the chi square test. The second phase of data analysis involved a test of the explanatory value of sex relative to other sociodemographic factors, in accounting for differences in those dependent variables found to be significantly related to sex. An hierarchical regression procedure was used for this purpose. Sex and a set of sociodemographic factors (age, marital status, education, income and employment status) were entered successively into an hierarchical regression. In other words, the sociodemographics were not entered until the explanatory power of sex had been exhausted. The residual, that is, the proportion of variance in the selected indicators left unexplained, was then entered into the analysis as the dependent variable. This procedure allowed an assessment of the explanatory value of sex alone and a comparison with the proportion of variance explained by other sociodemographic factors [49]. In the third and final phase, an analysis of variance was performed in order to explore sex/gender differences in the dependent measures. This involved a closer assessment of two variables that are related to sex and reflect associated behavioural patterns (i.e. gender role). These variables are employment and marital status. The data were examined for differences in health, illness and sick role behaviour, by marital status and employment status for women as compared to men, and for women only. In short, inter- and intra-group differences were explored. GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR: THE EMPIRICAL EVIDENCE
The first step of the data analysis was to assess whether there were significant differences between the two groups (males and females) other than sex. This involved a comparison of age, marital status, education, employment status and income distributions for men and women. As indicated in Table 1, there
Gender differences in health related behaviour Table I. Distribution
of male/female respondents graphics
by selected demo-
Male Variable
n
Female %
n
Age 18-24 25-34 35-44 45-54 5544 65 yr Total
yr yr yr yr yr or older
33 53 54 21 37 28 232 x2 = 7.785, df = 5, P
1991 copyright
0
0277-9536191 $3.00 + 0.00 1991 Pergamon Press plc
GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR: SOME UNANSWERED QUESTIONS* MARY-ANNEKANDRACK Department of Sociology & Anthropology,
Carleton University, Ottawa, Ontario, Canada KlS 5B6
KAREN R. GRANT and ALEXANDERSEGALL Department of Sociology, University of Manitoba, Winnipeg, Manitoba, Canada, R3T 2N2
Abstract-To date, no single explanation has accounted for discrepancies between male and female morbidity rates and health care ultilization patterns. The sociomedical approach to sex/gender differences in health related behaviour has generated a variety of hypotheses. However, despite extensive study, many unanswered questions remain. The findings of this study fall short of offering conclusive evidence as to the causes of variations in morbidity and health services use between women and men. However, an effort is made to identify the salience of social role and related social status characteristics (e.g. labour force participation) in accounting for variation in health, illness and sick role behaviour. This paper utilizes data from the 1983 Winnipeg Area Study. Findings of this study raise questions about the adequacy of current concepts and measures for studying sex/gender differences in health related behaviour. The study concludes with a critical discussion of conceptual, methodological and theoretical issues which must be considered in our efforts to advance our understanding of why women experience greater longevity, but experience greater morbidity and make more extensive use of health services. Key words-gender,
health, behaviour
INTRODUCT’ION In the past decade, the issue of sex/gender differences in health and illness has gained popularity as a research topic among medical sociologists. It has been observed, for example, that women live longer than men and have lower mortality rates for most causes of death [l-8]. Nathanson [9, p.57 states that “the sex differential in mortality statistics has usually been explained by women’s constitutionally greater resistance to both infections and degenerative disease.” However, epidemiological evidence contradicts this explanation. Studies of sex/gender differences in health and illness indicate an excess of female morbidity as compared to males [lO-151. The problem then becomes one of accounting for the discrepant pattern of morbidity data [16]. Clearly, constitutional resistance and/or susceptibility cannot adequately explain all health and illness differences between men and women. The study of sex/gender differences has been approached from a number of perspectives. For example, a biomedical approach essentially tests illness hypotheses, whereas a sociomedical approach tests illness behaviour hypotheses. According to the biomedical approach, it has been hypothesized that observed sex differences in morbidity are the product of biologically based inherited risks. If being female means greater risk of disease, and disease may cause death, then this should be reflected in sex differences in mortality rates. Given the inability of this perspective to account for differences in both morbidity and mortality *Revised version of a paper presented at the Western Association of Sociology and Anthropology Meetings in February, 1985, Winnipeg, Canada.
rates, an alternative approach has been formulated. The sociomedical perspective hypothesizes that these differences can be better explained in terms of social role obligations, acquired risks, health reporting behaviour, and/or the illness and preventive health orientations of women and men. In short, it is argued that differences in both morbidity and mortality are related to sociocultural and social-psychological factors (i.e., gender). The findings of research informed by sociomedical hypotheses are, unfortunately, often confusing and inconsistent. To complicate this area of study even further, it is readily acknowledged that health related behaviour may also be- influenced by the interaction of biological and sociocultural factors. Yet, our ability to separate out the effects of nature vs nurture are extremely limited. How can one show where biology stops and environment starts? Many hypotheses that have been formulated both within the biomedical and sociomedical approaches have considerable descriptive value. This study, informed by a sociomedical perspective and using a Canadian sample, provides additional data on the subject of sex/gender differences in health related behaviour. In particular, the salience of role characteristics is explored in relation to health, illness and sick role behaviour. While circumspect in terms of the scope of the hypotheses tested, this study attempts to move the level of discourse from description toward an explanation of sex/gender differences. This is in recognition of the need for a synthetic theoretical model to integrate the diverse empirical findings which already exist, and to serve as a guide to future research on male/female differences in health and illness behaviour. 579
580
MARY-ANNEKANDRACK~~ AND HEALTH CARE UTILIZATION: DESCRIBING SEX/GENDER DIFFERENCES
MORBIDITY
Before reviewing some of the literature on sex/gender differences in health and illness, it is worthwhile to consider the conundrum which characterizes the study of sex/gender. Writers in gender relations typically differentiate between the concept of sex and the concept of gender, the former referring to a physiological attribute and the latter representing the sociocultural expression of sex [ 171.Although this is an important distinction, its utility is often lost at the empirical level, owing to our inability to capture these differences. Furthermore, as Eichler asserts, “no human activity exists that is not mediated through culture.. . , i.e., there are always aspects of gender in sex, if we use this language. In principle, therefore, the distinction cannot be clear” [18, p. 3601. In light of this point, it is worth remembering that while some writers speak of sex differences and others speak of gender differences, the line between the two is often quite fine. This is illustrated by the body of research evidence showing the greater longevity of women. In most industrialized nations, women outlive men by approximately 7 years. Though there has been a recent stabilization of the life expectancy differential, women’s apparent constitutional resistance continues. In 1976, Waldron posed the question “Why Do Women Live Longer Than Men?“. Waldron and others suggest that sex differences in life expectancy can be attributed to the biophysical differences between males and females; i.e. it is generally thought that males are less durable biologically than females [l-8, 19-221. Sex differences in life expectancy have also been related to occupation [23-281. In this case, the discussion becomes one of gender differences because the intervening variable (occupation) is a social role, which may systematically vary by sex, but is still not a biophysical fact. Similarly sex differences in morbidity may be examined more closely by exploring occupational and familial roles that have been linked to the health of men and women [6, 16,29, 301. Again, as soon as these behavioural dimensions are introduced, the discussion shifts to one of gender, and not simply sex differences in morbidity. In keeping with the task at hand, interpretations and explanations of sex/gender differences in health and illness behaviour must be sensitive to the problem of mistaking these two dimensions for each other. What does the existing empirical evidence reveal about the nature of sex/gender differences in health related behaviour? Health survey data indicate an excess of morbidity among females as compared to males [l, 2,4, 5, 7-9, 15, 16, 3 1, 321. This excess appears for both acute and chronic conditions with the qualification that the latter are more “consequential” and/or “incapacitating” for males [S, 7-91. In a summary of American and British evidence regarding sex differences in morbidity, Nathanson [9, p. 571 reports that “women have higher rates than men for almost all indices of morbidity and utilization of health services.” Clearly et al. [33, p. 1061 cite a variety of sources which suggest that “women consistently have been found to use outpatient medical services more frequently than
al.
men.” According to Verbrugge and Wingard, females utilize more medical and health services for curative as well as preventative, purposes. Thus, studies of utilization typically show an excess in female rates, which in turn “boost their rates of morbidity compared to males” (32, p. 291. While morbidity and mortality rates and health services utilization patterns are apparently interrelated, it is extremely difficult to establish the causal connection between these factors. There tends to be some agreement between the morbidity and mortality data for men and women, however, the linkages between these health status indicators are not always direct. When one considers health care utilization, the picture becomes even more complicated. The basic assumptions of the biomedical perspective suggest that higher morbidity is correlated with higher utilization, and that sicker individuals will more likely die than those who are healthy. Furthermore, it is argued that if individuals avail themselves of medical services, then improvements to health will follow. In other words, it is assumed that women may be healthier (i.e. have a greater life expectancy) in part because they are more active consumers of health care services. This line of reasoning obviously has some validity, although several critics of modern medicine point to the fact that the use of medical services is neither a necessary, nor a sufficient, condition for the production of health [34-371. While the debate regarding the efficacy of medical measures is beyond the scope of this paper, it would be naive to argue that medical services alone are responsible for improvements in health. Lalonde [38], among others, has pointed out that medical services are but one element in the health field. Human biology, the environment and individual lifestyle are also important to understanding the distribution of health, disease and death in any human population. Sociomedical studies of sickness have emphasized the important role of demographic characteristics (e.g. age, education, occupation) and socio-structural conditions (e.g. access to medical services and alternative types of health care) as possible mediating factors in the relationships between morbidity, mortality and utilization. To date, however, no single explanation had been able to account for the discrepancies between male and female rates of utilization (and the bearing of utilization on morbidity and mortality). THE SEARCH FOR AN EXPLANATION: RESEARCH HYPOTHESES ON SEX AND GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR
The selective evidence summarized in the preceding section demonstrates that male/female differences do indeed exist in the distribution of disease and death, and in the utilization of health services. Marcus and Seeman [16] suggest that it must now be asked why do women experience higher rates of milder conditions? In addition, it is important to seek an explanation for women’s greater utilization of health services and experienced disability days, in view of the finding that their excess morbidity is primarily of a mild form. One possible answer to these questions may be found in the nature of male
Gender differences in health related behaviour vs female biological risks of disease and death. Other answers focus on acquired risks, role occupancy and the psychosocial aspects of illness. Once the social psychological and sociocultural dimensions of health and illness are introduced, the central question becomes-why are there gender differences in health related behaviour? An exclusively biomedical approach to this question will not suffice. Although biological differences may explain certain symptoms or conditions, such differences cannot explain the overall pattern. For example, Waldron has noted that genetic and hormonal factors may account for some differences between males and females in their susceptibilities to congenital anomalies, infectious disease, atherosclerotic disease [l, 2,5]. It has been suggested, as well, that sex-specific conditions such as pregnancy account to some extent for females’ greater use of health services. However, it has been reported that “even when conditions associated with reproduction are excluded, women have higher rates than men for all indices of illness experience employed” [9, p. 581. Inherited health risks simply do not fully explain morbidity patterns, or the discrepancy between morbidity and mortality rates. It is clear, therefore, that a systematic explanatory model must incorporate the psychosocial and sociocultural aspects of the experience of illness. To date, the sociomedical approach to gender differences in health and illness has generated a wide variety of hypotheses, which focus on the following: illness behaviours [ 10, 39-421; reporting behaviours [39,43]; acquired risks related to social roles, lifestyles, and stress [16,21,25,20,41,42]; methodological artifacts [44]; and a sex bias in medicine [7,45]. Verbrugge [7], Nathanson [9] and Wingard [3] have each summarized the major hypotheses in their review articles. Since the present study extends some of the work previously done regarding social roles and illness behaviours, our review of the literature is restricted to these areas. Despite extensive study, there is no conclusive support or refutation to be found in the research investigating any of these hypotheses. Indeed, in some instances, conflicting findings have been reported [e.g. 11, 16,4042]. Marcus and Seeman [ 16, p. 18I] found that “fixed role obligations are particularly relevant for the study of sex differences in reduced activity-but may be less relevant for explaining male/female differences in symptom recognition.” Similarly, Verbrugge [7] has aptly noted that the salience of any given hypothesis tends to vary with the outcome measure under consideration. The present study examines self-reported health related behaviours in relation to social roles. The central research question is: how do role attributes (particularly, marital status and social status characteristics) influence individuals’ beliefs and actions related to health? The purpose of this analysis is to extend the empirical basis of the exploration of gender and health. TESTING GENDER DIFFERENCES IN HEALTH AND ILLNESS
The Winnipeg Area Study (WAS) is an annual cross-sectional, city-wide survey of a sample of WM3215-E
581
Winnipeg residents [46]. The 1983 WAS data were not collected for the explicit purpose of studying sex/gender differences in health related behaviour. However, this community survey included questions on gender and health and consequently the data can be used to illustrate some of the conceptual and methodological problems associated with testing hypotheses pertaining to illness behaviour and the acquired risks of social roles. Sample design and data collection
The 1983 WAS population was designated as all dwelling units that were listed in the 1982 assessment file for the City of Winnipeg, Canada. A simple random sample of 701 addresses was selected for personal interviewing from a computerized list of addresses compiled by the City Planning Department. Nursing homes and temporary residences were deleted from the sample and the household was the primary sampling unit. Within the household one eligible person was selected as the respondent for a 1 hr interview. A person was eligible to be the respondent if the dwelling was her/his usual place of residence and s/he was 18 years of age or older. Interviewers were instructed to try to obtain an equal number of male and female respondents within their given allotment of addresses. As adult males are generally more difficult to contact, set procedures were followed by the interviewers to assist them in obtaining their quota. If eligible respondents refused, the household was not replaced. This method for selecting interviewees in households has been used successfully in the WAS over the past decade, without producing evidence of biased samples. A total of 524 respondents were interviewed for a response rate of 75%. The sample was compared to the 1981 Census of Canada for Winnipeg and the results reveal that the sample is representative of the city on a number of key characteristics such as, household size, age and sex distributions. For example, the sample is 56% female compared to 53% of the city population (over 20yr of age). It may be of interest to note that on average the person answering the door in Winnipeg Area Studies was female in 57% of the households. There was no significant difference between the 1981 Census distribution of males and females (53%) in Winnipeg, the sex of the first contact within households (57% female), and the sex of the respondents who participated in this survey (56% female). Operationalization
Studies of sex/gender differences have varied considerably in the specific aspects of health/illness behaviour investigated. This study focused upon the relationship between gender and health status, health behaviour, illness behaviour and sick role behaviour. The major independent variable is, of course, gender. In measuring gender, the first step was to determine the respondents’ sex and then to explore the various roles (e.g. marriage, employment) occupied by each person. The intent was to try to learn about the role commitments of men and women which may be linked to health. The underlying assumption was that a better understanding of the link between male/female social roles and acquired
582
MARY-ANNE KANDRACK et
health risks may provide further insight into gender differences in health related behaviour. A number of social factors were measured including: age (actual age in years); marital status (married/common-law, single/never married, divorced/separated/widowed); level of education completed (elementary, junior high, high school, vocational, university); income (total individual income); and employment status (employed, unemployed, retired, in school, keeping house). Turning to the dependent variables, two indicators of self-reported health status were assessed. Respondents were asked: how would you describe your present health; and how healthy have you felt in the past twelve months. The response categories provided for these questions were: excellent, good, fair, and poor. The health locus of control belief scale was used in this survey to measure beliefs about health behaviour. Each respondent was presented with a list of twenty statements designed to gauge their views about an individual’s ability to control health matters. Responses ranged from strongly agree (5) to strongly disagree (1). An internal preventative subscale that reflects beliefs about one’s ability to avoid illness was derived from a factor analysis of responses to the locus of control items included in the interview schedule [47]. To assess one dimension of actual health behaviour, the respondents were asked about their use of informal social networks (i.e. spouse, friends, parents, children) as health resources. For example, each respondent was asked who s/he speaks to about health problems. A summary measure was examined for gender differences in the number of informal social network members consulted for discussions of health problems. A number of measures of illness behaviour were used to assess the respondents’ illness experience and their symptomatic responses. Measures focused upon the disability days experienced by each respondent during the past year including the number of days that an illness required a cut down in one’s usual activities, kept the individual from work, and/or kept him/her in bed. These were included as indicators of illness behaviour, as many studies have concentrated on sex/gender differences in the level of restricted activity resulting from illness. In addition, illness behaviour was gauged by the type of response (i.e. ignore, self-help, see doctor) made to a variety of conditions (e.g. dizziness, loss of appetite). Respondents were also asked to report the number of different medications presently being taken. Finally, sick role behaviour was assessed in terms of willingness to adopt the sick role and actual utilization of medical services. The interview schedule contained a number of Likert type attitudinal items which focused upon: (1) self-reliance, or one’s willingness to relinquish well-roles and become dependent upon lay others; and (2) medical skepticism, or one’s faith in physicians and willingness to become dependent upon doctors [48]. Included among the measures of actual medical care contact were having a regular doctor; number of different doctors visited in the past twelve months; recency of last general medical checkup and the total number of visits made to all of one’s doctors in the past twelve months.
al.
Some of the measures used here are less than ideal for the purposes of the present analyses. For example, measures of self-reported health may confound aspects of reporting with actual health status. Our measures cannot unambiguously capture the nuances of socially constructed meanings attached to health and illness for women and men. As well, measures of health services use are not sufficiently sensitive to differences in help-seeking related to preventive vs reactive care. Plus, we cannot differentiate between help-seeking related to reproduction and other reasons for frequenting medical practitioners. The latter point is particularly crucial in the study of gender differences in utilization, as has been noted previously. Thus caution is exercised in the interpretations of data offered here. At a later point, we will return to a consideration of these methodological issues, as they bear upon this field of research. Data analysis
The data analysis was done in stages. In the first step, a test of the statistical significance of relationships between sex and the selected set of indicators of health status, health, illness and sick role behaviour was conducted. These bivariate relationships were assessed with measures of association and the chi square test. The second phase of data analysis involved a test of the explanatory value of sex relative to other sociodemographic factors, in accounting for differences in those dependent variables found to be significantly related to sex. An hierarchical regression procedure was used for this purpose. Sex and a set of sociodemographic factors (age, marital status, education, income and employment status) were entered successively into an hierarchical regression. In other words, the sociodemographics were not entered until the explanatory power of sex had been exhausted. The residual, that is, the proportion of variance in the selected indicators left unexplained, was then entered into the analysis as the dependent variable. This procedure allowed an assessment of the explanatory value of sex alone and a comparison with the proportion of variance explained by other sociodemographic factors [49]. In the third and final phase, an analysis of variance was performed in order to explore sex/gender differences in the dependent measures. This involved a closer assessment of two variables that are related to sex and reflect associated behavioural patterns (i.e. gender role). These variables are employment and marital status. The data were examined for differences in health, illness and sick role behaviour, by marital status and employment status for women as compared to men, and for women only. In short, inter- and intra-group differences were explored. GENDER DIFFERENCES IN HEALTH RELATED BEHAVIOUR: THE EMPIRICAL EVIDENCE
The first step of the data analysis was to assess whether there were significant differences between the two groups (males and females) other than sex. This involved a comparison of age, marital status, education, employment status and income distributions for men and women. As indicated in Table 1, there
Gender differences in health related behaviour Table I. Distribution
of male/female respondents graphics
by selected demo-
Male Variable
n
Female %
n
Age 18-24 25-34 35-44 45-54 5544 65 yr Total
yr yr yr yr yr or older
33 53 54 21 37 28 232 x2 = 7.785, df = 5, P
