Low levels of leisure-time physical activity are associated with an increased risk of many chronic diseases including coronary heart disease, diabetes mellitus, some cancers, and osteoporosis.1 2 The prevalence of physical activity at recommended levels is low, particularly in older women.3 Reducing the number of people engaging in low levels of physical activity would lead to significant health benefits.4 Great benefit could be obtained by promoting physical activity in the least active individuals who are therefore most at risk of adverse health outcomes. Promoting physical activity in older individuals who are among the most sedentary in society is now generally acknowledged as being beneficial.5^–8

Behavioural risk factors in adulthood, including physical activity, vary according to socioeconomic position.3 9 10 Furthermore, there is increasing evidence that for disease outcomes associated with low levels of physical activity, in particular for cardiovascular disease, low socioeconomic position across the life course acts cumulatively to increase risk.11 As far as we are aware, few studies have examined the association between childhood socioeconomic position and adult physical activity.12^–15 Existing studies have used just one or two indicators of childhood and adult socioeconomic position, which are unlikely to capture fully the complete effect of socioeconomic inequalities.

Furthermore, the role of socioeconomic position in health outcomes in older adults is contested. In one study of Japanese adults,16 a “cross-over” in the effect of socioeconomic position with those who were least educated living longer among older adults was observed. The authors argued that overall, age was a leveller of social differences because biological processes associated with ageing assume dominance over social determinants.16 By contrast, findings from the United Kingdom suggest that socioeconomic inequalities in health do persist into old age.17 18 In addition to a possible levelling effect of age-related biological processes, it has also been suggested that greater welfare provision with increasing age may reduce socioeconomic inequalities in health outcomes, although there is evidence that the welfare provision for people aged 65 years and over may be insufficient to cover the minimum income for healthy living (including physical activity).8

Independent of personal circumstances, living in a deprived residential area may be related to low physical activity,19^–21 but studies reporting this have tended to take account of just one or two indicators of individual socioeconomic position. No studies have simultaneously evaluated the effect of a range of individual social circumstances throughout the life course as well as area-level deprivation. Understanding whether socioeconomic position in general is associated with physical activity in older individuals, whether childhood or adulthood socioeconomic position are the more important determinants of older adult levels of physical activity and whether there is an independent effect of area-level deprivation over and above individual socioeconomic position is important for developing health and social policy.

This study aims to assess the association between individual life-course socioeconomic position (comparing childhood and adulthood) and adult levels of physical activity, and to examine whether residential area-level deprivation, beyond the effect of individual life-course socioeconomic position, is associated with levels of physical activity in older British women.

METHODS

Participants

Data from the British Women’s Heart and Health Study were used; full details of the selection of participants and measurements have previously been reported.11 22 Women aged 60–79 years were randomly selected from general practitioner lists in 23 British towns. A total of 4286 women participated and baseline data (self-completed questionnaire, research nurse interview, physical examination and primary care medical record review) were collected between April 1999 and March 2001.

Physical activity

At baseline participants were asked to indicate their usual duration of activity in hours per week for several types of activity as well as their usual walking pace (“slow”, “steady average”, “fairly brisk” or “fast”).23 This information was used to calculate the hours of moderate or vigorous intensity physical activity per typical week in winter and summer. Activities that were considered to be of moderate or vigorous intensity were: walking for those who indicated that their usual pace was fairly brisk or fast, cycling, heavy gardening and physical exercise (aerobics, swimming, jogging, tennis, etc). Weekly hours spent in moderate/vigorous activity in winter and summer were highly correlated (0.97), so the mean of the two values was used in all analyses.

Socioeconomic position

Details of the longest held occupation of the participant’s father, husband and her own longest held occupation were requested in the self-completed questionnaire. For the participant’s father the women were asked “What job did your father do for the longest period of time?” (ie they were not specifically asked about their father’s occupation when they were a child). Adult social class was derived from the longest held occupation of the participant’s husband for married women and her own for single women. Childhood social class was derived from the longest held occupation of the participant’s father. Social class was categorised into social class I—professional to social class V—unskilled manual occupations; these are also grouped into two broad categories of manual and non-manual social classes using the Registrar General’s Standard Classification.24 Other indicators of childhood socioeconomic position were self-reported childhood household amenities (living in a house with a bathroom, living in a house with a hot water supply as a child, sharing a bedroom) all indicators of material circumstances,25 family access to a car as a child and age at leaving full-time education. For the childhood household amenities and car access the women were asked about these for the longest period of their childhood; they were not asked about a particular age. For example, they were asked “As a child, did the home you lived in longest have a bathroom?” Indicators of adult socioeconomic position were housing tenure (social housing, private rented, owner-occupied, other), car ownership and pension arrangements (state only, state and occupational, state and personal, other).

The geographical area used in this analysis was the electoral ward in which the women lived at the time of the baseline assessment. The postcode for each woman was used to locate her residence at the time of interview and these postcodes were mapped to electoral wards for each woman. The mean population of wards in Great Britain is 5700, but with a range from 100 to 33 000; they cover, on average, an area of 16 km². The Carstairs’ deprivation score (based on 2001 census data) was obtained for all wards in Great Britain (England, Scotland and Wales) from the census electronic files that are available to academic institutions (http://census.ac.uk). These scores were then mapped to residential ward codes for each woman. The Carstairs’ deprivation score is based on four variables derived from census data: male unemployment, household overcrowding, car ownership and the proportion of households in social classes IV and V.26 Weights applied to each of these four variables are determined by the proportion in the British population with each variable, and a Z score for the whole population is then calculated.26 A ward score of zero thus indicates that a ward has socioeconomic circumstances that are similar to the mean for the whole of Great Britain, a negative score indicates greater affluence compared with the average of Great Britain and a positive score indicates greater disadvantage.

Covariables

We considered that body mass index, smoking and pre-existing disease might confound any association between socioeconomic position and activity. Height (without shoes to the nearest millimetre) and weight (in light clothing and without shoes, to the nearest 0.1 kg) were assessed using standard procedures and were used to calculate body mass index (kg/m²). Smoking was categorised, as per the original study protocol, as never, ex, current (including those who had given up smoking in the past 6 months). A history of cardiovascular disease (angina, previous history of myocardial infarction or stroke) and respiratory disease (chronic obstructive pulmonary disease or asthma) was obtained from medical record reviews, research interviews and self-completed questionnaires as previously reported.11 22

Statistical analysis

We generated a life-course socioeconomic position score from the 10 indicators. To generate the life-course socioeconomic position score we dichotomised those indicators that were not binary as follows: adult and childhood social class into non-manual (I, II, III non-manual) and manual (III manual, IV, V); pension arrangements into state only or state plus other (employment or private pension); adult housing tenure into local authority (social housing) or other (owner occupied, private rental, living with a relative); and age at leaving full-time education into those leaving school at or younger than 15 years, or above that age. Two scores were developed, one in which equal weight was given to each indicator and another in which the inverse of prevalence weights was used. The first score has the advantage of being easy to understand because the score gives the actual number of adverse indicators. The score ranged from 0 (most advantaged position across the life course) to 10 (most disadvantaged position across the life course). Because there were small numbers in the 0 category and in the 10 category, the 0 category was combined with the 1 category and the 10 category with the 9 category. The second score, in which each indicator was weighted by the inverse of its prevalence, gave greatest weight to adverse indicators that were least prevalent and as such may be thought of as being more severe indicators of adverse socioeconomic position (for example, overall just 29% of the sample have no car access in adulthood and therefore this indicator is given more weight as an indicator of adverse socioeconomic position than having no access to a car in childhood, which was the situation for the majority (83%)). The resulting weighted score was highly positively skewed with a range from 0 to 28.9. We generated similar scores (both weighted and unweighted) for adult socioeconomic position (score from 0 to 4 or 5 for unweighted score and 0 to 9.66 for weighted score) and for childhood socioeconomic position (score from 0 to 4 or 5 and 0 to 24.5 for weighted score). The analyses by unweighted score did not differ substantively from those using the weighted score and therefore for the main analyses presented here we have used the unweighted scores, which are easier to interpret. We included years of completed education with the adult socioeconomic position score for the main analyses presented here. Analyses were repeated with education in the childhood score, or omitting education from either score, and the results were not substantively different to those presented here.

Weekly hours of moderate/vigorous activity was not normally distributed, having a range from 0 to 58 with a median and mode of 0 and interquartile range of 0 to 3. To maximise efficiency and test the hypothesis that socioeconomic position was related to levels of physical activity across its distribution in older women we used ordinal logistic regression. Hours of physical activity was split into four categories: 0; less than 2; 2 to 3.99; 4 or more and this four-level categorical variable was the outcome in the regression models. Proportional odds ordinal logistic regression was undertaken in which the parameter represents the combined odds ratio from the three possible hierarchical comparisons (ie 0 hours versus all others; 0 or less than 2 hours versus the other two categories; 0, less than 2, 2 to 3.99 hours, versus the highest level category). Formal tests of difference between effect estimates from these three comparisons did not show variation for any of the models. The resultant combined odds ratio from the ordinal logistic regression is interpreted as the effect per category of exposure (socioeconomic position measure) of being more active compared with less active. An odds ratio of less than one per unit of increasing socioeconomic position disadvantage indicates that more adverse socioeconomic circumstances are associated with less activity. To account for the clustered nature of the data and possible non-independence of women living in the same ward we used robust standard errors (using area ward as the clustering variable) in all analyses.

There were small amounts of missing data on some of the variables included in the analyses. For most variables this was less than 5%, but 12% of the women had missing data for the variable concerned with income arrangements on retirement. In the descriptive statistics (tables 1 and 2) each result is based only on women with complete data for the particular variable. In the multivariable analyses we used multiple multivariate imputation, using all other covariates, to impute a distribution of missing values for those women with some missing data.27 We used switching regression in Stata as described by Royston,27 and carried out 20 cycles of regression switching, and generated 10 imputation datasets. There was no evidence of statistical heterogeneity between the datasets generated and the findings from this multivariate multiple imputations analysis were essentially the same as those from the complete case analysis. All analyses were undertaken using Stata version 9.0 (Stata Corp, College Station, Texas, USA).

RESULTS

Socioeconomic circumstances generally improved from childhood to adulthood. For example, whereas 77% of the women reported occupations for their fathers that belonged to manual social classes, occupational social class based on their husbands’ or their own occupation in adulthood placed 53% in manual social classes. Similarly, 83% reported that their family had no access to a car when they were a child, compared with just 29% reporting no car access as an adult. Despite this, the associations of childhood and adulthood socioeconomic position indicators based on the same indicators showed similar associations with physical activity. For example, the age-adjusted odds of being more active comparing those without access to a car in childhood was 0.69 (0.59 to 0.81) and comparing those without access to a car in adulthood was 0.64 (0.55 to 0.73), the p value for difference in these associations was 0.87. These similarities of effect explain why the weighted and unweighted socioeconomic position scores (see Statistical analysis section above) gave similar results. Study participants resided in 457 electoral wards. The wards in which the women lived were those in the middle of the distribution of all wards for Great Britain, with a mean (range) population size of 6889 (1753 to 15 372). Ward-level Carstairs’ scores for study participants were positively skewed and ranged from −5.13 to 18.68 with a median (interquartile range) of −0.11 (−2.22 to 2.34). Of the 4103 women with physical activity data, 2186 (53%) did not take any moderate or vigorous physical activity, 488 (12%) took up to 1.99 hours per week, 503 (12%) took between 2 and 2.99 hours per week, and 926 (23%) took 3 or more hours per week.

Table 1 shows the age-adjusted prevalence of each indicator of socioeconomic position by physical activity categories. For all indicators of socioeconomic position from across the life course the proportion with the most adverse socioeconomic position was greatest in those who were least active. There was a lower proportion of women in the most adverse socioeconomic groups as the level of physical activity increased. Similarly, median area-level deprivation was greater in those undertaking higher levels of physical activity. Greater levels of physical activity were also associated with a lower prevalence of cardiovascular and respiratory disease. Current smoking prevalence and mean body mass index were lower in those who reported greater levels of activity.

There was evidence of marked geographical variations in physical inactivity. Across the 23 towns the prevalence of inactivity varied from 25.2% (95% CI 19.6% to 31.6%) in Ipswich in the south of England to 43.3% (95% CI 35.7% to 51.1%) in Merthyr Tydfil in Wales. In general, towns in the south east of England had the lowest prevalence of inactivity and those in Scotland and Wales had the highest. Table 2 shows the prevalence of inactivity and other characteristics of the women across fifths of ward-level Carstairs’ scores. The prevalence of physical inactivity increased linearly with worsening area-level deprivation and each individual indicator of socioeconomic position from across the life course was also strongly associated with area-level deprivation.

Table 3 shows the multivariable associations of childhood, adult and total life-course socioeconomic position and ward-level deprivation with weekly hours of activity. In age-adjusted analyses, all three indicators of childhood and adulthood socioeconomic position and area-level deprivation were inversely associated with physical activity. With mutual adjustment for the socioeconomic position/deprivation scores (model 2), the associations attenuated, but each of childhood and adult socioeconomic position and area deprivation remained independently associated with physical activity. Further adjustment for adult covariables had little effect on the association of childhood socioeconomic position but some attenuation of the associations of adult socioeconomic position and area deprivation. In the final model, all three of childhood and adult socioeconomic position indicators and area deprivation remained independently associated with physical activity.

There was a cumulative effect of life-course socioeconomic position on physical inactivity, with the proportion who undertook no moderate or vigorous activity per week increasing linearly with each additional indicator of life-course socioeconomic position; p<0.001 for linear trend (see fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Prevalence (%) of women reporting no moderate or vigorous physical activity by the number of individual-level adverse socioeconomic position indicators from across the life course.

DISCUSSION

Among older British women, we found that those from poorer socioeconomic circumstances in childhood, adulthood, and who live in areas with greater deprivation spend fewer hours per week in moderate or vigorous activity. These effects were independent of each other and also of potential confounding factors, suggesting that their individual effects are not completely explained by the associations of life-course socioeconomic position with adult disease, which might in turn then reduce the ability to remain active.

Study strengths and limitations

A strength of this study is that, to our knowledge, it is the first to assess the effect of a range of childhood and adult socioeconomic position indicators, together with adult area deprivation, on levels of physical activity. We have relied on self-report of characteristics of individual life-course socioeconomic position, and there may be some misclassification bias for these covariables. Self-report of childhood socioeconomic circumstances in particular may be affected by reporting bias.28 It is, however, unlikely that recall inaccuracy of socioeconomic position would be affected by physical activity, and therefore any misclassification would most likely be non-differential; the expectation resulting from non-differential misclassification is underestimation, but any observed result may be biased in either direction. For the self-report of childhood household amenities and car access we asked participants about these during the longest period of their childhood rather than at a specific age or age range. The advantage of this is that these will provide a better indicator of exposure over most of the childhood period and are likely to be more accurately reported than asking adults retrospectively about a particular age or age period. By contrast, for father’s occupation we did not specify “childhood”, but simply asked about the longest job that their father had held. This may thus refer to a job that their father held when the participant was in adulthood rather than childhood. The percentage of participants whose fathers were in manual social classes is more consistent with pre-World War II distributions for Great Britain than distributions from the 1960s–80s, and therefore suggests that on average we were obtaining a valid measure of childhood occupational social class. Furthermore, for other outcomes (eg coronary heart disease) the results of the association of father’s occupational social class obtained retrospectively and based on the father’s longest held occupation in the British Women’s Heart and Health Study29 are similar to those from a study in which father’s occupational social class was obtained prospectively in childhood,30 suggesting that our measure of childhood social class is relatively accurate.

Physical activity was also self-reported, which may lead to some misclassification. If this misclassification was purely random then the expected association with socioeconomic position would be attenuated, although our actual observed association may be biased in either direction. If there was reporting bias and, for example, participants who overreported physical activity were from a higher socioeconomic position, the observed association would be inflated. The magnitude of this potential misclassification bias is unclear.

This study is cross-sectional; therefore, two potential limitations are reverse causality and survival bias. For childhood socioeconomic position reverse causality is not possible. For adult socioeconomic position and area deprivation, it seems implausible that becoming physically inactive would lead to a fall in socioeconomic position. Survivor bias is a possibility because individuals originating from the lower socioeconomic position groups are more likely to have died at a younger age, so it is possible that women alive in their 60s and 70s, who are in the lower socioeconomic position groups, represent the fittest individuals within that group. This would have the tendency to result in an underestimation of the true association of socioeconomic position with physical activity. Nonetheless, further prospective studies of the role of life-course socioeconomic position in physical activity levels in older individuals would be useful to confirm our findings. We were only able to examine area-level deprivation in adulthood with physical activity and could not examine childhood or adolescent area deprivation with the outcome because such measures were not determined for Great Britain in the 1930s–50s, when the study population were children. We used Carstair’s measure of area-level deprivation because this has been derived for the whole of Great Britain. We have, however, obtained other available measures of area deprivation for this study sample. These other measures are strongly correlated with the Carstair’s score (Townsend–Carstair’s correlation 0.87 and Index of Multiple Deprivation–Carstair’s correlation 0.83) and therefore our results are likely to be very similar if we had used one of these alternative measures of area-level deprivation. This study does not have data on known psychosocial correlates of physical activity (eg social capital) that may explain the relationship between socioeconomic position and physical activity. Finally, our study is in women and the results are not necessarily generalisable to men.

Comparisons with other studies and implications of findings

Studies of the life-course effect of socioeconomic position on adult physical activity in older women are sparse. A study of the Whitehall II cohort found women’s adult socioeconomic position to be associated with physical activity but there was no association with childhood socioeconomic position.13 Similar observations have been reported for men. In a prospective cohort study of over 7000 men, those whose fathers were in manual social classes in childhood were less likely to be physically active in adulthood, although this difference attenuated when the participants’ own social class in adulthood was controlled for.12 31 Baseline results from a prospective cohort study of 25–74 year olds, recruited in the Netherlands in 1991, show that childhood socioeconomic position is associated with physical activity after adjustment for the current socioeconomic position in women but not men.32 In a 1966 Finnish birth cohort, adolescent social class was associated with adolescent physical activity but not physical activity at the age of 31 years.14 A United Kingdom birth cohort of 3500 men and women, found that women who played sport at the age of 36 years were better educated, had better educated mothers, had fewer childhood health problems and were good at school sports compared with less active women.33 Most of those studies based childhood socioeconomic position on father’s social class and typically only personal social class and educational attainment were used to capture adult socioeconomic position. As mentioned earlier, such analyses may not fully account for all aspects of individual-level socioeconomic position. Furthermore, differences in the study population, particularly the period of childhood and the measure of physical activity, may explain some of the differences between our findings and those of other studies.

The results of this study suggest that older women are less physically active if they had a poor start in life, even if their personal circumstances improve or they live in a more affluent neighbourhood as adults. Women from a low childhood socioeconomic position may have had less active parents and report receiving less encouragement to be active from their parents compared with high socioeconomic position women.34 In addition, women of low socioeconomic position parents report childhood physical activity as a necessity (for transport) rather than for pleasure, whereas women from high socioeconomic position parents report enjoyment from more organised physical activities and sports.34 Such early negative experiences and lack of support may be compounded by less physical ability to be active.

Compared with low socioeconomic position women, high socioeconomic position women in adulthood in this study reported being physically active even if they did have a poor start in life and currently live in a deprived neighbourhood. Barriers to physical activity such as poor health, lack of money and lack of transport are less frequently reported by men and women in high socioeconomic position groups compared with low socioeconomic positions.35 There is also some evidence that compared with low socioeconomic position women, high socioeconomic position women set aside specific time for organised physical activity indicating a more social role for physical activity.34 Furthermore, compared with high socioeconomic position women, low socioeconomic position women may value sedentary behaviours such as TV viewing more.34

Living in an affluent rather than a deprived neighbourhood was associated with a higher level of physical activity in women, even if both childhood and adult socioeconomic positions were low.

Area socioeconomic position may be a sign of access to physical activity resources,36 37 social norm towards physical activity or perceptions of personal safety. It may also be true that women with a propensity to be physically active selectively migrate into supportive environments.38

The cumulative effect of low socioeconomic position throughout the life course suggests that the different indicators of socioeconomic position influence physical activity through different pathways, of which we know little. Reducing inequalities across the life course may have a positive effect on physical activity levels in older women. Further research is required to understand why women who experience adverse socioeconomic circumstances report low levels of physical activity.

CONCLUSIONS

Adverse socioeconomic position across the life course is associated with an increased cumulative risk of low physical activity in older women. Further studies are required to understand what underlies physical activity inequalities.