The aim of this study was to evaluate the effectiveness of Devon Active Villages, a community-level physical activity intervention delivered to rural villages. The Devon Active Villages intervention had no effect on the proportion of people active at recommended levels, and there was uncertainty regarding the true size of the increase in the number of MET-minutes per week reported, as reflected in the 95% confidence interval for the mean difference. It is possible that the intervention was effective at the individual level, but the low levels of population penetration prevented any observable effect at the village level.
Ensuring sufficient penetration and reach across a community to attain a population-level impact is one of the most difficult aspects of community-level interventions . Although few studies have reported population participation rates, one review found that the highest exposures were obtained for public information and screening activities rather than more intensive interventions, and that population penetration rates ranged from 4-60% . In the Devon Active Villages intervention, there was only a limited budget for promotion activities, which may have contributed to the low levels of participant awareness in the research study. However, in rural areas with an ageing population, it is arguably more difficult to find effective ways of communicating new physical activity opportunities to sedentary individuals, because most methods rely on participants seeing an advertisement in the local area. Media activities (e.g., television, radio) can achieve greater levels of reach, but can also be expensive for localised community-based interventions, such as the Devon Active Villages intervention. Despite the intervention incorporating a local needs-led approach, the budget only allowed for 1-2 activities per age group per village. Thus, it is possible that the provided activities did not appeal to all residents who were aware of the intervention.
Baker et al.  conducted a systematic review of community-level physical activity interventions and found that only three out of the 25 included studies reported positive changes in physical activity behaviour –. Jiang et al.  conducted an intervention in urban communities within Beijing, finding a reported increase in regular physical activity in the intervention group (adjusted relative risk 1.20, 95% CI 1.09 to 1.31). However, the intervention achieved substantial penetration within the community (73% participation), through ‘door-to-door’ hand-outs and individualised counselling by health practitioners. In the Finnmark Intervention study , a sport and activity-based intervention in a small artic community in Norway, males reported a significant increase (p = 0.047) in physical activity behaviour six years after the initial baseline measurement. No change was found in the female population, however. Similar to the Beijing study, the Finnmark Intervention reached large segments of the population, through community engagement, mass media, and individual counselling. The only other study in the review to find an increase in physical activity was the Rockhampton 10,000 Steps Project , where the proportion of females who met the recommended guidelines increased significantly from baseline to post-intervention. The study found no evidence of physical activity behaviour change in males. Again this intervention involved a large number of components, including social marketing, pedometers, individual counselling, partnering with local organisations, and environmental changes.
In contrast, the studies that reached a smaller proportion of the population, either through low cost or low activity, found no intervention effect on physical activity . For example, the low cost of one intervention in rural municipalities in Denmark limited the amount of intervention activities that took place, resulting in the intervention being purely mass-media . Simon et al.  was one example of a low reach intervention, aimed at school communities in France. Although the intervention initially aimed to reach the whole community, in actuality, the vast majority of the intervention activities were targeted at one specific section of it. This was similar to Devon Active Villages, where many of the intervention activities were targeted at a specific group within the community (i.e., basketball for primary school children, or armchair aerobics for older adults). From the population penetration rates achieved by Devon Active Villages, it is clear that the intervention would be classed as ‘low reach’. Therefore, the results of the present investigation are in line with previous research, where interventions with low reach failed to have an effect on physical activity behaviour .
Despite the above, the intervention was associated with stronger activity habits, suggesting that those in the intervention mode perceived themselves to be physically active, but did not report a greater level of physical activity than controls. Physical activity habits was the only outcome for which there was evidence of an effect. We are not aware of any other community interventions that have reported physical activity habit as an outcome.
The majority of reported intervention opinions were positive, suggesting that the intervention was well received by the small proportion of participants who were aware of its existence.
Strengths and limitations
Strengths of the study include the large sample size (>10,000) and the large number of participating villages. Incorporating multiple data collection periods into the research meant that it was possible to analyse both whether the intervention had an immediate effect on physical activity that later subsided, or whether the intervention effect was delayed. Each village acted as its own control, meaning communities were not subjected to “best-fit” matching with control communities. Another strength is that the period in which villages first received the intervention was randomly allocated, eliminating any selection bias. Indeed, in a recent review of community-level physical activity interventions , only one study out of 25 used randomisation to allocate communities .
This study fills a gap in the literature by being the first to use a stepped wedge cluster randomised trial design to evaluate a physical activity intervention. Examples of previous stepped wedge investigations include examination of the efficacy of Hepatitis B vaccinations , the effect of housing improvements on respiratory health symptoms , and different tuberculosis treatments on number of disease episodes . The stepped wedge trial design was the most appropriate study design for this intervention for three reasons: first, there was a necessity to deliver the intervention in waves due to limited resources; second, once the intervention was implemented it was never fully taken away; and third, the intervention was delivered to all eligible communities of a certain size within the county . Despite the stepped wedge trial design requiring greater data collection and longer trial duration , it was successfully able to evaluate a pragmatic community-level physical activity intervention.
Despite being better than anticipated, and comparing well with other survey studies from the United Kingdom (15.9% , 17% ), the response rate was low (32.2%). Non-response bias often occurs in survey studies, where non-responders may differ in some way from those who do respond . The participants in the present research were similar to the wider population in terms of IMD score and the population density of the village they resided in. Compared to the wider population, however, the survey respondents tended to be older, with a greater proportion being female. Previous research suggests females and older adults are often over-represented in health surveys . Survey respondents also tend to report being healthier and doing more physical activity than the general population . Two-thirds of the present research population reported meeting the recommended guidelines, suggesting that those of higher activity levels were over-represented. However, previous research suggests that the IPAQ-SV has a tendency to over-report time spent doing physical activity –, with one review finding that the IPAQ-SV over-reported physical activity on average by 106% (Range 36-173%) . Nevertheless, if the more physically active are over represented in the study it could be that the intervention effect is smaller for these people than those who did not respond and who might not normally engage with physical activity.
Participants may have over-reported exposure to the Devon Active Villages intervention events because they believed this response to be favourable to the researchers . However, the high level of consistency between the reported participation and participation according to village registrations suggests that such reporting bias was not present in this study. In addition, while the generally positive intervention opinions may have been an accurate representation of how well the intervention was received, participants may have reported overly positive opinions in an attempt to stop any intervention funding from being withdrawn .
The main limitation of this research is the use of self-reported data. Self-reported outcome measures of physical activity tend to include bias due to social desirability and may lead to some misclassification, with some participants finding it difficult to recall activities from the past seven days. Nevertheless, there is no reason to believe that any misclassification was systematically different with regard to intervention or control group. Furthermore, established and validated measures were used where possible (e.g., the IPAQ-SV to measure physical activity). Another potential limitation is the nature of the study sample. Although the intervention was available to all age groups, the study focused on adults, because child and youth physical activity comprises a separate body of literature with different guidelines and understanding about what constitutes physical activity behaviour in this age group.
Repeated cross-sectional samples of participants were used in this research in order to measure the community-level impact of the intervention on physical activity levels, rather than follow individuals over time to detect individual changes in behaviour. Although it is possible that the repeated cross-sectional samples included people new to the village who were not exposed to the intervention, it is perhaps more likely that there was contamination due to people in control villages participating in neighbouring village intervention activities. Both of these factors would have attenuated intervention effects . Finally, it may be that the reach, intensity and duration of the intervention were insufficient to achieve a population-level impact.
The results of this research indicate that unless community-level physical activity interventions can reach a substantial proportion of the target population they are unlikely to be able to change the population prevalence of physical activity. This research also demonstrated that it is possible to rigorously evaluate pragmatic community-level physical activity interventions using novel research techniques. This research is also the first to use a stepped wedge cluster randomised trial design to evaluate a community-level physical activity intervention. The stepped wedge design was suitable for evaluating the Devon Active Villages intervention, because it was by necessity delivered in waves, administered to all eligible communities in the population, and, once a community received the intervention, it was never fully taken away. This study also adds to the limited research available on physical activity in rural communities from England.
It is advocated that future evaluation studies consider the use of the stepped wedge cluster randomised trial design for evaluating health interventions, especially for community-level physical activity interventions. Additionally, more rigorous evaluations of community-level physical activity interventions are needed to help understand what works in altering population prevalence. In order to improve validity and reliability, these intervention evaluations should include objective measurements (e.g., accelerometry data). Finally, more research is warranted on how to achieve greater community penetration/engagement in community-level physical activity interventions.