Inequalities in participation and time spent in moderate-to-vigorous physical activity: a pooled analysis of the cross-sectional health surveys for England 2008, 2012, and 2016

Data came from the Health Survey for England (HSE): this dataset is used to monitor progress on numerous national health objectives, including PA [17, 18]. Details about the HSE sample design and data collection are described elsewhere [19]. Briefly, the HSE annually draws a nationally-representative sample of people living in private households in England using multistage stratified probability sampling with postcode sectors as the primary sampling unit and the Postcode Address File as the household sampling frame. All adults in selected households are eligible for interview. Fieldwork takes place continuously through the year. Trained interviewers measured participants’ height and weight and assessed their demographic characteristics, self-reported health, and health behaviours including PA using computer-assisted personal interviewing. We used the most recent surveys (2008, 2012, 2016) that included the adult Physical Activity and Sedentary Behaviour Assessment Questionnaire (PASBAQ).

The household response rate ranged from 64% in 2008 to 59% in 2016. This study is restricted to adults (i.e. aged 16 years or over). Participants gave verbal consent for interview. Relevant committees granted research ethics approval for the survey. Overall, 31,399 adults participated in the three surveys, of whom 31,183 had valid PA data. Of these, 6301 had missing income data, leaving an analytical sample of 24,882 adults with complete data.

PASBAQ data is used to monitor adherence to UK PA recommendations [17, 18] and for other epidemiological research [20, 21]. The PASBAQ has demonstrated moderate-weak convergent validity in comparison with non-synchronous accelerometry [22]. PASBAQ assesses frequency (number of days in the last 4 weeks) and duration (of an average episode of at least 10 min) in four leisure-time domains [23]:

“Heavy” domestic and manual activities were classed as moderately-intensive. Walking intensity was assessed by a question on usual walking-pace (responses: slow, average, fairly brisk, or fast); moderate-intensity was classed as a fairly brisk or fast pace. Intensity of sports/exercise was determined as indexed in the metabolic equivalent (METs) compendium [24, 25] and a follow-up question on whether the activity had made the participant “out-of-breath or sweaty”.

In addition to leisure-time PA, participants engaged in any paid or unpaid work answer questions on occupational PA. Our analyses classed three activities – walking, climbing stairs or ladders, and lifting, carrying, or moving heavy loads - as moderate-intensity PA for participants working in occupations identified a-priori as moderately-intensive [17].

Time spent in domain-specific MVPA was calculated as the product of frequency and duration, converted from the last 4 weeks to hours/week. For sports/exercise, time in vigorous-intensity activities was multiplied by two when combined with moderate-intensity activities to calculate ‘equivalent’ hours/week as specified in MVPA guidelines [26]. Total MVPA was calculated by summing across the five domains (four leisure-time plus occupational), and was truncated at a maximum of 40 h/week to minimise unrealistic values.

Household income was our chosen marker of socioeconomic position (SEP). The household reference person reports annual gross household income via a showcard (31 bands ranging from ‘less than £520’ to ‘£150,000+’). Household income is equivalised (McClements scale [27]), and grouped into tertiles. Age (in ten-year bands), current smoking (current, ex-regular, never), self-rated health (‘very good/good’, ‘fair’, or ‘bad/very bad’), and BMI were chosen as potential confounders of the SEP and MVPA associations [6]. We computed BMI as weight in kilogrammes (kg) divided by height in metres squared (m²), classifying participants into four groups according to the World Health Organization (WHO) BMI classification [28]: underweight (< 18.5 kg/m²), normal-weight (18.5–24.9 kg/m²), overweight (25.0–29.9 kg/m²), or obese (at least 30.0 kg/m²).

Data was pooled over the three surveys to increase precision (prior analyses revealed no change in associations over time). Differences in age, self-rated health, current smoking, and BMI were estimated by income, using Rao-Scott tests for independence [29]. For total and domain-specific MVPA, we computed descriptive estimates for four outcomes:

Outcomes MVPA and MVPA-active represent unconditional and conditional (on participation) means, respectively. We decided, a-priori, to conduct gender-stratified analyses due to expected differences in inequalities as reported in the literature [7, 8, 30]. Income-specific estimates were directly age-standardised within gender using the pooled data as standard. Pairwise differences between income groups (low-income households as reference) were evaluated on the absolute scale using a linear combination of the coefficients [31].

To handle continuous MVPA data with excess zeros and positive skewness, we used the hurdle model proposed by Cragg, which comprises two parts: a selection/participation model and a latent model [15]. The former determines the boundary points of the continuous outcome (a selection variable equals 1 if not bounded and 0 otherwise), whilst the latter determines its unbounded values (a continuous latent variable which is observed only if the selection variable equals 1). In our analyses, the selection model assessed the influence of income on the binary outcome of participation (any versus none), whilst the latent model assessed its influence on the amount of time spent active, conditional on participation (MVPA-active). We specified a probit model for the former and an exponential form for the latter. Each model contained income (as a three-category variable) and the confounders listed above.

Based on the model estimates, three sets of marginal means by income were calculated, evaluated at fixed values of the confounders. These sets correspond to different definitions of the expected value of MVPA [32]: (i) the probability of doing any, (ii) the average hours/week MVPA for all participants (the unconditional mean), including those who did none; and (iii) the average hours/week MVPA conditional on participation (MVPA-active). Inequalities after confounder adjustment (average marginal effects: AMEs) were quantified by computing the absolute difference in the marginal means (low-income as reference).

Dataset preparation and analysis was performed in SPSS V20.0 (SPSS IBM Inc., Chicago, Illinois, USA) and Stata V15.0 (College Station, Texas, USA), respectively. All analyses accounted for the complex survey design by applying sample weights (including correction for non-response) and incorporating the clustering of participants in postcode sectors (the primary sampling unit in the HSE series) via the “svy” package in Stata. HSE datasets are available via the UK Data Service (http://​www.​ukdataservice.​ac.​uk) [33‐35]; statistical code is available from the corresponding author.

Information on confounders by income is presented in Additional file 1. Poorer self-rated health and higher smoking levels were evident among adults in low-income households (both P < 0.001). BMI status also varied by income (P < 0.001 for both genders), with higher obesity levels especially among women in low-income households (Additional file 1).

Tables 1 and 2 show the descriptive estimates for total and domain-specific MVPA for all adults and by income for men and women, respectively. Overall, 85% of men (n = 9254) and 81% of women (n = 10,947) did any MVPA; 66% of men (n = 7120) and 56% of women (n = 7537) were ‘sufficiently’ active (Table 1 men; Table 2 women). Men and women spent on average 9.7 and 6.8 h/week respectively in total MVPA (Table 1 men; Table 2 women); however, these distributions showed excessive zeros and positive skewness (Fig. 1). Among those doing any MVPA, men and women spent on average 11.5 and 8.4 h/week respectively in total MVPA (Table 1 men; Table 2 women). The largest difference between MVPA and MVPA-active means was for occupational PA; among men, these were 2.5 and 15.2 h/week respectively (Table 1 men; Table 2 women).

Inequalities were evident in descriptive analyses in each aspect for total MVPA and for sports/exercise. Differences between high-income versus low-income households in total MVPA were 2.2 h/week among men (95% CI: 1.7, 2.8; P < 0.001) and 1.8 h/week among women (95% CI: 1.3, 2.2; P < 0.001); the same pattern, but with narrower effect sizes, was found for total MVPA-active (men: 0.9 h/week, 95% CI: 0.3, 1.6; P = 0.004; women: 1.0 h/week, 95% CI: 0.6, 1.5; P < 0.001) (Table 1 men; Table 2 women). Likewise, differences in sports/exercise MVPA (i.e. including those who did none) for men and women in high-income versus low-income households were 1.9 h/week (95% CI: 1.6, 2.2; P < 0.001) and 1.5 h/week (95% CI: 1.3, 1.7; P < 0.001), respectively (Table 1 men; Table 2 women). Differences in sports/exercise MVPA-active were 1.2 h/week among men (95% CI: 0.7, 1.7; P < 0.001) and 1.1 h/week (95% CI: 0.7, 1.5; P < 0.001) among women (Table 1 men; Table 2 women).

Results for other domains were heterogeneous. Inequalities were evident in the unconditional outcomes (any; sufficient activity; MVPA) for walking, yet the time spent walking amongst those who did any walking was higher among men in low-income versus high-income households (levels were similar by income among women). Men in high-income versus low-income households did less occupational PA (MVPA: P = 0.021; MVPA-active: P = 0.019); whilst men (MVPA-active: P < 0.001) and women (P < 0.001 for MVPA and MVPA-active) in high-income households did less domestic activity (Table 1 men; Table 2 women).

Table 3 shows the AMEs from estimated hurdle models corresponding to the absolute difference in the marginal means for the binary outome of participation, and the continuous outcomes of MVPA and MVPA-active (AMEs are graphically shown in Fig. 2).

Higher MVPA in high-income versus low-income households was robust to confounder adjustment for total MVPA and for sports/exercise (P < 0.001 for all outcomes and both genders; except P = 0.003 for sports/exercise MVPA-active among men) (Table 3). For example, at fixed values of the confounding variables, differences between high-income versus low-income households in sports/exercise MVPA were 2.2 h/week among men (95% CI: 1.6, 2.8) and 1.7 h/week among women (95% CI: 1.3, 2.1); differences in sports/exercise MVPA-active were 1.3 h/week (95% CI: 0.4, 2.1) and 1.0 h/week (95% CI: 0.5, 1.6) for men and women, respectively (Table 3).

Heterogeneity in associations was observed for other domains. Participants in high-income versus low-income households were more likely to do any walking (men: 13.0% (95% CI: 10.3, 15.8%); women: 10.2% (95% CI: 7.6, 12.8%)). Among all adults (including those who did no walking), the average hours/week spent walking showed no difference by income. Among those who did any walking, adults in high-income versus low-income households walked on average 1 h/week less (men: − 0.9 h/week (95% CI: − 1.7, − 0.2); women: − 1.0 h/week (95% CI: − 1.7, − 0.2)) (Table 3).

Women in high-income versus low-income households were less likely to do any (− 4.0%; 95% CI: − 6.0, − 1.9%, P < 0.001) and spent less time in domestic activity (P < 0.001 for MVPA and MVPA-active) (Table 3). Lower levels of occupational PA for men in high-income versus low-income households were robust to confounder adjustment (P = 0.001 and P < 0.001 for MVPA and MVPA-active) (Table 3).

Differences in financial resources (especially for sports/exercise) [5] [39], health status [40], psychological or cultural characteristics [40, 41], and the built environment [40, 42, 43], including those driving inequalities in access to highly walkable neighbourhoods [44, 45], are key determinants of inequalities in physical activity. Reducing the inequalities presented here for sports/exercise will require policy actions and interventions to move adults in low-income households from inactivity to activity, and to enable those already active to do more. For example, removing user charges from leisure facilities in northwest England has had some success in increasing overall activity levels and in reducing inequalities [46]. Having world-class sports facilities that are free for anyone to use – as is the case in several Latin American cities – would reduce inequalities [47]. In contrast, our results suggest that interventions to promote walking should focus on reducing the sizeable income gap in the propensity to do any walking; such interventions could positively impact PA levels and reduce inequalities through increasing activity in the most sedentary [48] as well as the elderly and those in poorer health. A recent systematic review and meta-analysis [49] examined the effectiveness of interventions such as individual counselling [50], group training sessions [50, 51] and behavioural informatics [52] that were targeted at changing physical activity behaviour among low-income adults. The results showed a small positive intervention effect among those that focused on PA only as opposed to those targeting multiple behaviours [49]. However, evidence suggests that PA interventions are less effective in low-income groups, potentially widening rather than reducing inequalities [49]. Worryingly, a recent systematic review identified that there is insufficient evidence to allow for firm conclusions to be made regarding the impact of PA interventions on inequalities [53]. According to the WHO, effective national action to reduce disparities in PA requires a strategic combination of population-based policy actions aimed at tackling the “upstream” determinants that shape the equity of opportunities for participation (such as encouraging non-motorised modes of travel through improved provision of cycling and walking infrastructure, improved road safety, and creating more opportunities for PA in public open spaces and local community settings [54]) and those policy actions that are focused on “downstream” individually-focused (educational and informational) interventions, implemented in ways consistent with the principle of proportional universality (i.e. greatest efforts directed towards those least active) [55].

Our analyses used novel modelling methods to assess inequalities in MVPA. Although it is well-known that MVPA distributions typically contain excess zeros and positive skewness, no epidemiological studies to date have applied hurdle models to assess inequalities. Such models avoid the loss of information and power that occurs when practitioners typically categorise a continuous variable into a binary or ordinal variable [9]. Precision of our estimates was increased by pooling standardised PA data across survey years. Caution is required, however, when interpreting our findings. First, self-reported PA data has well-known limitations such as recall and reporting (social desirability) bias [56, 57]. Secondly, the dataset contained a sizeable amount of missing data for income and BMI (~ 20%); among HSE participants, the probability of having missing income data varies systematically across groups [58], which we minimised to some extent through applying non-response weights. The software routine for estimating hurdle models does not currently permit multiply imputed data, and so our findings may be statistically underpowered to some extent. Thirdly, the choice of potential confounders was limited to some extent by data availability; furthermore, we were unable to account for ethnic differences due to small numbers. As in all studies, our findings could have been influenced by unmeasured confounders. Fourthly, our findings are contingent upon HSE data collection, including the minimum duration of 10 min (in accord with the contemporaneous UK guidance but differing from recent UK [59] and US [60] guidelines, which acknowledge that PA of any duration enhances health), a specific subset of occupational PA for a selected group of occupations, and the inability to distinguish between walking for leisure and active travel. We acknowledge that different definitions may have led to different conclusions. Finally, we cannot draw causal inferences, as this was a descriptive study based on cross-sectional data.

As mentioned previously, more evidence on the equitable impact of PA interventions is needed to ascertain ‘what works’ best to increase PA levels among low-income groups [49]. Given the aforementioned limitations of cross-sectional data on self-reported PA collected within large-scale national health examination surveys, it is imperative that innovative studies such as those using smartphones with built-in accelerometry to measure PA on a global scale [61] be used to shed light on inequalities and their interaction with aspects of the built environment such as walkability. However, maximising the potential for such research to inform policy-makers and practitioners will require efforts to minimise the potential bias of such data towards younger, more affluent, and more active populations [61]. Finally, as emphasised in this study, whatever the source of data, separate model equations should be used to assess inequalities in participation and in duration.