Background
Physical exercise in childhood is important for children’s current health and wellbeing, and for their future health outcomes [
1]. Lower rates of physical activity have been associated with cardiovascular and cardiometabolic risk factors [
2,
3], and poorer mental health outcomes [
1,
4]. Lack of moderate to vigorous physical activity (MVPA) has also been linked to overweight and obesity in children [
5], although there are challenges in identifying the direction of causation [
6]. In the UK, the National Child Measurement Programme (NCMP) has identified increasing numbers of overweight and obese children in England, with enduring inequalities by area deprivation and region in all countries of the UK [
7,
8]. There are challenges in reliably recording rates of MVPA in children [
9], but indicators suggest that few children in the UK meet recommended targets for MVPA [
10,
11]. There are then, good grounds for developing and promoting interventions which can increase the amount of physical activity, particularly MVPA, that children undertake.
The Daily Mile (TDM) is one promising intervention for increasing physical activity in children. Originating in Stirling, Scotland, in 2012, the TDM initiative requires school teachers to take schoolchildren out of their classroom to run for 15 min per day, which equates to a distance of approximately one mile. TDM is a scheme promoted by The Daily Mile Foundation, a non-profit organisation funded by INEOS, a multinational petrochemicals company. According to the Foundation website, the aims of TDM are to improve the ‘physical, mental, emotional, and social health and wellbeing’ of children [
12]. Teachers can implement it at any time of the day, and in varied weather conditions, without any need for special equipment. It is therefore designed to be a simple, free (in principle) and sustainable intervention which is inclusive for all children, including ‘children with mobility difficulties [who] should be fully supported to take part as well’ [
12]. The scheme has been acceptable to schools, parents and children, and has now been taken up increasingly across the UK, and beyond, with considerable policy support [
13]. As of February 2019 there are over 7000 schools and nurseries taking part, with over 4000 of those schools in the UK [
12]. TDM is in principle easy for schools to implement requiring limited time out of class, no special clothing, and no staff training. Indicative results from a pilot study of TDM in Stirling found that children in an intervention school had increased the number of minutes per day of MVPA, increased physical fitness and decreased skinfold measures compared to children in a non-intervention control school [
14]. One RCT of TDM is already underway in Birmingham [
15], which will provide invaluable evidence on impact - including BMI, quality of life, wellbeing, and academic attainment - albeit in the context of a trial in one region.
There is, then, good evidence that increases in MVPA will improve children’s current and future health [
1], and that taking part in TDM can contribute to such an increase in principle [
14]. However, there is no robust evidence to date of whether rolling out TDM in schools more generally is likely to have a positive impact on the public health and what the un/intended impacts of the intervention outside of trial studies/RCTs might be. This study contributes to this evidence base by examining how TDM is implemented in a ‘real world’ setting. Examining public health impact in ‘real world’ settings is crucial as pilot trials or RCTs are often unrepresentative of everyday practice: there may be increased input to achieve fidelity; participants in trials may be more committed to the intervention, and context plays an important role in shaping the intervention [
16]. Impact is here thus understood through ‘real world’ settings of adoption (e.g. schools), as well as the implementation consequences, both the intended and unintended ones, and the ongoing efforts to maintain or sustain the intervention, within the contexts in which they are situated [
17]. Often RCTs and pilot trial data cannot account for the ‘moderating factors’ of interventions in real-world settings, which can create difficulties for the transferability of RCTs to practice settings [
18].
There has been some scepticism from the public health community on both the likely effects and sustainability of TDM, citing concerns about the minimal impact of an additional 15 min of activity, concerns about it displacing more effective forms of exercise in the school day (such as active play), and the risks of putting children off future participation in sports and exercise if the activity does not meet their needs for meaningful activity [
19]. Despite its perceived simplicity and the fact that it is promoted as ‘free’, the scheme has a number of components (time, organisation, staff input, a safe accessible space to run), which must be assembled for the intervention to be implemented. As in other school-based interventions [
20], there is likely to be considerable variation in implementation across school settings, and it may be adapted in various ways by individual class teachers and participating pupils. Schools not adopting TDM may be taking part in other, similar schemes. Individual children taking part may or may not be also participating in other forms of physical activity, or broader healthy schools initiatives. These factors present significant evaluation challenges in identifying the effects of TDM as a single intervention, and in identifying what the essential components of an effective intervention might be.
The focus of this study is to understand how TDM is being implemented in practice, and to provide a firmer foundation for future evaluations of the public health impact of this and similar interventions. Our aims thus were to: identify factors that impact on whether TDM is adopted in particular settings (e.g. schools and classes); to describe how TDM is being implemented; to identify intended and unintended consequences; and to describe factors that affect whether implementation is maintained. A secondary aim of the study was to identify potential design considerations for future evaluation.
Methods
Below, we outline the study design, and provide an overview of study participants and our analysis. We use the COREQ checklist [
21] to guide reporting.
Study design
To understand how TDM operates as a public health intervention in a naturalistic setting, we undertook a study in 2018 (with fieldwork conducted between May and December) of what happened when the scheme was promoted for primary schools (for children aged 5 to 11) across one local authority area, the south London borough (LB) of Lewisham. To understand adoption and implementation of TDM in Lewisham, we undertook a rapid ethnographic assessment [
22]. This is a pragmatic, focused and mixed-method approach, using analysis of data from mixed-methods (observations, interviews and focus groups, as well as secondary data analysis) to generate evidence for evaluation [
22,
23].
Setting and population
Lewisham is a diverse borough of south London, UK, with 29.6% of children living in income deprived households [
24]. The borough includes 69 primary schools in the state sector educating pupils aged between 5 and 11 years. These schools range from small stand-alone primary education providers to larger federations of schools which include primary years. The Health & Wellbeing Strategy in Lewisham has ‘achieving a healthy weight’ as a priority area for action, and its Public Health Report of 2016 [
25] set out a whole system approach to obesity which included supporting three key initiatives: Sugar Smart (a campaign to reduce sugar in diets); greater use of parks within the borough; and uptake of TDM across its primary schools. This intervention (TDM) was therefore being supported by the public health directorate, who were encouraging schools to sign up.
Sampling strategy and procedure
To examine experiences in south London we used a purposive sampling approach to identify and select stakeholders and cases (i.e. schools) from the locality who had varied experiences with TDM adoption and implementation. This included Lewisham public health practitioners (
n = 3), who were interviewed about their experience implementing the project in the borough. Lewisham Public Health also provided data about school characteristics, TDM adoption, percent of pupils from Black and Minority Ethnic (BME) communities and percent of pupils eligible for free school meals. Recruitment was undertaken through invitations sent through a monthly public health newsletter to schools, direct emails to head teachers, as well as recruitment at a local event focused on physical education and sports in schools. Of those schools who indicated an interest we initially selected 6 schools, with varying levels of success in rolling out the intervention. However, one school chose not to participate, with 5 schools in total participating in this study. The schools included a range of school types (e.g. faith schools, community schools) with varied levels of TDM implementation (see Table
1).
Table 1
Schools in the Qualitative Component of the Rapid Ethnographic Assessment
School01 | 3 years or less, Whole School | 17.1 | 89.8 | Yes |
School02 | 2 years or less, Select Classes | 9.5 | 63.1 | No |
School03 | 2 years or less, Select Classes | 16.4 | 87.1 | Yes |
School04 | 2 years or less, Select Classes | 25.1 | 77.5 | No |
School05 | Ad hoc, Select Classes | 22.9 | 73.5 | No |
Qualitative data generation
Qualitative data were collected at each school. This included in-depth interviews (n = 22); focus groups (n = 11) with 41 participants; and participant observation of 49 Daily Miles across 12 classes in the five schools. Interviews and focus groups were recorded and lasted for approximately 30–45 min in length. All participants self-selected to take part and provided their own and, where applicable, parent/guardian consent. Participants from the 5 schools did not know the researcher prior to the commencement of the study.
Of the 63 participants who participated, in-depth interviews were undertaken with public health practitioners (n = 3), headteachers (n = 2), assistant head teachers (n = 2), school teachers (n = 5), and pupils (n = 10). Focus groups took place in school settings with teaching staff (assistant head teacher (n = 1), school teachers (n = 2)), parents and carers (n = 3), and pupils (n = 35). Focus groups were made up of participants of the same gender (n = 5), as well as mixed gender (n = 6).
Of the qualitative sample, in total 56% of participants were female and 44% male. The majority of teaching staff (n = 11) and public health practitioners (n = 2) identified as female. All parents and carers (n = 3) identified as female. There were slightly more male children (n = 25), than female children (n = 20). Children ranged in ages from 7 to 8 (16), 9–10 (17), and 10–11 (n = 12). All teaching staff interviewed were white, whilst 20 children, and 1 parent were BME.
Interview/focus group schedules were semi-structured and covered similar themes with all participants. This included questions related to their involvement in TDM, what worked well, any challenges and costs associated with the implementation of the programme, and, (for public health professionals and school staff) what medium-long term plans were in place to sustain the intervention (if any) (See Additional file
1 for copies of the interview/focus group schedules).
Observational data included 61 h of observations across all schools. Field notes recorded transitions from classroom to TDM, the implementation of TDM, and transition back into the classroom afterwards by the lead author (BH). Field notes were guided by both the implementation intended outcomes, and topics that emerged from in-depth interviews and focus groups. Field notes were recorded before, during and after the implementation of each TDM across a two-three week period in each of the 12 classes to document what happened in practice.
Quantitative data sources
Routine data for schools in Lewisham were collated from existing records, which included secondary data held by Lewisham Council and data they collected as part of the roll out of TDM in the Borough. Existing data on school characteristics of all schools included size of the school (number of pupils); per cent of pupils with free school meal entitlement; and per cent of pupils who identified as from BME populations. The Department of Public Health also recorded date of adoption of TDM.
Analysis
Interview and focus group data were transcribed verbatim. All qualitative data, transcribed interview and focus group data, as well as fieldnotes, were analysed by researchers using thematic content analysis [
26], using NVivo software.
For analysis of routine data on school uptake we calculated summary statistics in the two school groups (‘daily mile implemented’ and ‘no daily mile in school’). All data were continuous and these were therefore summarised with group means and standard deviations. The distributions of each variable were clearly non-normal (number of pupils was bimodal, percent pupils with free school meals was positively skewed, and percent pupils BME was negatively skewed). To compare the distributions of these characteristics in adopting and non-adopting schools, we used two-sided Mann-Witney U tests since t tests would not be valid with bimodal and skewed data. We drew density plots to depict the overlap of the distributions of characteristics by school group. The statistical data analysis was conducted using STATA 14.
To describe the variation in implementation, we used the template for intervention description and replication for public health and policy interventions (TIDieR-PHP) [
27]. This was developed from the existing TIDieR checklist [
28] to improve the reporting of key features of interventions, such that they can be replicated. Items 9.1 and 9.2 of this checklist refer to fidelity: how well this was assessed and maximised (item 9.1) and how well it was actually achieved in delivery (item 9.2). These are perhaps most relevant for trial designs, rather than observational designs in contexts such as this, where the roll out of TDM aimed to encourage as many schools to sign up as possible, rather than to ensure that they were adhering to the core principles promoted by TDM Foundation. Issues of fidelity are also problematic in observational designs such as rapid assessments, where the aim is observing what happens in practice in non-trial settings, rather than necessarily intervening to maximise fidelity. We therefore used item 9.2 from the TIDieR checklist as a way of summarising how TDM was implemented in practice in schools across Lewisham.
Discussion
Interventions such as TDM are difficult to evaluate in traditional public health terms with RCT or quasi-experimental designs. Pathways between ‘intervention’ and ‘outcome’ are long and not always linear [
23]. This has a number of implications for evaluation. First, the health benefits that might accrue from an intervention may take many years to manifest. Second, effects of interventions may result from interactions between its components or those components and the context [
31] making it difficult to attribute causality. Third, a key challenge for designing evaluations that are meaningful to potential evidence users is that primary outcomes may not be those that are most salient. Trial evaluations are typically powered on one primary outcome, whereas for many school staff a key benefit of this scheme, for example, was the way in which it integrated a number of perceived health and educational needs for their pupils.
However, given the local commitments to rolling out TDM as part of a broader strategy of obesity, and the focus of the TDM Foundation on particular health outcomes, it is perhaps not surprising that these health benefits were the anticipated outcomes mentioned most frequently. However, our findings also point to other benefits (e.g. cross school year peer relationships, teacher-student relationships, connection to other parts of the curriculum) that were important outcomes of the intervention. This assemblage of wellbeing benefits often facilitated variations in adoption and implementation. Few participants mentioned any potential negative effects, though the erosion of curricula time was a concern that was articulated in all schools. Where this was of particular concern, this, along with spatial constraints and, at times the weather, acted as key barriers that prevented TDM from taking place, and resulting in students not engaging in planned PA.
Of particular interest here in terms of outcomes is the focus, at least in this locality, on obesity. This is perhaps the outcome where there is less robust evidence for likely impact, given that TDM may increase PA without demonstrable effects on obesity, at least in the short to medium term, given the limited potential impact of an additional 15 min of PA for children. A systematic review that found ‘moderately’ strong evidence of the effect of obesity programmes in schools drew largely on US data, but with small effect sizes [
32]. Recent high quality UK studies also reported similar small effect sizes but these were not statistically significant for outcomes including BMI, dietary and physical activity outcomes, as well as psychological measurements [
33,
34]. Understanding this lack of apparent success is hampered by limitations in process evaluation literature, which tend to focus on what participants think about interventions [
35]; and on a limited number of components, such as reach, dose and measures of fidelity [
36]. Process evaluations on obesity have been less informative on what works (to encourage sustainability, as well as on outcomes) and why [
37].
More generally, an intervention such as TDM can be thought of as an event within multiple interlinked complex systems [
38] such as schools and local authorities: in this setting, this was explicitly recognised by the local authority who were taking a ‘whole system’ approach to obesity, which, as described earlier, involves a focus on PA, sugar intake and use of green space in the borough to promote healthy living. The limitations of RCT evidence for examining causality in such settings have been widely documented [
23,
39], and are clear in this case. Disaggregating essential components to identify ‘what works’ may not be possible through cluster RCT methods, given the wide variety of methods of take up we have identified, and the complex interactions between school settings, the intervention and health states. This variation was evident at a number of levels. The local authority were supporting the scheme in their locality: others might not be, or might be using other materials or methods of support. At school level, many schools were implementing the scheme within broader programmes of healthy schools initiatives, including addressing food sales or lunch provision, as well as the adoption of other physical activity initiatives in some schools. There was also a sense from head teachers that schools had limited capacity to take part in multiple initiatives: if TDM was adopted, other potential schemes might not be adopted, which in turn may mean, for instance, that the focus is on PA, and less on food-related initiatives, such as Sugar Smart. In addition, there were concerns raised in each school about how the intervention would fit within existing curriculum and timetables that are ‘too full’, which has been identified elsewhere as a barrier [
29]. Understanding if and how these decisions are made and what trade-offs are made at the level of the school is important for understanding intervention implementation.
There was considerable variation across classes in schools that had implemented the scheme. Even where teachers were taking out whole classes most days of the week, individual pupils were not necessarily getting an additional 15 min MVPA, with many not running, or not running unless encouraged in ‘competitive’ ways. The intervention, both in terms of what is being promoted and what is implemented, is also likely to change over time. TDM Foundation are amending their web site and informational material in part as a response to ongoing research findings, and are engaged in resourcing co-ordinators to assist schools with maintaining fidelity to core principles. There is also discussion around mandating the scheme in the light of evidence of its effectiveness for increasing PA in children. This would of course change the intervention: a scheme that schools are required to do might have different effects than one they choose, with a potentially higher likelihood of it displacing other activities.
There is increasing recognition of the ways in which context shapes the effects of interventions in complex systems [
39,
40]. These effects are likely to have implications for health inequalities at a number of levels, each of which might (separately) mitigate or exacerbate existing inequalities in health outcomes. First, an intervention such as TDM might have differential effects on different population groups because of inequalities in uptake. We found no evidence of this within one borough, at least for the indicators available of deprivation levels and population mix of school populations (per cent eligible for free school meal, per cent in BME groups): schools that did agree to take part had no significant differences from those who did not. It is pertinent to note however that in a borough considered within the 20% most deprived in England [
24], TDM uptake was slightly greater, though not statistically significant, across schools with a higher proportion of pupils having free school meals. Second, exposure by population group might differ if the intervention is implemented differently across the population. Although our rapid ethnography could not measure differences by gender, ethnicity or socio-economic status, we did identify some important ways in which these might shape the effects of the intervention in practice. For instance, at school level, schools with more constrained outdoor facilities (which in many settings include those in more disadvantaged areas) might be less likely to be able to sustain adherence; and at pupil level, we observed systematic gender differences in how much PA children were gaining from the intervention, whereby female students would be more likely to walk arm in arm with friends, compared to male students; and male children were more likely to define the outcomes of the intervention in terms of strength, and/or avoidance of weakness than females. Third, at the level of the individual, there is considerable evidence that the effects of specific activities are likely to differ depending on context, the subjective meaning attributed to the activity, and physiological status [
41,
42]. Should these differences be systematically associated with social and demographic factors (such as the meaning of competitive activity for different genders, or the potential differential effects of exercise on children’s bodies who have not had access to breakfast [
43]) the impact of the same exposure might be different. These factors are under-researched, but may account for the limited evidence for behavioural interventions on health equity [
44]. Far more research is needed on the ways context interacts with interventions within complex systems, and the ways in which inequalities might emerge from those systems.
Recent guidance [
16] suggests that descriptions of context are vital for reporting on public health interventions, to better understand how they work and why impacts may vary. We have described the context in one setting, a diverse London borough, and suggested that each school, and class, also has its own context. Methods for addressing the public health effects of interventions such as TDM, which will vary considerably by context, are underdeveloped. Experimental designs often do not illuminate how different causal conditions may operate in differing contexts to produce the intended outcomes: the successful implementation of an intervention. We suggest approaches that bring together quantitative and qualitative methods. This will help elucidate the configurations of conditions needed for implementation to be successful. To this end Qualitative Comparative Analysis (QCA) [
45] might hold promise here as one way to explain causal complexity in context. QCA is a set-theoretic method, which examines casual complexity across a medium to large number of cases (between 10 and 60 +), whilst also being able to generalise across those cases [
45,
46]. QCA analysis involves using Boolean algebra and formal logic to ask what conditions (alone or in combination with other conditions) across contexts are necessary or sufficient to produce some outcome [
46]. QCA has been shown to be well placed to address concerns around complexity [
47], though its use in the field of public health remains nascent [
48].
Strengths and limitations
This study was restricted to one borough in London and a relatively small number of schools. However, our sample of schools was relatively diverse, and sufficient to identify considerable variation in implementation. Our aim was to map this variation, and a rapid assessment was sufficient for this, but cannot provide an in-depth understanding of how and why interventions are adopted. More detailed ethnographic work would be needed to understand in detail what happened in schools when they took up the intervention. Our sample of schools and classes was also purposively rather than systematically selected, and drawn from those who volunteered. It is therefore less likely to represent schools uninterested in TDM, or in broader PA initiatives. We are therefore likely to have underreported challenges in implementation. The TIDieR-PHP checklist, although designed for use in describing interventions in evaluation studies, was a useful framework for structuring this study of implementation.
Conclusions
The Daily Mile is, in principle, easy to adopt and has potentially positive impacts on children’s current and future health. However, evidence to date of its effects is likely to shed little light on what happens in practice in non-trial settings. This is important, as we have identified a variety of implementation practices in a non-trial setting, which have implications for future population health. This includes both variations in physical movements of the children taking part in each context, as well as variations in implementation, which result from, for example, various spatial (e.g. playground area) and temporal (e.g. time of year, time of the day and/or weather) constraints, as well as the presence or absence of other physical activities implemented within the school. All these adaptations are likely to shape the specific benefits and disbenefits for participants.
For complex interventions in public health, there have been calls for the development of more appropriate methods for both describing and evaluating interventions. This study found the TIDieR-PHP framework a useful checklist for describing TDM. In terms of future evaluation, TDM is typical of interventions for improving the health of the public, in that it is often implemented in the context of wider strategies, and in variable ways. Experimental and quasi-experimental methods have well-documented shortcomings for evaluating the causal effects of such complex interventions [
23,
49]. Future evaluation of TDM, and similar interventions, could usefully draw on approaches such as QCA to explore both successful uptake and impact. Understanding this variability of intervention adoption and implementation is crucial to understand how it becomes adapted and whether or not we might define these adaptations as meeting the initial criteria of the intervention. Assessing interventions in this way acknowledges complexity in practice and has important implications for assessing intervention efficacy outcomes.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.