Impact of free access to leisure facilities and community outreach on inequalities in physical activity: a quasi-experimental study

Background There are large inequalities in levels of physical activity in the UK, and this is an important determinant of health inequalities. Little is known about the effectiveness of community-wide interventions to increase physical activity and whether effects differ by socioeconomic group. Methods We conducted interrupted time series and difference-in-differences analyses using local administrative data and a large national survey to investigate the impact of an intervention providing universal free access to leisure facilities alongside outreach and marketing activities in a deprived local authority area in the northwest of England. Outcomes included attendances at swimming and gym sessions, self-reported participation in gym and swim activity and any physical activity. Results The intervention was associated with a 64% increase in attendances at swimming and gym sessions (relative risk 1.64, 95% CI 1.43 to 1.89, P<0.001), an additional 3.9% of the population participating in at least 30 min of moderate-intensity gym or swim sessions during the previous four weeks (95% CI 3.6 to 4.1) and an additional 1.9% of the population participating in any sport or active recreation of at least moderate intensity for at least 30 min on at least 12 days out of the last four weeks (95% CI 1.7 to 2.1). The effect on gym and swim activity and overall levels of participation in physical activity was significantly greater for the more disadvantaged socioeconomic group. Conclusions The study suggests that removing user charges from leisure facilities in combination with outreach and marketing activities can increase overall population levels of physical activity while reducing inequalities.


Interrupted time series.
Ln(gymswim)= B 1* re:fresh +B 2 time1+B 3 time2 +quarter Where gymswim is the total number of attendances at leisure centres in Blackburn with Darwen for gym and/or swim activities in each quarter.
re:fresh is a dummy variable that is 0 before 3 rd quarter of 2008 and 1 after.
Time1 is a time trend term for before the 3 rd quarter of 2008, (set to zero after) Time1 is a time trend term for after the 3 rd quarter of 2008, (set to zero before) Quarter is a set of dummy variables for the four quarters of the year.
As is shown in web Figure 3 there is some evidence of autocorrelation in the data. The regression was therefore estimated with Newey-West standard errors. We used the automatic lag selection procedure outlined by Newey and West 1 to set the maximum lag order of autocorrelation, this identified a maxim lag of 8 as appropriate. In practice as the effect size if very large this made very little difference to the findings, sensitivity analysis using other maximum lags from 1 to 8, gave results that were identical for the first two decimal places, and all p values were <0.001.

Analysis of residuals distribution and autocorrelation.
Web Appendix 2. Alternative model specifications.
Web Table 1 Estimated increase in swim and/or gym activity associated with the introduction of re:fresh estimated from the ITS regression analysis. % increase in activity estimated as the difference in logged number of attendances.

Outcome
Realtive increase in activity associated with introduction of re:fresh (RR)

Difference in Differences.
Outcome ikt =B 1 interv k +B 2 After t +B 3 After k *interv t +B 4 SES ikt + B 5 SEX ikt + B 6 ethnicity ikt + B 7 AGE ikt + B 8 AGESQ ikt + B 9 Year t Where interv is a dummy variable indicating respondents in Blackburn with Darwen and is 0 otherwise.
After is a dummy variable that is 0 before 2008 and 1 after.
After k *interv t is the interection between the two -B 3 if therefore the DiD parameter.
Model included survey weights to adjust for non-response. We estimated robust standard errors clustered at the local authority level to allow for within LA correlation due to sampling design.

SES is a set of dummy variables for each socioeconomic group
Ethnicity is a set of dummy variables for each ethnic group AGE ikt is the age of respondent I in local authority k at time period t.

AGESQ is the square of AGE
Year is a continuous variable indicating the survey year.
Alternative difference in differences analyses.

Separate analysis for gym and swimming participation.
Web Figure 7 . Estimates of the effect of the introduction of re:fresh from the difference-in-differences analysis on (1) % participating in gym activity and (2) % participating in swim activity at least once in the past month Results for all socioeconomic groups in Blackburn with Darwen and separately for 3 socioeconomic groups. Effect sizes indicate the additional percentage of the population participating due to the intervention.