Weekend and weekday associations between the residential built environment and physical activity: Findings from the ENABLE London study.

BACKGROUND: We assessed whether the residential built environment was associated with physical activity (PA) differently on weekdays and weekends, and contributed to socio-economic differences in PA.
METHODS: Measures of PA and walkability, park proximity and public transport accessibility were derived for baseline participants (n = 1,064) of the Examining Neighbourhood Activities in Built Living Environments in London (ENABLE London) Study. Multilevel-linear-regressions examined associations between weekend and weekday steps and Moderate to Vigorous PA (MVPA), residential built environment factors, and housing tenure status as a proxy for socio-economic position.
RESULTS: A one-unit decrease in walkability was associated with 135 (95% CI [28; 242]) fewer steps and 1.2 (95% CI [0.3; 2.1]) fewer minutes of MVPA on weekend days, compared with little difference in steps and minutes of MVPA observed on weekdays. A 1km-increase in distance to the nearest local park was associated with 597 (95% CI [161; 1032]) more steps and 4.7 (95% CI [1.2; 8.2]) more minutes of MVPA on weekend days; 84 fewer steps (95% CI [-253;420]) and 0.3 fewer minutes of MVPA (95%CI [-2.3, 3.0]) on weekdays. Lower public transport accessibility was associated with increased steps on a weekday (767 steps, 95%CI [-13,1546]) compared with fewer steps on weekend days (608 fewer steps, 95% CI [-44, 1658]). None of the associations between built environment factors and PA on either weekend or weekdays were modified by socio-economic status. However, socio-economic differences in PA related moderately to socio-economic disparities in PA-promoting features of the residential neighbourhood.
CONCLUSIONS: The residential built environment is associated with PA differently at weekends and on weekdays, and contributes moderately to socio-economic differences in PA.

Introduction

Physical activity (PA) is a protective factor for a wide range of physical and psychological disorders [1-3]. Current population levels of PA are too low in the UK, with 37% of adults aged 16 years or more not meeting recommended levels of activity of at least 150 minutes of moderate-intensity per week [4]. Individuals from more disadvantaged socio-economic groups are less physically active [4], both in terms of daily steps [5] and moderate to vigorous physical activity (MVPA) [6, 7]. Recently, epidemiological research has increasingly incorporated socio-ecological models that acknowledge the role of the built environment, especially the local residential environment [8], in influencing PA behaviours [9]. The extent to which the residential built environment contributes to individual socio-economic differences in PA has been explored by drawing on two hypothetical pathways: (i) a deprivation-amplification effect [10], whereby disadvantaged individuals are less exposed to health-promoting facilities in their residential neighbourhood, and (ii) a moderating effect, by which socio-economic groups use the physical activity facilities available in their neighbourhood differently [11]. Regarding the deprivation-amplification pathway, some UK studies have reported that the most affluent urban areas have the poorest accessibility to recreational PA facilities [12-15], but others the most deprived [16]. As for the moderating effect pathway, some studies have reported that the association between the residential built environment and PA was moderated by the socio-economic position [17, 18], whereas others did not find such evidence [19, 20]. These mixed findings likely reflect diverse local realities, that may owe to local population specificities or regional policy interventions. More research is needed to depict a comprehensive, overarching view of how built environment, socio-economic status and PA behaviours interrelate, and, in turn, deliver effective ‘contextually sensitive’ policy interventions [21].

Residential neighbourhood characteristics positively associated with PA include increased walkability [22], better access to greenspace [23, 24] and public transport [23]. However, these UK findings rely on a limited number of studies where the evidence is mixed. For instance, a study from Combes et al. [24] shows a decrease in odds of achieving PA recommendations associated with increasing distance to green space from home adult (16+ years) residents in Bristol (UK). Yet, Foster et al. found a null association between proximity to the nearest public open/green space and PA levels in adults aged 45–74 years from the European Investigation into Cancer study (EPIC) Norfolk study [25]. More work is required to establish whether specific features of the residential built environment shape PA behaviours in the UK, and whether these features might help explain socio-economic differences in PA.

Associations between residential PA facilities and PA have been shown to vary by time of the day and day of the week [26]. Daily variations in number of steps and minutes of MVPA have been reported in English adults, with more PA on weekdays compared to weekends [5, 27]. UK time use surveys highlight how individuals spend more time travelling and in paid work or studying on weekdays, and more time in leisure activities and domestic work on weekends [28, 29]. Use of the residential built environment may therefore be expected to be different on weekdays and weekends. This suggests we should also consider weekday and weekend variations in PA when seeking to better understand the contribution of the residential built environment to PA and the role of socio-economic status within it.

Aims

The aim of this London (UK) based study is twofold. First, to assess whether the residential built environment is associated with the number of daily steps, and the amount of daily MVPA (min) accumulated, on weekdays and weekend days. Second, to explore two pathways in which the built environment may contribute to household-level socio-economic differences in PA levels on weekdays and weekend days: (i) a moderating effect, whereby different socio-economic groups relate differently to the physical activity facilities available in their neighbourhood; (ii) a deprivation-amplification effect [10], by which reduced access to PA-promoting facilities for disadvantaged individuals translate into lower levels of PA.

Methods

Study design and participants

The Examining Neighbourhood Activities in Built Living Environments in London (ENABLE London) study is a controlled before and after natural experiment aimed at establishing whether PA behaviours improve among individuals relocating to East Village [30]. East Village, formerly the London 2012 Olympics Athletes’ Village, is a planned mixed-use residential neighbourhood development designed to both encourage active living and promote social mixing through three types of housing tenure: “social”, “intermediate” (including affordable rent, shared ownership and shared equity), and “market-rent”. In this paper we use baseline measurements of PA and sociodemographic data from the ENABLE London study on adults prior to any relocation into East Village.

Participants were recruited from those seeking “social”, “intermediate” and “market-rent” accommodation in East Village. The sample consisted of 1,278 individuals aged 16 years and over clustered in 1,006 households, with response rates of 52%, 57% and 58% for social, intermediate and market-rent housing seekers, respectively. Baseline assessments were carried out at the participant’s place of residence between January 2013 and December 2015, prior to any potential move to East Village. Participants were asked to complete a questionnaire and to wear a hip-mounted accelerometer (ActiGraph GT3X+) for one week, that they were to remove only when in water or sleeping. Full ethical approval was obtained from the City Road and Hampstead Ethical Review Board (REC Reference 12/LO/1031). All participants provided written informed consent.

Of the 1,278 study participants, we excluded those who lived outside Greater London and without sufficient accelerometer data (see flow diagram of participation in Fig 1). The 214 individuals who were excluded from the analytical sample had similar characteristics to the 1,064 who were included, in terms of housing status (Pearson Chi-square, p = 0.172), age (Pearson Chi-square, p = 0.273), sex (Pearson Chi-square, p = 0.06) and ethnicity (Pearson Chi-square, p = 0.181) (S1 Table). Of the 1,064 participants clustered in 842 households retained for analyses, 442 were seeking relocation into a social, 436 into intermediate, and 186 into market-rent accommodation; 1,053 participants had PA data on weekdays, 848 on weekends, and 837 on both weekdays and weekends.

Fig 1

Flow diagram of participation.

Variables

Accelerometer-derived PA outcomes

Participants were asked to wear a hip-mounted ActiGraph GT3X+ accelerometer for 7 consecutive days during waking hours. This provided daily measures of steps and time spent in moderate-to-vigorous physical activity (MVPA), based on the standard threshold of ≥ 1952 counts per minute [31]. We excluded days of accelerometer data where the registered wear time was less than 540 minutes. Multilevel linear regression models were fitted to allow for repeated measurements of PA, level 1 was day within individual and level 2 was individual. Daily steps (minutes of MVPA) was regressed on day-order-of-wear, day-of-week and month-of-wear, and an unbiased estimate of average daily steps (minutes of MVPA) obtained for each participant for weekdays and weekend days separately [5].

Environmental variables

Participants were geocoded to the centroid of the footprint of their building of residence, using Ordnance Survey (OS) AddressBase Premium (March 2015) to match participants’ declared residential addresses with XY coordinates. Residential locations were then used to derive a range of built environment factors hypothesised to be associated with PA, as follows.

Residential neighbourhood walkability. Walkability scores were calculated by summing the z-scores of three GIS-derived variables aggregated within a 1km-street network home-centred buffer: street connectivity, land use mix and net residential density. Street connectivity was defined as the number of 3 or more branch road junctions per street-kilometre. Land use mix was derived as the evenness of distribution of square footage of residential, commercial, office, entertainment and institutional building footprints, based on existing literature [22, 32]. Net residential density was defined as the unique residential addresses per squared kilometre of building footprint devoted to residential use (i.e. residential building and attached gardens). Land use mix and net residential density variables were log-transformed before deriving the z-scores due to their skewed distributions. The choice of 1km-street-network-home-centred buffers was motivated by two considerations. First, destinations that are within 1km from home have been defined as reachable by foot in the literature [33]. In the absence of more personalised measures of the residential neighbourhood, 1km was thus judged as a good aggregation unit to encompass the opportunities available in the residential area of each participant. Second, a 1km buffer has been used in many studies that found significant associations between features of the built environment and PA outcomes (e.g. [34, 35]). Additionally, a 1-km street network is equivalent to a 10–15 minute walk from home/centre of the buffer to its boundaries.

Proximity from home to parks. Using data from Greenspace Information for Greater London (GiGL) [36], proximity to three types of park ‒ “metropolitan”, “district”, and “local” ‒ was calculated on the basis of street-network distance from the home address to the nearest entrance for each park type. Park type is derived from the Greater London Authority (GLA) London Plan March 2016 [37] and is based on park size and the number and type of facilities they provide. “Metropolitan” parks are the largest and have the most amenities, and “local” parks are the smallest and least well-equipped of these three types (S2 Table). Where there were missing entrance points to parks in the GiGL database (n = 22, i.e. 2.9%), they were manually geocoded based on visual data drawn from Google Maps.

Public transport accessibility. Each ENABLE London participant was assigned a PTAL (Public Transport Accessibility Level) score based on the closest location to their place of residence where a PTAL value was made available by Transport for London (TfL) [38]. PTAL is an averaged measure of the densities of, and the frequency and reliability of service at, the London public transport access points (trains, buses, underground, Docklands Light Railway (DLR), trams) [39], that is being commonly used in London [40]. It is classified into six-value ranges (lower scores reflecting poorer accessibility), whose scores are available for the centroid of each 100m by 100m cell of a grid covering the whole of Greater London [39]. We collapsed the six-value ranges into three categories (Low: PTAL scores 0, 1a, 1b; Intermediate: PTAL scores 2,3,4; High: PTAL scores 5, 6a,6b) in order to increase the number of participants per category, especially for those residing in areas with the lowest accessibility scores.

Data sources and versions used for computing these environmental variables are detailed in S3 Table.

Covariates and moderators

Covariates included sex (female, male), age group ([16-24y], [25-34y], [35-49y], [50y+]), ethnicity (White, Black, Asian, Mixed/Other), and the type of housing tenure participating households sought to relocate to in East Village (“social”, “intermediate”, and “market-rent”). Aspirational housing tenure was used as a marker of household socio-economic status, with “social” and “market-rent” referring to the most deprived and affluent groups, respectively. This was for two reasons. First, housing tenures are allocated largely based upon individual and household social and economic characteristics (S4 Table). Housing tenure has therefore commonly been used as a proxy for socio-economic status in the wider epidemiological literature [41]. In our sample, 67% of those seeking social housing were currently living in social housing accommodation, with the remainder being mostly on social housing waiting lists. Of those seeking intermediate or market-rent accommodation, 64% were currently living in privately rented accommodation, with the remainder largely living with relatives or friends. Considering aspirational housing tenure as a socio-economic marker is supported by the observation that ENABLE London participants seeking relocation into social housing are also more likely to be unemployed, less educated and from ethnic minorities (i.e. classic markers of socio-economic vulnerability) compared to those seeking relocation into affordable and market-rent housing (Table 1). Second, literature has highlighted associations of both current housing tenure and aspirational housing tenure with a variety of health outcomes [5, 30, 42, 43] (e.g. BMI, mental health, morbidity and mortality), including PA [44]. Seekers of social housing were used as the reference group, because it enrolled a larger number of participants.

Table 1

Baseline demographic, residential built environment and daily physical activity by aspirational housing tenure in the London-ENABLE study.

	Total		Social		Intermediate		Market rent
N	1064		442		436		186
	n	(%)	n	(%)	n	(%)	n	(%)	p-value
Sex: female	621	(58%)	330	(75%)	210	(48%)	81	(44%)	<0.001
Age									<0.001
16–24	222	(21%)	95	(21%)	77	(18%)	50	(27%)
25–34	464	(44%)	112	(25%)	254	(58%)	98	(53%)
35–49	310	(29%)	198	(45%)	92	(21%)	20	(11%)
50+	68	(6%)	37	(8%)	13	(3%)	18	(10%)
Ethnic group									<0.001
White	511	(48%)	83	(19%)	301	(69%)	127	(68%)
Black	270	(25%)	210	(48%)	46	(11%)	14	(8%)
Asian	172	(16%)	90	(20%)	61	(14%)	21	(11%)
Mixed / Other	111	(10%)	59	(13%)	28	(6%)	24	(13%)
Qualifications									<0.001
Degree or equivalent	612	(58%)	101	(23%)	364	(84%)	147	(79%)
Intermediate qualification	318	(30%)	243	(55%)	51	(12%)	24	(13%)
Other/None	132	(12%)	97	(22%)	20	(5%)	15	(8%)
NS-SEC 1									<0.001
Higher managerial/professional	488	(46%)	50	(11%)	309	(71%)	129	(69%)
Intermediate occupations	158	(15%)	54	(12%)	70	(16%)	34	(18%)
Routine/manual occupations	139	(13%)	107	(25%)	24	(6%)	8	(4%)
Unemployed/Economically inactive	270	(26%)	225	(52%)	30	(7%)	15	(8%)
Residential Built Environment Factors
Walkability, mean (95% CI)	0.04	(-0.12, 0.20)	-0.51	(-0.70, -0.31)	0.21	(-0.05, 0.46)	0.94	(0.48, 1.39)	<0.001
Distance to metropolitan parks (km), median (IQR)	2.16	(1.21, 3.48)	2.53	(1.43, 3.71)	1.90	(0.95, 2.99)	1.85	(1.04, 3.15)	<0.001
Distance to district parks (km), median (IQR)	2.21	(1.35, 3.08)	2.39	(1.72, 3.32)	2.04	(1.07, 2.88)	2.00	(1.39, 2.75)	<0.001
Distance to local parks (km), median (IQR)	0.75	(0.44, 1.18)	0.62	(0.38, 0.98)	0.88	(0.48, 1.35)	0.86	(0.48, 1.37)	<0.001
Public transport accessibility, n (%)									<0.001
Low	96	(9%)	47	(11%)	37	(8%)	12	(6%)
Intermediate	606	(57%)	294	(67%)	219	(50%)	93	(50%)
High	362	(34%)	101	(23%)	180	(41%)	81	(44%)
Physical activity 1,2	mean	(95% CI)	mean	(95% CI)	mean	(95% CI)	mean	(95% CI)
Daily steps on week days (n = 1,053)	9,126	(8,919, 9,333)	8,618	(8,247, 8,990)	9,516	(9,170, 9,862)	9,409	(8,895, 9,923)
Daily steps on weekend days (n = 848)	8,448	(8,170, 8,725)	6,909	(6,390, 7,428)	9,385	(8,925, 9,846)	9,540	(8,874, 10,206)
Daily minutes of MVPA on week days (n = 1,053)	61.0	(59.4, 62.6)	56.8	(53.9, 59.7)	63.3	(60.6, 66.0)	65.6	(61.6, 69.6)
Daily minutes of MVPA on weekend days (n = 848)	55.5	(53.2, 57.7)	45.6	(41.4, 49.8)	60.8	(57.1, 64.5)	64.1	(58.8, 69.5)

Total number = 1064, data collected 2013–2016.

1 Daily steps and minutes of MVPA are adjusted for sex, age group, ethnic group and housing group as fixed effects and household as a random effect in a multi-level model.

2 Differences between Social and both Intermediate and Market-rent groups were statistically significant, p<0.01.

Because household income, occupational status, presence of children in the household and education level were part of the eligibility criteria for housing in East Village (S4 Table) and/or highly correlated to aspirational housing status, these variables were not adjusted for in the final regression models to avoid over-adjustment. Other hypothesised covariates, including whether a car was available for use in the household and attitude towards PA, were excluded after initial consideration, either because they were not found to be associated with exposures and outcomes in bivariate analyses or because their addition to fully adjusted models did not alter estimated coefficients appreciably.

Aspirational housing tenure was also tested as a moderating factor of the associations between residential built environment factors and PA outcomes.

Statistical analyses

Multi-level linear regression models were used to examine the association between daily PA (steps and minutes of MVPA) and residential built environment variables (walkability, distance to parks and accessibility to public transport) on weekdays and weekend days separately. Average daily steps (minutes of MVPA) were regressed on each built environment variable separately with further adjustment for (i) household as a random effect; (ii) sex, age group, ethnicity and aspirational housing group as fixed effects and household as a random effect; (iii) remaining built environment variables. Differences in weekday and weekend physical activity associations with built environment characteristics from the separate models were formally tested using Z tests. Residuals from all models were checked for assumption of normality to confirm the analytic approach as appropriate.

Further models examined two ways in which built environment variables may contribute to socio-economic differences in PA levels. First, to examine the moderating effect of housing group on the association between the built environment and PA levels, interaction terms for housing group and built environment variables were added to the models and Stata post-estimation commands (testparm) were used to assess their statistical significance. Second, to examine whether differences in access to PA facilities across housing groups translate into differences in levels of PA, attenuation in PA outcomes for each housing group was examined following adjustment for each built environment variable.

Sensitivity analyses examined whether associations remained when the sample was restricted to (i) 837 participants who had data on both weekday and weekend days; (ii) 1,029 participants aged 18+. We also examined whether adjustment for self-reported access to a car or van made any material difference to the findings.

All analyses were carried out in Stata v15.1.

Results

Analytical sample

Descriptive statistics are shown in Table 1. Females comprised 58% of the analytical sample, which was largely White (48%). Women, older people and those belonging to ethnic minorities were more prevalent among social compared to intermediate and market-rent housing seekers (p<0.001); the socio-demographic characteristics of intermediate and market-rent housing seekers were similar. Participants seeking relocation to social housing were less physically active compared to other participants. They took 8618 (95%CI [8247; 8990]) steps on weekdays and 6909 (95%CI [6390; 7428]) on weekends, when intermediate housing seekers took 9516 (95%CI [9170; 9862]) and 9385 (95%CI [8925; 9846]), and market-rent housing seekers took 9409 (95%CI [8895;9923]) and 9540 (95%CI [8874;10206]), respectively. Social housing seekers resided in less walkable areas (walkability scores: -0.51, 95%CI [-0.70; -0.31]) compared with both intermediate (0.21, 95%CI [-0.05; 0.46]) and market-rent (0.94, 95%CI [0.48;1.40]) housing seekers. They also lived further away from metropolitan parks (median: 2.53km, IQR [1.43;3.71] for social vs 1.90km, IQR [0.95;2.99] for intermediate and 1.85km, IQR [1.04;3.15] for market-rent housing seekers). Though they lived closer to their nearest local park (median 0.62km, IQR [0.38;0.98] for social vs 0.88km, IQR [0.48;1.35] for intermediate and 0.86km, IQR [0.47;1.37] for market-rent housing seekers). They also had poorer accessibility to public transport (only 22.9% have high accessibility compared to 41.3% for intermediate and 43.5% for market-rent housing seekers).

Participants seeking market-rent housing relocation had a similar level of PA to participants seeking intermediate housing both on weekdays and weekends. They also resided in more walkable areas (walkability scores: 0.94, 95%CI [0.48; 1.40] for market-rent vs 0.21 95%CI [-0.05; 0.46] for intermediate housing seekers), but lived at a similar distance to their closest metropolitan, district and local parks, and had similar accessibility to public transport.

Built environment and PA-levels associations

Associations between built environment factors and the number of daily steps and amount of MVPA taken on weekdays and weekends are presented in Table 2 and displayed graphically in S1 Fig.

Table 2

Regression estimates for the association between residential built environment variables and physical activity outcomes (daily steps and daily minutes of MVPA) in the London ENABLE study.

		Model 1 1		Model 2 2		Model 3 3		Model 1 1		Model 2 2		Model 3 3
		β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)
		Outcome: daily steps on weekdays (n = 1053)						Outcome: daily steps on weekend days (n = 848)
Walkability		52	(-30, 134)	-5	(-86, 76)	38	(-57, 132)	255	(139, 371)	135	(28, 242)	144	(19, 269)
Distance to closest
	metropolitan park (km)	-295	(-446, -143)	-206	(-354, -58)	-264	(-422, -107)	-363	(-582, -144)	-125	(-323, 74)	-144	(-355, 66)
	district park (km)	-93	(-265, 79)	7	(-160, 174)	-64	(-240, 113)	-370	(-617, -124)	-180	(-402, 42)	-139	(-373, 96)
	local park (km)	170	(-174, 514)	-84	(-420, 253)	-185	(-529, 158)	1,147	(669, 1,624)	597	(161, 1,032)	598	(155, 1,040)
Accessiblity to public transport
	Low	442	(-362, 1,247)	767	(-13, 1,546)	1,186	(296, 2,076)	-1,352	(-2,516, -188)	-608	(-1,658, 442)	60	(-1,136, 1,256)
	Intermediate	-482	(-948, -15)	-152	(-609, 305)	8	(-487, 503)	-1,128	(-1,804, -453)	-365	(-981, 252)	-8	(-670, 654)
	High (reference group)
		Outcome: daily minutes of MVPA on weekdays (n = 1053)						Outcome: daily minutes of MVPA on weekend days (n = 848)
Walkability		0.6	(0.0, 1.3)	0.1	(-0.6, 0.7)	0.4	(-0.3, 1.2)	2.2	(1.2, 3.1)	1.2	(0.3, 2.1)	1.5	(0.4, 2.5)
Distance to closest
	metropolitan park (km)	-2.5	(-3.7, -1.3)	-1.8	(-2.9, -0.6)	-2.2	(-3.4, -0.9)	-3.0	(-4.7, -1.2)	-1.1	(-2.7, 0.5)	-1.2	(-2.9, 0.5)
	district park (km)	-0.6	(-2.0, 0.7)	0.3	(-1.0, 1.6)	-0.2	(-1.6, 1.2)	-2.4	(-4.4, -0.4)	-0.9	(-2.7, 0.9)	-0.5	(-2.3, 1.4)
	local park (km)	2.0	(-0.8, 4.7)	-0.3	(-3.0, 2.3)	-1.0	(-3.7, 1.6)	9.0	(5.2, 12.9)	4.7	(1.2, 8.2)	4.9	(1.3, 8.5)
Accessiblity to public transport
	Low	2.7	(-3.7, 9.1)	5.8	(-0.3, 11.9)	9.7	(2.7, 16.6)	-9.5	(-18.9, -0.1)	-4.0	(-12.5, 4.5)	2.7	(-7.0, 12.4)
	Intermediate	-4.5	(-8.2, -0.8)	-1.7	(-5.3, 1.9)	-0.2	(-4.1, 3.7)	-8.2	(-13.7, -2.8)	-2.7	(-7.6, 2.3)	0.7	(-4.7, 6.0)
	High (reference group)

Total number = 1064, data collected 2013–2016. Effect estimates highlighted in bold are statistically significant, p<0.05.

1 Model 1 adjusts for household as a random effect to allow for clustering at the household level (referred to as “minimally adjusted model” in the text).

2 Model 2 additionally adjusts for sex, age group, ethnic group, aspirational housing group as fixed effects.

3 Model 3 additionally adjusts for all residential built environment variables as fixed effects.

On weekdays

In models adjusted for sociodemographics (models 2), both daily steps taken and MVPA accumulated were negatively associated with the distance from home to the closest metropolitan park (mean difference in daily steps (-206, 95%CI [-354;-58]) and in daily minutes of MVPA (-1.8, 95%CI [-2.9;-0.5]) per km of distance to the closest metropolitan park), indicating that the more distant the metropolitan park, the smaller the number of steps taken and the shorter the time spent in MVPA (Fig 2). Associations between accessibility to public transport and daily steps and MVPA were borderline statistically significant in models adjusted for sociodemographics (mean differences in daily steps (767, 95%CI -13,1546) and in daily minutes of MVPA (5.8, 95%CI [-0.3; 11.9]) for those with a low accessibility compared to those with high accessibility), and reached statistical significance after further adjustment for the other residential built environmental factors (models 3) (mean differences in daily steps: 1186, 95%CI [296;2076], mean difference in daily minutes of MVPA: 9.7, 95%CI [2.7;16.6]). Although adjustment for other residential built environmental factors led to an inflation of the regression estimate and a widening of the confidence interval compared with models adjusting only for sociodemographics, Variance Inflation Factor (VIF) values did not exceed 2, downplaying concerns over multicollinearity.

Fig 2

Daily steps and MVPA on weekends and on weekdays plotted against quintile of residential built environmental factors.

Mean (95%CI) daily steps and mean (95%CI) daily MVPA (min) on weekdays (n = 1,053) and on weekends (n = 848) plotted against A) walkability scores in the residential area at each quintile (median value), B) distance to the closest local park at each quintile (median value), C) distance to the closest district park at each quintile (median value), D) distance to the closest metropolitan park at each quintile (median value), E) accessibility to public transport. Means are adjusted for sex, age group, ethnic group, aspirational housing tenure and clustering at household level.

On weekend days

In models adjusted for sociodemographics (models 2), daily steps and MVPA were positively associated with residential walkability (mean difference in daily steps and daily minutes of MVPA per unit of walkability: 135, 95%CI [28; 242], and 1.2, 95%CI [0.3; 2.1], respectively). Daily steps and MVPA also showed a positive association with the distance from home to the closest local park (mean difference in daily steps and daily minutes of MVPA per km of distance to the closest local park: 597, 95%CI [161; 1032], and 4.7, 95%CI [1.2; 8.2], respectively). This suggests that the more walkable the residential environment the higher the number of steps taken and the greater the time spent in MVPA, and that the further the closest local park, the higher the number of steps taken and the greater the time spent in MVPA on weekend days (Fig 2). Associations remained consistent after further adjustment for the other residential built environmental factors (models 3). Although in models adjusted only for household clustering (models 1) daily steps and MVPA were negatively associated with the distances to the nearest metropolitan and district parks, and with accessibility to public transport, these associations failed to reach significance after adjustment for sociodemographic factors (models 2).

Comparison of weekday versus weekend day effects

There was formal evidence of a difference between weekday and weekend day levels of physical activity in sociodemographic models (Model 2), for (1) increased steps on a weekend day compared with weekday per unit increase in walkability (p = 0.041), (2) increased steps and MVPA at weekends versus weekdays per km increase in distance to local park (p = 0.015, p = 0.026 respectively), and (3) increased steps on a weekday compared with a weekend day associated with lower public transport accessibility (p = 0.02). All other differences between weekend versus weekday physical activity associations with the built environment were not formally statistically significant (p>0.05).

Pathway 1: Moderating effect of aspirational housing group on the association between the built environment and PA levels

For mean daily steps on weekday, an interaction between low accessibility to public transport and social vs intermediate housing seekers was observed (p = 0.042). However, a post-estimation Wald test suggested the interaction term did not add significantly to the model (p = 0.06), indicating that the residential built environment relates to PA behaviours in a similar way across the three housing groups.

Pathway 2: Attenuation of the associations between aspirational housing tenure and PA levels by built environment factors

To investigate the degree to which the residential built environment could explain socio-economic differences in PA, we tested the association between aspirational housing tenure and PA levels using various models of built environment factor adjustment (Table 3). Adjustment for walkability decreased differences in daily steps and MVPA on weekend days between social and intermediate housing seekers by 1.5% and 2.6%, respectively (i.e. decrease of 36 steps and 0.4min accumulated in MVPA in the gap between social and intermediate housing seekers), adjustment for distance to the closest local park decreased differences in daily steps and MVPA on weekend days between social and intermediate housing seekers by 4.7% and 5.9%, respectively (i.e. decrease of 112 steps and 0.9min accumulated in MVPA), and adjustment for distance to the nearest metropolitan park reduced differences in steps taken on weekdays between social and intermediate housing seekers by 9.9% and 11.9%, respectively (i.e. decrease of 81 steps and 0.7min accumulated in MVPA). These findings suggest that aspirational housing tenure differences in PA relate to the differences in neighbourhood walkability and distance to local and metropolitan parks observed across the three housing groups.

Table 3

Effect of adjustment for residential built environment factors in the differences in daily physical activity by housing group in the London-ENABLE study.

				Base model with further adjustment for
		Base model 1		Walkability		Distance to metropolitan park		Distance to district park		Distance to local park		Accessibility to public transport		All BE variables
	Housing group	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)	β	(95% CI)
Daily steps
Weekday, n = 1053	Social (reference group)
	Intermediate	898	(348, 1448)	900	(349, 1451)	817	(266, 1369)	900	(347, 1452)	914	(360, 1469)	884	(330, 1,438)	814	(255, 1,373)
	Market-rent	791	(119, 1463)	797	(118, 1476)	723	(51, 394)	792	(119, 1466)	808	(132, 1483)	798	(122, 1,473)	727	(45, 1,409)
Weekend n = 848	Social (reference group)
	Intermediate	2477	(1721, 3232)	2441	(1689, 3194)	2436	(1678, 3194)	2427	(1671, 3184)	2365	(1607, 3122)	2409	(1646, 3172)	2242	(1477, 3007)
	Market-rent	2631	(1737, 3526)	2474	(1575, 3373)	2598	(1702, 3494)	2596	(1702, 3490)	2513	(1618, 3409)	2545	(1642, 3448)	2280	(1375, 3186)
Daily minutes of MVPA
Weekday n = 1053	Social (reference group)
	Intermediate	6.6	(2.3, 10.9)	6.6	(2.3, 10.9)	5.9	(1.6, 10.2)	6.7	(2.4, 11.0)	6.7	(2.3, 11.0)	6.4	(2.1, 10.7)	5.8	(1.5, 10.2)
	Market-rent	8.9	(3.6, 14.1)	8.8	(3.5, 14.1)	8.3	(3.0, 13.5)	8.9	(3.7, 14.2)	8.9	(3.7, 14.2)	8.8	(3.6, 14.1)	8.1	(2.8, 13.5)
Weekend n = 848	Social (reference group)
	Intermediate	15.3	(9.2, 21.4)	14.9	(8.9, 21.0)	14.9	(8.8, 21.0)	15.0	(8.9, 21.2)	14.4	(8.3, 20.5)	14.8	(8.6, 20.9)	13.6	(7.4, 19.8)
	Market-rent	18.6	(11.4, 25.8)	17.2	(9.9, 24.4)	18.3	(11.0, 25.5)	18.4	(11.2, 25.7)	17.7	(10.4, 24.9)	18.0	(10.7, 25.3)	15.8	(8.4, 23.1)

1 The base model is adjusted for sex, age group, ethnic group as fixed effects and household as a random effect to allow for clustering in a multi-level model.

Sensitivity analyses performed on the subsample of 837 participants with PA data on both weekdays and weekends showed consistency with the main findings presented above (S5–S7 Tables). Analyses performed on the subsample of 1,029 participants aged 18+ year (i.e. excluding 35 participants) and with adjustment for self-reported access to a car or van (in 460/1,064) also showed similar associations (results not shown).

Discussion

This study aimed to measure the association of a range of residential built environment factors with PA levels on weekdays and weekend days. It further explored whether the residential built environment contributes to socio-economic differences in PA.

In step with our first aim, we found that participants from more walkable neighbourhoods took more steps and accumulated more MVPA on weekends compared to those from less walkable neighbourhoods. These findings add to the growing evidence, both worldwide [45, 46] and in the UK [22], that the more walkable an urban environment, the more attractive for active modes of travel. For instance, Stockton et al., using data from Londoners aged 35 to 55 years from the Whitehall II study, similarly found that those residing in the most walkable neighbourhoods spent significantly more time walking than those in the least walkable ones [22]. Our findings however highlighted a stronger association with neighbourhood walkability on weekends. Research has shown that the time spent in the residential neighbourhood modifies the association between neighbourhood walkability and PA levels, with the longer the time spent, the stronger the relationship [47]. Our findings of a stronger relationship on weekends could suggest that ENABLE London participants spent more time in the home vicinity on weekends. This interpretation is consistent with a UK time survey highlighting that people spend more time on leisure and domestic work on weekends [29].

We also found evidence of association between park proximity and PA levels. Increased number of steps and amount of MVPA were associated with increased distance to the nearest local park on weekends, but with decreased distance to the nearest metropolitan park on weekdays. Local parks are on average within walking distance from home [37], rendering plausible the assumption of physically active commutes. Local parks are also small, and provided with none or few sport facilities [37]. The journey to rather than activities within local parks may therefore contribute to PA levels. This would explain why the number of steps and amount of MVPA increase with distance to local parks. Conversely, metropolitan parks are often situated beyond the commonly admitted walking distance. Interest in reaching them may therefore decrease with increased distance. Metropolitan parks are also large and well provided with sport facilities [37]. Increased number of sport facilities within a park has been associated with increased use of that park for PA [48]. Activities undertaken within metropolitan parks, rather than the journey to reach them, may therefore contribute to PA levels. This would explain that the number of steps and amount of MVPA increase when distance to the park decreases. Like others highlighted [49], park characteristics such as distance, size and enclosed facilities are important to consider when exploring the role of greenspaces on PA levels. Overlooking these aspects may help explain inconsistent findings in the literature on access to greenspaces and PA, which reports both null [25, 35] and beneficial [23, 24, 26, 49–51] associations. Interestingly, we found that distance to local parks related to PA levels only on weekends, whilst distance to metropolitan parks only on weekdays. Local parks are common features of the immediate home vicinity, which individuals may make greater use on the weekend, when they are assumed to spend more time at home. Because metropolitan parks are further from the home, their use may be associated with other weekdays activities, like work, which often takes individuals beyond their local neighbourhood. These assumptions would require further testing.

Positive associations between accessibility to public transport and PA levels have been found worldwide [52], including in the UK [23], suggesting that interest in use of public transport may decrease with decreasing accessibility. For instance, Sallis et al. found that accessibility to public transport, measured as the density of public transport access points within a 1-km buffer centred on the home address, positively relates to adult PA levels in 14 cities worldwide, including Stoke-on-Trent, UK [23]. Yet, our findings did not fully align with this conclusion, as the positive association between accessibility to public transport and steps and MVPA found on weekends failed to reach significance after adjustment for sociodemographic factors. More surprisingly, the association becomes negative on weekdays, with participants experiencing a low accessibility to public transport taking more daily steps compared to those with high accessibility. To our knowledge, such findings have not been reported in the literature and raise several questions. Participant’s heavy reliance on public transportation for inescapable weekday activities, like work, may explain why those with poor accessibility to public transport walk more on weekdays than those who live closer. Examining whether this negative association between public transport accessibility and PA on weekdays holds for individuals who have alternative ways of travelling (e.g., car) offers an avenue for further research. On a different note, our measure of public transport accessibility is not strictly comparable to density-based measures (simple count of public transport stops within a given area) used in other studies, in that it also accounts for frequency and reliability of service. Yet, our variable reflects a concept of public transport accessibility not that different to more basic density-based variables and is unlikely to be responsible for inverting the direction of the association. Overall, opposite directions of the associations on weekend and weekdays may suggest that the use of public transport may depend on utility and travel function: an inescapable commute to work on weekdays and leisure and recreation on the weekends.

Put altogether, our findings suggest that the residential built environment is associated with PA, but that this varies according to day of the week. Features in the immediate vicinity of the home, like local parks and walkability captured within 1-km home-centred buffer, were found to be associated with PA on the weekend. Conversely, features like metropolitan parks that were beyond the immediate home vicinity, and public transportation that take individuals beyond the limit of their residential neighbourhood, were found to relate to PA levels on weekdays. The intertwinement of spatial and temporal patterns of human activities unveiled here suggests that PA behaviours are shaped by broad contextual forces that go beyond the simple assessment of what is physically accessible from home [53]. Others have highlighted that space and time constraints in adults’ daily activities are important factors that shape the impact of residential neighbourhood attributes on PA. For instance, Cerin et al. found that time of the day and day of week were significant moderators of residential built environment-MVPA associations [26]. The combined use of GPS and accelerometer data appears as a promising avenue for research to deepen our understanding of the spatio-temporal determinants of PA behaviours [54]. Such combined data would, for instance, have allowed us to confirm/overturn our assumptions that the association between metropolitan parks and PA on weekdays owe to PA done within the metropolitan parks. By identifying the activities that precede and follow a visit to the metropolitan park, we could further have possibly identified some time specific constraints that encourage individuals to use metropolitan parks on weekdays rather than weekends.

Finally, in line with our second aim, we explored two plausible pathways by which the built environment may contribute to socio-economic differences in PA: (i) differences in the way the housing groups interact with the features supportive of active behaviours in the residential environment, and (ii) differences in the availability of these features across the housing tenure groups. We found little evidence for the former pathway, with an absence of interactions between residential built environment factors and housing tenure group in relation to PA levels on weekdays and weekends. This suggests that the residential built environment relates to PA levels in a similar way across the three socio-economic groups, regardless of whether PA occurs on weekdays or weekends. This is in line with some studies [55], but disagrees with others like Jones et al., who found a strong park distance decay effect on park use for the most affluent neighbourhoods, but not for the most deprived ones in Bristol, UK [56]. As for the second pathway, we found that social housing seekers lived in less walkable environments, closer to local parks and further away from metropolitan parks compared to intermediate and market-rent housing seekers. In turn, decreased neighbourhood walkability, shorter distance to local parks, and increased distance to metropolitan parks were all associated with a decreased number of daily steps and time spent in MVPA. These findings align with the idea that disadvantaged individuals are distributed in neighbourhoods with less health-promoting facilities [16]. There is mounting evidence from various countries of socio-economic disparities in the distribution of PA-promoting features in the residential neighbourhood [57, 58], and in PA levels [5-7]. Yet, comparatively, there is little evidence of a link between these two disparities. We found that the association between socio-economic status and PA behaviours on weekdays was attenuated by disparities in distance to metropolitan parks (decreased differences in daily steps and MVPA between social and intermediate housing seekers of 9.9% and 11.9%, respectively). The associations between socio-economic status and PA behaviours on weekends were also slightly attenuated by disparities in neighbourhood walkability (1.5% and 2.6% for steps and MVPA, respectively) and distance to local parks (4.7% and 5.9% for steps and MVPA, respectively). These findings support the notion that disadvantaged people are less physically active because they live into less health-promoting neighbourhoods [41, 59]. Only few empirical studies have so far provided evidence that the residential built environment contributes to socio-economic differences in PA. And like ours, these studies found that only a limited amount of the socio-economic differences in PA relate to differences in environmental features in the residential area [14, 60]. More research is needed to confirm the weight of the residential built environment on socio-economic differences in PA.

Strengths and limitations

Strengths of this paper include the use of validated objective measures of PA [61], assessing weekday/weekend variations in PA, and exploring the contribution of the residential built environment in explaining socio-economic differences in PA levels. Sensitivity analyses further strengthened our findings by showing consistent results in both our inclusive analytical sample and the more restricted sample of participants with PA data on both weekdays and weekends, and participants 18 years or above. Durations of adequate wear-time to define habitual activity vary. Previous studies have often used a minimum of 3 to 5 days of accelerometry data [62], with as little as 480 minutes [63] per day to 600 minutes [64] per day of daily recording. We chose to include all days of ≥540 minutes of registered accelerometry time during at least one day to lessen attrition bias and maximise inclusion of hard to reach groups, i.e., those from social housing who are likely to have lower compliance and record less PA data. This decision was part of the study design and was in our a priori Statistical Analysis Plan approved by the ENABLE London Steering Committee [30]. There are several limitations which warrant consideration. This study draws on the baseline sample of a natural experiment, where recruits were deemed eligible if they were seeking to relocate to East Village. Such sampling strategy limits external validity, as the participants may not be truly representative of the London population as a whole. Moreover, the cross-sectional study design is prone to selection bias (i.e., those living in less walkable neighbourhoods are intrinsically different to those who do not), and therefore restricts interpretations about direction of effects.

Moreover, partly because we overlooked non-residential exposures to the built environment, we were not able to align with calls for addressing the Uncertain Geographic Context Problem (i.e., the extent to which areal units of measurement deviate from the geographic context truly experienced by individuals) [53]. Although the spatial unit used to derive our environmental exposure complies with the wide majority of studies that have looked at the way the residential built environment relates to health behaviours (i.e. using an home-centred buffer within a walkable distance radius), focusing on the residential neighbourhood only may have led to misestimation of the association between the residential environment and health behaviours [65].

Conclusions

This study adds to the evidence that the residential built environment could act as a potential lever for increasing physical activity levels in adults. Interestingly, individuals from lower socio-economic status may disproportionately benefit from greater availability of physical activity resources, because they tend to reside in urban environment less supportive of PA. Policymakers should however be sensitive to that possibility that interventions might have different impacts on physical activity on weekends and weekdays. More evidence, drawn from longitudinal studies and relying on the combined use of GPS and accelerometer data is needed though, to confirm the direction and interpretation of these associations.

Sociodemographics of ENABLE London participants included in (n = 1064) and excluded from (n = 214) the analyses.

(DOCX)

Click here for additional data file.

Description, number and size of metropolitan, district and local parks.

(DOCX)

Click here for additional data file.

Data sources and versions used for computing residential built environment variables.

(DOCX)

Click here for additional data file.

Criteria for allocation of housing tenures in East Village.

(DOCX)

Click here for additional data file.

Baseline demographic and daily steps of 837 ENABLE-London participants who provided daily physical activity data for weekdays and weekend days.

(DOCX)

Click here for additional data file.

Regression estimates for the association between residential built environment variables and physical activity outcomes (daily steps and daily minutes of MVPA) in the London ENABLE study for 837 participants who provided PA data on weekdays and weekend days.

(DOCX)

Click here for additional data file.

Effect of adjustment for residential built environment factors in the differences in daily physical activity by housing group in the London-ENABLE study in 837 participants who provided PA data on both weekdays and weekend days.

(DOCX)

Click here for additional data file.

Legend: Mean (95%CI) daily steps and mean (95%CI) daily MVPA (min) on weekdays (n = 1,053) and on weekend days (n = 848) plotted against A) walkability scores in the residential area at each quintile (median value), B) distance to the closest local park at each quintile (median value), C) distance to the closest district park at each quintile (median value), D) distance to the closest metropolitan park at each quintile (median value), E) accessibility to public transport. Means are adjusted for sex, age group, ethnic group, aspirational housing tenure and clustering at household level.

(DOCX)

Click here for additional data file.

16 Apr 2020

PONE-D-19-33687

Weekend and weekday differences in associations between the residential built environment and physical activity: findings from the ENABLE-London Study

PLOS ONE

Dear Prof. Owen,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by May 31 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Adewale L. Oyeyemi, Ph.D

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at http://www.plosone.org/attachments/PLOSOne_formatting_sample_main_body.pdf and http://www.plosone.org/attachments/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for including your ethics statement: Full ethical approval was obtained from the relevant Multi-Centre Research Ethics Committee (REC Reference 12/LO/1031). All participants provided written informed consent.

Please amend your current ethics statement to include the full name of the ethics committee/institutional review board(s) that approved your specific study.

Once you have amended this/these statement(s) in the Methods section of the manuscript, please add the same text to the “Ethics Statement” field of the submission form (via “Edit Submission”).

For additional information about PLOS ONE ethical requirements for human subjects research, please refer to http://journals.plos.org/plosone/s/submission-guidelines#loc-human-subjects-research.

3. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

4. Please amend either the abstract on the online submission form (via Edit Submission) or the abstract in the manuscript so that they are identical.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Thank you for the opportunity of reviewing this interesting study. I hope these comments and suggestions help strengthen this manuscript.

This study aimed to determine the association of the residential built environment with physical activity outcomes (daily steps and daily MVPA), stratifying by type of day (weekend vs. weekdays). Next, authors explored potential effect measure modification of these relations by socioeconomic status.

Overall strengths are a large sample size, objective measures for physical activity, and a decent response rate for these types of studies. The authors explore important research questions (the difference in the effect of built environment variables on physical activity outcomes by type of week-day, and the role of the BE in the SES-PA relationship).

INTRODUCTION

• There is a mention of the prevalence of physical inactivity in the UK based on old recommendations of accumulating at least 150 mins/wk of MVPA within 10-minute bouts. Please cite evidence (prevalence of inactivity) based on current guidelines, without the stipulation of the 10-minute bouts.

• Regarding the phrase: “recently, epidemiological research has increasingly incorporated socio-ecological models that acknowledge the role of the built environment, especially the local residential environment, in determining PA behaviors”. Please change the term “determinants” to “influencing”, or, state “as potential determinants of PA behaviors”. The term “determinants” implies strong evidence of causality, which for the most part is still lacking for the role of built environments on health behaviors.

• In the first paragraph, the authors seem to make a case about findings being mixed/inconclusive with regards to the role of the built environment on socioeconomic disparities. They then cite multiple studies with divergent findings. To me, it is apparent that more than having mixed findings, we have evidence of context-specific differences. In some places, low-income neighborhoods are deprived of adequate BE resources for PA. In other settings, it is precisely in low-income neighborhoods where one finds the most BE resources for PA (and, because virtually all evidence is from cross-sectional studies, this might be as a response from local governments to equalize the playing field for these economically disadvantaged areas). This could be much better framed acknowledging the role of context in these relations.

METHODS

Sampling

• This is the baseline sample of a natural experiment, where recruits were deemed eligible if they were people seeking to relocate to a new residence in the short term. I believe a more extensive comment on the limitation of such a sampling strategy for the aims of this specific analysis is warranted. This isn’t a truly representative sample of the population, yet the data are being treated as though this were a typical cross-sectional survey. While the sample serves the purpose of the parent study (a natural experiment) well, there are limitations to conducting a cross-sectional analysis with its baseline data, mainly due to selection bias and external validity limitations, which require further acknowledgement.

Variables

• Accelerometer-derived PA outcomes: when describing the outcome variables, authors state that daily steps were “adjusted for day of the week, day order of recording and month of data collection”. I assume this is referring to the regression models per se? Or, are the authors implying that the actual variables are somehow weighted for these characteristics? This is not typically seen in accelerometry studies, and I believe requires further explanation. If this is something done at the modeling step, I recommend removing this information from the variable description section, as this can confuse the reader, and simply include it in the modeling section.

• MVPA variable: The use of a single day of valid accelerometry data (and of only 9 hours) certainly seems a bit low. Usual standards call for a minimum of 4 days with at least 10 hours of valid data per day. However, authors justify this by referring to a sensitivity analysis performed against complete data, which they define as that from participants with at least 4 days of 9 hours of data or more (instead of 10 hours). A single day of valid accelerometry data is thought to be sufficient for surveillance of populations (when truly representative samples are used), but I do worry that for this dataset, which is composed of a non-representative sample of London residents, and is not done for surveillance purposes, this may be a bit of a stretch.

• Land-use mix and net residential density variables: can the authors please provide further detail on the basis for the log-transformation of these variables? I haven’t seen this done before in other studies from other global settings. They state it is to “fit a comparable scale”. Comparable to what?

• When explaining the rationale for the 1-KM network buffers, it might be good to state the average walking distance/time this represents (e.g., about a 10-15 minute walk from home in any direction for most people).

• Proximity to parks variable: other studies have found limited utility in these distance-based variable, including IPEN-adult, a 12 country study, where park density within 1-KM buffers was deemed a more useful predictor of physical activity. Did the authors consider using a buffer-based variable for parks too? Why was this discarded?

• I find it very interesting that car ownership was not deemed as a relevant confounder in this analysis, as this contradicts what we know about this variable in other parts of the world.

Statistical Analysis

• The first phrase is not clear. Authors mention a series of variables “were examined”. This could mean anything. Please be clear on the actual analytic procedures applied for “examining” these variables.

• Was a test for interaction actually performed to determine of day of the week (weekend vs. weekday) actually modifies the effect observed for some of these BE variables on the PA outcomes?

• The use of multilevel models seems adequate, but I am surprised at such a high ICC (0.3, wow!). It would be important to report what percentage of the sample shared a household with another sample participant. Most studies of this type only accept one participant per household to avoid this level of clustering. Also, did the authors explore and, if needed, account for clustering at the neighborhood level (e.g., people residing in the same zip-code or small administrative geospatial unit?).

• Could the authors explicitly state their hypotheses for the two different type of analyses proposed for examining the role of the BE in explaining PA-SES differences? I can follow the rational for the effect-measure modification analysis (testing for interaction terms of BE X SES variables). The simply adjustment by BE variables of the effect of SES on PA seems a repetition from the first modeling step, and implies confounding vs. effect measure modification, when, per the introduction, it seems that the authors have an effect modification hypothesis to begin with.

• Also, the authors don’t mention which test for interaction and alpha level they used in their analysis (hopefully this was not just based on the p-value of the interaction term in the regression model – but was the result of a formal test for interaction – please clarify).

• Outcome variables in models (steps and mins of MVPA/day): were these transformed in any way for achieving normality? Physical activity data are usually not normally distributed, hence the use of linear regression models usually requires a data transformation step. Alternatively, one could consider the use of logistic regression models (multinomial, politomous, or binary).

• Did the authors examine potential multicollinearity in their models? This is usually a concern with some built environment variables in a single model. Please outline the methods via which this was examined.

RESULTS

• Table 2 title: Should read: “Regression estimates for the association between residential built environment favors with physical activity outcomes (daily steps and mins/wk of MVPA), among… XXX [insert study name, year of data collection, etc.)”. It is customary to list the independent variable first and the dependent variable last, not the other way around.

• For all table/figure titles: add information on the study and sample, and year data were collected, as tables and figures should be “stand-alone” items. Figure 2 is very nice and helpful for the reader.

• I like the way the results for the examination of the role of the BE on the PA-SES relation are presented by two potential pathways. However, it would have been much more helpful to lay out the two hypothesized pathways since the introduction and methods (Statistical analysis) section, so the readers understand why both approaches (test for effect modification vs. BE as a confounder of the SES-PA relation) were explored. Otherwise the “two pathways” seem more a bit of an afterthought. For this purpose, it might be good for the authors to present a figure with the hypothetical DAGs of the two pathways they are exploring, a priori (i.e., before the results section).

DISCUSSION

• I enjoyed reading the interpretation of the results regarding the counter-intuitive finding on longer distance to small parks being associated to higher levels of MVPA on weekends. This makes a lot of sense and it is not something the readers could have derived just by seeing the estimates.

• The discussion on the public transit results should be re-framed. I would suggest not opening with a statement on how they found a positive association with access to transit and PA, given that in reality once the models were adjusted this was not significant (the authors do state this, but I just don’t see the need to open with a statement of a positive relation). I think a deeper discussion is warranted to explain why their findings are inconsistent with most other reports of transit and PA. It is counterintuitive that those with lower availability to public transit have higher levels of PA, in a model that is also adjusted for SES. I wonder if adjusting for car ownership would have made a difference for this specific variable [access to transit] (?). The discussion on this topic should also include some reflection on the difference in measures of access to transit. Most other studies have relied on transit stop density measures (count of public transit stops within a buffer), and thus, are not truly comparable to the measure used in this study.

• Limitations paragraph:

o The use of a single day of accelerometry data with at least 9 hours is not in line with previous studies and should be acknowledged.

o The transit access measure, although apparently more sophisticated than what others use, is not comparable to what most researchers use when examining the role of transit infrastructure on PA.

o The Uncertain Geographic Context Problem and this study’s inability to address it should be listed as a limitation.

o Another major limitation is the authors simply assumed differences in the relation of the BE and PA by weekday (effect modification by type of weekday), which is reasonable, but a stronger study analysis would have included a confirmation of this with a test for interaction, followed by the stratified analysis.

o Finally, selection bias due to a) the response rate, and b) the eligibility criteria (this is not a representative sample of London residents, rather, it targets people with a desire to move) needs to be fully disclosed, with a description of how these factors are likely limiting the external validity of the study. In particular, because a reason for moving might be the BE features of the new neighborhood where people are hoping to move to.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Deborah Salvo

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

8 Jun 2020

Further Editorial queries (dated 28th May 2020)

E1. Thank you for including your ethics statement: Full ethical approval was obtained from the relevant Multi-Centre Research Ethics Committee (REC Reference 12/LO/1031). All participants provided written informed consent.

Please amend your current ethics statement to include the full name of the ethics committee/institutional review board(s) that approved your specific study.

Response to Editorial comments

We have amended the ethics statement in both the paper and submission form to indicate that full ethical approval was obtained from the City Road and Hampstead Ethical Review Board (12/LO/1031).

E2. Are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

Response to Editorial comments

While we are fully supportive of providing anonymised data, unfortunately this was not part of the original ethics approval (see response to E1 above), and permission to share anonymised data was not included in the ENABLE London participant consent form. This raises issues with General Data Protection Regulations. This is why we have said that data from this study are available upon request, as we hope these would be considered by our data protection team on an individual basis.

Reviewer 1

1. Thank you for the opportunity of reviewing this interesting study. I hope these comments and suggestions help strengthen this manuscript.

This study aimed to determine the association of the residential built environment with physical activity outcomes (daily steps and daily MVPA), stratifying by type of day (weekend vs. weekdays). Next, authors explored potential effect measure modification of these relations by socioeconomic status.

Overall strengths are a large sample size, objective measures for physical activity, and a decent response rate for these types of studies. The authors explore important research questions (the difference in the effect of built environment variables on physical activity outcomes by type of week-day, and the role of the BE in the SES-PA relationship).

Response to reviewer

Thank you for these positive comments about our work, and for all the suggestions that have helped us strengthen the manuscript.

2. Introduction

There is a mention of the prevalence of physical inactivity in the UK based on old recommendations of accumulating at least 150 mins/wk of MVPA within 10-minute bouts. Please cite evidence (prevalence of inactivity) based on current guidelines, without the stipulation of the 10-minute bouts.

Response to reviewer

We thank the reviewer for this suggestion. We have amended the text (page 3, lines 44-46), citing the most recent evidence available (Sport England 2018/19 report published in April 2020), as follows: “Current population levels of PA are too low in the UK, with 37% of adults aged 16 years or more not meeting recommended levels of activity of at least 150 minutes of moderate-intensity per week (1)”.

3. Introduction

Regarding the phrase: “recently, epidemiological research has increasingly incorporated socio-ecological models that acknowledge the role of the built environment, especially the local residential environment, in determining PA behaviors”. Please change the term “determinants” to “influencing”, or, state “as potential determinants of PA behaviors”. The term “determinants” implies strong evidence of causality, which for the most part is still lacking for the role of built environments on health behaviors.

Response to reviewer

As suggested, we have amended the sentence on page 3 (lines 47-50) as follows: “Recently, epidemiological research has increasingly incorporated socio-ecological models that acknowledge the role of the built environment, especially the local residential environment, in influencing PA behaviours”.

4. Introduction

In the first paragraph, the authors seem to make a case about findings being mixed/inconclusive with regards to the role of the built environment on socioeconomic disparities. They then cite multiple studies with divergent findings. To me, it is apparent that more than having mixed findings, we have evidence of context-specific differences. In some places, low-income neighborhoods are deprived of adequate BE resources for PA. In other settings, it is precisely in low-income neighborhoods where one finds the most BE resources for PA (and, because virtually all evidence is from cross-sectional studies, this might be as a response from local governments to equalize the playing field for these economically disadvantaged areas). This could be much better framed acknowledging the role of context in these relations.

Response to reviewer

We agree with this suggestion. We have reframed this introductive paragraph to highlight that mixed findings can reflect diverse local realities, and to enhance the role of context in the BE and PA associations. We have also used this opportunity to clarify the way the built environment contribution to socio-economic differences in PA is explored in the literature. Overall, this introductive paragraph now aligns better with the aims of the study as described at the end of the Introduction and partly addresses the issues raised in the reviewer’s comment #15. The reframed paragraph on page 3 (lines 50-65) now reads:

“The extent to which the residential built environment contributes to individual socio-economic differences in PA has been explored by drawing on two hypothetical pathways: (i) a deprivation-amplification effect (2), whereby disadvantaged individuals are less exposed to health-promoting facilities in their residential neighbourhood, and (ii) a moderating effect, by which socio-economic groups use the physical activity facilities available in their neighbourhood differently (3). Regarding the deprivation-amplification pathway, some UK studies have reported that the most affluent urban areas have the poorest accessibility to recreational PA facilities (4-7), but others the most deprived (8). As for the moderating effect pathway, some studies have reported that the association between the residential built environment and PA was moderated by the socio-economic position (9, 10), whereas others did not find such evidence (11, 12). These mixed findings likely reflect diverse local realities, that may owe to local population specificities or regional policy interventions. More research is needed to depict a comprehensive, overarching view of how built environment, socio-economic status and PA behaviours interrelate, and, in turn, deliver effective ‘contextually sensitive’ policy interventions (13).”.

5. Methods, Sampling

This is the baseline sample of a natural experiment, where recruits were deemed eligible if they were people seeking to relocate to a new residence in the short term. I believe a more extensive comment on the limitation of such a sampling strategy for the aims of this specific analysis is warranted. This isn’t a truly representative sample of the population, yet the data are being treated as though this were a typical cross-sectional survey. While the sample serves the purpose of the parent study (a natural experiment) well, there are limitations to conducting a cross-sectional analysis with its baseline data, mainly due to selection bias and external validity limitations, which require further acknowledgement.

Response to reviewer

We agree with the reviewer that the limitations related to our sampling strategy should be better acknowledged. We have added the following paragraph in the Limitations section of the manuscript (page 22, lines 488-494): “This study draws on the baseline sample of a natural experiment, where recruits were deemed eligible if they were seeking to relocate to East Village. Such sampling strategy limits external validity, as the participants may not be truly representative of the London population as a whole. Moreover, the cross-sectional study design is prone to selection bias (i.e., those living in less walkable neighbourhoods are intrinsically different to those who do not), and therefore restricts interpretations about direction of effects.”.

6. Methods, Variables

Accelerometer-derived PA outcomes: when describing the outcome variables, authors state that daily steps were “adjusted for day of the week, day order of recording and month of data collection”. I assume this is referring to the regression models per se? Or, are the authors implying that the actual variables are somehow weighted for these characteristics? This is not typically seen in accelerometry studies, and I believe requires further explanation. If this is something done at the modeling step, I recommend removing this information from the variable description section, as this can confuse the reader, and simply include it in the modeling section.

Response to reviewer

We agree that this could have been clearer and we have amended the text to clarify different aspects of the accelerometry assessment. Participants were asked to wear the accelerometer for 7 consecutive days from waking to going to bed. This provided daily step counts and daily minutes spent in moderate to vigorous physical activity (MVPA). Days where recorded wear time was less than 540 minutes were dropped from the analysis. Average daily step count and average daily MVPA were obtained using multi-level linear regression models (level 1 was day within individual and level 2 was individual). Daily steps (minutes of MVPA) were regressed on day-order-of-wear, day-of-week and month-of-wear. This method provides unbiased estimates of physical activity levels for each individual, which were used to examine associations with residential built environment variables (as outlined in the Statistical Analysis section). We have amended the text as below to demarcate between derivation of the physical activity variable, and the statistical methods used to relate the residential built environment to these derived measures of physical activity.

In the Methods section, under the ‘Accelerometer-derived PA outcomes’ subheading (page 7, lines 135-142) we have added the following: “Participants were asked to wear a hip-mounted ActiGraph GT3X+ accelerometer for 7 consecutive days during waking hours. This provided daily measures of steps and time spent in moderate-to-vigorous physical activity (MVPA), based on the standard threshold of ≥ 1952 counts per minute (14). We excluded days of accelerometer data where the registered wear time was less than 540 minutes. Multilevel linear regression models were fitted to allow for repeated measurements of PA, level 1 was day within individual and level 2 was individual. Daily steps (minutes of MVPA) were regressed on day-order-of-wear, day-of-week and month-of-wear, and an unbiased estimate of average daily steps (minutes of MVPA) obtained for each participant for weekdays and weekend days separately.”.

In the Statistical Analysis (page 11, lines 222-242) section we have altered the text to be clearer about the multi-level models used, as follows:

“Multi-level linear regression models were used to examine the association between daily PA (steps and minutes of MVPA) and residential built environment variables (walkability, distance to parks and accessibility to public transport) on weekdays and weekend days separately. Average daily steps (minutes of MVPA) were regressed on each built environment variable separately with further adjustment for (i) household as a random effect; (ii) sex, age group, ethnicity and aspirational housing group as fixed effects and household as a random effect; (iii) remaining built environment variables. Residuals from all models were checked for assumption of normality to confirm the analytic approach as appropriate.

Further models examined two ways in which built environment variables may contribute to socio-economic differences in PA levels. First, to examine the moderating effect of housing group on the association between the built environment and PA levels, interaction terms for housing group and built environment variables were added to the models and Stata post-estimation commands (testparm) were used to assess their statistical significance. Second, to examine whether differences in access to PA facilities across housing groups translate into differences in levels of PA, attenuation in PA outcomes for each housing group was examined following adjustment for each built environment variable.

Sensitivity analyses examined whether associations remained when the sample was restricted to (i) 837 participants who had data on both weekday and weekend days; (ii) 1,029 participants aged 18+.”.

7. Methods, MVPA variable

The use of a single day of valid accelerometry data (and of only 9 hours) certainly seems a bit low. Usual standards call for a minimum of 4 days with at least 10 hours of valid data per day. However, authors justify this by referring to a sensitivity analysis performed against complete data, which they define as that from participants with at least 4 days of 9 hours of data or more (instead of 10 hours). A single day of valid accelerometry data is thought to be sufficient for surveillance of populations (when truly representative samples are used), but I do worry that for this dataset, which is composed of a non-representative sample of London residents, and is not done for surveillance purposes, this may be a bit of a stretch.

Response to reviewer

Although some participants only provided one valid day of accelerometry data, the use of multi-level models using all available valid data from all participants provides unbiased estimates of daily MVPA and steps for each participant. These models adjusted for day of the week and day order of wear, both of which can affect steps and MVPA. Thus, we are not using the original data provided by the participant, but an adjusted average based on all available data from all participants. This method has been used successfully in other physical activity studies carried out by the authors and others (15). Durations of adequate wear-time to define habitual activity vary. Previous studies have often used a minimum of 3 to 5 days of accelerometry data (16), with as little as 480 minutes (17) per day to 600 minutes (18) per day of daily recording. As we outline in the paper, we chose to include all days of ≥540 minutes of registered accelerometry time during at least one day to lessen attrition bias and maximise inclusion of hard to reach groups, i.e., those from social housing who are likely to have lower compliance and record less PA data. This decision was part of the study design and was in our a priori Statistical Analysis Plan approved by the Steering Group (19). All physical activity related papers emanating from the ENABLE London study have followed this methodology (20-22). Moreover, as the reviewer correctly recognises, the sensitivity analysis referred to using 4 days of 9 hours in preference to 1 day, found no effect on the associations between PA and socio-demographic characteristics observed. In the current study, sensitivity analyses using 4+ days are not feasible as we analyse week days and weekend days separately, hence the need for inclusion of fewer days. We have added text to the Discussion (page 22, lines 481-488) to further justify the use of at least 1 day of 540 minutes.

8. Methods, Land-use mix and net residential density variables

Can the authors please provide further detail on the basis for the log-transformation of these variables? I haven’t seen this done before in other studies from other global settings. They state it is to “fit a comparable scale”. Comparable to what?

Response to reviewer

Walkability was derived by adding the z-scores of land use mix (LUM), residential density and street connectivity. LUM and residential density distributions were skewed. Because z-scores inform how far from the mean a data point is, it made sense to transform these 2 variables so that they better fit the normal distribution before deriving the z-scores. We have deleted the term “to fit a comparable scale” in the ‘Environmental variables’ section of the Methods (page 8, lines 161-162, and replaced this with “Land use mix and net residential density variables were log-transformed before deriving the z-scores due to their skewed distributions.”.

9. Methods

When explaining the rationale for the 1-KM network buffers, it might be good to state the average walking distance/time this represents (e.g., about a 10-15 minute walk from home in any direction for most people).

Response to reviewer

We have added that “a 1-km street network is equivalent to a 10-15-minute walk from home/centre of the buffer to its boundaries” to the ‘Environmental variables’ section of the Methods (page 8, lines 168-169).

10. Methods

Proximity to parks variable: other studies have found limited utility in these distance-based variables, including IPEN-adult, a 12 country study, where park density within 1-KM buffers was deemed a more useful predictor of physical activity. Did the authors consider using a buffer-based variable for parks too? Why was this discarded?

Response to reviewer

We agree with the reviewer that this is an important consideration. Because deriving built environment variables is time-consuming, we had to take a priori decisions as to which variables to create over others. The decision to use a distance-based rather than a density-based measure for park accessibility was based on the following.

First, a distance-based measure of park accessibility was likely to provide greater variability across our analysis sample than a density-based measure. In urban areas, where parks are a relatively sparse facility, a large portion of participants may not have had any parks falling within their 1km street-network home-centred buffer, leaving us with a considerably large group of participants with “no access”. The use of a distance-based measure overcomes this issue by attributing a quantitative value of accessibility even to those who do not have a park within the close vicinity of their home.

Second, London parks have a great deal of heterogeneity in size (from an average of 5 hectares for local parks to an average of 53 hectares for metropolitan parks) and in the type of facilities they provide. Therefore, there is a great deal of variation on functionality and utility. Size and facility types are known to influence how far individuals are ready to travel to use these facilities. Hence, a density-based measure of accessibility using a single buffer for the 3 types of parks would have not captured this attractivity dimension. A distance-based measure with models fitted separately for the 3 types of park appeared a better strategy than density-based measures for this setting.

11. Methods

I find it very interesting that car ownership was not deemed as a relevant confounder in this analysis, as this contradicts what we know about this variable in other parts of the world.

Response to reviewer

London has a highly developed and well-integrated public transport system with 64% of all trips undertaken by walking, cycling or public transit (TfL Travel in London Report 11, 2018 – http://content.tfl.gov.uk/travel-in-london-report-11.pdf). In addition, it has high street connectivity and high residential density. Hence, car use is much lower compared to elsewhere and we did not consider car ownership to be a major confounder. However, to respond constructively to the reviewer’s comment, we have carried out a sensitivity analysis to examine the effect of adjustment for car ownership on the main findings and can confirm that this has no material effect. We have added a comment to this effect (Statistical analyses, page 11, lines 242-244; Results, page 16, lines 353-354).

12. Statistical analysis

The first phrase is not clear. Authors mention a series of variables “were examined”. This could mean anything. Please be clear on the actual analytic procedures applied for “examining” these variables.

Response to reviewer

Thank you for highlighting this. On reflection, we feel this sentence is not needed and have removed it. We have also used this opportunity to re-write the Statistical Analysis section of the Methods (page 11) to make this clearer and more concise (see Response to comment #6 above).

13. Statistical analysis

Was a test for interaction actually performed to determine of day of the week (weekend vs. weekday) actually modifies the effect observed for some of these BE variables on the PA outcomes?

Response to reviewer

Literature suggests that patterns of PA are different on weekdays and weekends, and built environment variables may also have different effects on PA on weekdays compared with weekend days. Hence, average daily steps and MVPA were calculated separately for weekday and weekend days. We assessed if there was evidence of difference in associations on weekdays and weekend days by carrying out a Z-test to compare the coefficients on weekdays and weekend days. We have added the following to the Statistical Analysis section on page 11 (lines 227-229):

“Differences in weekday and weekend physical activity associations with built environment characteristics from the separate models were formally assessed using Z-tests.”.

On page 14 (lines 315-322), we have added the following paragraph to the ‘Built environment and PA-level associations’ section in the Results, after consideration of ‘On weekdays - ’ and ‘On weekend days - ’ effects:

“Comparison of weekday versus weekend day effects – There was formal evidence of a difference between weekday and weekend day physical activity associations with the built environment in some of the sociodemographic adjusted models (Model 2). Specifically, (1) increased steps on a weekend day compared with weekday per unit increase in walkability (p=0.041), (2) increased steps and MVPA at weekends versus weekdays per km increase in distance to local park (p=0.015, p=0.026 respectively), and (3) increased steps on a weekday compared with a weekend day associated with lower public transport accessibility (p=0.02). All other differences between weekend versus weekday physical activity associations with the built environment were not formally statistically significant (p>0.05).”.

14. Statistical analysis

The use of multilevel models seems adequate, but I am surprised at such a high ICC (0.3, wow!). It would be important to report what percentage of the sample shared a household with another sample participant. Most studies of this type only accept one participant per household to avoid this level of clustering.

Also, did the authors explore and, if needed, account for clustering at the neighborhood level (e.g., people residing in the same zip-code or small administrative geospatial unit?).

Response to reviewer

The participants invited to take part in the study were individuals and families who were looking to move to East Village. Thus, households with more than one participant were a feature of the study and we encouraged more than one member of a household to take part. At baseline, we recruited 1278 adults from 1006 households, with 40% sharing a household with another participant. Multi-level models were necessary to allow for clustering at household level, this approach was driven by the study design rather than the ICC, and was determined by our a priori statistical analysis plan, which was approved by our Steering Committee. The suggestion of clustering at the neighbourhood level is an interesting point but we did not explore this in our data, partly because controlling for neighbourhood effects by allowing for area clustering, might potentially dampen local influences of the built environment.

15. Statistical analysis

Could the authors explicitly state their hypotheses for the two different type of analyses proposed for examining the role of the BE in explaining PA-SES differences? I can follow the rational for the effect-measure modification analysis (testing for interaction terms of BE X SES variables). The simply adjustment by BE variables of the effect of SES on PA seems a repetition from the first modeling step, and implies confounding vs. effect measure modification, when, per the introduction, it seems that the authors have an effect modification hypothesis to begin with.

Response to reviewer

Following the reviewer’s suggestion, we have clarified in the last paragraph of the Introduction the aims of the study (pages 4-5, lines 88-95). Especially, we have better highlighted that the two pathways explored to examine the role of the BE in explaining PA-SES differences are (i) a moderating effect and (ii) a deprivation-amplification effect. The text now reads:

“The aim of this London (UK) based study is twofold. First, to assess whether the residential built environment is associated with the number of daily steps, and the amount of daily MVPA (min) accumulated, on weekdays and weekend days. Second, to explore two pathways in which the built environment may contribute to household-level socio-economic differences in PA levels on weekdays and weekend days: (i) a moderating effect, whereby different socio-economic groups relate differently to the physical activity facilities available in their neighbourhood; (ii) a deprivation-amplification effect (2), by which limited access to PA-promoting facilities for disadvantaged individuals translate into lower levels of PA.”.

16. Statistical analysis

Also, the authors don’t mention which test for interaction and alpha level they used in their analysis (hopefully this was not just based on the p-value of the interaction term in the regression model – but was the result of a formal test for interaction – please clarify).

Response to reviewer

Post-estimation commands were used to look at the effects of adding interaction terms. We have edited the Statistical Analysis section on page 11 (lines 233-236), as follows:

“First, to examine the moderating effect of housing group on the association between the built environment and PA levels, interaction terms for housing group and built environment variables were added to the models and Stata post-estimation commands (testparm) were used to assess their statistical significance.”.

17. Statistical analysis

Outcome variables in models (steps and mins of MVPA/day): were these transformed in any way for achieving normality? Physical activity data are usually not normally distributed, hence the use of linear regression models usually requires a data transformation step. Alternatively, one could consider the use of logistic regression models (multinomial, politomous, or binary).

Response to reviewer

We have checked for normally distributed data before running our linear regression models. As we limited our analysis to those with days of ≥540 minutes of physical activity data, the distributions of average daily steps and average daily MVPA appeared normally distributed. We also checked the distribution of the residuals, which were also normally distributed. Hence, we feel confident that our methods of analysis are appropriate. We have added a sentence to the statistical methods section (lines 229-230) to confirm that the analytic approach was appropriate, as follows: “Residuals from all models were checked for assumption of normality to confirm the analytic approach as appropriate.”.

18. Statistical analysis

Did the authors examine potential multicollinearity in their models? This is usually a concern with some built environment variables in a single model. Please outline the methods via which this was examined.

Response to reviewer

Multicollinearity was tested in our models using the Variance Inflation Factor. VIF values did not exceed 2, downplaying concerns over multicollinearity. We have amended the Results section (lines 287-290) of the manuscript where concerns over multicollinearity had already been raised, as follows: “Although adjustment for other residential built environmental factors led to an inflation of the regression estimate and a widening of the confidence interval compared with models adjusting only for sociodemographics, Variance Inflation Factor (VIF) values did not exceed 2, downplaying concerns over multicollinearity.”.

19. Results, Table 2

Table 2 title: Should read: “Regression estimates for the association between residential built environment favors with physical activity outcomes (daily steps and mins/wk of MVPA), among… XXX [insert study name, year of data collection, etc.)”. It is customary to list the independent variable first and the dependent variable last, not the other way around.

Response to reviewer

Thank you for this suggestion. We have changed the title for Table 2 to ‘Regression estimates for the association between residential built environment variables with physical activity outcomes (daily steps and daily minutes of MVPA) in the ENABLE London study’ as suggested. Numbers and dates of data collection are now in the footnote.

20. Results

For all table/figure titles: add information on the study and sample, and year data were collected, as tables and figures should be “stand-alone” items.

Response to reviewer

We have changed the titles of the tables to add “ENABLE London study” at the end of the titles and added footnotes to indicate numbers of participants and years of data collection.

21. Results

Figure 2 is very nice and helpful for the reader.

Response to reviewer

Thank you for this comment.

22. Results

I like the way the results for the examination of the role of the BE on the PA-SES relation are presented by two potential pathways. However, it would have been much more helpful to lay out the two hypothesized pathways since the introduction and methods (Statistical analysis) section, so the readers understand why both approaches (test for effect modification vs. BE as a confounder of the SES-PA relation) were explored. Otherwise the “two pathways” seem more a bit of an afterthought. For this purpose, it might be good for the authors to present a figure with the hypothetical DAGs of the two pathways they are exploring, a priori (i.e., before the results section).

Response to reviewer

We thank the reviewer for this suggestion. We have amended the Methods section (lines 232-239) so that the analyses described here relate more directly to the aims they serve. The paragraph now reads: “Further models examined two ways in which built environment variables may contribute to socio-economic differences in PA levels. First, in order to examine the moderating effect of housing group on the association between the built environment and PA levels, interaction terms for housing group and built environment variables were added to the models and Stata post-estimation commands (testparm) were used to assess their statistical significance. Second, in order to examine whether differences in access to PA facilities across housing groups translate into differences in levels of PA, attenuation in PA outcomes for each housing group was examined following adjustment for each built environment variable.”.

23. Discussion

I enjoyed reading the interpretation of the results regarding the counter-intuitive finding on longer distance to small parks being associated to higher levels of MVPA on weekends. This makes a lot of sense and it is not something the readers could have derived just by seeing the estimates. The discussion on the public transit results should be re-framed. I would suggest not opening with a statement on how they found a positive association with access to transit and PA, given that in reality once the models were adjusted this was not significant (the authors do state this, but I just don’t see the need to open with a statement of a positive relation). I think a deeper discussion is warranted to explain why their findings are inconsistent with most other reports of transit and PA. It is counterintuitive that those with lower availability to public transit have higher levels of PA, in a model that is also adjusted for SES. I wonder if adjusting for car ownership would have made a difference for this specific variable [access to transit] (?). The discussion on this topic should also include some reflection on the difference in measures of access to transit. Most other studies have relied on transit stop density measures (count of public transit stops within a buffer), and thus, are not truly comparable to the measure used in this study.

Response to reviewer

Following the reviewer’s suggestion, we have now revamped the discussion on accessibility to public transport (see lines 400-423). Especially, we have (i) better put our findings in light with previous literature findings, (ii) discussed whether accessibility to a motor vehicle may partly explain the association, and (iii) acknowledged that our variable is reflecting the concept of accessibility in a slightly different way to more simple density-based measures commonly used, as follows:

“Positive associations between accessibility to public transport and PA levels have been found worldwide (23), including in the UK (24), suggesting that interest in use of public transport may decrease with decreasing accessibility. For instance, Sallis et al. found that accessibility to public transport, measured as the density of public transport access points within a 1-km buffer centred on the home address, positively relates to adult PA levels in 14 cities worldwide, including Stoke-on-Trent, UK (24). Yet, our findings did not fully align with this conclusion, as the positive association between accessibility to public transport and steps and MVPA found on weekends failed to reach significance after adjustment for sociodemographic factors. More surprisingly, the association becomes negative on weekdays, with participants experiencing a low accessibility to public transport taking more daily steps compared to those with high accessibility. To our knowledge, such findings have not been reported in the literature and raise several questions. Participant’s heavy reliance on public transportation for inescapable weekday activities, like work, may explain why those with poor accessibility to public transport walk more on weekdays than those who live closer. Examining whether this negative association between public transport accessibility and PA on weekdays holds for individuals who have alternative ways of travelling (e.g., car) offers a further avenue for research. On a different note, our measure of public transport accessibility is not strictly comparable to density-based measures (simple count of public transit stops within a given area) used in other studies, in that it also accounts for frequency and reliability of service. Yet, our variable reflects a concept of public transport accessibility not that different to more basic density-based variables and is unlikely to be responsible for inverting the direction of the association.”.

24. Limitations paragraph

The use of a single day of accelerometry data with at least 9 hours is not in line with previous studies and should be acknowledged.

Response to reviewer

We have highlighted in the Strengths/Limitations section of the Discussion that our choice to include all days of ≥540 minutes of registered accelerometry time during at least one day was made to lessen attrition bias and maximise inclusion of hard to reach groups (i.e., those from social housing who are likely to have lower compliance and record less PA data). We have also presented this choice in light of previous studies that used accelerometer data. The other crucial point is that weekend estimates can be at best based on only two days of data. The text (lines 481-488) now reads: “Durations of adequate wear-time to define habitual activity vary. Previous studies have often used a minimum of 3 to 5 days of accelerometry data (16), with as little as 480 minutes (17) per day to 600 minutes (18) per day of daily recording. We chose to include all days of ≥540 minutes of registered accelerometry time during at least one day to lessen attrition bias and maximise inclusion of hard to reach groups, i.e., those from social housing who are likely to have lower compliance and record less PA data. This decision was part of the study design and was in our a priori Statistical Analysis Plan approved by the ENABLE London Steering Committee (19).”.

25. Limitations paragraph

The transit access measure, although apparently more sophisticated than what others use, is not comparable to what most researchers use when examining the role of transit infrastructure on PA.

Response to reviewer

We have raised this concern in the Discussion (lines 418-423), as follows:

“On a different note, our measure of public transport accessibility is not strictly comparable to density-based measures (simple count of public transport stops within a given area) used in other studies, in that it also accounts for frequency and reliability of service. Yet, our variable reflects a concept of public transport accessibility not that different to more basic density-based variables and is unlikely to be responsible for inverting the direction of the association.”.

26. Limitations paragraph

The Uncertain Geographic Context Problem and this study’s inability to address it should be listed as a limitation.

Response to reviewers

We have referred to the Uncertain Geographic Context Problem in the limitation paragraph (lines 496-504), as follows:

“Moreover, partly because we overlooked non-residential exposures to the built environment, we were not able to align with calls for addressing the Uncertain Geographic Context Problem (i.e., the extent to which areal units of measurement deviate from the geographic context truly experienced by individuals) (25). Although the spatial unit used to derive our environmental exposure complies with the wide majority of studies that have looked at the way the residential built environment relates to health behaviours (i.e. using an home-centred buffer within a walkable distance radius), focusing on the residential neighbourhood only may have led to misestimation of the association between the residential environment and health behaviours (26).”

27. Limitations paragraph

Another major limitation is the authors simply assumed differences in the relation of the BE and PA by weekday (effect modification by type of weekday), which is reasonable, but a stronger study analysis would have included a confirmation of this with a test for interaction, followed by the stratified analysis.

Response to reviewers

As highlighted in our response to comment #13, models were fitted separately for weekends and weekdays based on literature searches suggesting varying patterns of PA over the week. We assessed if there was evidence of difference in associations on weekend days and weekdays by carrying out a Z-test to compare the coefficients on weekend days and weekdays. This has been added to the ‘Statistical Analysis’ section of the Methods (page 11, lines 227-229).

28. Limitations paragraph

Finally, selection bias due to a) the response rate, and b) the eligibility criteria (this is not a representative sample of London residents, rather, it targets people with a desire to move) needs to be fully disclosed, with a description of how these factors are likely limiting the external validity of the study. In particular, because a reason for moving might be the BE features of the new neighborhood where people are hoping to move to.

Response to reviewers

As highlighted in our response to comment #5, we have acknowledged this in the Limitation paragraph (lines 488-494), as follows:

“This study draws on the baseline sample of a natural experiment, where recruits were deemed eligible if they were seeking to relocate to East Village. Such sampling strategy limits external validity, as the participants may not be truly representative of the London population as a whole. Moreover, the cross-sectional study design is prone to selection bias (i.e., those living in less walkable neighbourhoods are intrinsically different to those who do not), and therefore restricts interpretations about direction of effects.”.

Supporting references

1. Ding J, Wai KL, McGeechan K, Ikram MK, Kawasaki R, Xie J, et al. Retinal vascular caliber and the development of hypertension: a meta-analysis of individual participant data. J Hypertens. 2014;32(2):207-15.

2. Macintyre S, Ellaway A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Soc Sci Med. 2002;55(1):125-39.

3. Frohlich KL, Corin E, Potvin L. A theoretical proposal for the relationship between context and disease. Sociology of health & illness. 2001;23(6):776-97.

4. Ogilvie D, Lamb KE, Ferguson NS, Ellaway A. Recreational physical activity facilities within walking and cycling distance: Sociospatial patterning of access in Scotland. Health & Place. 2011;17(5):1015-22.

5. Higgs G, Langford M, Norman P. Accessibility to sport facilities in Wales: A GIS-based analysis of socio-economic variations in provision. Geoforum. 2015;62:105-20.

6. Pliakas T, Wilkinson P, Tonne C. Contribution of the physical environment to socioeconomic gradients in walking in the Whitehall II study. Health Place. 2014;27:186-93.

7. Barbosa O, Tratalos JA, Armsworth PR, Davies RG, Fuller RA, Johnson P, et al. Who benefits from access to green space? A case study from Sheffield, UK. Landscape and Urban Planning. 2007;83(2-3):187-95.

8. Hillsdon M, Panter J, Foster C, Jones A. Equitable access to exercise facilities. Am J Prev Med. 2007;32(6):506-8.

9. Forsyth A, Oakes JM, Lee B, Schmitz KH. The built environment, walking, and physical activity: Is the environment more important to some people than others? Transportation research part D: transport and environment. 2009;14(1):42-9.

10. Manaugh K, El-Geneidy A. Validating walkability indices: How do different households respond to the walkability of their neighborhood? Transportation research part D: transport and environment. 2011;16(4):309-15.

11. Van Dyck D, Cerin E, De Bourdeaudhuij I, Salvo D, Christiansen LB, Macfarlane D, et al. Moderating effects of age, gender and education on the associations of perceived neighborhood environment attributes with accelerometer-based physical activity: The IPEN adult study. Health Place. 2015;36:65-73.

12. Winters M, Barnes R, Venners S, Ste-Marie N, McKay H, Sims-Gould J, et al. Older adults' outdoor walking and the built environment: does income matter? BMC Public Health. 2015;15:876.

13. Cummins S, Curtis S, Diez-Roux AV, Macintyre S. Understanding and representing 'place' in health research: a relational approach. Soc Sci Med. 2007;65(9):1825-38.

14. Freedson PS, Melanson E, Sirard J. Calibration of the Computer Science and Applications, Inc. accelerometer. Med Sci Sports Exerc. 1998;30(5):777-81.

15. Harris T, Kerry SM, Limb ES, Victor CR, Iliffe S, Ussher M, et al. Effect of a Primary Care Walking Intervention with and without Nurse Support on Physical Activity Levels in 45- to 75-Year-Olds: The Pedometer And Consultation Evaluation (PACE-UP) Cluster Randomised Clinical Trial. PLoS Med. 2017;14(1):e1002210.

16. Trost SG, McIver KL, Pate RR. Conducting accelerometer-based activity assessments in field-based research. Med Sci Sports Exerc. 2005;37(11 Suppl):S531-43.

17. Miller GD, Jakicic JM, Rejeski WJ, Whit-Glover MC, Lang W, Walkup MP, et al. Effect of varying accelerometry criteria on physical activity: the look ahead study. Obesity (Silver Spring). 2013;21(1):32-44.

18. Katzmarzyk PT, Champagne CM, Tudor-Locke C, Broyles ST, Harsha D, Kennedy BM, et al. A short-term physical activity randomized trial in the Lower Mississippi Delta. PLoS One. 2011;6(10):e26667.

19. Ram B, Nightingale CM, Hudda MT, Kapetanakis VV, Ellaway A, Cooper AR, et al. Cohort profile: Examining Neighbourhood Activities in Built Living Environments in London: the ENABLE London-Olympic Park cohort. BMJ Open. 2016;6(10):e012643.

20. Limb ES, Procter DS, Cooper AR, Page AS, Nightingale CM, Ram B, et al. The effect of moving to East Village, the former London 2012 Olympic and Paralympic Games Athletes' Village, on mode of travel (ENABLE London study, a natural experiment). Int J Behav Nutr Phys Act. 2020;17(1):15.

21. Nightingale CM, Limb ES, Ram B, Shankar A, Clary C, Lewis D, et al. The effect of moving to East Village, the former London 2012 Olympic and Paralympic Games Athletes' Village, on physical activity and adiposity (ENABLE London): a cohort study. Lancet Public Health. 2019;4(8):e421-e30.

22. Nightingale CM, Rudnicka AR, Ram B, Shankar A, Limb ES, Procter D, et al. Housing, neighbourhood and sociodemographic associations with adult levels of physical activity and adiposity: baseline findings from the ENABLE London study. BMJ Open. 2018;8(8):e021257.

23. Li F, Harmer PA, Cardinal BJ, Bosworth M, Acock A, Johnson-Shelton D, et al. Built environment, adiposity, and physical activity in adults aged 50-75. Am J Prev Med. 2008;35(1):38-46.

24. Sallis JF, Cerin E, Conway TL, Adams MA, Frank LD, Pratt M, et al. Physical activity in relation to urban environments in 14 cities worldwide: a cross-sectional study. The Lancet. 2016;387(10034):2207-17.

25. Kwan M-P. The uncertain geographic context problem. Annals of the Association of American Geographers. 2012;102(5):958-68.

26. Chaix B, Duncan D, Vallee J, Vernez-Moudon A, Benmarhnia T, Kestens Y. The "residential" effect fallacy in neighborhood and health studies: formal definition, empirical identification, and correction. Epidemiology. 2017.

Submitted filename: PloSOneComments_Response_update (280520).docx

Click here for additional data file.

27 Jul 2020

Weekend and weekday associations between the residential built environment and physical activity: findings from the ENABLE-London Study

PONE-D-19-33687R1

Dear Dr. Owen,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Adewale L. Oyeyemi, Ph.D

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors addressed all of my comments in a thoughtful manner. I am satisfied with this new version of the manuscript. Thank you again for the opportunity to review this important paper.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Deborah Salvo

3 Aug 2020

PONE-D-19-33687R1

Weekend and weekday associations between the residential built environment and physical activity: findings from the ENABLE-London Study

Dear Dr. Owen:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Adewale L. Oyeyemi

Academic Editor

PLOS ONE