Research into relations between social roles and women’s health has fairly consistently shown that women who occupy multiple roles tend to be healthier than those who enact fewer roles.^{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16} However, most work in this field has been cross sectional in nature so that it has been impossible to determine the direction of these relations. Does this body of work provide evidence for a causative relation between multiple role occupation and good health, or are women with good health more likely to occupy multiple roles? This question of selection has plagued research on women’s health for decades.

As positions in the social structure that are based on enduring relations with other people,^17–20 social roles provide both a sense of identity and behavioural guidance. Thoits²⁰ has used the concept of role identity as a determinant of mental health. In addition, poor quality roles may ultimately influence physical health, either directly through the immune and neuroendocrine systems,^21–23 or indirectly through coping behaviours such as smoking, excessive alcohol, drug, or food consumption.^24–26 Alternatively, it is possible that health determines the social roles that people occupy. Some cross sectional studies have attempted to address the issue of causal direction by taking account of the presence of chronic illness when analysing relations between roles and reported health. Bartley and colleagues³ found that adjustment for longstanding illness somewhat attenuated the relation between women’s employment status and the number of symptoms reported in the past month. However, Macran and colleagues¹² found that full time homemakers were more likely to report poor health independent of limiting longstanding illness. Results from a study of relations between roles and health in the Netherlands suggested that the positive relation seen between paid employment and health as well as that between parenthood and health was partly the result of health selection; women who participated in the labour force or who had had children were less likely to have a limiting longstanding illness than non-employed or childless women.⁵

The few longitudinal studies to have examined social roles at baseline and later health have been limited in two ways in their ability to answer the question of selection. Firstly, few longitudinal studies have included measures of health earlier in the life course and so have been unable to investigate the influence of health prior to social role occupation. For example, a two year follow up study of the relation between social role occupancy and health status in Canada included no measure of prior health status.⁷ Also, studies that have examined relations between roles and mortality have lacked measures of prior health.^27–29 Secondly, previous longitudinal studies of roles and health have relied on social role measures at one or, at most, two points in time. For example, Janzen and Muhajarine⁷ examined the interaction between age and social roles, hypothesising that role effects would be stronger for older women, presumably because they had spent more time exposed to their particular roles. They found no interaction effect, but only had one measure of social roles (and age) for each respondent. Hope and colleagues investigated whether psychological distress at age 23 explained the increased psychological distress of lone mothers at age 33, and found that it only had a modest impact.⁶ In their 15 year follow up study of social roles and mortality, Hibbard and Pope found that non-employed women had a 70% greater chance of death than employed women after adjusting for self reported health two years before baseline, but their study also relied on a single social role measure from one point in time.¹¹ The only study to include social role measures at more than one point over the life course examined whether psychological distress at age 23 explained marital transitions between ages 23 and 33.³⁰ The study found that the increased psychological distress of divorced men and women involved both selection and causation. In addition, most of the studies in this field have used self reported general health to measure physical health. This study is able to examine relations between long term role histories and both self reported health and obesity in mid-life using data on women from the Medical Research Council (MRC) national study of health and development (NSHD), a longitudinal cohort study of British men and women born in 1946 and followed up throughout life. Self reported health, body mass index (BMI), and mental health in adolescence and early adulthood are then included to investigate whether relations between adult role histories and health in mid-life are explained by health selection.

METHODS

Sample

The MRC NSHD originated as a study of 5362 people who were born in Great Britain during the first week of March in 1946.^31,32 Of these, 2547 were girls. In adulthood, data have been collected in face to face interviews at ages 26, 36, 43, and 53, with postal follow ups between some waves. By age 53, 7.7% of the original cohort had died, 10.3% had emigrated, and 12.1% had permanently refused, resulting in an eligible sample of 1781 women. Altogether 87.8% (n = 1563) of this eligible female sample was interviewed by a survey nurse at age 53. Wadsworth and colleagues³³ compared the weighted cohort sample responding at ages 43 and 53 with data from the 1991 British census. They found that full time employed and widowed women were over-represented while never married, separated, and divorced women were under-represented at both years. The NSHD received ethical approval from multicentre and local research ethics committees across the UK. Informed consent was obtained from all participants.

Measures

Health outcomes in mid-life

Health in mid-life was examined using self reported health at age 54 and obesity at age 53. Obesity was chosen as a measure of physical health partly because of the social meaning it has for women and girls in particular. The inverse gradient between adult obesity and social position (in both childhood and adulthood) in industrialised countries is more consistent for women than for men.^34–37 Several studies have found that obesity is predictive of education, earnings, and social mobility for women but not for men, and also that obese women are less likely to marry than women who are not overweight, while the same is not true for men.^38–40 Obesity was also chosen because it is possible to hypothesise psychosocial pathways in which long term occupation of particular roles leads to dissatisfaction, frustration, or stress, then to unhealthy eating habits (and/or inactivity) as a coping mechanism, and eventually to obesity.

Height and weight were measured during face to face interviews in respondents’ homes. One height measurement was taken using portable stadiometers with shoes removed, the cohort member stretching to their maximum height, and their head positioned in the Frankfort plane. The reading was recorded to the nearest 0.5 centimetre. Weight was measured using a scale and asking cohort members to remove their shoes and heavy clothing. In accordance with the World Health Organisation international standards for adult obesity,⁴¹ women with a BMI of 30 or more were considered to be obese (25% of women at age 53, which is typical for English women at this age⁴²).

Self reported health at age 54 was collected as part of a women’s health series of postal questionnaires that were sent to women in the cohort every year between ages 47 and 54. The self reported health question was included in the eighth year of this series, and so valid responses were fewer than for obesity at age 53. Response categories for self reported health were excellent, good, fair, and poor. For analysis, women who reported fair or poor health were categorised as having poor reported health (nearly a third at age 54).

Social role histories

Information about women’s work, marriage, and parental status collected at ages 26, 36, 43, and 53 was used to create a social role history variable with six categories. Women who held all three of the main social roles of employee, spouse, and mother over the long term were designated as the reference category. Women in this Multiple roles (n = 555) group were married once and remained married at age 53, had children, and had relatively strong long term ties to the labour market. Women in this group were employed at three or four of the four ages of data collection, with about half entering the labour market after age 26. Three categories of role histories were defined by the absence of one of the main roles of employee, spouse, or mother over the long term. Childless women (n = 162) did not have children. Homemakers (n = 151) were full time homemakers at three or four of the four ages of data collection. Lone mothers (n = 254) were never or previously married mothers (mostly the latter). Finally, two role history categories were not defined by the absence of a main role, but made up substantial groups in their own right. These were Remarried mothers (n = 185) and Intermittent employed married mothers (n = 150) who were married once and had children, but whose labour market ties were weaker than those of Multiple roles women and stronger than those of Homemakers. The Homemaker group excludes 30 lone mothers and four childless women who followed the Homemaker employment pattern, but were included in the Childless and Lone mother groups. Twenty five remarried mothers who followed the Homemaker employment pattern were included in the Homemaker group rather than among Remarried mothers. Women who were unemployed at one or two ages were categorised into role history groups based on their employment status at the alternative two or three ages of data collection, while women who were unemployed at three or four of the four ages of data collection (n = 73) were excluded from analysis. Table 1 provides a brief description of the role history groups for easier reference.

Many,^3,5,12,43 but not all,² cross sectional studies of social roles and health have found that women who were employed part time had better reported health than women employed full time. Preliminary analysis in this dataset showed no relation between a longitudinal measure of full time compared with part time employment and self reported health at age 54 or obesity at age 53. As a result, full time and part time employment were treated equally as employed at each age of data collection.

Prior health, temperament, and childhood social class

A measure of self reported health was collected at age 26 that was identical to that asked at age 54. Women who reported fair or poor health (12% at age 26) were categorised to having poor reported health for analysis purposes. A summary measure of mental health episodes between ages 15 and 32 was used to examine early mental health. This measure distinguished between those with a major mental health episode (defined as being admitted to a psychiatric hospital, being in outpatient psychiatric care after the age of 15, having four or more episodes of illness covering more than a year between ages 15 and 32, or having illness for more than one year’s duration between the ages of 15 and 32), and those who reported experiencing minor mental illnesses, as well as those who reported no such illnesses between these ages. As neuroticism is known to be associated with marital quality^44,45 and divorce^46,47 as well as perceptions of poor health,⁴⁸ personality type is another factor that might explain relations between roles and self reported health. A neuroticism scale (range 0–12) from the Maudsley personality inventory,⁴⁹ collected at age 26, was also examined.

Obesity was rare at early ages (2% at age 26, 7% at age 36, 13% at age 43, and negligible at age 15), so mean BMI at ages 15, 26, 36, and 43, rather than obesity, was used to investigate whether relations between role histories and obesity at age 53 could be explained by selection into particular role history groups according to prior body mass. At ages 15, 36, and 43, height and weight were measured in the same way as has been described for age 53. At age 26, BMI was calculated from self reported height and weight.

Childhood social class influences both adult work roles,^50–55 and adult health.^{31,40,56–58} Therefore, the confounding effect of childhood social class on relations between role histories and health was investigated. Childhood social class was measured using the registrar general’s social class of respondents’ father’s occupation when cohort women were 11 years old, categorised into four groups: classes I and II combined, III non-manual, III manual, and classes IV and V combined. If father’s occupation at age 11 was unknown, similar measures at age 16 or 4 years were used.

Analytical techniques

The sample used for analysis of obesity in mid-life includes all women with a valid social role history and a valid BMI measure at age 53 (n = 1433). The sample size is reduced for models including measures of BMI at earlier ages. The sample used for analysis of self reported health in mid-life includes all women with a valid social role history and a valid measure of self reported health at age 54 (n = 1171).

A series of logistic regression models tested relations between role histories and both self reported health at age 54 and obesity at age 53, first unadjusted, then adjusted for prior health measures, and, finally, adjusted for both prior health and childhood social class. To investigate the effects of role histories on BMI across adult life, first, F tests were conducted to test whether mean BMI at each age (15, 26, 36, 43, and 53) varied by role history group. Next, generalised estimating equations were used to take account of the correlation among the repeated measures of BMI for a given subject.⁵⁹ We then tested whether the effect of roles on BMI changed over time by adding role by age interaction terms for each of the social role history categories.

RESULTS

Social role histories and self reported health at age 54

In the unadjusted model (table 2), women in each of the groups lacking one of the main roles of employee, spouse, or mother were significantly more likely to report poor health than women in the Multiple roles group. Long term Homemakers were most at risk of reporting poor health followed by Lone mothers and Childless women. Neither self reported health at age 26 nor mental health between ages 15 and 32 were associated with role histories (self reported health: p = 0.3, mental health: p = 0.7). Table 2 shows that, although both self reported health at age 26 and early mental health were significantly associated with self reported health at age 54, their inclusion influenced relations between role histories and self reported health at age 54 very little. Table 2 also shows that these relations were not confounded by childhood social class.

Neuroticism was also significantly associated with reporting poor health at age 54 (p<0.001), but not with role histories (p = 0.8) and did not influence relations between role histories and self reported health at age 54 (not shown).

Social role histories and obesity at age 53

Table 3 shows that there was significant variation in obesity at age 53 by social role histories. Women with weaker ties to the labour market over the long term were more likely to be obese at age 53 than those with stronger labour market ties. Homemakers were the most likely to be obese (38%) and women who occupied Multiple roles were the least likely (23%).

There was significant role history variation in mean BMI at ages 26, 36, and 53 (age 26: p = 0.03, age 36: p = 0.03, age 53: p = 0.001). Homemakers had the highest mean BMI at these ages (not shown). In models with BMI as a repeated outcome, BMI was modelled to increase linearly with age, and there was a significant age by Homemaker interaction (p = 0.01). The mean BMI of long term Homemakers increased significantly faster with age than it did for women in the Multiple roles group (fig 1).

Table 4 shows that inclusion of BMI at age 26 reduced the odds of being obese at age 53 for Homemakers compared with the Multiple roles group slightly, but did not explain differences between women in the two groups. Findings were similar adjusting for BMI at 15 (not shown). Childhood social class did not influence these relations (table 4).

Nearly a third (32%) of women who had had a major episode, and 29% of those who had had a minor mental health episode between ages 15 and 32 were obese at age 53, compared with under a quarter of women who had had no episode. Early mental health episodes were not significantly associated with role histories (not shown) and did not explain the increased likelihood of being obese at 53 of Homemakers (table 5).

DISCUSSION

Our findings suggest that relations between social roles and health are generally not explained by health selection into employment and parenthood, although some health selection may occur for obesity. Using role information from four points between ages 26 and 53, this study showed that women who occupied multiple roles over the long term reported relatively good health at age 54—a relation that corresponds with previous cross sectional results^{1–3,5,8,9,10,11,12,13,14,15,16}—and this was not explained by early health. Early self reported health, early mental health and neuroticism each significantly predicted reported poor health at age 54, but none of these factors were significantly associated with role histories and none explained the higher risk of reporting poor health at age 54 among women who did not occupy multiple roles over the long term. Also, childhood social class did not influence relations between role histories and health.

In addition, this study has shown an increased likelihood of obesity in mid-life among women with weak long term ties to the labour market compared with women who occupied multiple roles. Examination of BMI from age 15 showed that long term homemakers had a higher mean BMI from age 26, showing some degree of selection out of the labour market in relation to women’s body weight. However, homemakers’ mean BMI increased significantly more with age than that of women with stronger ties to the labour market, and BMI at age 26 did not account for homemakers’ increased likelihood of being obese at age 53, supporting a causation hypothesis alongside any initial selection effect. Recent work in a younger cohort has shown persistent obesity to be associated with never having been gainfully employed, as well as not having a current partner, among women, but not among men.⁶⁰

The fact that this study is limited to women and has been conducted within a specific cohort of women must be taken into account when considering its generalisability. The social environment in which this particular cohort of women formed their families and began employment was highly structured and traditional. British women born in the 1940s had remarkably similar age at marriage patterns, which were the youngest ever recorded since civil registration began. Over the 20th century, the frequency of marriage reached its height in 1960s Britain.⁶¹ In light of this the measure of self reported health at age 26 may not be early enough to pre-date the establishment of family roles in this cohort of women.^61,62 (Most women in this cohort became employed after age 26, A McMunn, unpublished PhD thesis, May 2004). However, our findings showed that neither self reported health at age 26 nor mental health between ages 15 and 32 were associated with social roles histories between ages 26 and 53. The extent to which the results shown here are sex and generation specific given the rapid diversification of social roles in more recent cohorts is an area for future investigation.

In summary, our results suggest that good health is more likely to be the result, rather than the cause, of multiple role occupation. The tendency of studies to focus on role occupancy rather than on the context and quality of particular role situations has been a criticism of research into social roles and health.^7,63 Having established that lifetime social roles are associated with later health among women in this cohort, and that this relation cannot be dismissed as being the result of health selection, research currently underway in this same dataset turns to the question of “why?”. The next step is to better understand what it is about particular work and family roles that influences people’s health.