In addition to the factors that affect women’s access to health care, its delivery, and its quality, the list of possible health-influencing factors includes a range of socioeconomic and behavioral topics. The workshop and this chapter cover geography, socioeconomic status, education, employment, and two behavioral factors, substance use and mental health.
Jennifer Karas Montez (Syracuse University) introduced her presentation as the result of collaborative work with Mark Hayward and Anna Zajacova over the last several years with the aim of explaining the large inequalities in women’s mortality in the United States and, specifically, the inequalities at the state level. In her presentation, she first addressed the overall situation and then reported on a recently completed innovative project that attempted to explain those inequalities.
Montez emphasized that life expectancy varies markedly not only between the United States and other comparable nations (see Chapter 1), but also among the states in this country. The inequalities by state are really striking: some states have a life expectancy similar to very low-income countries around the world. For example, Minnesota has high life expectancy, on par with the United Kingdom. At the other extreme, Mississippi has a life expectancy on par with Syria.
The range in life expectancy across the states exceeds the range on life expectancy across comparable high-income countries. This is true both for
life expectancy at birth (the range at birth is 7.4 years for U.S. states and 4.7 years for comparable high-income countries) and for life expectancy at age 50, though to a lesser extent (the range at age 50 is 4.4 years for the states and 4.1 years for comparable high-income countries) (Wilmoth et al., 2010).
The situation is not improving, Montez said, and there is no sign of the states converging toward one common U.S. life expectancy. Comparing female life expectancy at age 50, a small handful of states, including Massachusetts, have seen some impressive gains in women’s life expectancy, while other states, including Oregon and Mississippi, have shown only modest gains, and still others, including West Virginia and Wyoming, have actually seen women’s life expectancy decline. (Life expectancy has not declined for men in any state.) The inequalities have been growing since the 1980s, and they have been growing more for women than they have for men.
Montez presented two hypotheses that have been suggested to explain the large differences across states, counties, and other geographic areas. The hypotheses are referred to as “people versus place” and “composition versus context.” The hypotheses attempt to sort out whether the cause is women’s characteristics or state characteristics.
Montez explained that it is only recently that the hypotheses have been subject to testing because of data limitations. Most public-use datasets that have mortality data do not contain geographic information. As a result, the small number of studies that have attempted to explain disparities have been focused on a “people” explanation. However, both individual and place characteristics affect how people live.
There are a wide range of place characteristics that could influence their residents’ mortality, Montez suggested. One example is income-tax policy in that taxes affect residents’ economic well-being, as does Medicaid eligibility rules. A state’s abortion laws affect access to health care. At a more institutional level, corporate tax incentives determine how enticing it is for employers to move into a state, and state tobacco control strategies can shape health behaviors.
In addition to looking at specific characteristics, it is important to test the hypotheses over time. Montez noted. The large inequalities have grown since the early 1980s during a time when federal aid to states declined, and states have been granted more discretion over policies and programs.
Montez reported that it is now possible to test these hypotheses because, as of 2013, data from the National Longitudinal Mortality Study (NLMS) are available by state of residence for respondents. The NLMS was developed for the purpose of studying the effects of demographic and socioeconomic characteristics on differentials in U.S. mortality rates.
It is based on a random sample of the non-institutionalized population of the United States, comprising data from U.S. Census Bureau’s Current Population Surveys and annual social and economic supplements and a subset of 1980 census data. These data are combined with death certificate information to identify mortality status and cause of death. The NLMS currently consists of approximately 3.8 million records with more than 550,000 identified mortality cases with socioeconomic variables.1
The hypotheses tested focused on women aged 30-89. Over the course of the study, women in this age cohort experienced almost 26,000 deaths. The study gathered data on some fundamental characteristics that might shape mortality, including race, education, and income. Montez said that she and her colleagues also collected information on several characteristics of states, such as economics, politics, and the tobacco environment. With these data in hand, the team estimated a series of multilevel models. Montez summarized the findings—not yet published—which showed that the variation in women’s mortality across states reflects differences in both people and place, that is, both composition and context.
Montez concluded that inequalities in women’s mortality reflect more than individual choices, characteristics, and behaviors. States seem to play an important role in creating and sustaining the inequalities in women’s mortality and morbidity. Research agendas for women’s health should focus on the role of differences in state environments in women’s mortality trends overall, as well as the differences within the United States. She suggested, given the amount of inequality that is related to geography, the usefulness of tracking health only at the national level as an average is questionable: comparative analysis based on percentiles or ranges of risk rather than averages might be more useful.
A participant noted that the research shows why it is so important to stratify and to have access to new data sources and wondered about the effect that this information would have for the shaping of women’s health policy. Montez responded that the results have not yet been discussed with a dedicated policy audience and that it is too early to identify any possible policy changes from the research
Another participant seconded the importance of undertaking analysis of local-area differences: How much of the residual would be reduced by adjusting for more granular place-based policies and conditions? Montez responded that other work, such as that by David Kindig and Erika Cheng (e.g., Kindig and Cheng, 2013), at the county level, shows differences within states. Some potential explanatory factors are better con-
ceptualized as operating at more local levels. However, the local-level analysis is limited by the available data.
Continuing on the topic of local analysis, a participant wondered how the variables might change when incorporating local data. Certainly, the variable set will expand when assessing differences at the local level, Montez responded. Perhaps some of the variables that are being conceptualized as state-level variables will be reconceptualized as local–level variables. Factors that would be important at the local level are likely to be environmental issues. Additional detail on space and place would become critical, Montez said.
The research on socioeconomic status (SES) and health was summarized by Sarah Burgard (University of Michigan). In her introduction, she noted that this is a broad topic, and some key aspects will be covered by other panelists. Multiple aspects of SES are associated with health and survival, and there is a tremendous amount of information and evidence on this topic. The consensus conclusion is that social factors appear to be quite important in the overall poor performance of U.S. women’s health relative to other wealthy counties.
Burgard began her talk with reference to a recent study, The Growing Gap in Life Expectancy by Income (National Academies of Sciences, Engineering, and Medicine, 2015), which contained estimates of remaining life expectancy at age 50 for men and women based on their quintiles of average Social Security earnings in their 40s. The report found that low-income adults in the United States are losing ground relative to their wealthier peers.
The study also found a stark differential by income quartiles for men and women. Among men, their remaining life expectancy at age 50 is basically flat in the bottom 40 percent of the income distribution for both 1930 and 1960, while gains have been made in the top 60 percent of the income distribution, strengthening the income gradient in life expectancy. But among women, there appear to be actual losses in expected life expectancy at age 50 in the bottom two income quintiles, no progress in the middle or fourth quintile, and gains only among the top quintile.
Burgard observed that these are shocking findings for many Americans. They point to the need to understand how socioeconomic status differences like this can emerge and how to interpret them. The understanding starts with focusing on two possible explanations—causal explanations and explanations based on health selection.
In a causal framework, SES embodies an array of factors, including such resources as money, knowledge, credentials, and power or beneficial
social connections. These in turn influence where people live, where they work, their earnings, and their working conditions as well as their stress levels and the way people cope with them; access to health care; and, ultimately, health and survival.
In a health selection framework, the explanation is that people with poor health tend to move down the socioeconomic hierarchy: that is, it is poor health that leads to the poorer outcomes. To understand health selection requires understanding of the relationship between indicators of SES, such as employment status and earnings, and health status. For example, poor health in early life health could impair educational attainment, which could influence subsequent earnings and then subsequent health prospects.
Both causal influences and health selection can be at work, and understanding them is very important for understanding SES gradients and health. For example, causation could be influenced by gender if women have less access to higher SES standing and the resources that it provides. Gender differences and selection processes could be important if health is less likely to be an impediment for women than for men, who more commonly occupy blue-collar occupations (as well as high status and less strenuous ones). However, women suffer from greater morbidity throughout their lives, so health selection may be a salient aspect in many dimensions of SES accumulation.
Burgard presented two heuristics that can help understanding of the ways SES is implicated in U.S. women’s health. First, one can think about explaining the difference in the health of two groups by addressing the possibility that the distribution of SES resources varies across the groups, so that they have differential access to the health-promoting benefits of those resources. Average levels of income could be different, or income inequality could be different across groups, or one group might have a heavy concentration at the high or the low end of the income distribution. So, for example, U.S. women may have lower average SES on some key dimensions than men, or they may be more likely to be clustered below the poverty line. Similarly, U.S. women may have more years of education or be more likely to have completed a postsecondary degree than women in other wealthy nations, but they may be less likely to be employed full time.
Second, SES can be moderated by context when making these comparisons, for example:
- Does having a postsecondary degree “buy” the same amount of income for men and women in the United States?
- Do expectations about appropriate female behavior lead to differences in the way men and women perform their work or family roles, even if they are at the same SES level?
- Does having a low income mean the same thing across all countries, given wide variation in welfare state supports?
Using these heuristics, Burgard said, it is possible to make useful comparisons to isolate the way that SES operates to influence U.S. women’s relative health standing. Comparing U.S. men and women directly targets the influence of gender, but there may be physiological as well as social differences to acknowledge. This is, women’s reproductive capacity strongly shapes their social roles and resources and interactions between their biological and social lives. By contrast, distributional differences in SES and social factors (the way societal context might moderate or modify the ways that women can use their resources) can be better understood if the comparisons are made among women across peer nations. Finally, comparisons between U.S. women with high and low SES are more likely to isolate distributional differences in and the effects of consequences and mechanisms of the way SES works in this country. With all these approaches, the influence of a person’s place in the social hierarchy in relation to health can be understood.
These alternative ways of considering the relationship of SES and health are a challenge to the design of future research. Burgard said that it is really important to consider the kinds of questions that can be best answered about women’s health with each of these different kinds of comparisons and then to communicate better across the different research streams that look at these comparisons.
She noted that the past half-century has marked enormous progress for U.S. women in terms of socioeconomic resources and achievements, along with changes in other areas of life. From the 1960s to the mid-1980s, women attended college in greater numbers than they had previously, and rates of labor force participation grew almost continuously, even among mothers of young children. The gender gap in earnings has slowly eroded, and women have contributed more to family income. Overall, they have more resources to enhance their own health and the health of their families. Over the same period, fertility fell, although divorce and single parenting increased.
These are important social trends, Burgard contended, but while the changes in SES have been positive, women’s health may not be benefitting at the same pace. For example, although the rising education of women is important, this trend does not seem to result in the same increase in income and assets (or control in the workplace) and other important resources as does the rising education trend for men. Even when they possess the same credentials as men, women face gendered norms about appropriate titles and types of jobs and careers, and they may face employer discrimination and other factors that make it difficult for them
to advance in their careers, leading to occupational segregation and flatter, less well-paying career ladders.
Women have lower labor-force participation rates, Burgard noted. One factor leading to that difference has to do with women’s much greater obligations for unpaid household production work (see below). The difference in work histories has other effects. In addition, although men and women both benefit from social welfare programs in the United States, in some cases the social welfare programs are tied to labor-force participation histories, and women, who tend to have shorter work histories due to child-rearing responsibilities, may not benefit as much as men from those benefits.
Another dimension affecting the differential health experiences of men and women, even when they have the same level of SES, is patterns of time use. There are gender differences in use of time for leisure, exercise, sleep, and other health-enhancing behaviors. Burgard noted that men are penalized because they are more apt to work full time and, full-time workers have the least time for sleep. But in other ways, time use is modified by gendered expectations and structures that affect women’s health more than men’s health.
Women do more housework and have more interruptions in their personal time. Their time is not as “sacred” as men’s, even when they have the same roles and the same SES. Burgard reported on her analysis (unpublished) of social and demographic characteristics and SES, as well as other aspects of time use (with date from the American Time Use Survey). She found that males had a major advantage in leisure time, but women actually slept a few minutes more than men at the same stage of their life cycles. Their sleep may not be of the same quality as men’s sleep, however. Women tend to get up out of sleep to provide care to a child or another person more than men do. While this difference is understandable among those with very young children, the same gender gaps show across all different kinds of family arrangements. The quality of sleep is a health issue. Animal studies in which rats are awakened frequently show that they suffer high mortality rates.
Burgard then turned to analysis of the comparisons across groups of women in other wealthy countries. Noting that the international comparison studies have shown that U.S. women are losing ground relative to women in peer countries, she reported that part of the reason is the different levels of critical SES resources between the countries. Though average U.S. incomes are high relative to other countries, relative poverty is also the highest in the United States. Other reasons for the U.S. standing may be the higher U.S. female labor force participation, the rising rates of single parenting, and lesser access to steady working careers. Different patterns in women’s parental support and caregiver support could be
particularly important for women’s relative standing among peer counties, she noted.
Another perspective on the SES issue is offered by comparing U.S. women at different points on the SES ladder. Burgard stressed that there are clear differences in the lives of low-SES women in the past several decades that are especially concerning. She referred to an editorial in the American Journal of Public Health by Montez and Zajacova (2014) that presents two contrasting explanations of why low-SES women are doing particularly poorly: low SES causes poor health or the U.S. women with very low education are a special group that faces particular barriers to health.
Burgard stated that to better understand the role of SES, better measurements are needed. Although there are good measures of education and income in many studies, there are few measures of debt, income volatility, or assets—which could be hiding a tremendous amount of heterogeneity within a group of people with the same educational credentials, for example.
The measures need to take a life-course approach and consider a range of SES indicators that vary in importance as adults age. Analysis over the life course could help to identify for what stages it would be most productive to propose interventions and which aspects of SES could yield the largest returns to women’s health. SES analyses need to pay attention to health conditions that particularly burden women, including arthritis and musculoskeletal disorders, other disabling disorders, and the disabilities that are influenced by both biological and sociological mechanisms.
Burgard suggested that an SES research agenda should consider such questions as how recessions perturb many aspects of SES; how the Affordable Care Act will affect gender differences health in the United States; and how interventions in workplaces could make them more family friendly and could both reduce inequality among men and women and also reduce inequality across high- and low-SES women.
Mark D. Hayward (University of Texas at Austin) discussed the relationship between education and other SES indicators and mortality. In his presentation, he reviewed the empirical evidence documenting the dynamic nature of this association for men and women. He emphasized that the relationship between education and health is dynamic and pervasive. For example, as discussed below, over the past two decades, women with less than a high school education experienced increased mortality but there were extremely rapid declines in mortality among women with a college education or more. Mortality for men declined across the board.
He argued that the association between health and education has been strengthening in recent decades, and he ascribed these trends as a result of the acceleration in the pace of social change.
Hayward argued further that education can have massive consequences in terms of marriage, social relationships, economic consequences, where one lives, who are one’s friends, and a sense of agency. All of these mechanisms can affect health. The core of Hayward’s presentation was a review of trends in mortality rates and life expectancy and how they differ by gender and by race.2
There are three critical issues in understanding the effects of education on health, Hayward said. First, the associations between education and health are endogenous to larger societal changes in technology, the political economy, and changing demography. Second, the associations between education and health have changed in fundamentally important ways in recent decades, but not for everyone, and not in the same way. Third, knowledge about mechanisms is changing and is likely to change more in the future. The relationships are not in equilibrium.
A most important finding of several studies (Hayward et al., 2015; Montez et al., 2011, 2012; Olshansky et al., 2012: Sasson, 2014) is that, for non-Hispanic white women, the increasing gradient appears to be the consequence of two trends: increases in mortality for women with less than a high school education and extremely rapid declines in mortality among women with a college education or more. The first trend is in contrast with the experience for non-Hispanic white men: mortality for those with less than a high school education has remained relatively stable. The second trend is the same for men: mortality has declined rapidly for those with a college education or more.
Hayward next discussed the paper by Montez and colleagues (2011) that assesses the trends over three different time periods: 1986 to 1992, 1993 to 1999, and 2000 to 2006. Over all three time periods, mortality has increased for non-Hispanic white women with less than a high school education. In contrast, for black women, there have been some declines in mortality, especially among those with less than a high school education. Both of these trends—for white and black women—are in contrast to the experience for black men, who have enjoyed a rapid decline in mortality in the last decade or so. Hayward emphasized that mortality estimates, especially those based on vital statistics, depend on the assumptions that are made in the analyses. Even with simple epidemiologic data, it is important to be transparent in the assumptions that are made and how the estimates are obtained.
2The analysis covered only blacks and whites because there are insufficient data on Hispanics.
Turning next to the Sasson (2014) paper, Hayward reported that for white women for 1990-2010, there was a decline in life expectancy at age 25 for women with less than a high school education, a quite stable life expectancy for women with a high school education, and an increase in life expectancy for women with 16 years of education (college) or more. The combinations of the gains and losses in life expectancy expanded the educational gradient.
For men, too, advanced education is related to improvements in men’s life expectancy at age 25. In contrast, however, life expectancy was relatively stable for men with less than a high school education. Thus, the overall growth in the educational gradient in life expectancy at age 25 was driven largely by the gains in life expectancy for men with 16 or more years of education.
By further decomposing the change in the life expectancy at age 25 between 1990-2010 by sex, age group, and years of schooling, it is possible to identify which groups are gaining or losing life expectancy for the 20-year period. Women with less than a high school education lost life expectancy primarily in the age ranges below 60, although losses in life expectancy occurred above age 60 as well. Women with a high school education experienced some losses in life expectancy in the age groups to about age 55 or 60; at older ages, women experienced improvements in life expectancy. For women with 16 or more years of education, there were improvements in life expectancy for all age groups and dramatic improvements at the older ages.
In summary, Hayward said, the improvements for educated women are being experienced at a variety of ages, which suggests multiple causes for these trends. At the other end of the spectrum, different causes are also involved with regard to the losses in life expectancy experienced by less educated women. A research agenda examining the association between education and women’s mortality should focus on understanding how different mechanisms may influence mortality for different education groups.
A cohort analysis by Masters and colleagues (2012) yielded similar results, Hayward reported. The authors used random-effects models to simultaneously measure age, period, and cohort patterns of mortality risk between 1986 and 2006 for non-Hispanic white and black men and women with less than a high school education, a high school education, and more than a high school education. The analysis looked at mortality risk from all causes and separately for those from heart disease, lung cancer, and unpreventable cancers. Again, their results show that the life expectancy increases with increased education and that there is a clear cohort phenomenon and a weak period phenomenon. This study suggests that each new cohort goes through a different set of circumstances and
that advanced education is becoming more important in recent cohorts in reducing mortality risks.
Hayward said that these findings are illustrated in a paper by Montez and colleagues (2012), which estimated how much each additional year of educational attainment was associated with the risk of mortality. Previous studies had identified that mortality risk dropped at discrete points in the education distribution—12 and 16 years of education–and there were no changes in the risk before or between the points. The more recent study, however, shows that the basic association has changed: especially noteworthy were the very sharp declines in mortality risk associated with each year of additional year of education after grade 12, with no floor effects. This change in functional form pointed to the growing importance of advanced education for reducing women’s mortality risk. Moreover, the change in functional form occurred in a very short time span—approximately 10 years. An update to the study (Hayward et al., 2015) documented that the trend was accelerating, with advanced education being even more strongly associated with low-mortality risks among women.
In summary, Hayward said, there is a growing literature that points to an increasingly stronger link between education and health in the United States over the past several decades. The reasons for that relationship are being explored. These changes are endogenous to larger societal changes in technology, the political economy, and changing demography. Technology, for example, increasingly defines the key activities of daily life. The market for health care, as well as the complexity of health care, has dramatically changed. And demographic changes, such as the trend for well-educated people to marry well-educated people, may be concentrating resources among the best educated groups in the population. These macrolevel changes have the reinforced what Fogel and Costa (1997) have termed “techno-physiological evolution”—a synergistic process in which technological change is tied to improved human physiology through humans’ ability to gain control over their bodies. Education may be particularly important in responding to rapid social change and improvements in technology that result in health advantages.
Hayward introduced a conceptual framework for understanding the dynamic nature of the association between education and adult health in the United States: although it may not help to understand international comparisons, it is relevant for explaining U.S. trends. Hayward said that, in his framework, there is no inherent causal association between education and adult mortality. It is clear that the relationship is becoming increasingly important and is different for different parts of the educational distribution and for different gender and socioeconomic groups.
A participant asked about the influence of technology and the relationship of technology to health. The issue is whether there is a gender
difference in the use of the new technology and, if so, whether there is any age stratification to the use of technology. Hayward replied that he was not aware of any scientific evidence that there are differences in men’s and women’s use of digital devices.
Another participant asked what the data suggest about the differential health trends for men and women. Hayward responded that males start out with more of a health disadvantage as measured in terms of life expectancy so that the improvement reflects their low starting point. It is also useful, he added, to look at measures other than life expectancy. Using measures such as modal ages of death, well-educated women with 16 years of education could be characterized as “maximized,” in sharp contrast to women with low educational levels. The long life expectancy for these highly educated women is due to the combination of resources accruing from stable marriages, excellent economic prospects in adulthood, healthy life-styles, and friends and neighborhoods that provide a range of social resources and the kind of health care that is available. At the other end of the educational distribution are women who possibly lack all of these resources.
Another participant commented that there is a lot of compositional change in educational attainment over recent decades. The prevalence of people with less than a high school education declined by 50 percent, from slightly more than 20 percent to slightly more than 10 percent of the population. This group is a differentially selected population, and the selection processes may help to explain the decline in life expectancy. It would also lead to understating the improvements in life expectancy in those other groups, as the other groups have grown. The participant also commented that a big factor in differentiating by education is conscientiousness rather than cognition. Education helps to know enough to do the right thing and having the practice of doing the right thing at the right time and place.
Montez responded that not much is known about how much of the increase in women’s mortality is due to selection rather than causation. Some arguments work against selection. For example, selection would imply the same trend for low-educated men since they graduate high school at slightly lower rates than women. However, men do not share the some negative mortality trends
Another participant suggested that wealth accumulation and family status might be factors in health outcomes. Wealth accumulation is related to debt. More women are going to school and increasing their educational attainment, but they also incur greater debt that, in turn, leads to lower wealth. Hayward responded that, in a family, education is a resource. If a household has heterogeneous educational composition, the more educated people will help the less educated people in terms of health benefits.
A participant asked Hayward to comment on the social context of women’s health. How much do the events of the past decades contribute to the higher mortality rate for either men or women and why the difference? Hayward responded that institutional factors at the federal and state level are important. However, the research has not yet been able to attribute women’s health trends to macrolevel phenomena.
Nancy L. Marshall (Wellesley College) reviewed the latest research on employment and women’s health in the context of the changing economy and changing family lives of the 21st century. Historically, she said, the argument has been that higher education would damage women’s health and that employment would interfere with women’s roles as mothers and wives or lead to rising health risks as women become “like men” and therefore at risk for cardiovascular disease and other “men’s diseases.” She noted that the context of employment for women’s health has changed dramatically: women are now almost as likely to be employed as are men and so equally vulnerable to the effects of poor working conditions. However, she said, women’s position in the economy is often different from that of men from similar backgrounds. In addition, women continue to have greater responsibility for caring for children and extended family, which creates additional demands on women.
Marshall pointed out that women’s participation in the labor force has changed dramatically in the last 60 years, rising from 34 percent in 1950 to 45 percent in 1974 and to 57 percent in 2014 (Smith and Bachu, 1999; Bureau of Labor Statistics, 2015a). The largest gain in labor-force participation rates was by women with children under 6, from a participation rate of 39 percent in 1975 to 65 percent in 2013 (Bureau of Labor Statistics, 2014). In 2014, more than 68 million women in the United States were employed. Of these, more than 50 million (74%) were employed full time (Bureau of Labor Statistics, 2015b).
Marshall summarized two competing views of the role of employment on women’s health: (1) women’s employment conflicts with their family responsibilities and creates role overload, which would negatively affect their health and well-being; and (2) women’s employment provides them with an additional arena in which to develop competencies, self-esteem, and social connections, and the combination of roles would enhance women’s health. She stated that a review of the research between 1950 and 2000 on the relation between employment status and health found that employment either had no effect on women’s health or had positive effects (Klumb and Lampert, 2004). But, she cautioned, state-
ments about what is true for “women” need to be followed by questions about “which women,” and “under what circumstances.”
Working conditions for women may also play a role in health. Several key aspects of the organization of work are related to job stress and lower job satisfaction, such as heavy workloads, little control over work, lower levels of substantive complexity, and little work-related social support. Trends in the economy—including downsizing and outsourcing of core functions, increasing use of contingent labor, flatter management structures and lean production technologies—have contributed to reduced job stability and increased workloads for many workers, and these factors give rise to high levels of job stress. Marshall pointed out that job stress has been found to be associated with cardiovascular disease and other illnesses, as well as psychological distress and depression, and that women are more likely to be employed in jobs with higher levels of stress (Vermeulen and Mustard, 2000).
Women are differentially exposed to particular health risks, Marshall noted. Some occupations with large proportions of women pose significant health risks. For example, the education and health services industries, which account for more than one-third of all employed women, have higher than average rates of nonfatal occupational injuries and illnesses (Bureau of Labor Statistics, 2010). Occupational health risks include low-back pain (nurses, child care workers), asthma (health-related industries and teaching), noise exposures that can contribute to reduction in hearing sensitivity and increased stress (teaching), and exposure to infectious, biological, or chemical hazards (nurses, child care workers) (McGrath, 2007).
Marshall also noted that sexual harassment has a direct health effect on the well-being of women (and men). It has been argued that sexual harassment is a stressful condition, and there is evidence that it reduces psychological well-being (self-esteem and life satisfaction) and increases psychological distress (depression, anxiety, symptoms of post-traumatic stress disorder) (Chan et al., 2008). Individuals often report what Selye (1993) calls “diseases of adaptation,” such as headaches, gastrointestinal disorders, and sleep disturbances.
Although employment has many benefits for women, as it does for men, including more positive perceptions of health and improved physical functioning, combining work and family can lead to experiences of work-family conflict that may negatively affect health (Byron, 2005). Work-family conflict is more common among employed mothers than among employed fathers, Marshall said, which partly reflects the fact that mothers continue to bear greater responsibility for day-to-day parenting despite fathers’ increased involvement with their children (Marshall and Barnett, 1993). Work-family conflict, in turn, is associated with overall physical health and, in particular, with hypertension and
high blood pressure (Frone et al., 1996). Overall, Marshall said, a large body of research has found that the working conditions—such as job demands and autonomy—are another significant contributor to work-family conflict, such that more stressful jobs are associated with greater conflict between the demands of work and family.
Work schedules are also a factor. When the schedules of paid work and family demands are incompatible, mothers of young children may choose nonstandard work schedules to facilitate combining work and family, often working evenings or nights while the fathers work days. Shift work, in general, has been associated with greater work-family conflict for both women and men (Haynes and Feinleib, 1980).
Finally, Marshall said, workplaces also vary in their norms and expectations of workers’ behavior in negotiating the borders between work and family. Some workplaces view borders as rigid: family needs should not interfere with work responsibilities. Other workplaces view borders as temporally or spatially flexible: workers may select semipermanent employment schedules that fit their family needs or may use flexplace options (working from home). Still other workplaces view borders as permeable, allowing workers considerable day-to-day flexibility in managing the needs of family and work. One study found that flexible work arrangements are particularly important to women with a lot of family responsibility and that flextime was more strongly linked to reducing work interference with family life than was flexplace (Shockley and Allen, 2007).
Marshall reported that research has begun to examine variations in work-family conflict associated with women’s life stages. Not all women have children or move through the same combinations of employment and family trajectories in the same way, and research has identified some important variations across the life span. For example, although combining employment and family is positive for many women, women with young children are more likely to report greater work-family conflict than are mothers of older children (Higgens et al., 1995).
This conflict is potentially most acute for mothers of infants, Marshall noted, and most new mothers return to work by the time their baby is 3 months old (Bureau of Labor Statistics, 2014). Research on postpartum health has identified health challenges faced by women, including physical recovery from childbirth, postpartum blues or depression, stresses in the marital relationship, as well as health problems of the newborns. Research on occupational health in the postpartum period has shown that longer maternity leave (time off from work) has a positive relation to maternal health and quality of life (Chatterji and Markowitz, 2005; Gjerdingen and Chaloner, 1994; Hyde et al., 1995; McGovern et al., 1997). A study of more than 700 working mothers of infants found that, while
employment provided these working mothers and their families with important income and other benefits, women in jobs with poor working conditions or who experienced greater work-family conflict because they were single mothers or caring for infants who were sick more often than other infants reported poorer emotional health (Marshall and Tracy, 2009).
Women with school-age children have the highest labor-force participation rate of women across all of the life stages, Marshall noted. At this stage of life, women either have reentered the labor force after having children or have several years of work experience; some women have advanced in their careers to positions that, while potentially more demanding, may also offer better pay and benefits. The parenting needs of school-age children are also different from, and less labor-intensive than, those of children under 3. These characteristics of this life stage contribute to the findings that women with school-age children report lower levels of work-family conflict than do women with preschool-age children (Martinengo et al., 2010).
Older women workers with grown children face different work-family challenges, with both growing numbers of older workers and growing numbers of workers with increasingly old parents. Some 15 percent of people 65 and older need assistance with one or more activities of daily living, and many more need assistance with chores, errands, or transportation (Treas, 1995). Most of this assistance is provided by family members, and, as been discussed earlier in this workshop, most of these caregivers in the United States are women (Barrah et al., 2004).
In summary, Marshall said, on the basis of several decades of research, it is clear that women react to stressful working conditions and to work-family conflict in much the same way that men do in terms of health outcomes. However, there are important differences in the levels and severity of stress women face, which are the result of differential exposures to stressful working conditions, to occupational hazards, to sexual harassment, and to work-family conflict.
A workshop participant asked about Marshall’s views on whether these work and other life-course conditions in the United States are different from the conditions experienced by women in other countries. Marshall replied that two major differences between U.S. women and women in comparable countries are income and wealth disparity (inequality affects both women and men) and the relative lack of government policies that support working parents in the United States. Another participant asked if it is appropriate to think about redefining women’s relationship to work rather than to employment, on the grounds that there is work at home in addition to work outside the home. Some have suggested defining work for women in terms of transportation, emotions, the burden of raising a family, and all the other demands on them. Marshall agreed, with a caveat:
she said that it is important to include consideration of caregiving and other unpaid family and relational labor, which are more likely for women than for men, but these work activities should be considered in the general context of paid work because more than one-half of all women are in the labor force, and three-fourths of them are working full time.
A participant asked about the status of women who do not have children. The health issues of these women are not linked to employment and family. Marshall replied that working conditions, occupational segregation, and sexual harassment in the workplace are important issues for all women, including women who do not have children, and have an influence on health.
Another participant noted that recent work has examined women in authority positions and the stresses they experience as they move up towards the “glass ceiling.” The consequences are higher rates of depression than their male colleagues, and they also have high rates of breast cancer. There are indeed differences for men and women as they move up in their careers, Marshall responded. Both men and women take on more responsibility and work longer hours as the move into management and professional jobs. However, men have a return in terms of greater flexibility and more control over their work. Women do not have that return and therefore may be more exposed to job stress and related illnesses.
Christine E. Grella (University of California, Los Angeles) focused on the behavioral health disorders—specifically, substance use and mental health disorders—and gender differences and their risk for morbidity and mortality among women. She said that it is important to consider these effects from a life-span perspective because the issues vary over the life course, from adolescence to older ages. Differential exposures to risk and caregiving also have an influence on morbidity and mortality.
Grella discussed the influence of biology on the etiology, development, and prevalence of substance use and mental health disorders. The biological responses to psychoactive substances vary by substance. Research on biological responses to alcohol has found strong gender differences in how men and women respond to alcohol, how they metabolize it, and its physiological effects in the body. These differences make women more vulnerable to organ-related damage and to morbidity associated with alcohol consumption at lower levels of consumption than men. Women tend to develop more increased symptoms at a faster pace—a phenomenon that has been labeled telescoping.
Telescoping has been identified in responses to alcohol, stimulants, and opioids, Grella noted. Women respond much faster biologically
and develop problems of greater severity to the use of these substances than do men (Hernandez-Avila et al., 2004). The science of understanding the biological responses to psychoactive substances is termed pharmacokinetics—the study of how psychoactive substances are distributed and metabolized throughout the body. Research has identified gender differences in the metabolic effects related to how substances are processed by the body (Greenfield et al., 2010). The gender differences in severity of alcohol-related morbidity, for example, are largely attributed to the difference in body mass between men and women. Women have less body mass so a smaller quantity of alcohol has a bigger effect.
Research has documented that men and women also have different enzymatic reactions to different psychoactive substances, Grella said. There are neurobiological differences in how the brain responds to psychoactive substances in ways that lead to women’s greater sensitivity to the substances and faster development of problems. The evidence shows that women’s response sensitivity is affected by different neural mechanisms that lead to analgesic effects—that is, the sedating effects of these substances.
Other effects, such as menstrual-cycle effects, have also been documented (Greenfield et al., 2010). It is important to understand how hormones influence the reinforcing experience of different substances in order to treat substance use disorders, Grella said. She noted that the very large gender differences, for instance, in the experiences of craving and withdrawal that are related to hormonal influences. These differences are seen in studies of nicotine addiction: women experience strong subjective reactions in terms of a craving response at different points in their menstrual cycles, which make it more difficult to treat tobacco dependence.
The response to stressors is an important aspect of the development of substance abuse, Grella noted. The sensitivity to substances may be influenced by a neuroendocrine response to stress, which may lead to changes in the neuroendocrine system. Prolonged exposure to environmental stressors as early as childhood may lead to emotional dysregulation—not having healthy responses to stressful situations, which in turn leads to a greater vulnerability to substance use disorders. There are gender differences in the extent to which emotional dysregulation leads to greater sensitivity, heightened responses, and more severity of disorders at a quicker period of onset.
The social–environmental context also plays a role in response to substance use. Grella summarized the results of a pivotal study examining sex differences in prevalence of use and opportunity to use for various substances (Van Etten and Anthony, 1999). Using data from a household survey, the examination of the general prevalence of use of cocaine, hallucinogens, heroin, and marijuana found that males in the population have higher prevalence of use of all of these substances. But substance
use is related to opportunity, and males also have a greater opportunity to procure these substances: when use is adjusted controlling for opportunity to use, the gender differences virtually disappear.
Grella stressed that it is important to understand the circumstances in which drug use is initiated. There are very different social factors and influences on men and women that stem from gender role expectations. With growing gender parity in many different domains, there is a greater likelihood that women will encounter greater opportunity, greater access, and less social inhibition to using alcohol and drugs.
The prevalence of drug use over people’s lifetimes also varies by gender. Grella presented data from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), conducted by the National Institute on Alcohol Abuse and Alcoholism in 2001 and 2002. The survey found that there is roughly a 2-to-1 ratio of male to female use across the different types of substances, including alcohol, marijuana, sedatives, and opioids; the ratio is slightly less for amphetamines (Conway et al., 2006). Marijuana use is particularly dominant among men. The higher prevalence of substance use among men has led to more visibility associated with more male substance abusers, particularly in terms of interactions with the criminal justice system, violence, and the kinds of externalizing behaviors that come to the attention of society. Grella commented that the greater attention paid to male substance abuse has led to neglect of these issues among women, who have a greater susceptibility of developing problems once they initiate use of alcohol and drugs.
Grella next turned to the results of research studies on mental health disorders, which show a different pattern of prevalence (Grant et al., 2004). Women have higher rates of mood and anxiety disorders than men, while men have higher rates of antisocial personality disorder and substance use disorders. However, women tend to have a higher rate of both mental health and substance abuse problems than do men, thus increasing the complexity of treating women’s substance abuse.
The life-span perspective is critically important in understanding the development of substance use and mental health disorders, Grella argued. She discussed the results of research (Cotto et al., 2010), based on the National Survey on Drug Use and Health (NSDUH), which provides national- and state-level data on the use alcohol and drugs (including nonmedical use of prescription drugs) and mental health disorders in the United States. The study looked at the rates of substance dependence in the past year among adolescents and young adults who reported any use of a particular substance. The outcome measure was the proportion who manifested dependence.
Among adolescents (aged 12-17) there is a parity of boys and girls who use alcohol and develop dependence. For marijuana users, there is a
higher rate of dependence on for boys than girls. Among adolescents who have used cocaine, girls have a higher rate of developing dependence. For prescription medications, girls have higher rates of dependence than boys. This difference for prescription medications may be the effect of a gender difference in opportunity: girls are more likely to be referred to mental health services than boys and thus may have more access and more opportunity to abuse prescription medications.
Among 18- to 25-year-olds, the differences continue. The gender differences are significant across all four drugs, with greater rates of dependence on cocaine and prescription medications among young women who have used those substances than among young men who have used those substances. Males use different substances at higher rates, but when girls and women use those substances, they develop problems more quickly and have more severe problems than men.
Similar results were found when measuring the symptoms of dependence for men and women by the number of days of cocaine use in the past year (Chen and Kandel, 2002). The findings of this study show that, among men and women who are dependent, women are using cocaine at higher levels with more frequency. This study added to the accumulating evidence that women’s patterns of substance use are more severe, and the consequences are more rapid.
Other data showing years from first use of drugs to the onset of a disorder illustrate the telescoping phenomenon. The data show that women have a much quicker onset of dependence for alcohol, nicotine, marijuana, and all of the illicit drugs (Costello et al., 1999). Girls may have lower overall prevalence of use, but those who do initiate use become dependent more quickly than boys. The greater complexity, comorbidity and severity of disorders that women experience present complications for the health care system, Grella said. The system has not been designed to address the complexity of substance use and mental health disorders combined with physical health disorders.
Comorbidity is the crux of the issue of the greater severity among women when they initiate substance use and have comorbid mental health disorders, Grella stressed. Comorbidity is greater for women, even at lower levels of substance use. For example, depression is the most prevalent of the mental health disorders in the population, and women who have both depression and substance use disorders, particularly of alcohol, typically report that their symptoms of depression preceded the onset of alcohol disorder. Although temporal ordering is not clear evidence of causality (there could be a common third genetic factor), it is notable that women typically report initiation of alcohol use subsequent to symptoms of depression. In contrast, men tend to report alcohol initiation, then alcohol dependence, and then the onset
of depressive disorders, which often will resolve following a period of abstinence.
The interlinkage between the substance use and mental health disorders makes treatment for women who manifest these disorders complex. The challenge for the treatment system is that both internalizing disorders, such as depression and anxiety, and externalizing disorders, such as attention deficit hyperactivity disorder, conduct disorder, and antisocial personality disorder, are more severe and pronounced with comorbid substance use disorders. One striking example is antisocial personality disorder: it is relatively rare among women—affecting less than 5 percent of women—but the rates of substance use among women with the disorder are much higher than for men with the disorder (Alegria et al., 2013).
Grella discussed another study of opioid users that looked at gender differences in patterns of comorbidity (Grella et al., 2009). It found that, among opioid users, women were twice as likely as men to have co-occurring mood disorders (major depression, dysthymia, manic disorder, and hypomanic disorder) and anxiety disorders (panic, social, specific and generalized anxiety disorder), but they were less likely than men to have antisocial personality disorder and alcohol disorder.
At she noted earlier, Grella said that studies of juvenile populations indicate that the onset of substance use disorders among juvenile girls progresses rapidly from initiation of use. For girls in the juvenile justice system, their psychiatric disorders tend to persist (Abram et al., 2015).
Trauma is a critical issue among women with substance use disorders. A study of a cohort of women in a California prison (Grella et al., 2013) compared this sample of women in a prison-based substance-abuse treatment program with women in the general population who were matched on sociodemographic characteristics. The study found that women substance abusers who had been incarcerated had much higher rates of lifetime trauma exposure: more than 50 percent of the women in the prison sample reported seven or more types of traumatic exposures. They were from two to four times more likely to have suffered from a variety of types of traumatic exposure than their counterparts in the general population.
In understanding morbidity and mortality, Grella said, it is important to consider treatment access and utilization for women with substance use disorders. Using data from the NESARC population survey, she and a colleague found that women with substance dependency had low rates of help-seeking (about 24%), even lower than men (30.5%) (Grella and Stein, 2013; Grella and Otiniano Verissimo, 2015). The reasons for not seeking help can be attributed to stigma, financial reasons, structural barriers, and fear of the consequences of entering the substance abuse system, which women are more likely to report than men as barriers to seeking help.
Female substance abuse is strongly associated with morbidity. Grella reported on a study involving a sample of heroin users in which their age of death was compared with the age of death for women in the general population by computing standardized mortality ratios (Grella and Lovinger, 2011). The study found that substance-abusing women have a much younger age of death than women in the general population, with five times the risk of death (controlling for age and race and ethnicity). They also had a higher risk of death than substance-abusing men in the study sample. Similarly, they lose more years of life and have much higher rates of chronic health problems than did the male heroin users in the same study sample.
Another way to examine morbidity and mortality due to substance use and mental health disorders is by using the concept of the global burden of disease. The global burden of disease is calculated by aggregating data across regions, combining years of life lost and years lived with disability into an aggregate statistic, disability adjusted life years (DALYs). Using DALYs as the measure, the Global Burden of Disease study found that boys under the age of 10 have a bigger burden of disease due to mental health and substance abuse problems, primarily because of behavioral disorders, but that females over the age of 10 have a greater burden of disease at all age groups from the combined burden of substance use and mental health disorders (Whiteford et al., 2013). These global data portray a greater burden of disease for women over the life span stemming from substance use and mental health disorders.