A randomised controlled trial and cost-effectiveness evaluation of "booster" interventions to sustain increases in physical activity in middle-aged adults in deprived urban neighbourhoods

Participants

To identify potentially eligible study candidates we worked with NHS Sheffield (Sheffield Primary Care Trust), the health service organisation responsible for commissioning health services for the local population (until April 2013). Between May 2009 and June 2011 NHS Sheffield sent letters with postage-paid reply cards to 70,388 people inviting them to enrol in a programme to help them become more physically active. Six general practices were also given a total of 305 marked recruitment packs to distribute. These packs were identical to those sent from NHS Sheffield with the exception that the covering letter was not personalised. An unknown number of packs were also given to four community centres to be distributed by health trainers and health champions. Mail-outs and responses were allocated to output areas (OAs) according to the National Statistics Postcode Look-up table [November 2010 version; see http://geoconvert.mimas.ac.uk/help/documentation/10nov/nspd-version-notes-november-2010.pdf (accessed 21 November 2013)]. The response rate for each OA was calculated as the total number of allocated responses divided by the total number of allocated mail-outs. We performed a post hoc analysis to investigate whether response rate was related to population transience. Transience rate was calculated using information obtained from the 2001 census (small area statistics, Table CAS008), specifically the outgoing population divided by the outgoing population plus the static population. 21 An unweighted Spearman’s rank correlation test between response rate and transience was performed in R version 2.15.0. (R Foundation for Statistical Computing, Vienna, Austria; see www.R-project.org/).

The physical activity programme involved a ‘brief intervention’ combining an interactive DVD and area-specific written information about local facilities and opportunities for physical activity. The DVD was developed by a team at Sheffield Hallam University using MI principles, which are consistent with NHS guidance on physical activity and behaviour change interventions. 10,14 The content and development of this DVD are provided more fully in Appendix 1.

Research assistants (RAs) telephoned respondents and administered the Scottish Physical Activity Questionnaire (SPAQ). 22 Those eligible to receive the brief intervention (DVD and information sheet) were (1) residents of the 55 most economically deprived neighbourhoods in the city of Sheffield (out of 100), (2) those aged 40–64 years and (3) those not achieving the recommended activity level (30 minutes of moderate activity on at least 5 days) assessed using the SPAQ and wishing to have support to become more active. RAs telephoned those sent a DVD 3 months later to assess their eligibility for participation in the trial. Eligible candidates (4) had increased their physical activity level by at least 30 minutes of moderate or vigorous activity per week (assessed using the SPAQ) over the 3-month brief intervention (DVD) period and (5) were capable of giving written informed consent for trial participation. Individuals with chronic conditions who could benefit from physical activity were not excluded unless their condition significantly impaired their ability to exercise.

The validated SPAQ contains a series of questions assessing physical activity behaviour over the previous 7 days and also asks whether or not this level of weekly activity is typical. When observed weekly activity was reported as atypical, SPAQ asks participants to clarify what constitutes a more typical week in terms of an increased or reduced number of minutes of physical activity. Before and after the brief intervention period the RAs were asked to interpret the observed minutes of physical activity (as assessed using SPAQ) and to include any additional minutes of activity that were felt to be typical for study candidates who reported that the previous week had been atypical.

Interventions

Candidates who were assessed as eligible during the telephone assessment described in the previous section were invited to attend a baseline assessment at a community venue. Those who consented were randomly allocated (see Randomisation and blinding) to one of three groups:

1. a ‘full booster’ group receiving two face-to-face physical activity consultations provided in a MI style, 1 and 2 months after randomisation

2. a ‘mini booster’ group receiving two telephone-based physical activity consultations provided in a MI style, 1 and 2 months after randomisation

3. a control group who received no intervention after randomisation.

Both booster interventions are fully described in the study protocol, which can be found in Appendix 1. The full booster involved two face-to-face consultations, intended to last between 20 and 30 minutes, which took place in community venues. The consultations replicated a brief MI method designed for time-limited consultations in medical settings and which had already been successfully employed to change health-related behaviours. 23,24 During the full booster consultations, strategies were worked through at a pace dictated by the participant and the menu used to structure information exchange without being prescriptive.

The mini booster involved two physical activity MI consultations delivered by telephone, each intended to last approximately 20 minutes. The telephone consultations followed a script of known efficacy that has been implemented in previous physical activity promotion studies delivered by members of this research team. 25,26 It is thought that telephone counselling can provide an alternative to face-to-face contact that is relatively inexpensive in time and financial terms. 14 In previous studies in adult populations, telephone-based approaches have increased physical activity participation at 6 months compared with no telephone support and to a greater extent than standard reading materials. 27,28

Objectives

The primary objective of the main study, a parallel-group randomised controlled trial (RCT), was to determine if physical activity assessed using an Actiheart device (CamNtech Ltd, Cambridge, UK) 3 months after randomisation (6 months after a brief intervention) increases in participants allocated to two intervention groups (receiving two booster physical activity consultations, delivered in a MI style, either by telephone or face to face) compared with participants allocated to a control group (receiving no further contact after the baseline assessment). Secondary objectives were to:

1. determine whether physical activity 9 months after randomisation (12 months after the brief intervention) is significantly increased in participants allocated to the two intervention groups compared with participants allocated to the control group

2. compare physiological measures of fitness (12-minute walk test29) and self-reported physical activity (SPAQ instrument) between allocated groups

3. compare health-related quality of life (HRQoL), resource use (including health and social care contacts) and economic costs between allocated groups

4. investigate whether the impact of the intervention may be modified by gender, ethnicity or the types of physical activity undertaken (including use of community facilities for physical activity)

5. undertake a process evaluation to identify, using both quantitative and qualitative methods, psychosocial and environmental factors that may mediate or modify the effectiveness of the intervention.

Outcomes

Table 1 shows the timing of assessments and interventions, all made at community venues. The primary end point was the level of physical activity measured at 3 months post randomisation using Actiheart [specifically, the mean total energy expenditure (TEE) in kcal per day over a 1-week period]. Secondary end points were:

1. objective measures of physical activity including:

   1. TEE in kcal per day from 7-day accelerometry and heart rate monitoring using Actiheart (at 9 months)

   2. physical activity counts (PACs) per week

   3. minutes of moderate/vigorous physical activity per day

   4. meeting the current physical activity recommendation of at least 30 minutes per day (continuous or in bouts of at least 10 minutes of at least moderate intensity) for at least 5 days a week (yes or no)

2. self-reported moderate or strenuous physical activity using the SPAQ, which records type and duration of activities in the previous week

3. HRQoL using the Sheffield version of the 16-item Short Form health survey instrument (SF-12v2 plus 4)

4. self-reported use of community facilities for physical activity

5. self-reported health and social care contacts (see Methods for the health economic analysis for the analysis plan)

6. self-determination using the Behavioural Regulation in Exercise Questionnaire (BREQ-2)30

7. body weight and height [to allow calculation of body mass index (BMI)]

8. physiological measure of fitness (12-minute walk test).

The primary outcome was measured at 3 months post randomisation whereas the secondary outcomes were measured at 3 and 9 months post randomisation.

A decision was made to abandon the analysis of self-reported physical activity based on a questionnaire adopted from the HTA-funded Exercise Evaluation Randomised Trial (EXERT) trial. 31 This was because completion errors remained high even after repeat training of those administering it. The decision to abandon the use of this form was made in consultation with the trial steering committee and the HTA programme manager and the EXERT team and is fully documented elsewhere. 32

Sample size

The sample size was originally based on a primary outcome that was subsequently superseded by the use of Actiheart, that is, a physical activity measure based on the mean physical activity levels from the 7-day accelerometric assessment (recorded as counts per week) at 3 months post randomisation (6 months after initial contact). Before progression to the main trial, an internal pilot was undertaken to estimate the variability of the outcomes and the minimum clinically important difference (MCID) based on one-third of the standard deviation (SD) of the primary outcome from the observed data and power estimation (conditional on the initial proposed sample size of 600 subjects). From the feasibility phase, the estimated effect size based on one-third of the SD was 34,464.7 PACs per week and 101.5 kcal per day TEE. 32 When re-estimating the sample size using data from an internal pilot study the revised sample size estimate either stays the same or increases (it cannot be less than the original estimate). 33,34

Assuming a mean difference in TEE of 101.5 kcal per day between the intervention group and the control group as the smallest difference of clinical and practical importance that is worth detecting, then with 450 subjects (300 intervention, 150 control) the trial was originally determined to have 92% power to detect this mean difference or greater between the ‘booster’ arm and the control arm (assuming a SD of 304.6 kcal per day) as statistically significant at the 5% (two-sided) significance level using a two independent samples t-test. With 300 subjects in the booster intervention (150 mini booster, 150 full booster) the trial would also have had approximately 82% power to detect a similar mean difference in TEE of 101.5 kcal per day between the two booster arms as statistically significant at the 5% (two-sided) significance level using a two independent samples t-test. Assuming an approximate 25–35% loss to follow-up by 3 months post randomisation, we proposed to recruit and randomise 200 subjects per intervention group to give a total sample size of 600 participants, giving the study power of between 87% and 92% to detect a mean difference in TEE of 101.5 kcal per day.

Randomisation and blinding

A Sheffield Clinical Trials Research Unit (CTRU) statistician, not on the trial team, used a simple randomisation procedure to generate the randomisation sequence, with each participant having a one-third probability of being allocated to one of the three intervention arms. We used a block size of 200 with no stratification. Eligible participants were randomised to one of the three arms using a central web-based randomisation service delivered by Sheffield CTRU after patient eligibility and informed consent were confirmed by a RA. Participants and outcome assessors were not blind to treatment allocation because of the practical nature of the intervention. However, the primary outcome was objectively assessed using the Actiheart device. Most other outcomes were self-reported. Study statisticians and the principal investigator were blinded to the treatment allocation codes until after the final analysis.

Statistical methods

Survey

Survey questionnaires were sent to participants before their 3-month research assessment and they were asked to completed the questionnaire before the assessment and return it in a sealed envelope. If the survey was not completed participants were asked if they would be willing to complete it at the 3-month research assessment, after which the participant sealed it in an envelope and handed it back to the RA. Because of delays in regulatory approvals, questionnaires were sent out from April 2010 only and the first 47 randomised participants were not invited to complete it.

The survey questionnaire asked participants about the type and location of physical activity that they had undertaken during the previous 3 months, reasons for staying physically active, factors that influenced their physical activity behaviour and social support from significant others. The versions of the questionnaire sent to participants in the full and mini booster arms of the trial also contained questions on the intervention received. Participants were asked why they chose to participate; their preferred format for such an intervention; their expectations of the intervention and the extent to which these were met and whether they found the intervention easy, convenient, non-judgemental and non-confrontational. Participants were also asked their opinion about the amount of contact time with the project worker, the extent to which they felt encouraged to set their own goals for physical activity and the extent to which they felt that the intervention had helped them to resolve their barriers to physical activity, expand their knowledge of physical activity, increase their awareness of local facilities and opportunities and increase their confidence to stay active. Finally, participants were asked whether they had become more physically active than they were before participating and what had helped them achieve this. The questionnaire sent to those in the full booster arm of the trial can be found in Appendix 4. The questionnaires distributed to the mini booster and control arms are available from the team on request.

In-depth interviews

Those receiving a booster intervention who also responded to the survey questionnaire (see Survey) were given the option of participating in an in-depth interview. Because of the poor response there was no scope for purposive sampling; as a result, we interviewed a sample comprising all of the 26 people who volunteered. We did not elicit reasons for declining a research interview. Three RAs performed the interviews: Andrew Hutchison PhD (male) and Kimberly Horspool MSc and Sue Kesterton MSc (both female). KH and SK had both studied qualitative research techniques as part of their MSc but were novice interviewers. AH was more experienced having conducted a number of qualitative research interviews as part of his doctoral research. None of the interviewers delivered the intervention to the interviewees but interviewers may have been involved in collecting baseline data for the RCT component from some interviewees. Interviewees would have known that interviewers were on the research team and were from Sheffield Hallam University and may have associated them with exercise science and delivery of the intervention. The interviewers were asked to withhold their own opinions and to make it clear that this interview was separate from the intervention motivational interviews. No field notes were taken and no repeat interviews were undertaken.

Semistructured interviews lasted between 9 and 32 minutes (median 21 minutes) and were conducted over the telephone or face to face in a quiet room at a community venue, according to each participant’s choice. For most interviews no one was present except for the participant and the researcher. In one case a participant chose to conduct the interview on a mobile phone and, for part of the time, in a public place.

A topic guide was provided to interviewers (see Appendix 5); this was not pilot tested. This guide included questions on participants’ levels, choice and prioritisation of physical activity as well as the benefits and costs associated with staying physically active. It also included questions on participants’ experiences of the booster sessions and why they had or had not helped them to stay active and why participants felt that the booster sessions were or were not a good way to give them the support that they needed.

Interviews were digitally recorded and transcribed verbatim. Transcripts were not returned to participants for comment or correction. Daniel Hind conducted the initial data analysis in NVivo version 10 (QSR International, Southport, UK) using a constant comparative method to identify themes. We used a ‘framework’ approach to analysis in which a priori and emergent themes were identified using the following stages: familiarisation, identifying a thematic framework, indexing, mapping and interpretation (charting was not undertaken). 50 For instance, a theme of a priori interest was the perceived effectiveness of the booster sessions; subthemes within this category were derived inductively from familiarisation with the transcripts. 50,51 The results were used to explore insights into the mechanisms that may have contributed towards the quantitative findings and to identify any other emerging issues or factors that may have influenced the uptake of the boosters and which had not previously been documented. 52 Data saturation was achieved53 with no substantively new themes emerging in the last 10 interview transcripts. Participants were not asked to provide feedback on the themes.

Having indexed transcripts using our own thematic framework, we undertook a rapid review of the literature to find existing frameworks to evaluate dimensions of (1) prior conditions experienced by the participants; (2) barriers to and facilitators of adoption of new behaviours or technologies; and (3) the acceptability/appropriateness of the interventions. For prior conditions (see Chapter 4, Prior conditions) we used dimensions described by Rogers (p. 172). 54 We adopted, with modifications, the dimensions of the Motivators of and Barriers to Health-Smart Behaviors Inventory, developed by Tucker and colleagues55 (see Chapter 4, Barriers to physical activity and Physical activity: motivators). Our dimensions of intervention acceptability are based on those described by Nastasi and Hitchcock56 (see Chapter 4, Motivational interviewing: perceived effectiveness, Motivational interviewing: consistency with perspectives or world views, Motivational interviewing: perceived feasibility and Motivational interviewing: perceived importance).

Background

Although a small number of studies assessing the efficacy of physical activity counselling have reported the content, frequency and duration of training of those delivering the physical activity counselling intervention, the majority do not, and it is not uncommon for most clinical trials to fail to even report the content of the counselling intervention. 57,58

Although physical activity counselling based on MI has been rolled out across the UK through the Let’s Get Moving education programme,59 it remains unclear whether those delivering the training and those delivering the intervention to patients are doing so according to the approach intended. This failure to embed assessments of competence has raised questions over the value of short-term workshops with little or no ongoing supervision and professional practice reflection. It is clear that programmes such as this offer a potentially valuable additional education framework but few studies are currently being published that have clearly assessed the fidelity of those delivering the intervention.

It has therefore been suggested that behavioural interventions should clearly report (1) the content of or elements of the intervention, (2) the characteristics of those delivering the intervention, (3) the characteristics of the recipients, (4) the setting [e.g. Physical Activity Referral Scheme (PARS)], (5) the mode of delivery (e.g. face to face or by telephone), (6) the intensity and contact time (e.g. number of sessions) and (7) participant adherence to delivery protocols. 60 The booster trial embedded these principles into its design along with treatment fidelity frameworks intended to provide standardisation of the behavioural intervention without losing innovation and flexibility within the two intervention arms. 17,18 Furthermore, the action planning and maintenance phases in both experimental arms used behaviour change techniques, as recommended by Michie and colleagues,60 which are thought to enhance self-regulation towards change. The assessment of the existing competence and subsequent training of those delivering the MI interventions was evidence based and built on recent reviews of training in MI. 61

Motivational interviewing content and delivery

The MI component was delivered by RAs trained by a member of the Motivational Interviewing Network of Trainers and followed existing frameworks for MI in physical activity contexts. 62,63 The key phases and content of the intervention are provided in Table 2 and follow the phases of MI. 64 The relational aspect (or ‘spirit’) of MI is pivotal to the approach and emphasises participant ‘autonomy’ as opposed to ‘imposing authority’; ‘evocation’ rather than ‘education’; and ‘collaboration’ instead of ‘confrontation’. Once underpinned with the relational approach, the technical skills of MI were delivered, which included open-ended questions, affirmations, reflective listening and summarising. 13 Those delivering the intervention received 6 days of formal training over the first 12 months of the study in addition to follow-up supervision for the remainder of the study using audio recordings for reflective feedback.

Treatment fidelity assessment and methods

To ensure that the criteria for treatment fidelity were met, those delivering the MI physical activity booster interventions (RAs) were assessed for their competence in delivering MI before and throughout the intervention period. The assessment of practitioner competence used was the Motivational Interviewing Treatment Integrity (MITI) assessment. 65 Minimum practitioner levels will be based on the levels of ‘competence’ as stated in the MITI coding system. To account for practitioner competence ‘drift’ (post training), follow-up reviews of practitioner MI competence were carried out at appropriate intervals (approximately every 9 months).

The MITI assessment was used to measure the interventionist application of key facets of MI, which included ‘global ratings’ of evocation, collaboration, autonomy/support, direction and empathy. In addition, ‘behaviour counts’ were recorded, which included giving information, MI adherent behaviours (e.g. asking permission, affirming, emphasising personal control), MI non-adherent behaviours (e.g. advising, confronting, directing), open compared with closed questions and simple and complex reflections. The calculations for MITI were based on existing standards,65 as seen in Table 3.

The relationship between motivational interviewing fidelity and levels of physical activity: statistical methods

We employed analysis of variance (ANOVA) to test the null hypothesis that physical activity measured by mean TEE at 3 months was the same across all of the RAs who delivered the MI intervention. We plotted the means of mean TEE with their associated 95% CIs stratified by the RA who delivered the MI intervention to show how physical activity varies across RAs ranked by their global proficiency ratings. A further ANOVA model was fitted with RAs with the same global proficiency rating grouped together. We dropped from the analysis RAs who delivered very few sessions. In addition, for the few sessions in which MI was delivered by two RAs, we allocated the session to the RA who delivered more sessions.

Aim

The aim of this substudy was to explore whether access to green space and leisure facilities influenced the effectiveness of the intervention. We used network distance analysis to produce a set of variables for each of the 282 trial participants, which represents his or her pedestrian access to municipal green space and any relevant leisure facilities.

Network distance analysis

Euclidean (straight line) distance is the simplest measure of distance between two points. However, in a city it is rarely possible to follow a straight line between two points. Moreover, rivers, railway lines and sometimes roads force pedestrians to take routes that may deviate considerably from a straight line. To gauge the realistic walking distance between two points in a city, it is necessary to build a network representation of the pedestrian-accessible routes within the city and surrounding area. The shortest route between any two points on the network can then be calculated mathematically.

The most labour-intensive step is creating the network. The Integrated Transport Network^™ (ITN) in OS MasterMap^® [Ordnance Survey, Southampton, UK; see www.ordnancesurvey.co.uk/oswebsite/products/os-mastermap/itn-layer/index.html (accessed 10 October 2012)], can be used for this purpose, but unfortunately it is not possible to download a single portion of the ITN covering all of Sheffield and the surrounding area. Instead, we downloaded the data for this area from OpenStreetMap (see www.openstreetmap.org). In a certain respect, OpenStreetMap is actually more useful to us than the ITN; it is created (in part) by local volunteers, walking on foot and recording data with global positioning system (GPS) devices, and therefore provides information on pedestrian access that is more detailed than a typical OS map.

Roads and footpaths, etc. in OpenStreetMap are stored in a simple vector format that makes it relatively easy to extract the information required to build a network. The only problem is that the labelling of pedestrian accessibility for roads (especially trunk roads) is not consistent and is occasionally inaccurate. It was therefore necessary to manually exclude the following sections of road:

• A61 Dronfield bypass

• Sheffield Parkway

• Mosborough Parkway

• A57 (section linking Mosborough Parkway to the M1)

• A616 Stocksbridge bypass

• Park Square roundabout

• Tinsley roundabout and Tinsley viaduct.

Crossing points for non-pedestrian roads, for example bridges and underpasses, are well detailed in OpenStreetMap.

Although the shortest distance between points within the network can be calculated very accurately, there is the potential for error when calculating the network distance between features that do not lie within the network. Unfortunately, the centroids of postcode areas, and the polygons representing municipal green space boundaries, do not lie within the network. Here it is necessary to spatially ‘join’ these features to an appropriate point on the network. This is best done manually; however, because of the large amount of work that this would require it was necessary to automate the joining process, as follows.

Each postcode centroid was linked to the closest point on the network and the Euclidean distance between the centroid and the network point added to the network distance.

Each green space polygon was linked to all network points falling inside it, with the minimum network distance to any of these points deemed to be the minimum network distance to the polygon itself. As pedestrian entry points to green space are well detailed in OpenStreetMap, this means that the network distance mostly takes the entry points into account.

The major limitation of the automated joining process is that sometimes the network point that the feature is linked to is not always the most appropriate. For example, a postcode area may include adjacent houses that face onto different streets so that an address may be attributed to a network point on a street from which there is actually no access to the property. Similarly, for green space, points on the network outside the green space polygon may sometimes be closer to the actual entry points than those inside the polygon. This can result in network distance errors of ≥ 100 m, in some cases substantially more if the difference in network distance between the automated choice of point and the optimum choice of point is large. Fortunately, very large anomalies appear only in rural areas where the road network is sparse and therefore were not a significant concern in this analysis.

Green space measures

All municipal green spaces in Sheffield were considered, excluding certain types of land use inappropriate for exercise. Sheffield City Council classifies these spaces as ‘city’, ‘district’ or ‘local’ according to their catchment areas. We used ‘city’ as a proxy for high-quality space and ‘district’ as a proxy for medium-quality space in terms of attractiveness. Local spaces were too ubiquitous for a meaningful spatial analysis. Of municipal green spaces belonging to other authorities, only Rother Valley Country Park was included (designated high quality).

Digitised boundary data for municipal green spaces in Sheffield were provided by the Sheffield City Council Parks and Countryside team. As well as boundary information, these data included a classification of the usage and catchment area categorisation of each space. These are explained fully in a Sheffield City Council Parks and Countryside report66 but are summarised here as follows:

• Usage type:

  • parks

  • gardens

  • sports sites (e.g. tennis courts)

  • playing fields

  • playgrounds

  • playground/open space

  • woodlands

  • moorland/heathland

  • open spaces

  • allotments

  • churchyards/cemeteries

  • other site types – specified:

    • golf courses

    • farms

    • show grounds

    • depots

    • ancillary sites to other leisure facilities (e.g. car parks).

• Catchment area categorisation:

  • city wide

  • district (up to 1.3 km)

  • local (up to 0.4 km).

We decided to disregard the following usage types: playgrounds (because our participants are middle-aged), allotments (as nearby allotments are relevant only to the minority of local residents who rent them), churchyards/cemeteries (less likely to be appropriate places for exercise), farms, depots and ancillary sites.

Inclusion of the catchment area categorisation information is more problematic than inclusion of usage type as it is not clear on what research, if any, it is based. However, disregarding the categorisation is equally problematic as ‘local’ municipal green space is ubiquitous. For example, our preliminary study considered all municipal green spaces in Sheffield of the appropriate usage type and the size of a football pitch or greater and found very few postcodes in the city that were further than 1 km from such a space, with roughly 50% falling within 500 m and roughly 30% falling within 300 m. This tallies with the findings of Barbosa and colleagues. 67 Also, most distances > 500 m fell in the most affluent parts of the city, which were not included in the booster trial (this also tallies with the findings of Barbosa and collegaues67). Because of random errors inherent in our network distance measurements (see Network distance analysis), such small network distances would contain an unacceptable amount of uncertainty. It was also felt that it was important for us to take some account of the attractiveness of municipal green spaces.

Therefore, we decided to make use of the catchment area categorisation as a proxy for the quality/attractiveness of the space, using an ordinal scale in which citywide represents the highest quality, district represents medium quality and local represents the lowest quality. Disregarding all green spaces with unsuitable usage types (see earlier) we measured the shortest network distance to:

1. high-quality municipal green spaces of appropriate types

2. high- or medium-quality municipal green spaces of appropriate types.

We excluded Sheffield’s three municipal golf courses; these have citywide catchment areas but this is clearly based on their attraction as a leisure facility (as defined in leisure facility measures) offering golf, rather than as a green space, and the booster participant questionnaire responses indicated that very few participants played golf.

One potential problem with using the green space data for Sheffield is that many of the booster participants live on the edge of the city and could be using municipal green spaces belonging to surrounding authorities (Barnsley Metropolitan Borough Council, Rotherham Metropolitan Borough Council, Derbyshire County Council or North East Derbyshire District Council). This was a matter of sufficient concern for us to request green space boundary data from these authorities. However, because of the green belt around Sheffield, there are actually few important municipal green spaces belonging to these authorities that are close to the borders of Sheffield. The exception to this is Rotherham, with the two conurbations forming a continuous urban area. However, this area is heavily industrialised and there are few important municipal green spaces. One very important exception is Rother Valley Country Park, which borders south-east Sheffield; a boundary polygon for this park was added manually to the set of green spaces having a citywide catchment area.

All green space measures in this study are shortest distance measures. So far we have decided not to use a gravity model for green space as the need to include an arbitrary distance decay parameter would be a potential source of bias. However, this remains an option for future work. Regarding the more sophisticated floating catchment area gravity model, this may actually be inappropriate for green space: unlike a capacity-constrained service such as a health practitioner, a green space that is heavily used (and thus full of people) may be more attractive and more likely to be perceived as appropriate for recreational exercise or as a walking route than one that is less well used.

Leisure facility measures

Leisure facilities, as distinct from green spaces, are defined here as places where a physical activity is facilitated by some kind of organisation, usually in return for payment. For example, we define a tennis club as a leisure facility but an unsupervised tennis court within a park as part of a green space. Similarly, a publicly accessible playing field is considered a green space rather than a leisure facility, unless it is part of a sports club. Data on leisure facilities are easier to collect than data on green spaces because leisure facilities can be treated as point locations. Even though these facilities may cover a large area, there is usually an office where people must go to pay, and this is typically also the building to which the address of the facility relates. Therefore, noting the limitations of geocoding accuracy explained in Network distance analysis, it is credible to use the postcode centroid of the facility’s address as a point location.

The chief problems in collecting data on leisure facilities were twofold:

• deciding which types of leisure facility were relevant to the booster participants

• ensuring that all leisure facilities of these types, located within a reasonable distance of the study area, were included.

To address the first point we had the benefit of the questionnaires completed by the booster participants (see Outcomes and Survey). These data suggest that the main physical activities relevant to the participants were walking, gardening, swimming and gym-based activities. Walking relates to municipal green space rather than leisure facilities and gardening relates to neither, so we considered only leisure facilities offering a gym and/or swimming.

To address the second point we used the Sport England online ‘active places’ database,68 which provides an authoritative list of sports facilities, with each facility listed as offering one or more types of activity. Two of these types of facilities, ‘health and fitness suite’ and ‘swimming pool’, correspond directly to gyms and swimming respectively. The postcode of each facility is also included on the database, as is an indicator of how the public can access the facility (those open only to sports clubs or community associations, rather than individuals, were excluded). For each facility postcode we calculated the network distance to every other postcode within Sheffield, and for each Sheffield postcode we recorded the shortest network distance to:

• a unisex gym

• a female-only gym

• a swimming pool.

As with green space, we did not implement any of the more sophisticated ‘gravity’ models for leisure facilities. The basic gravity model is additive, meaning that having three gyms a 500-m network distance from one’s home would be scored three times better than having one gym 500 m away. This is clearly inappropriate as gyms are often paid for on a membership basis, and it is unlikely that a participant would join several gyms. Although swimming may be offered on a pay-per-session basis more frequently than gym facilities, it is still questionable whether having three local pools is exactly three times better than having one local pool.

Here, the floating catchment area gravity model is potentially more useful than the basic gravity model as it links the usefulness of local facilities to local demand. 69 A gym may be less desirable if it is frequently crowded and one has to wait to use particular exercise machines, and the same may also apply to lane swimming in pools. In this context having multiple local facilities can be beneficial, as greater provision of facilities means that they are less likely to be crowded. Unfortunately, the problem with interpreting this model is determining local demand; although the size of the local population can be easily obtained from census data, people often choose leisure facilities close to where they work rather than where they live, so using population data could be misleading, particularly in the city centre.

Given these issues, and for consistency in the analysis methods, we have used straightforward minimum distance analysis.

With leisure facilities there is also the issue of affordability. Some facilities are expensive and may not be realistically accessible to the more deprived communities targeted by the booster recruitment strategy. However, pricing information is not included in the Sport England database and it was not possible within the timescale of this project to obtain pricing information separately for each facility. Restricting the analysis to municipal facilities was considered but in Sheffield some privately run leisure facilities are less expensive than municipal ones. It is therefore important to recognise that affordability may be an additional barrier to accessing local leisure facilities, even when they are geographically close, which this analysis does not address.

Statistical methods

We constructed crude scatter plots of mean TEE per day (kcal) at 3 months against potential moderators generated from the geographical information systems (GIS) analysis and other continuous baseline potential moderators to explore any univariable linear relationships. In addition, we stratified these plots by intervention arm to explore whether the univariable linear relationships were consistent within the intervention arms. Univariable linear regression models were fitted with amount of physical activity measured by mean TEE per day (kcal) as the response and potential moderator as an explanatory variable to explore whether a potential moderator was a predictor of physical activity at 3 months.

To explore the moderation effect, we fitted multivariable linear regression models on the same response variable with intervention group, potential moderator and interaction between treatment group and potential moderator as explanatory variables (as given by Equation 3). Plots of the fitted regression line stratified by booster treatment group were constructed to explore any potential interactions. In addition, hypothesis testing on the interaction term was also undertaken using Equation 3 to test the moderation effect of GIS variables on physical activity. Variables were classified as being moderators of physical activity if they had a significant interaction effect with treatment at 3 months.

Introduction

As part of the trial data collection, participants answered questions on their use of NHS facilities at 3 months and 9 months post randomisation. Participants also completed the SF-12v2 plus 4 HRQoL questionnaire, allowing SF-6D HRQoL scores at 3 and 9 months to be produced. 70 Together with an estimate of the intervention cost, such data allow a simple estimate of the effect of the intervention on mean HRQoL and average cost to be produced by comparing costs and utility scores at the end of the intervention with those before the intervention.

A potential problem with using this approach in the case of the booster intervention is illustrated in Figure 1. Figure 1a illustrates a typical pharmaceutical RCT. The patient population is recruited on the basis of suffering from a particular disease, which has a progressive effect on their HRQoL. The intervention reduces disease progression and so a difference in HRQoL emerges between the intervention arm (dashed line) and the control arm (solid line) by the end of the trial. Figure 1b indicates the potential effect on HRQoL of a lifestyle intervention. Both the control arm and the intervention arm are generally in good health initially but because of low levels of physical activity both are at an increased risk of developing a range of diseases and conditions related to sedentary behaviour. As the intervention is not intended to modify a disease but instead modify a lifestyle factor that predisposes people over the long term to increased morbidity and mortality risks, a straightforward comparison of before-and-after levels of resource use and HRQoL may be misleading as it may have been too soon for the mediating effect of physical activity on health to have become apparent.

FIGURE 1.

Timing of HRQoL benefits. (a) Curative intervention; and (b) preventative public health intervention.

For this reason, the cost-effectiveness of the interventions was assessed using two distinct modelling approaches. Alongside a short-term model comparing participants’ self-reported HRQoL and NHS resource use at 3 and 9 months, a long-term epidemiological model was also developed in which differences in the primary physical activity measure recorded in the trial, TEE, were mapped onto effects on mortality reported in epidemiological literature. This approach better allows the complex mediating effects of physical activity on health to be formally represented but also increases the dependence of the modelling results on strong assumptions about this physical activity–health relationship.

In the rest of this section overviews of the methods used to produce the two models are provided.

Cost-effectiveness modelling overview

Two types of cost-effectiveness model were developed, which used different approaches and sources of data to estimate the health effect, in quality-adjusted life-years (QALYs), of the interventions. Both types of model are described in more detail below.

Short-term questionnaire-based model

A short-term cost-effectiveness model has been constructed that incorporates trial-based estimates of the effect of the two interventions – mini booster and full booster – on participants’ use of NHS resources during the trial period. It also uses trial-based estimates of the effect of the interventions on participant utility using responses from participants who completed the SF-12v2 plus 4 HRQoL questionnaire at baseline and 9 months. 70 Approximate costs of the interventions are also incorporated in the model alongside the estimates of the effect of the interventions on resource use.

Long-term epidemiological model

As stated previously, interpretation of the short-term cost-effectiveness model can be problematic as the intervention is preventative rather than curative and the participants are not selected from a population characterised by a particular disease. Because of this, it may be more valid to consider the effect of the intervention over a much longer time horizon than the trial duration and to assume that any potential QALY benefits of the intervention are mediated through the clinically measurable health benefits of increased physical activity. This is the rationale for presenting a long-term epidemiologically based model alongside the short-term questionnaire-based model described in the previous section. Unfortunately, adopting this approach requires making a number of strong assumptions about how the physical activity measures recorded within the trial translate into an impact on population health. The long-term model was an individual sampling model constructed in the R statistical programming language (version 2.14.2). The approach taken to develop this model, together with the assumptions made, are described in the following section.

Assumptions about the causal relationship between physical activity and mortality hazards

A number of sources of epidemiological research exist which suggest that a monotone relationship exists between how physically active people of working age are and their all-cause mortality rates. 75–80 However, no source could be identified that related directly to the population eligible for the intervention (those aged between 40 and 64 years inclusive), those who participated in the intervention (who tended to be drawn disproportionately from the older ages within the pool of those eligible) or the particular physical activity measures recorded by the Actiheart system.

Because of this a number of assumptions had to be made to relate the epidemiological results to the outcomes reported in the booster trial. The main epidemiological source used to inform the assumed relationship between physical activity measures and adjusted mortality risks was a US-based study published in the New England Journal of Medicine (NEJM) in 2002. 81 In this study the peak exercise capacity, as measured in METs, for older men of working age (mean age 59 years) was assessed through standardised treadmill-based testing. The paper presents a figure showing the relationship between relative risk (RR) of death and quintile of exercise capacity. The risk ratio for the highest exercise capacity quintile was chosen as the reference category. These data are presented in Table 6, as well as values adjusted to have the third quintile as the reference category.

TABLE 6 - Reported relationship between exercise capacity quintile and RR of death

Quintile 5 was the reference category.

Quintil 3 was the reference category.

Source: Myers and colleagues. 81

Quintile	Risk ratioa	Centralised risk ratiob
1 (lowest)	4.5	2.5
2	2.5	1.4
3	1.8	1.0
4	1.4	0.8
5 (highest)	1.0	0.6

Although initially it was assumed that the risk ratio to MET relationship could be mapped directly from the NEJM results to the Actiheart data, as both the data in the NEJM article and the booster trial data were categorised according to METs, it was recognised that this was not appropriate. This is because the contexts in which these METs were recorded were not equivalent. In the case of the NEJM article, METs were of physical exercise capacity and involved trying to identify an upper physical limit to patients’ physical performance over a short period of time. In the case of the Actiheart data the METs were of typical activity over a period of 1 week. Instead, an indirect approach to mapping the relationship was used. This involved making the additional assumption that the relationship between quintiles held, even if the unit of measurement was different.

The outcome measure used from the booster trial to estimate the relationship between exercise quintile and RR was mean daily TEE, which was recorded using the Actiheart device at both 3 months and 9 months. The median TEE score from both time periods combined was identified and all scores recorded were converted to a proportion of this median score. The central proportional TEE score within each of the five quintiles was calculated by sorting all values from lowest to highest and identifying the values 10%, 30%, 50%, 70% and 90% of the way across this distribution, to identify proportional TEE values associated with each quintile. The relationship between TEE (as a proportion of the median TEE) and exercise quintile is presented in Table 7. This shows that someone whose TEE was approximately 80% of the median is in the lowest physical activity quintile and someone whose TTE was approximately 30% more than the median was in the highest quintile. Participants’ TEE levels at 3 months and 9 months are also presented graphically in Figure 2.

TABLE 7 - Mapping between quintile and TEE as a proportion of the median

Quintile	Mean TEE per day (kcal)	TEE as proportion of the median
1 (lowest)	1703	0.77
2	1998	0.90
3	2224	1.00
4	2434	1.09
5 (highest)	2839	1.28

FIGURE 2.

Relationship between mean daily TEE at 3 months and mean daily TEE at 9 months by intervention group. Grey lines indicate the centre of each quintile.

Data from Tables 6 and 7 were combined to produce a regression equation mapping proportional TEE against RR of death. A good fit was found by assuming a power law relationship between these variables. The data and mapping equation identified are shown in Figure 3. The best fit equation identified was y = 1.056x^−2.951, where y is RR and x is TEE as a proportion of the median.

FIGURE 3.

Mapping equation between assumed RR of death and TEE as a proportion of the median.

Problems with inferring causal effects from the data

There are two fundamental challenges to the validity of the long-term model: first, the main physical activity outcome recorded in the booster trial, TEE, is a surrogate or proxy from the perspective of the long-term model, in which the outcome is QALYs; second, it is difficult to judge what is the counterfactual or baseline, that is, what would have happened to those participants who were assigned to the intervention arms of the trial if they were assigned to the control arm.

The previous section has shown the degree of potential model dependence on a number of assumptions that had to be made to link TEE with mortality. Some of these assumptions are strong, such as the assumption that quintiles of exercise capacity in one population correspond directly to quintiles of physical activity in another population. The validity of such assumptions therefore has a knock-on effect on the validity of the model outputs.

The issue of estimating the counterfactual is a fundamental challenge in the case of the booster trial for a number of reasons:

1. Actiheart measurements including TEE and PACs were not recorded at the start of the trial (‘baseline’). Instead, the first measures were recorded at 3 months post intervention. Additionally, subjective measures of physical activity such as the SPAQ were known to have only very slight correlation with the objective measures so cannot be reliably used to impute the objective scores at baseline. 82

2. The level of attrition within the trial was high, meaning that the sample of participants who completed the trial may not be representative of those who started the trial, and so there is a significant potential bias issue.

3. The number of participants who had valid and complete Actiheart measures at 3 months and 9 months post randomisation is significantly smaller than the number who completed the booster trial.

Because of the high level of attrition and low number of participants with complete Actiheart data, it may be wrong to assume that the population mix in each arm is similar at 3 months and 9 months after the start of the trial, because of patterns of selective attrition. This means that it may be problematic to assume that differences in the outcome measures between the control arm and the treatment arm observed at 3 months and 9 months are due to the treatment effect.

For this reason, a differences-in-differences approach was used. This involves comparing change over time in the outcome in the treatment group with change over time in the outcome in the control group. As no TEE measurements were recorded at baseline, the 3-month scores were used as proxies for the baseline levels and the 9-month scores were assumed to be the longer-term post-intervention levels.

It may be argued that using the 3-month post-intervention scores as the baseline scores is a very strong assumption because a significant treatment effect may be present at 3 months post intervention. In theory, a highly effective intervention could appear to be less effective than a less effective option because the more effective intervention is most effective at 3 months whereas the less effective option is more effective at 9 months than at 3 months. This is illustrated in Figure 4, in which the more effective intervention is indicated by the solid line and the less effective intervention by the dashed line. Because of this potential problem with using the 3-month scores as the baseline scores in the differences-in-differences approach, separate analyses using a simple differences approach were also carried out. These analyses involved comparing the TEE levels at 3 months in the intervention arms with the TEE levels at 3 months in the control group, and the 9-month TEE levels in the intervention arms with the 9-month TEE levels in the control group.

FIGURE 4.

Illustration of a scenario in which it would be wrong to attribute differences in differences (9 months vs. 3 months) between the treatment arm and the control arm to a treatment effect.

The fundamental problem with the simple differences approach is that differences between groups at either 3 months or 9 months could be due to differences between the groups that were present at randomisation, rather than differences as a result of the intervention. This is illustrated in Figure 5 in which the intervention had no effect on the physical activity levels of either arm (the solid or dashed lines), but comparing scores in arms at either 3 months or 9 months would suggest a treatment effect. In this study, although the participants were randomly assigned to each of the three arms and so on average can be assumed to have similar baseline characteristics at randomisation, because of the degree of attrition in the study the composition of the three arms cannot be assumed to be the same at 3 months or 9 months because the attrition may have been selective. For example, participants who were least inclined towards additional physical activity may be more likely to leave the treatment arms but may stay in the control arm.

FIGURE 5.

Illustration of a scenario in which it would be wrong to attribute differences between the treatment arm and the control arm at 3 months or 9 months to a treatment effect.

Primary and secondary models

Because both the simple differences approach and the differences-in-differences approach require that problematic assumptions be made, neither the simple difference models nor the differences-in-differences models should be considered preferable to each other. Instead, they should all be considered equal primary models.

In addition to these primary analyses, a secondary model was also produced. This model involved assuming that the changes in physical activity levels recorded between 3 months and 9 months in each arm were equally distributed. This means that a mean increase in physical activity between 3 and 9 months will translate into a mean increase in survival and longevity. Because the primary long-term models are individual-level models and make use of a separate single observation from the trial and involve mapping changes in physical activity to changes in annual risks, this does not necessarily occur in these models. The approach used in this ‘value-added’ model is described in more detail in the following section.

The ‘value-added’ model

If F_9i defines the level of activity recorded in the individual in the full booster arm at 9 months and F_3i defines the level of activity recorded in that same individual at 3 months, then in the differences-in-differences model the physical activity level change in each individual at the end of 9 months is calculated as F_9i – F_3i and so is different for each individual. In the ‘value-added’ model, however, an equal increase is assumed for all participants in each arm. This increase, Δ_F, is the mean difference between 9-month and 3-month scores observed in individuals in the full booster group. The increases in the control group, Δ_C, and the mini booster group, Δ_M, are calculated similarly.

‘Value added’ means that the direction of effect in terms of estimated effect on mean utility should always be in the same direction as the estimated effect on mean activity level and so may appear to produce more plausible estimates when trial results indicate a mean increase in physical activity but the economic model indicates a mean decrease in estimated utility. The cost of doing this, however, is to not make use of some available individual-level data relating level of physical activity at 3 months to level of physical activity at 9 months.

Additional scenario analyses

Subgroup analyses

Gender, ethnicity and access to community facilities (self-reported use vs. no use of community facilities) were predefined as subgroups that we wished to test for evidence of effectiveness in an exploratory analysis. As the majority of participants with valid 3-month primary outcome data were white British [89% (143/160)], we could not test for an ethnicity subgroup effect. We found no reliable statistical evidence of any (gender or access to community facilities) subgroup effects or interactions between the booster group and control group at 3 and 9 months post randomisation (Figures 19–24 and see Appendix 6, Tables 51–54). In addition, there was no reliable statistical evidence to suggest that the timing of the initial mail-out had an impact on the response of participants at 3 months (see Appendix 6, Tables 55 and 56). However, there was some indication (with respect to direction of the effect) that participants who were approached to take part in the summer or the spring were performing slightly better in the full booster arm than their counterparts who were approached in the winter or the autumn, although there was huge uncertainty in the estimated effect (see Appendix 6, Table 55).

FIGURE 19.

Subgroup evaluation for the primary outcome at 3 months stratified by gender (n = 160).

FIGURE 20.

Interaction effect of the intervention and gender at 3 months. Interaction = –18.2 (95% CI −260.6 to 224.3; p = 0.882).

FIGURE 21.

Interaction effect of the intervention and gender at 9 months. Interaction = 49.2 (95% CI −262.6 to 361.0; p = 0.755).

FIGURE 22.

Forest plot of the effect of the intervention stratified by use of community facilities in the last month at 3 months.

FIGURE 23.

Interaction plot of the intervention and use of community facilities in the last month at 3 months. Interaction = −160.2 (95% CI −469.1 to 148.7; p = 0.307).

FIGURE 24.

Interaction plot of the intervention and use of community facilities in the last month at 9 months. Interaction = 100.3 (95% CI −264.8 to 465.3; p = 0.586).

The sample

In total, 239 people were sent and 75 returned the survey questionnaire. The baseline characteristics of the survey population are compared with the baseline characteristics of trial participants who did not take part in the survey in Appendix 7, Table 57. The only variable that shows a difference between groups worth highlighting was the change in SPAQ score between −3 months and baseline, with non-responders reporting larger changes than survey responders. Formal significance testing for baseline comparability was not performed.

Closed questions

The responses to the process evaluation closed questions are presented in Appendix 7, Table 58. Over the 3 months before they were surveyed, 64 respondents (84.2%) reported taking part in recreational/leisure activities (e.g. gardening, cycling), 10 (13.2%) in competitive sports/exercise, 26 (34.2%) in structured exercise (e.g. an exercise class) and 54 (71.1%) in active commuting (e.g. walking/cycling to work). In total, 37 (48.7%) respondents reported undertaking physical activity at home, 41 (53.9%) in a local open space (e.g. a park), 34 (44.7%) in a facility (e.g. a gym, pool, community centre or track), 52 (68.4%) as part of their daily activities (e.g. in work, shopping, walking the dog or commuting) and 11 (14.5%) in other places. The most common reasons given for staying physically active were to improve health and get fitter or to stay fit, with 59 respondents (77.6%) reporting that this was very much the case. The most commonly reported influence on being able to perform their chosen activity was a respondent’s own health (n = 56, 73.7%). In total, 34 respondents (44.7%) reported that they did physical activity with other people, of whom most (n = 29, 38.2%) said that they found this fairly or very useful. The most commonly reported partners for physical activity were a spouse (n = 19, 25.0%), friend (n = 15, 19.7%) or other family member (n = 10, 13.2%). If further physical activity ‘booster’ advice had been available in the future, more people would have preferred it to be delivered face to face (n = 29, 58.0%) than over the telephone (n = 13, 26.0%) or by written advice (n = 12, 24.0%).

Overall, 43 (86.0%) of those who received physical activity ‘booster’ advice said that it met their expectations, with more in the face-to-face full booster group reporting so (n = 24, 96.0%) than in the telephone mini booster group (n = 19, 76.0%). The majority (n = 38, 76.0%) agreed or strongly agreed that the ‘booster’ counselling/advice sessions fitted easily into their daily schedule, with no obvious difference between the mini booster (n = 19, 76%) and full booster (n = 19, 76%) participants. The majority of respondents who were full booster recipients (n = 21, 84.0%) agreed or strongly agreed that the ‘booster’ counselling/advice sessions had been conducted at a convenient location.

In total, 41 respondents (82.0%) who received booster interventions agreed or strongly agreed that the project worker was non-judgemental and 39 (78.0%) agreed or strongly agreed that the booster counselling/advice sessions were non-confrontational in nature. A total of 44 participants (88.0%) agreed or strongly agreed that, throughout the ‘booster’ counselling/advice sessions, the project worker understood what they were saying. Overall, 22 respondents (44.0%) said that they spoke more than the project worker and 21 (42.0%) spoke roughly the same amount as the project worker. In total, 45 respondents (90.0% of those receiving the intervention) said that the amount of contact time with the project worker was about right. The majority (n = 42, 84.0%) agreed or strongly agreed that they were encouraged to set their own goals for physical activity and 31 (62.0%) agreed or strongly agreed that, as a result of the ‘booster’ counselling/advice sessions, they had been able to resolve their barriers towards physical activity. Overall, 33 (66.0%) agreed or strongly agreed that, as a result of the ‘booster’ counselling/advice sessions, they now knew more about the benefits of physical activity. The same number agreed or strongly agreed that, as a result of the ‘booster’ counselling/advice sessions, they now knew more about the risks associated with living an inactive lifestyle. In total, 32 (64.0%) agreed that, as a result of the ‘booster’ counselling/advice sessions, they were now more aware of available physical activity facilities and opportunities and 41 (82.0%) agreed or strongly agreed that, as a result of the ‘booster’ counselling/advice sessions, their confidence to stay active had increased.

Open questions

The sample

In total, 26 people participated in the in-depth interviews between December 2010 and February 2012. The baseline characteristics of the interview population are compared with the baseline characteristics of the trial participants who did not take part in the interviews in Appendix 7, Table 59. Participants in full-time employment were under-represented (15% in the interview population compared with 35% of those not interviewed). As with the survey population, increases in self-reported physical activity in the 3 months before baseline were lower for those interviewed than for those not interviewed.

Actiheart data were not available for six interviewees. Of the remaining 20, 17 contributed 7 days of data, one contributed 4 days of data and two contributed 3 days of data (see Appendix 7, Table 60). We have no evidence that any of the participants met the national guidance of at least 30 minutes of moderate/vigorous exercise on at least 5 days per week, although it is highly likely that one participant would have done so as he did > 5 hours of at least moderate activity on each of the 3 days over which he contributed data. Two people met the target of 30 minutes of moderate/vigorous exercise on 3 days; two people met the target on 2 days; three people met the target on 1 day; and the remaining 13 people did not meet the target on any day for which they provided accelerometry data.

Participants noted that they preferred the following physical activities: walking, swimming, cycling, gardening, walking the school run, football, running, yoga, gym, badminton, physical jobs, playing on the Wii and dancing. Some participants also declared a strong dislike for the gym and one declared a strong dislike for jogging.

Previous conditions

Barriers to physical activity

Physical activity: motivators

Motivational interviewing: perceived effectiveness

Only three people said that they thought that the booster sessions had no impact on them. All expressed the sentiment that they continued to do what they would have done anyway. Two of these achieved 0 and 5 minutes of moderate levels of activity during 7 Actiheart days. The other did not contribute Actiheart data.

nothing that was said over the phone or through interview made the slightest bit of difference.

Married male, 48 years

Seven people expressed ambivalence about whether the booster sessions were effective, through convoluted, contradictory, evasive or indirect answers:

Well it’s ticking boxes, yes . . . there’s things I mentioned to [motivational interviewer] which I haven’t quite got round to doing yet . . . I don’t know if I can get down to it. I think my problem is, I’m not really a great fan of gyms.

Single male, 51 years

Of these seven, three contributed no Actiheart data and three achieved 0, 3 and 9 minutes of moderate/vigorous activity over 7 Actiheart days. The last had 3 days where they met (43 minutes) or almost met (27 and 28 minutes) the national target of 30 minutes of moderate/vigorous activity a day.

In total, 15 people said that they thought that the booster sessions had been effective. Of these, 11 met the national target of 30 minutes of moderate/vigorous activity on at least 1 day. Three exhibited fairly sedentary activity levels on all days.

Motivational interviewing: consistency with perspectives or world views

At least four people, were, at best, ambivalent about MI. At least two explicitly said that they had hoped for an exercise programme or some kind of physical activity provision:

I would have expected more of a, shall we say, physical challenge? . . . I thought there would have been more physical activity to be measured and said, ‘Well, yeah, you did this when you came on the programme, you were able to do this by the time you’d got to the end of the programme’.

Co-habiting male, 65 years

Two disliked the autonomy-supported communication style used in MI (see Chapter 8, Views on physical activity and motivational interviewing ‘boosters’, for discussion):

Just two people chatting really, you know, I’m trying to give her some answers and she’s doing the best to give me some advice . . . it’s just like same old same old.

Married male, 48 years

I wanted to be told what the opinion was of my state of fitness. I want some results, you know, and being told off or being told that, ‘Well, it’s what you actually do in your day to day work keeps you reasonably fit, if necessary, or . . .’ . . . I wanted somebody to say, ‘Well, we’ve done your blood pressure’ or ‘We’ve measured you, we’ve weighed you and so we think’, blah blah blah.

Divorced male, 50 years

However, at least 10 people expressed enthusiasm for the autonomy-supported communication style. Common themes that emerged were that people liked the question-and-answer format and the informality. They found it businesslike and characterised the approach as encouraging, supportive and to do with fact finding rather than being ‘pushy’ or judgemental:

Well I was encouraged, but not pushed . . . So it’s obviously a lot better than virtually being told, what about this or no sorry, do this or this or this. So it was open and it enabled me to do it at the pace I wanted it to.

Widowed female, 65 years

At least 10 people made comments to the effect that they found the advice appropriate and flexible to their context. Clients frequently made statements to the effect that MI was thought-provoking or that it helped them focus on the importance of and barriers to physical activity. It gave them ideas that they could not have come up with themselves and pushed them to think about how much exercise they actually did. For many, the assistance in setting realistic goals was critical:

We set little goals for me to achieve . . . sort of saying, build your exercise up from 15 minutes to 30, 40 and I managed to get up to an hour a day on some days . . . he sort of set me targets for me to achieve, not too big, and goals and I found that because I knew that I was meeting him again . . . I made myself do them. I enjoyed the exercise once I got into it . . . I didn’t want to go and say, oh I didn’t achieve what we’d set out because the goals weren’t big goals and I think that’s the key, not to set a goal too big that it’s impossible to reach, so you think well I’m never going to do it, so I’ll not bother. You know, you can give up if the goal is too far ahead . . . I knew that I could achieve it. If I thought it was impossible to do, then human nature being as it is, I thought well I’m never going to do that, so I’m not even going to try.

Married female, 52 years, achieving recommended activity levels on 3/7 Actiheart days

It’s gradually making the changes . . . it’s about not drastic changes, it’s about taking one step at a time and gradually make the changes towards setting the goals that you want . . . make the changes to achieve the goals that you want to set yourself. So it’s helping people to achieve realistic goals.

Single female, 47 years

Motivational interviewing: perceived feasibility

For the vast majority of participants, the timing, duration and periodicity of booster sessions was highly convenient. Complaints were infrequent but focused on continuity of care (‘I don’t think I’ve seen the same person twice have I?’) and the need for more sessions. Suspecting that lack of funding might explain the limited provision (two sessions), one participant suggested that introducing clients into a peer-support (‘self-help’) group would be better than the current ‘abrupt’ end to the programme.

Seven out of 13 people interviewed who received MI by telephone believed that face-to-face booster sessions would have been preferable. One man with his own business and children found it difficult to protect time either at work or at home:

because I work for myself, you know, if someone walks in, you know, I’ve got to see to them straight away . . . so it’s easier for me to make a time [for a face-to-face meeting] and, you know, I’ll be there . . . at 6 o’clock, when I get home, it’s like ‘Daddy, Daddy’ and . . . I’m being followed into every room, can’t go anywhere!

Divorced male, 49 years

Those receiving face-to-face MI often said that they thought that the physical presence of the person made the MI more effective (see Motivational interviewing: perceived importance and The importance of monitoring and feedback):

Actually seeing the individuals who come and talk to you, so it’s like a one-to-one communication, the physical contact I think is very important . . . What difference would it make to me – somebody just over the telephone, ticking a box? Why can’t they just meet me?

Single female, 47 years

Motivational interviewing: perceived importance

A majority of people expressed the view that they felt that MI was important in some way, for instance because it met some acknowledged need. Sixteen people used language to the effect that the MI gave them encouragement, motivation, ‘a push’, impetus, a boost to their confidence or similar. For some, it helped them keep motivated in a situation of social isolation or disliking exercise or preferring alternatives:

Because I live on my own, I find it more difficult to keep motivated, cos I don’t really like exercise. I could . . . I’m one of these people that could just sit and watch telly all day and enjoy it.

Divorced female, 65 years

Many people said that they needed help with ideas for alternative physical activity strategies or help with overcoming barriers or thinking through life changes, especially retirement. People frequently claimed that the sessions helped them ‘focus’ or reinforced knowledge or goals:

I did start to try and look for ways and means of filing my time in, after I’d retired. Most of my immediate family were worried that I wasn’t going to survive this winter with not being able to go to work, so I was looking to broaden my horizons.

Co-habiting male, 65 years

At least four people mentioned that they enjoyed the social contact of the booster session. Sometimes this was because they were able to engage in a depth of conversation about an issue that was important to them that they did not feel able to do (or would not be productive) with a friend or family member:

Well it’s good to have a chat with someone. I don’t get much opportunity at home . . . I mean, it’s a welcome break to my day, having a conversation along those lines. It’s a sort of a change . . . having to think about something different.

Single male, 51 years

Because your family and your friends know you too well and they have got a preconceived idea of you, whereas [the motivational interviewer] hadn’t, ya know, it was just me.

Married female, 62 years

In summary, participants claimed that the MI met a number of needs, including providing encouragement and the opportunity to think laterally about problems and solutions with a professional removed from their normal social context. However, as the next section discusses, interviewees often framed these views in terms of altering their behaviour because of an awareness that they were being observed:

It’s little things that you can talk through which you probably wouldn’t have thought of and it seems to be important, but who would you talk to about it? Like, how do I fit in the jobs that I’ve got and fit in what I want to do, and is the diet I’ve got important and should I be doing this, and things like, ‘Oh for goodness sake, I can’t be bothered with it’, but because I’ve had to think about it, because I’m going to talk to someone about it and then . . . It appears to have worked very well.

Married female, 65 years; our emphasis

The importance of monitoring and feedback

Around half (n = 13) of the participants expressed enthusiasm not for the motivational interview per se but for the opportunity to be monitored and have feedback on their performance. For some, this involved the expected pleasure of receiving objective feedback from the researcher or the Actiheart on their level of activity over a week or during particular activities, as well as how that compared with national guidance or other people:

to be honest, one of the reasons I participated in this is because I did want some feedback. I wanted to be told what the opinion was of my state of fitness.

Divorced male, 50 years

I just wish I’d been able to carry on with the monitor, just to see what the heart monitor showed through my normal week . . . I was quite disappointed I couldn’t just keep it on.

Single female, 62 years

I wanted to be able to see that result, to see what kind of exercise I did over that time, how much calorie did I burn, you know, things like that. I would love to have something like that. I think that generate a greater interest to go out and do something.

Single female, 47 years

Others reported that arranging to meet a researcher in 3 months’ time, for the research study primary outcome follow-up, motivated them to increase their physical activity. This may indicate possible confusion between research and intervention procedures and possible contamination of the intervention by the research procedures (see Strength and weaknesses of the study). Some thought that they would feel ashamed to admit it if they had not achieved their goals. Others actively looked forward to reporting achievements that they had made:

the fact that you know you’re going to a session . . . you think I’ve got to do something because she’s going to say, ‘what have you done since last time?’ If you sit there and say, ‘well, actually, nothing’ . . . [it] increases the guilt value [laughs].

Married male, 64 years

It’s sort of saying, I realise that I remember you saying I’d do this and I haven’t done that yet, sort of thing, and I’m always disappointed with myself when I say things and I don’t do them.

Single male, 51 years

The very fact that I actually told [the motivational interviewer] I was going to go swimming meant I had to do it. There was no pressure on me to do it from her but the very fact that I actually told somebody I was going to do it meant I had to do it . . . because I’d actually told her I was going to do it, it made me want to do it.

Married female, 56 years

in the nicest possible way, I was having to report to somebody and . . . I know me: that is what I need. I don’t mean that I was being told off or anything like that . . . Just to have to report that, yes I’m doing this, or I’ve started doing that and getting a bit of praise. I think we all like praise don’t we, deep down, and that helped.

Widowed female, 65 years

you keep that pressure on by saying, ‘Right, you’ve got an appointment in 4 weeks’ time and we’re going to put this on, and blah-di-blah’, so you know, people think, ‘Ah, I must do it, I must do it’.

Widowed female, 65 years

Of these, a number of people expressed how important it was that the person holding them to account was a relative stranger rather than a friend:

It could’ve been a friend, but it’s not as easy with a friend to do it . . . And it kept my mind focused, just the fact that I was talking to someone, someone who obviously knew what they were talking about as opposed to, I mentioned a friend before, where a friend wouldn’t know the same, wouldn’t have the same knowledge.

Widowed female, 65 years

Yeah, yes, cos day-to-day people, they’re with you all day, they’re either bored of it or they can see it for themselves, or you know, you might catch them when they’re not particularly interested. You know, so it’s not quite so easy, whereas with somebody else who is a stranger as well, I quite liked that, rather than, sort of, if it hadn’t gone well, it’s not, you know, somebody judgemental.

Married female, 53 years

The prevalence of optimistic self-assessment

We have already noted earlier that self-reported physical activity from the SPAQ correlated poorly with objective physical activity (see Figure 17) and the low numbers of people meeting physical activity targets of 30 minutes of moderate to vigorous physical activity on any particular day (see Appendix 6, Table 50). However, some interviewees made very explicit and detailed claims to the effect that they were very physically active when the Actiheart data suggested otherwise:

Well don’t forget, my work involves me running around after a tower crane up buildings that were anything from 6 to 18 floors up that had no lifts in them, and you know so that’s kind of pretty physical all by itself.

Married male, 48 years, who met physical activity guidance on 1/7 Actiheart days; median (range) minutes per day of moderate activity = 0 (0–40)

The relationship between motivational interviewing fidelity and levels of physical activity

There is moderate evidence to suggest that MI fidelity is associated with physical activity as measured by mean TEE per day (kcal) at 3 months (p = 0.027), that is, the level of physical activity of participants at follow-up was associated with the overall fidelity of delivery of the RA who delivered their MI intervention.

Figure 25 shows the means of mean TEE per day at 3 months with their associated 95% CIs stratified by RA ranked by their global fidelity rating. It should be noted that RA1, RA2 and RA3 had a similar global proficiency rating of 3.5. Grouping these RAs together showed a stronger association between MI fidelity and mean TEE per day at 3 months (p = 0.003). As observed from Figure 25, RAs with a higher global proficiency rating are associated with higher physical activity levels at 3 months. These results should be interpreted with caution, however, given that some RAs delivered a lower number of sessions, as observed by huge uncertainty around their estimated means. Moreover, the sessions were brief and it was not possible to demonstrate all MI processes in some RA sessions that were coded. RA6 does not appear in Figure 25 because she delivered very few sessions and was subject to MITI recording only over a short period after initial training.

FIGURE 25.

Means of mean TEE per day (kcal) stratified by RA who delivered the MI sessions.

Resource use analysis

An analysis of the NHS resources consumed by participants in each of the trial arms indicated that participants in the intervention arms had greater increases in NHS resource consumption at 9 months compared with baseline than participants in the control arm, as indicated in Table 10.

Cost of the intervention

Estimates of the cost of the intervention were arrived at taking into account a range of factors including number of sessions per completer; number of completer sessions per RA; numbers of completers at 3 and 9 months in the mini and full booster intervention arms; numbers of participants in the mini and full booster arms; estimates of the duration of sessions in the mini and full booster arms; RA training and monitoring costs; venue hire (full booster) and telephone call (mini booster) costs; and the ratio of participants who completed a full or mini booster course to those assigned to the course.

Following discussion with RAs, the average duration of a mini booster session was estimated to be 20 minutes and the average duration of a full booster session was estimated to be 30 minutes. The cost per minute of a mini booster session was estimated to be lower than the cost per minute of a full booster session as the latter required venue hire and RA travel time. Rates of attrition were similar in both intervention arms at 3 months but were higher in the mini booster arm at 9 months. Because of this the average total cost per completer (someone who attended two booster sessions and provided valid Actiheart data) was estimated to be slightly higher in the mini booster arm at 9 months than in the full booster arm at 9 months. Because of the high level of uncertainty in estimating the cost of the intervention, eight different intervention cost estimates were produced for both the mini booster arm and the full booster arm using different plausible assumptions. The assumptions made to produce these estimates are shown in Appendix 7, Table 62; 2011 prices were assumed.

The range of estimates appeared to follow a log-normal distribution with a mean (SD) cost per completer of £216 (£88) for the mini booster and £205 (£76) for the full booster. Because of the similarity of these numbers the booster intervention was assumed to cost approximately the same irrespective of whether it was the full booster or the mini booster. The two pairs of eight estimates were combined to produce a mean (SD) cost per completer of £211 (£91), also assuming a log-normal distribution. Within the deterministic analyses the mean estimate of £211 was used. Within PSA 1000 values were drawn from an appropriately parameterised log-normal distribution.

Health-related quality of life data used

The trial outcome data used to populate the short-term model are patient-estimated utilities based on SF-12v2 plus 4 questionnaire data recorded at baseline and at the end of the trial. These data are presented in Figure 30 for participants who provided valid 9-month Actiheart data (‘completers’).

FIGURE 30.

Estimated utility scores reported at 3 and 9 months by allocation group (trial completers only).

Within scatter plots of this form, an overall trend over time would be indicated by the scatter deviating from the diagonal line: above the line in the case of an upwards trend and below the line in the case of a downwards trend. A difference between groups would be apparent if the three sets of scatter, indicated by the three plotting symbols, tended to cluster in different places. It is seen from this plot that there does not appear to be either a trend over time or a difference between groups.

Table 11 shows the mean (SD) change in HRQoL from baseline to 9 months by allocation group. Differences-in-differences estimates, comparing the control group with either the mini booster group or the full booster group, are also presented. These differences are very close to zero in all cases but very slightly favour the control group over either of the intervention groups.

This approach was adopted for each of the analyses.

Simple threshold analysis

Given an estimated cost of the intervention of £211 per participant, and assuming a willingness-to-pay threshold of £20,000 per QALY, the intervention would have to provide an additional 0.01055 QALYs to be considered cost-effective.

Short-term scenarios

Figures 31–42 present scatter plots, CEAFs and tables summarising mean scores and ICERs for each of the 12 scenarios described in Table 9.

FIGURE 31.

Scenario 1. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 32.

Scenario 2. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 33.

Scenario 3. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 34.

Scenario 4. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 35.

Scenario 5. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 36.

Scenario 6. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 37.

Scenario 7. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 38.

Scenario 8. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 39.

Scenario 9. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 40.

Scenario 10. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 41.

Scenario 11. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

FIGURE 42.

Scenario 12. (a) Scatter plot of difference in costs (£) against differences in QALYs; (b) CEAF; and (c) mean costs, QALYs (to two decimal places) and ICERs.

Summary of the short-term model results

The results of the separate scenarios for the short-term, questionnaire-based cost-effectiveness model suggest that the estimated cost-effectiveness of the intervention is subject to a high degree of structural uncertainty and depends on factors such as which data are used and which time periods are compared. In the majority of the scenarios, the control intervention, no booster, appears both the optimal choice and the option with the highest probability of cost-effectiveness at all willingness-to-pay thresholds considered. However, if the assumptions involved in scenarios 5 and 6 are made, there is an indication that the full booster may be the optimal choice when assuming a willingness-to-pay threshold of ≥ £20,000.

Looking at the estimated mean costs and mean QALYs for each arm in each scenario, it is apparent that the differences in costs and QALYs are marginal for all scenarios. For QALYs, the differences in mean values between arms are typically around or < 0.01 QALYs. The cost of the booster intervention is just one of a number of costs to the NHS and society that the participant population incurred and, compared with the costs of other NHS resources accessed by the participants over the trial period, it is not large. As the mean costs and QALYs observed in all arms are very similar, and the ICER is a ratio of two numbers, even slight decreases in costs or increases in QALYs could lead to very different indications of the cost-effectiveness of either intervention and so all results presented are potentially very dependent on statistical noise.

As previously discussed, the approaches taken to estimate the causal effect of the intervention on mean resource use and mean utility should be considered cautiously. This is partly because of the small sample sizes involved but mainly because the population considered does not as a rule suffer from any particular NHS resource-consuming and quality of life-reducing disease that the intervention is specifically designed to treat. Because of this, changes in HRQoL or NHS resource consumption over the relatively short time horizon of the trial are unlikely to be directly related to the effect of the intervention. As the intervention is more preventative than curative, the effect of the intervention on NHS resource consumption and HRQoL is likely to be relatively indirect, mediated by the effect of increased physical activity on lifelong morbidity and mortality risks, and to operate over a much longer time horizon. The long-term model results presented in the following section attempt to take these factors into account.

Individual sampling model results

In the long-term model, 9-month and 3-month mortality RRs associated with individuals within each intervention arm are based on TEE scores sampled directly from individual participant trial data. Because of the power law mapping equation used to associate individual TEE scores with mortality RRs, and slight differences in the age and gender distribution of the trial arms, the mean incremental effects of the intervention on HRQoL may differ from the effects on TEE. This is particularly likely to be the case when the distribution of baseline levels of physical activity differs by group, as the mapping equation assumes that the most sedentary individuals have much higher mortality RRs than slightly less sedentary individuals.

As discussed earlier, a number of separate primary scenarios were considered to investigate the impact of structural uncertainty on the model results. These are shown in Table 12, which shows mean incremental life-years and QALYs in each arm, and Table 13, which shows, to two decimal places, estimated incremental differences in effectiveness in the intervention arms compared with the control arm in each of the scenarios.

As Table 12 indicates, all estimates for the number of additional life-years lived and QALYs accumulated are very similar, with all values differing by < 1 life-year and < 0.5 QALYs. As Table 13 indicates, this in turn leads to very small estimates for incremental differences, of < 1 life-year and < 0.33 of a QALY. Because of the stochastic qualities of the models, and the effect of initial ages on QALYs and annual QALY increments, the direction of the incremental differences in life-years and QALYs is different for some of the scenarios, highlighting how marginal the differences between the trial arms were and the resulting influence on random variation of the estimated results.

Secondary analyses

In addition to the main long-term model results, two supplementary analyses were also conducted. In the first series of analyses, the levels of physical activity of all participants were set to fixed values. These fixed values were varied from the first (lowest) quintile to the fourth quintile observed in the trial. The mean additional utility that resulted from shifting up by one quintile was estimated for each of these baseline levels of physical activity so that the relationship between additional physical activity and baseline activity could be explored. The second series of analyses adopted a similar approach but used a level of physical activity gain based on the mean differences-in-differences estimates for TEE in the full booster group compared with the control group.

Scenarios assuming gains of one quintile

Given the power law relationship that appears to exist linking physical activity with mortality RRs, it is important to consider the effect of the baseline level of physical activity in estimating the cost-effectiveness of an intervention. Given that the least physical active quintile has the highest mortality risk, it can be assumed that a given improvement in physical activity is likely to be disproportionately effective in terms of reduced mortality in this population compared with less sedentary baseline populations. Within the scenario analysis it was assumed for simplicity that the intervention led to an improvement in physical activity of one quintile for 2 years. The effect of this temporary increase in physical activity was modelled when assuming that the entire population was initially in the first (most sedentary) quintile, the second quintile, the third quintile and then the fourth quintile. The results of this analysis are shown in Table 14 and Figure 43. The maximum acceptable intervention cost for each of these quintiles is presented, assuming a standard willingness-to-pay threshold of £20,000 per QALY.

TABLE 14 - Shift in physical activity quintile

SE, standard error.

Quintiles moved between	Mean utility gain	SE	Maximum acceptable intervention cost (£)
1 (most sedentary) to 2	0.122	0.0119	2430.70
2 to 3	0.046	0.0102	914.36
3 to 4	0.043	0.0094	853.83
4 to 5 (most physically active)	0.032	0.0088	649.66

FIGURE 43.

Shift in physical activity quintile.

Recruitment and retention

The principal shortcoming of this study was its failure to recruit 600 participants and retain 450 within the allotted time and budget. This is a common experience as only 31% of Medical Research Council- and HTA-funded trials that recruited participants between 1994 and 2002 met their recruitment target on time and within budget, with 45% closing to recruitment having reached < 80% of their target and 54% requiring a time extension. 104 Furthermore, there is considerable evidence that recruitment to trials evaluating preventative interventions is considerably more difficult than recruitment to those evaluating therapeutic interventions, with typical rates of those randomised as a proportion of those screened cited of 1–5% and 20–27%, respectively, according to one overview. 105 High levels of deprivation are known to predict poor consent rates in trials evaluating behavioural interventions for sedentary people. 106,107 It may be particularly difficult to recruit healthy participants in deprived communities for an intervention that may have significant costs to participants, including both the costs of behaviour change and the time and travel costs of attending for the intervention as well as recruitment, baseline data collection and follow-up. We were also recruiting participants from an age group (40–65 years) in which many people have both work and family responsibilities, which may make it particularly likely that perceived costs will outweigh the perceived benefits of participation.

Our study needed to recruit patients who had already received and responded to a brief intervention but was hampered by the absence of access to patients already being given a brief intervention in the local primary care setting whom we could recruit. For this reason the study team had to devise and distribute a brief intervention (see Chapter 2, Participants) and deliver it in partnership with the local primary care authorities but outside of the routine NHS care setting. The principal method of approaching participants was through personalised letters, electronically signed by the local director of public health, inviting people to contact the study team. This method generated 4964 responses from the 70,388 letters sent, a response rate of 7.1%. It is possible that we could have generated more responses by recruiting GPs to carry out the mail-outs. However, this method would have been much more expensive and there is little evidence that it would have been more effective. The Food and Immunity Trial (FIT), another South Yorkshire-based public health study, recruited people aged 65–85 years using mail-outs from general practices. 108 Its response rate was 528 out of 7482 (7.1%), identical to our own. The FIT study team used a wide range of other methods to recruit participants but GP mail-outs accounted for 90% of consented participants. It is worth noting that there is no good-quality evidence for the superiority of direct face-to-face recruitment compared with mail-outs in this population. 109 Even had there been it is unlikely that we would have had the reach or the resources required to reach full accrual using face-to-face accrual. We attempted to recruit through a range of other stakeholders, including GPs, health trainers and health champions, but this was not successful (see Chapter 3, Recruitmant of trial participants). This suggests that direct recruitment through primary care and community networks may not have improved recruitment yields. Previous research comparing targeted mail-outs with recruitment through general practices has found that mail-outs are more effective and cost-effective, even when there are strong collaborative links with GPs. 110

The brief intervention devised by Sheffield Hallam University reached 1934 (2.7%) of the 70,388 people who were invited to apply for it, although a further 568 who contacted and were screened by us were excluded from using the brief intervention, mostly because they were already physically active (see Chapter 3, Recruitmant of trial participants). Of those who received the brief intervention, 840 (43%) were not contactable at 3 months, 538 (28%) were contactable but had not increased their self-reported physical activity levels sufficiently to be eligible for the trial and 274 (14%) who were eligible could not be randomised (half of whom actively refused consent; the other half did not attend consent appointments). The concept of the ‘funnel effect’, in which a pool of potential patients who are contacted about entering a clinical trial becomes progressively smaller as it passes through successive screens and the informed consent process, is relevant here. 105 It is widely believed that the lower the percentage recruitment yield, the more questionable it is to generalise from the study findings to the target population. 111 We would assert that the results of this study are generalisable. Although we found it difficult to identify participants who had benefited from receipt of the brief intervention (the key eligibility criterion), we randomised 51% (282/556) of those we were able to identify (see Figure 6). The issue is the willingness to take up and respond to a brief intervention of those who could benefit, as supported by the published evidence. Of the four published RCTs evaluating brief interventions delivered to UK populations,112–115 none reported statistically significant improvements in self-reported physical activity. For the same reason it is fair to say that the MI boosters cannot be seen as public health interventions, because they cannot be rolled out at a community level. Rather, individuals will identify whether the offered intervention is appropriate and acceptable to them, and only a minority of sedentary individuals are likely to take such an intervention up.

Impact of area deprivation on recruitment

We aimed to recruit people from deprived neighbourhoods. We knew from the outset that there were pockets of affluence within the predominantly deprived areas that we surveyed. In a few areas people from LSOAs that were less deprived were recruited to the study (see Figure 9). However, the majority of the study participants were from relatively deprived LSOAs. Out of 147 LSOAs targeted, 102 were in the bottom 20% of national IMD 2010 overall scores.

The contact details for each recipient were confirmed as current by the NHS in the week before each mail-out. Despite this we received 1117 letters (1.6%) returned to sender because of the addressee being unknown at that address. There were undoubtedly more letters that were not returned to us, suggesting that a sizable minority of invitations were not received and reflecting the inaccuracy of the primary care databases that depend on patients reliably informing their current general practice when they move. Registers will be less accurate for more transient populations and this population characteristic may have had an impact on recruitment in more deprived neighbourhoods. We tested the hypothesis that the impact might be more severe in areas with transient populations but an exploratory GIS analysis using a simple linear model found no evidence that response rate was correlated with indicators of population transience (see Chapter 2, Participants, and Chapter 3, Baseline characteristics of participants). However, as noted in Chapter 3, there was a statistically significant difference in the response rates from different neighbourhoods, with fewer responses from more deprived LSOAs. It is important to stress that none of the areas approached was affluent in national terms and that this finding indicates relative resistance to recruitment from very deprived communities compared with only moderately deprived communities.

Of the randomised participants, 165 out of 282 were living in LSOAs in the bottom 20% of national IMD 2010 overall scores. This may indicate that, although we surveyed neighbourhoods that are normally understood to be deprived, the study population from these areas is likely to have been relatively more affluent. This is most visible in Figure 10 in which the plot for High Green, the geographical outlier to the north of the city, shows a ‘halo’ of participants from relatively affluent LSOAs.

Operationalisation of eligibility criteria

Four protocol violators (2.5% of the ITT population) were included in the ITT analysis (see Chapter 3, Baseline characteristics of participants). These participants did not meet the eligibility criterion of increasing self-reported physical activity by 30 minutes per week over the previous 3 months and did not signal, in the final SPAQ question, that the amount of physical activity that week was atypical. Academic opinion is divided on the utility of post-randomisation exclusions. For some, ‘the only safe way to deal with [protocol violations] is to keep all randomized patients in the trials’ (p. 464). 116 Others advise excluding ‘ineligible patients from analysis, provided that the eligibility criteria are absolutely clear and objective’ (p. 176). 117 Guideline E9 of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) also makes provision for the exclusion of participants if a ‘major’ eligibility criterion was violated. 118 Although ICH E9 fails to define ‘major’, commentators have suggested that it might include criteria that, the violation of which, produce ‘a less homogeneous set of patients (a possible loss of power)’ (p. 885). 119 With the benefit of hindsight we can see that the SPAQ was a far from objective screening tool: the reports of physical activity provided by those to whom it was administered correlated poorly with objective measures of physical activity (using the Actiheart device) during follow-up. Further, the use by RAs of negative answers to the final question (‘Was the last week typical of the amount of physical activity in minutes you usually do?’) to provide a basis for exemption from the eligibility criterion and an argument for randomisation was probably inconsistently and subjectively applied across randomised and non-randomised participants. For this reason it is unlikely that the presence of these four individuals seriously affects the homogeneity of the analysis set. A post hoc sensitivity analysis also suggests that the treatment effect was unaffected by their presence (see Chapter 3, Effectiveness of the booster intervention at 3 months). Although the concurrent use of the SPAQ as a screening tool is clearly problematic, it is not clear what the alternative might have been, given the need to screen 2502 unconsented members of the public at 3 months before randomisation to randomise just 282. The SPAQ also has a level of external validity that accelerometry might not, in that it, or self-report measures like it, are in use by the kind of local services to which our findings were intended to be generalisable.

Impact of the use of the Actiheart system on the collection of primary outcome data

The trial design team anticipated a participant attrition rate of up to 25%. In practice, the trial lost 81 participants (29%) to follow-up at the primary outcome assessment time point of 3 months (see Figure 6), despite deploying many of the preventative measures described by Brueton and colleagues. 120 Although this was disappointing, the loss of a further 18% of participants from the main analysis because of missing Actiheart data was entirely unexpected. The majority of early Actiheart studies were focused on the validation of the device for use in young people and healthy adults in controlled laboratory conditions. There are more recent studies in which Actiheart has been used in hospital settings, free-living environments and community settings. 121–123 However, few provide comments on its acceptability or suggest the kind of poor compliance observed in this study. The best available evidence on its acceptability comes from a recent study comparing instruments for the objective assessment of physical activity frequency and intensity. 122 In this study of Native Hawaiian participants, the Actiheart device was deemed more comfortable than ActiTrainer (an accelerometer and a Polar heart rate monitor; ActiGraph, Pensacola, FL, USA) and was worn for 6.5 days out of 7 for an average 12 hours per day. However, only 17 people wore the Actiheart in this study and two participants (12%) reported skin irritation. Although only a few people in our study reported irritation from the electrodes, many found that the electrodes did not adhere for the full week. As a result of this, and after trying several different brands of electrodes in addition to the ones recommended by the manufacturer, all participants were issued with spare electrodes so that they could replace them during the week as necessary. It is also interesting to note that the manufacturer of Actiheart, CamNtech, now offer a belt as an alternative to electrodes for holding the Actiheart in place. 124 As far as we are aware, this study is one of the first studies to suggest problems with the acceptability of the device in community settings and with repeated measures, with most published studies having asked participants to use the device on one occasion only. Our relatively large sample and limited involvement in placing/replacing the device means that this study provides a novel insight into the use of the Actiheart as a device for community-based assessment of physical activity.

The most problematic characteristic of accelerometer data is that activity is not measured over a uniform period each day and therefore determining the minimum number of monitoring days with valid accelerometry measurements is always a challenge in practice, especially in pragmatic trials. 38 Ideally, at least 7 days of physical activity monitoring using the Actiheart device is recommended to achieve estimates of physical activity in adults of approximately 90% reliability. 39 However, in practice, there is a balance between achieving the required degree of reliability in physical activity estimates and the number of days with valid accelerometry data required to optimise retention of study participants in the ITT population. To achieve 80% reliability with respect to activity counts and time spent in moderate to vigorous activity in adults, at least 3–4 days (of the 7 days) of activity monitoring are required. 39,125 In this regard, in this study, participants were required to have carried the Actiheart monitor for at least 4 days (of the 7 days) and have a minimum of approximately 7 hours (not lost > 1000 minutes) of accelerometry measurements each day.

This meant that a significant number of participants, despite completing the study, did not provide primary outcome data at 3 and 9 months and it is impossible to rule out information bias being introduced because of differences in physical activity levels between those who successfully used and returned the Actiheart monitor and those who did not.

Potential impact of trial participation on physical activity levels independent of the trial intervention

The phenomenon of behaviour modification caused by the act of being observed (the Hawthorne effect126) has been reported in some studies127,128 but not in others. 129

A specific concern that influenced the design of the booster trial was the perceived risk that the objective measurement of baseline activity (Actiheart) would have an effect on all participants’ physical activity levels because of their awareness of being monitored, potentially reducing the likelihood of observing an intervention effect in those also receiving booster interventions over and above this. In other words, the very act of measuring participants’ baseline activity levels, although an element of the research data collection process and not part of the intervention, might in itself act as an intervention, thus reducing the potential to have a ‘true’ control group with no intervention.

In theory we might expect this type of Hawthorne effect, but at least expect it to be similar between the booster group and the control group if Actiheart devices are administered in all arms. Therefore, we would not expect the estimated treatment effect, the difference in 3-month post-randomisation energy expenditure, to be affected by the Hawthorne effect in this scenario, mitigating any potential concern about an intervention effect masking the measurable impact of the face-to-face or telephone MI sessions.

However, it is possible that a Hawthorne effect did still occur during the study, as several participants reported in interviews that the expectation of seeing a researcher for their 3-month research study outcome assessment provided a motivation for staying physically active (see Chapter 4, The importance of monitoring and feedback). Although the same RA was never used for both the intervention and the research procedures, visual cues (such as Sheffield Hallam University Centre for Sports and Exercise Science logos on polo shirts) might have blurred the distinction in the minds of participants.

Lack of objective measurement of physical activity at baseline

The main rationale for not collecting baseline Actiheart data in our trial was to reduce the possibility of encountering the Hawthorne effect as discussed in the previous section. However, we acknowledge that baseline measurements could have had three potential advantages in this trial. First, the availability and utilisation of baseline Actiheart data would have improved the precision around the estimated treatment effect (the difference in outcomes between the booster group and the control group). Second, in hindsight, with a baseline assessment of energy expenditure we would have predicted problems with the Actiheart devices at an earlier stage of our study. We could have then developed strategies to minimise attrition and improve retention of participants with valid Actiheart data at 3 months. Third, baseline measurements of activity levels might have provided additional information about whether any differences between groups at follow-up were due to differential retention in the three arms, rather than intervention effects (which is always a potential source of bias when participants cannot be blinded to their intervention group).

Strengths and limitations of the process evaluation

Neither the process evaluation survey nor the topic guide for the interviews was piloted, although the response to questionnaire closed items was generally good. Another weakness was that, because of sickness and maternity leave, interviews were conducted by Sheffield Hallam University researchers who also delivered the intervention. However, no participant was interviewed by the same person who delivered their intervention and the analysis was conducted independently by University of Sheffield researchers not involved in the delivery of the intervention.

The use of qualitative research allowed for a nuanced interpretation of the quantitative elements and raised several supplementary questions that we were able to address through additional analyses, for instance the importance of relative autonomy in participants’ responses to the intervention.

Strengths and limitations of the trial intervention

The booster trial applied a humanistic counselling approach that attempted to action plan based on participant readiness, values and perceived resources. The approach was manualised and fidelity calibration of those delivering the intervention was attempted. This was a clear strength of the study intervention, which is rarely seen in physical activity counselling settings. 18 The theoretical approach (self-determination theory15) underpinned the cognitive and behavioural intervention in line with the treatment fidelity framework. This ensured a consistent rubric, based on empirical evidence, providing justification for the content and delivery of the counselling intervention (MI). Measurement of the cognitive effect of factors such as exercise motivation (measured using the BREQ-226) on behavioural adaptations to physical activity (measured using accelerometers) also provides evidence for the effectiveness of an MI approach. In the process evaluation the use of empathy and engagement was reported by the participants throughout, which is a fundamental part of MI. 89 These concepts are clearly distinct from a traditional medical model in which advice giving, warning and confronting with evidence are often found to generate the opposite result.

The MI counselling intervention, measurement of behavioural change using objective instruments (accelerometer) and measurement of epidemiological change (physical markers such as BMI) were again a strength of the trial in attempting to clearly assess and report the link between cognitive, behavioural and epidemiological interventions and outcomes. Although objective measures of behaviour change are difficult to achieve, this is likely to be improved as technology enables smaller and less intrusive instrumentation. Correlating the cognitive (or affect) and behavioural change pre–post allows inference of the change in cognitive affect (motivation) mediating a corresponding change in behaviour. Moreover, process evaluations can provide greater explanations of the reasons for this change. 130

The identification of the target behaviour (e.g. physical activity, diet or other lifestyle behaviours such as smoking) is fundamental to MI and yet is challenging to achieve and manage in a brief session (approximately 30 minutes) such as that delivered in the trial. Moreover, the duration of the full (face-to-face) and mini (telephone) booster sessions was more akin to the duration of a brief MI intervention, whereas longer and more frequent sessions are often found to be significant in other health settings. 131 The duration of the sessions was usually dictated by the participant, although the RAs’ limited use of deeper reflections and extended exploration of issues and participant values, as reflected in the MITI scores, could also have contributed to sessions being briefer than might be optimal. Multifaceted behaviours often emerged, which are rarely discrete, and combinations such as diet, alcohol, physical activity, smoking and medication adherence are challenging to manage in such a short period. Therefore, a longer period of time would be useful to explore the agenda, target behaviour and carry out action planning as a result. Lundahl and colleagues131 and Hettema and colleagues132 have suggested that MI interventions including up to six repeated sessions are the most significant and, although briefer interventions have been found to show positive results, these are more reliant on interventionist effects and influence. In the current trial the duration of the individual intervention sessions limited the opportunity to fully explore client values within the level of MI skillfulness of the RAs.

There was limited opportunity to explore maintenance and relapse prevention with participants within the mini and full booster MI sessions primarily because of the limited time available for the interaction. It is clear that action planning in physical activity contexts rarely embeds maintenance strategies (which include relapse prevention) and these need to be considered in future trials and community-based interventions. Although the action plan did attempt to account for setbacks and managing relapse this was not perhaps explicit enough and specific ‘active’ compared with ‘passive’ follow-up sessions would support this. 60 In addition to the value of specific cognitive–behavioural ‘maintenance’ strategies, a number of components were delivered (e.g. monitoring and normative feedback) although these were not made explicit enough in the intervention.

More generally, the complexity of the intervention could not be managed sufficiently in the short time available. ‘Adoption’ behaviours such as physical activity and diet change involve multifaceted lifestyle adaptations over a long duration and the current trial had a limited intervention duration with what could be described as a ‘passive’ follow-up. There was also limited opportunity for the use of explicit maintenance strategies that would have embedded cognitive–behavioural content such as adapting change goals, active follow-up and normative feedback. Future studies should make more explicit the use of maintenance strategies based on client readiness and changes in levels of motivation and opportunity for change.

Strengths and limitations of the geographical information systems analysis

The main strengths of the GIS analysis are the use of network distance rather than straight line distance and the use of the Sports England database for comprehensive coverage of local leisure facilities. Unlike some other studies in this field, which use Euclidean distance as a proximity measure (including a well-cited study by Maas and colleagues86), we have calculated walking distance using network information extracted from OpenStreetMap, which provides extensive coverage of pedestrian access routes.

The main weaknesses are the use of automated geocoding (rather than more precise but labour-intensive manual geocoding) and the use of a ‘distance to nearest’ measure rather than a more sophisticated ‘gravity’ distance model. A more sophisticated ‘gravity’ measure (such as that used in the study by Dai134) can take into account multiple facilities in the local area and it is possible that it could be used in future work; however, the required assumptions meant that it was not suitable for all of the facilities that we studied in this trial.

Strengths and limitations of the health economic modelling

Because of considerable structural uncertainty and difficulties in translating from the measures recorded in the booster trial to QALYs and costs, two alternative models were developed and within each model a wide range of scenarios was considered. The majority of these scenarios indicated that both the full booster intervention and the mini booster intervention appeared to be ruled out by simple dominance, although a small number of scenarios provide weak indications that the full booster may be the optimal choice assuming willingness-to-pay thresholds of approximately £20,000 per QALY or more. All scenarios were reflections of the evidence of the trial, indicating that in substantive terms there were only marginal differences between trial arms.

Although the short-term model has the benefit of being heavily based on evidence collected during the trial, it remains an open matter whether a lifestyle intervention such as the booster intervention is likely to produce a clinically significant effect in terms of either participant HRQoL or use of NHS resources over the duration considered. This is because the intervention is primarily preventative rather than curative and participants were not selected on the basis of suffering from a particular disease adversely affecting their HRQoL and NHS resource use. Because of this it may not be plausible to assume that increasing physical activity now will lead to systematic changes in HRQoL or resource use in the short term. Instead, positive health effects of the intervention may emerge over a much longer time horizon, reducing the risk of a range of diseases developing and so improving health outcomes indirectly.

It was because of this concern that the long-term model was developed to be considered alongside the short-term model. In this modelling approach, the HRQoL gains from increased physical activity result from the mediating effect of physical activity on relative mortality risks, meaning that a cohort of more physically active people should be expected to live for slightly longer on average than an otherwise identical cohort of slightly less physically active people and so experience slightly higher average HRQoL. The model implicitly incorporates the effect of morbidity on HRQoL, by using age- and gender-specific mean HRQoL scores, but does not incorporate an additional mediating effect of physical activity on either the onset or the severity of non-fatal diseases. As such, it is similar in its approach to the economic model developed by Brennan and colleagues. 135 The Brennan and colleagues’ model uses work by Andersen and colleagues75 to map physical activity levels onto relative mortality risks, whereas this model uses work by Myers and colleagues for the same purpose. 81 Both mapping exercises indicate qualitatively similar relationships between physical activity and relative mortality risk, suggesting that the greatest benefit of a given increase in physical activity is likely to be in people who are the most sedentary and that the marginal benefit of additional physical activity is less in people who are already relatively physically active.

The report by Brennan and colleagues135 contains a review of existing papers on the cost-effectiveness of physical activity interventions, as well as a brief review of existing model frameworks. Readers are directed to this publication for further information.

The economic model in this report does not yet directly consider the relationship between physical activity levels and morbidity risks. Morbidity is indirectly handled through the changing mortality risk ratios and also by adjusting HRQoL scores for age and gender (e.g. because of differences in disease burden at different ages). However, specific disease states are not considered in the model. A further iteration of the model could consider the additional cost and utility implications of different diseases by modelling annual probabilities of developing particular conditions and combining these with RR modifiers related to physical exercise levels. In theory, a physical activity intervention could then appear cost saving if the additional cost of the intervention is outweighed by the additional reduction in NHS resource use costs as a result of a reduced burden of disease. An important example of such a disease would be type 2 diabetes mellitus, whose onset is known to be linked to lifestyle factors and whose treatment represents a significant and increasing annual cost to the NHS. 136

Given the complex array of mediators and moderators involved in translating changes in physical activity into changes in mortality and morbidity risks, HRQoL and NHS resource use over the short, medium and long term, it is important that economic models are based on a solid clinical and epidemiological foundation. Models should incorporate current expert opinion on how physical activity interventions lead to sustained changes in lifestyle behaviour and through this to changes in health-related utility and costs, as well as the range of other mediating and moderating factors that need to be taken into consideration so that the model represents the patient experience with sufficient accuracy. The paths of causal influence linking individual- and ecological-level factors to short-term and long-term mortality and morbidity risks should be established through extensive consultation with clinical experts and used to establish flexible public health models that can be used to consistently model a wide range of interventions affecting different parts of the aetiological pathways. This process of attempting to understand the clinical and epidemiological consensus in understanding causal processes could be used to form directed acyclic graphs and influence diagrams, which can be used as the basis for economic models. 137 When it is identified that consensus between clinical experts does not exist with regard to certain causal pathways, the effect of different structural assumptions could be explored by constructing separate models that reflect alternative clusters of opinion.

Implications for practice and commissioning

The uptake of both the brief intervention and the booster intervention suggests that the potential impact on population-level change of these types of recruitment approaches and interventions is limited. The overall findings suggest that neither the mini booster nor the full booster is likely to be an appropriate population-level approach to promote physical activity in middle-aged populations in deprived areas. Better integration in local primary care services, such as GP clinics and IAPT (Improving Access to Psychological Therapies) programmes, might yield improved take-up and results, although even with better integration these types of interventions are unlikely to enable sufficient throughput of participants to have the reach of a true public health intervention.

The qualitative research component identified large numbers of people who wished to maintain physical activity levels in the hope of managing chronic disease symptoms. In particular, those with mental health issues (depression and anxiety) expressed a belief that physical activity would provide symptom relief, although the same people were often exercising alone in private spaces, were unenthusiastic about doing so and reported problems maintaining exercise levels. The social element of exercise may mediate some of the mental health benefits and those making exercise referrals should be aware of this. The needs of people with long-term injuries are different again; these people are typically seeking highly tailored exercise advice to suit their physical abilities. Referrers and advisors should be sensitive to these diverse needs.

The duration of both the mini booster and the full booster was approximately 30 minutes, which is more akin to a ‘brief’ MI session. Longer MI sessions have been found to be more effective and therefore a greater response might be expected from a higher dose, similar to the findings in contexts in which MI is more commonly applied, such as addictions settings and behaviour adoption (e.g. smoking and alcohol). 97 Although the current trial attempted to provide an ecologically valid intervention period, delivering a bespoke, person-centred intervention in this time frame that explored and set the client agenda, managed resistance and ambivalence and set appropriate goals with effective relapse prevention strategies was challenging, and delivering a greater number of sessions would be a realistic mechanism for managing this in the future. It is unlikely that delivering a greater number of sessions would be possible within existing care pathways, although the Let’s Get Moving programme59 proposes more than two sessions.

Face-to-face contact was more acceptable than telephone contact. More evidence is needed on relative costs and effectiveness as the results of this trial suggest that neither intervention is cost-effective at conventional levels of statistical significance and willingness-to-pay threholds. At present, PARSs are being decommissioned by a large number of primary care providers and their local council partners. PARSs provide a face-to-face physical activity promotion intervention that typically consists of up to 12 practical sessions, although not all of these are delivered on a one-to-one basis. Currently, the preferred option for physical activity promotion within the NHS is Let’s Get Moving, which involves multiple one-to-one contacts regarding physical activity that are either integrated into existing health-care contacts or delivered by the primary health-care team as standalone appointments. 59 This is arguably a cheaper way of delivering an intervention. It does, however, rely on individuals initiating contact with a member of their health-care team who is trained to provide the intervention.

Some people seemed to like the MI approach but others seemed to want, or had expected, a more didactic delivery approach. On the surface this seems counterintuitive and is at odds with the premise that MI can foster intrinsic motivation, which is linked to longer-term maintenance of activity, whereas more didactic advice-giving approaches would provide extrinsic motivation for being active. Many people we interviewed appeared to be extrinsically motivated, with few expressing thoughts about exercising for enjoyment, despite using a maintenance intervention designed to foster intrinsic motivation. This may be related to participants’ stage of readiness to be regularly active. Those in a less advanced stage such as contemplation or preparation (or even action) may be becoming active for extrinsic reasons such as to lose weight or improve health; a true appreciation of and desire for physical activity for its own sake may not come until individuals have been active for some time and entered the maintenance stage of behaviour change. By definition, the participants in this study have become active only in the 3 months preceding the delivery of the booster intervention and are therefore still in the action stage. Thus, although there is a focus on MI and related interventions that emphasise building intrinsic motivation for physical activity in the field of exercise psychology, it is possible that many people who have only just started to become more active may prefer to be told what they should be doing or participate in an exercise class. It is also possible that other differences aside from motivational style may influence people’s preference for a certain delivery approach. Whatever the reason, the implication here is that delivery of physical activity promotion should involve some form of assessment to ascertain each person’s preferred approach and then a relevant style of delivery should be selected from a toolkit of approaches. 138,139

The monitoring and feedback elements of the intervention were identified as important motivators by participants. However, this may be another indicator of the prevalence of extrinsic motivation amongst these participants, and psychological theories of behaviour change suggest that extrinsically (rather than intrinsically) motivated behaviour change might impact negatively on the sustainability of the treatment effect.

The role of a professional has traditionally been seen as one of diagnosis and prescription for both pharmacological and behavioural interventions. In recent years, this has given way to a more inclusive person-centred approach, which challenges the individual, with empathy, to take a more active role in their behaviour change. The focus of more health practitioner sessions, also described as a ‘directive approach’, can resolve people’s resistance to change. 140 However, the process evaluation suggested that some people have concerns about the client centeredness of MI consultations (see Chapter 4, Motivational interviewing: consistency with perspectives or world views). As in other studies involving low socioeconomic status groups, some of our participants said that they had hoped for more paternalistic and didactic communication approaches. 63 Although autonomy-supported communication is understood to enhance intrinsic motivation (whereas paternalistic approaches diminish it), it has been suggested that MI can be adapted to suit client communication preferences. 63–65 Therefore, if the motivational interviewer becomes aware that the client would prefer a more didactic approach, the consultation can be directed towards that need. This ‘way of being’ with the client is a key part of skilful MI and a flexible and adaptive approach is an important part of challenging with empathy in a bespoke manner.

The process evaluation also suggested that two key motivators for adherence to physical activity targets are the anticipated shame, when face to face with a professional monitor, if one does not achieve the agreed objectives, and the pleasure of receiving feedback on one’s performance (see Chapter 4, The importance of monitoring and feedback). Different methods of feedback are the subject of the Feedback, Awareness and Behaviour (FAB) study (ISRCTN92551397; funded by the National Institute for Health Research programme). 141 The FAB team aims to randomly allocate 500 people aged 30–55 years to either a control group (no feedback) or one of three types of personalised physical activity feedback (‘simple’, ‘visualised’ or ‘contextualised’) and complete repeat measures of self-rated physical activity and psychosocial correlates.

One concern for physical activity promotion and ongoing patient behaviour change support is that there is no clear pathway for physical activity promotion at the moment in the UK. Current provision is limited to GP advice to individual patients to become active or, at best, referral to a PARS. Although other community-based interventions such as ‘green gyms’ and walking programmes do exist, these are rarely integrated into the care pathway. The traditional approach to physical activity promotion delivered in health-care settings is one of information exchange, which can often be seen as provision of unsolicited advice. In 2009, the Department of Health launched Let’s Get Moving59 to be delivered by health-care professionals and complement the PARS. Let’s Get Moving is physical activity counselling intervention based on MI principles and designed to be delivered by health-care professionals within usual care settings. Early research questioned the feasibility of using Let’s Get Moving in a primary care setting. 82,142 Materials have recently been revised and were relaunched in March 2012. Aside from referral to a PARS and Let’s Get Moving, there are some nurse practitioners who do deliver diet and exercise counselling within primary care. This provision, however, is variable by nurse and practice.

It is therefore essential that local strategies for physical activity promotion and local commissioning by clinical commissioning groups and local authorities is in future informed by the wider evidence base in relation to the broad range of potential community and individual interventions. As resources are limited and the potential for population benefits large, a systematic approach, prioritising the most cost-effective interventions and targeting the groups with the greatest baseline risk and ability to benefit from supportive interventions, is required.

Sponsor

Gill Wells, Research Development Manager, the University of Sheffield, New Spring House, 231 Glossop Road, Sheffield, S10 2GW.

Persons authorised to sign the protocol & amendments

Elizabeth Goyder, Director of the Section of Public Health and Reader in Public Health Medicine, ScHARR, University of Sheffield, Regent Court, 30 Regent Street, Sheffield, S1 4DA. Tel: (+44)(0)114 222 0783; Fax: (+44)(0)114 222 0791; e-mail : e.goyder@sheffield.ac.uk.

Partner institutions

Trial Manager

Daniel Hind, Research Fellow, Clinical Trials Research Unit, University of Sheffield, Regent Court, 30 Regent Court, Sheffield, S1 4DA. Tel: 0114 222 0707; Fax: 0114 222 0870; e-mail: d.hind@sheffield.ac.uk.

Trial Statistician

Stephen Walters, Reader in Medical Statistics, ScHARR, University of Sheffield, Regent Court, 30 Regent Court, Sheffield, S1 4DA. 0114 222 0730; Fax: 0114 222 0870; e-mail: s.j.walters@sheffield.ac.uk.

Trial Steering Committee

Trial Management Committee

Trial Summary

This study assesses whether it is worth providing further support, 3 months after giving initial advice, to those who have managed to do more physical activity. All participants will initially be given an interactive DVD, supported by advice from a trained facilitator. The facilitator will provide two telephone follow ups at one month intervals. Only those that have increased their physical activity at this point will remain in the study. These participants will receive a ‘mini booster’, a ‘full booster’ or no booster. The ‘mini booster’ consists of a two telephone calls one month apart to discuss physical activity and usage of the DVD. A ‘full booster’ consists of a face-to-face meeting with the facilitator at the same intervals. The purpose of these booster sessions is to help the individual to maintain their increase in physical activity. We will measure the differences in physical activity, quality of life and costs, associated with the booster interventions, 3 months and 9 months from randomisation. The research will be carried out in 20 of the most deprived neighbourhoods in Sheffield. These locations have large, ethnically diverse populations, high levels of economic deprivation, low levels of physical activity, poorer health and shorter life expectancy. Participants will be recruited through general practices and community groups, as well as by postal invitation to ensure the participation of minority ethnic groups and those with lower levels of literacy. Sheffield City Council and Primary Care Trust fund a range of facilities and activities to promote physical activity and variations in access to these between neighbourhoods will make it possible to examine whether the effectiveness of the intervention is modified by access to community facilities.

Rationale

There are a number of published systematic reviews of evidence for interventions that increase physical activity. 1–5 More recently the evidence base for brief interventions in primary care has been reviewed. 6 This review identified a sufficient evidence base for NICE to recommend the use of brief interventions to promote physical activity but also identified specific evidence gaps that this trial will be able to address, particularly in relation to the value of follow up beyond three months, for the longer term maintenance of physical activity.

Searches of the National Research Register and ClinicalTrials.gov for research in progress confirm that although there are a number of physical activity intervention trials in progress in specific patient groups and in older age groups or in children, there are few trials including ‘healthy’ middle-aged participants and no other trials specifically examining the value of further intervention after an initially successful ‘brief intervention’.

Investigational interventions

The trial will compare a ‘mini booster’ of two telephone physical activity consultations and a ‘full booster’ of two face-to-face physical activity consultations, offered four and five months after an initial brief intervention, to a standardised three month brief intervention alone. The purpose of these booster sessions is to help participants to sustain their physical activity levels and prevent relapse. The brief intervention will involve provision of an interactive DVD based on a MI approach that is directive, client-centred and replicates the style of other successful behaviour change programmes. 7,8 All interventions, including the initial brief intervention, will be delivered by trained facilitators (employed as research assistants and trained by the research team) to ensure consistent delivery.

Theoretical underpinning of interventions

Meta-analytical and systematic reviews of physical activity and behaviour change9,10 suggest that the transtheoretical model (TTM)11 is the most commonly adopted theoretical framework for promoting physical activity. The TTM has demonstrated effectiveness as an approach to increasing exercise adoption and adherence in adults. 10,12–14 The TTM describes how people modify problem behaviours or acquire positive new ones. 11,15,16 The TTM determines behaviour change as a process rather than a single event and offers practical suggestions for how individuals can change behaviour. The TTM consists of the following constructs: stages of change (describes when people change), processes of change (outlines techniques for helping people to change), decisional balance (weighing up the pro’s and con’s of change) and self-efficacy (increasing one’s confidence to change behaviour). 11 The TTM offers practitioners a common, validated framework for guiding participants through periods of change and proposes strategies for maintaining positive behaviours. We will also adopt a client centred approach to all interventions based upon the style of motivational interviewing.

Content of the Brief Intervention (Interactive DVD and telephone follow up)

Consistent with NICE guidance on physical activity interventions, the brief intervention will aim to help middle age adults consider, initiate and maintain physical activity behaviours. 6 The DVD represents an interactive tool that is based on the style of motivational interviewing and the principles of the TTM. The DVD offers individuals the opportunity to choose information on the following: the benefits (social, health, environmental) of physical activity; current physical activity recommendations; things to consider before starting; different types of physical activity; building confidence and efficacy to become physically active; myths and misconceptions about physical activity; staying motivated; sign-posting of opportunities to be physically active in Sheffield/South Yorkshire and example case studies.

A. NICE Physical Activity Guidance (March 2006)6

‘identify inactive adults and advise to aim for 30 minutes of moderate activity on 5 days of the week (or more)’. The DVD will allow individuals to assess their own physical activity relative to current physical activity recommendations that will be provided on the DVD. Information will be provided that will help individuals to understand what is meant by ‘active’ and ‘moderate intensity’.

‘When providing physical activity advice, primary care practitioners should take into account the individual’s needs, preferences and circumstances. They should follow them up at appropriate intervals over a 3 to 6 month period’. The DVD offers trainers a tool to support individuals in making healthier lifestyle choices regarding positive change in physical activity. The DVD is theoretically underpinned by the Trans-theoretical model of behaviour change and the Theory of Planned Behaviour – which both place emphasis upon the individual considering their perceived behavioural control, subjective norms and attitudes when initiating behaviour change – such as increasing physical activity. The DVD asks the individual to reflect upon their individuals needs, preferences and circumstances and this is done within an evidence-based stage matched approach – meaning that greater emphasis will be placed on those factors (such as circumstances and needs) that mediate behaviour change at the early stages of change adoption (pre-contemplation) and factors that mediate sustaining the new behaviour such as reinforcement management that will be more salient for those individuals who have already decided to take action but need help to implement change.

‘Practitioners should agree goals with them’. Goal setting, what it means and examples of goal setting for physical activity will be given in the DVD. This will put the potentially abstract theory of goal setting into a real life context. Appropriate language will be used to facilitate understanding and relevance.

‘They should also provide written information about the benefits of activity and the local opportunities to be active’. The DVD will provide a comprehensive list of physical activity opportunities within the Sheffield area with contact details. These will include activities that have already been found to be popular with minority ethnic groups eg. dance activities.

‘Local policy makers, commissioners and managers, together with primary care practitioners, should pay particular attention to the needs of hard to reach and disadvantaged communities, including minority ethnic groups, when developing service infrastructures to promote physical activity. ’ The information given within the DVD will have been discussed with local ethnic minority community groups to ensure it is culturally and contextually appropriate. The local strategic partnership will be consulted about current service provision for disadvantaged and ethnic minority groups.

B. NICE Behaviour Change Guidance (October 2007)31

‘Employ a range of behaviour change methods and approaches, according to the best available evidence. These concepts could be used to structure and inform interventions.’ Action planning, implementation-intention action plans, promoting user autonomy, providing tips and practical strategies to develop physical activity self-efficacy, adopting a client centred approach, information giving, expert based advice, peer modelling, engaging in reflective tasks, contingency planning are all features of the DVD intervention that comply with the list of concepts that are advocated by NICE to structure behaviour change interventions. Through engagement with the DVD, individuals will be able to devise their own personal physical activity plan, understand the benefits of physical activity, be given information regarding safe physical activity, how to monitor intensity, myths about activity, information re the difference between activity for health and fitness.

‘It should be taken into account behaviour is embedded in social, cultural & material circumstances. ’ The complexity of an individual’s personal circumstances, their socioeconomic and cultural context and their interactions with health behaviours will be considered in the content of the DVD and in ensuring associated advice is realistic and culturally appropriate.

Compliance with pragmatic practice

The brief intervention will be delivered as part of the local strategy for promoting physical activity in more deprived neighbourhoods within the Enhanced Public Health Programme which is based on a comprehensive needs assessment and targets specific neighbourhoods with poorer health outcomes. The use of DVDs is in line with the strategic approach and methods developed locally by Active Sheffield and supported by the partnership of bodies that deliver support for increasing physical activity across Sheffield. Use of DVDs is already being introduced in other areas with anecdotal success but without systematic evaluation. Health First, the specialist health promotion agency for Lambeth, Southwark and Lewisham, has recently produced a DVD designed to promote physical activity. This DVD includes similar content to that in our proposed intervention, tailored for a local population. The DVD ‘Choosing Physical Activity’ is in the public domain and can be viewed using http://video.google.co.uk.

The main reasons for not using existing community or primary care practitioners to deliver an initial brief intervention were both to ensure fidelity to the intervention protocol (so an ineffective intervention can be distinguished from inadequate delivery of an effective intervention) and also because in practice, if the intervention is effective and introduced into practice, it will need to be delivered consistently and efficiently across a large population with capacity to benefit by a range of health trainers and other community staff as well as primary care staff. The proposed DVD has the potential to reach a wide targeted audience at minimal cost compared to a one to one based intervention with a health professional.

Population

Men and women aged between 40 and 64 who have increased their physical activity by at least 30 minutes per day.

This trial will be conducted in compliance with the protocol, GCP and regulatory requirements.

Primary objective

To determine whether physical activity measured by accelerometry three months after randomisation (six months after a brief intervention) is significantly increased in participants allocated to two intervention groups (receiving two booster physical activity consultations, delivered in a motivational interviewing style, either by telephone or face-to-face) compared to participants allocated to a control group (receiving no further contact after the baseline assessment).

Secondary objectives

1. To determine whether physical activity nine months after randomisation (12 months after the brief intervention) is significantly increased in participants allocated to the two intervention groups compared to participants allocated to the control group.

2. To compare physiological measures of fitness (12 minute walk test) and self-reported physical activity (SPAQ instrument) between allocated groups.

3. To compare health related quality of life, resource use (including health and social care contacts) and economic costs between allocated groups.

4. To investigate whether the impact of the intervention may be modified by gender, ethnicity or the types of physical activity undertaken (including use of community facilities for physical activity).

5. To undertake a process evaluation to identify, using both quantitative and qualitative methods, psychosocial and environmental factors that may mediate or modify the effectiveness of the intervention.

Design

This is a three-arm, parallel group, randomised controlled trial with a feasibility study.

Feasibility study

In the first year a pilot trial will be undertaken to assess the feasibility of both trial recruitment plans and the proposed interventions39,40. A total of 3000 mailshots will be sent to the patients of a general practice situated in a ‘typical’ deprived ward (Manor Ward), at least 150 will receive the brief intervention and 60 randomised to the three trial arms. This will allow outcome measurement in 15–20 individuals in each study arm to estimate a mean and standard deviation for the primary outcome, total energy expenditure per day (averaged over a 7-day accelerometry assessment period in each group using the Actiheart Device). The main risks to trial success identified by reviewers that the feasibility trial will test are:

1. Recruitment targets for the brief intervention will not be met;

2. The brief intervention will not be effective enough to generate sufficient individuals eligible for the trial;

3. Insufficient eligible individuals will consent to participate in the trial.

These three issues in combination will determine whether the trial recruitment rate is adequate.

The success criteria for the feasibility study will therefore be:

1. At least 60 patients recruited to the pilot trial and 45 having 3 month follow-up measurements including accelerometry completed on the basis of an initial mailshot to 3000 individuals. (We will not use community recruitment at this stage since it may represent a more limited pool for recruitment that we can use to booster participants from ‘hard-to-reach’ groups as required in the main trial)

2. At least 70% of those randomised to booster interventions actually receiving the interventions per protocol

3. On the basis of the pilot primary outcome (accelerometry) data collected, the sample size for the main trial will be re-calculated. The trial will not proceed if the revised sample size calculation suggests a total sample size > 600 will be required. Assuming the protocol and intervention remain unchanged, the participants recruited during the feasibility phase will be included in the full trial population.

Main trial

The trial participants will be recruited from the 20 most deprived neighbourhoods of the city of Sheffield, based on Index of Multiple Deprivation (IMD) 2004 and health indicators. Average life expectancy is six years lower in lowest versus highest quintile of IMD (based on data from www.neighbourhoodstatistics.gov.uk). Up to 30 000 middle-aged residents, (aged 40 to 64 years at recruitment), will be invited to participate in a brief intervention to help them get more physically active. Up to 1500 residents who respond to the invitation will receive an initial assessment telephone call to determine their physical activity status. Eligible participants (i.e. those not already meeting current recommendations of 30 minutes moderate activity, 5 times a week) will then receive an interactive DVD and supporting written materials through the post. Follow up telephone contacts will be made one month and two months from initial contact. Active Sheffield, the organisation responsible for promoting physical activity in Sheffield, will ensure participants have access to a range of community facilities for exercise and will provide regularly updated information on current provision.

After three months, the Study Introduction letter will be sent to all DVD recipients and a further telephone assessment will establish whether they are eligible to participate in the booster trial (i.e. have increased their activity by at least 30 minutes per week and are willing to have further assessment and follow up). Eligible participants will then be invited to attend a baseline assessment appointment at a community venue and they will be randomly allocated to one of three groups:

1. a control group who will be assessed at randomisation, after three months and after nine months and receive no additional intervention between those assessments;

2. a ‘mini booster’ group also receiving an intervention comprising two telephone-based physical activity consultations, delivered in a motivational interviewing style, at one month and two months from randomisation;

3. a ‘full booster’ group also receiving an intervention comprising two face-to-face physical activity consultations, delivered in a motivational interviewing style, at one month and two months from randomisation.

Written consent to trial participation will be obtained at the start of the trial baseline assessment meeting. All randomised participants will be assessed at baseline, 3 months and 9 months from randomisation (i.e. 3 months, 6 months and 12 months from the initial contact for the brief intervention). Where possible, staff conducting assessments will not know participants’ group allocation and participants will be asked not to tell them.

Endpoints

1. Objective measure of physical activity including:

   1. – Total Energy Expenditure (TEE), in Kcal per day, from seven-day accelerometry and heart rate monitoring using Actiheart

   2. – Physical activity counts (PAC) per week;

   3. – Minutes of moderate/vigorous physical activity per day;

   4. – Meeting the current physical activity recommendation of at least 30 min per day (continuous or in bouts of at least 10 min] of at least moderate intensity) for at least 5 days a week (yes or no).

2. Self-reported moderate or strenuous physical activity using the Scottish Physical Activity Questionnaire (SPAQ) which records type and duration of activities in the previous week;

3. Health-related quality of life using the Sheffield Version of the 16-item Short Form health survey instrument (SF-12v2 plus 4);

4. Self-reported use of community facilities for physical activity;

5. Self-reported health and social care contacts;

6. Psychological measures of motivation, intentions, attitudes, beliefs, social influences and self-efficacy towards physical activity, measured using the Theory of Planned Behaviour. 41 Exercise stages of change42, and self-determination will be assessed using Behavioural Regulation in Exercise Questionnaire43 and questions used in the HTA-funded Exercise Evaluation Randomised Trial (EXERT). 44 This will allow comparison with results from other physical activity trials including EXERT.

7. Body weight and height (to allow calculation of BMI)

8. Physiological measures of fitness (12 minute walk test)45

Design measures to avoid bias

The allocation schedule will be concealed through the use of a centralised web-based randomisation service. The randomisation sequence is computer-generated. Data analysts will be blind to treatment allocation, but the study manager, participants (who are also outcome assessors) will not be blinded. Analysis will be by intention-to-treat. Where individuals are lost to follow-up or data is missing, imputation methods will be employed, which will be described in the statistical analysis plan.

Randomisation codes and allocation concealment

The randomisation schedule is generated prior to the study by the Clinical Trials Unit Randomisation Service. On identification of an eligible volunteer, the study manager or data manager will randomise and inform the patient and their general practitioner on the treatment allocation.

Inclusion criteria for brief intervention

1. Residents of the 20 most deprived neighbourhoods in the city of Sheffield

2. Aged 40 to 64 years

3. Not achieving the current recommended activity level (30 minutes of moderate activity on at least 5 days) assessed using the SPAQ46 and wishing to have support to become more active

Additional inclusion criteria for booster trial:

1. Have increased their physical activity level by at least 30 minutes of moderate or vigorous activity per week (assessed using the SPAQ) since initial assessment of activity level

2. Capacity to give written informed consent to trial participation

Exclusion criteria

Individuals with chronic conditions who can benefit from physical activity will not be excluded unless their condition significantly impairs their ability to exercise. They will be asked to consult their GP if they have a condition that increases their risk of adverse events during exercise (i.e. chronic cardiovascular or pulmonary disease).

Criteria for withdrawal from trial treatment

Participants may withdraw from active participation in the study on request. If a participant experiences chest pain or severe breathlessness during the 12-minute walk test, then the researcher will advise the GP directly and immediately, and will also advise the participant to make an appointment with their GP at their earliest convenience.

If analysis of Actiheart (accelerometer) readings suggests pre-existing arrhythmias, this information will be shared with the participant and their GP.

Subjects removed from active participation will not be replaced and, with their consent, will be followed up for all outcome information.

Procedures required at screening or before randomisation

Letter 1. Up to 30 000 middle-aged residents will be sent a letter and business response envelope inviting them to enroll in a programme to help them get more physically active.

Scottish Physical Activity questionnaire. The research team will send this out to potential participants with Letter 1. It will be administered face-to-face by a member of the research team at screening and sent out again by post on two subsequent occasions to participants (3 months and 9 months after randomisation).

Brief intervention

Interactive DVD and supporting written materials, delivered through the post. The research team will send this out to potential participants who respond to Letter 1.

Procedures required at initial follow-up

Follow up telephone contacts, one month and two months from initial contact to assess DVD usage and offer advice on physical activity. A member of the research team will contact the individual at home by telephone.

Procedures required at screening

After three months, the Study Introduction letter will be sent to all DVD recipients and eligibility to participate in the booster trial will be assessed by telephone. Eligible participants will be invited to attend a baseline assessment.

Procedures required before randomisation

Participant information sheet sent to potential participant at home. The research team will send this out to potential participants who are eligible and willing to participate in the trial (see above).

Visit to a community venue for informed consent, baseline assessment and randomisation. A member of the research team will meet and consent the individual, take baseline assessments and randomise them. A member of the research team will administer:

• Scottish Physical Activity questionnaire;

• Short-Form 12v2 plus 4 questionnaire, plus one wellbeing question;

• Behavioural Regulation in Exercise Questionnaire (BREQ-2);

• Exercise Evaluation Randomised Trial (EXERT) questionnaire ;

• Personal information questionnaire (Questionnaire 1).

• Questions about use of community facilities for physical activity, and about health and social care service contacts; and,

• Measurement of weight and height.

Procedures required at three month follow-up

A member of the research team will meet the individual, take baseline assessments and randomise them. A member of the research team will administer:

• Scottish Physical Activity questionnaire;

• Short-Form 12v2 plus 4 questionnaire, plus one wellbeing question;

• Behavioural Regulation in Exercise Questionnaire (BREQ-2);

• Exercise Evaluation Randomised Trial (EXERT) questionnaire’

• Questions about use of community facilities for physical activity, and about health and social care service contacts;

• Measurement of weight and height;

• Twelve-minute walk test; and,

• Seven-day accelerometry using Actiheart.

Procedures required at nine month follow-up

A member of the research team will meet the individual, take baseline assessments and randomise them. A member of the research team will administer:

• Scottish Physical Activity questionnaire;

• Short-Form 12v2 plus 4 questionnaire, plus one wellbeing question;

• Behavioural Regulation in Exercise Questionnaire (BREQ-2);

• Exercise Evaluation Randomised Trial (EXERT) questionnaire ;

• Questions about use of community facilities for physical activity, and about health and social care service contacts;

• Measurement of weight and height;

• Twelve-minute walk test; and,

• Seven-day accelerometry using Actiheart.

List procedures for attempted follow-up of patients ‘lost to follow-up’

Patients will be considered lost-to-follow-up if they fail to respond to questionnaires, one reminder letter and two telephone calls. There are no procedures for further follow-up.

Procedures required when closing a trial (premature or planned)

At the point at which all questionnaires have been collected (or participants have failed to respond despite reminders) and all data have been entered and cleaned, the management group will approve closure of the database. Further details will be presented in the data management protocol.

Procedures required to record serious adverse events

At each follow-up, participants will be asked if they have experienced any event or illness which:

• has required unscheduled hospitalisation; or,

• has resulted in persistent or significant disability/incapacity.

The details of serious adverse events will be confirmed with the participant’s general practitioner before classification.

Number of patients to be enrolled

600.

Reason for choice of sample size

The original sample size calculation assumed that physical activity would be measured using a simple hip-mounted accelerometer. It was also assumed that a mean difference of 400,000 PAC per week between the intervention and control groups at three months was the smallest clinically and practically important difference and that the SD of this outcome was 1.2 million counts/per week. Hence with 450 participants (300 intervention: 150 control), the main trial was determined to have 90% power to detect this mean difference or greater between the intervention and control arms as statistically significant at the 5% (two-sided) level using a two independent samples t-test. Assuming an approximate 25% loss to follow-up by three-months, it was proposed to recruit and randomise 200 participants per group giving total sample size of 600. The Actiheart accelerometer measures physical activity counts per week on a different scale of magnitude to a simple hip mounted accelerometer with a considerably lower mean and standard deviation.

When re-estimating the sample size using data from an internal pilot study the revised sample size estimate either stays the same or increases (it cannot be less than the original estimate). The original sample size calculation of 450 subjects with valid outcome data at 3 months post-randomisation, was based on detecting a standardised effect size of 0.33 or a mean difference of one-third of a standard deviation in the outcome measures between the intervention and control groups. This equates to an estimated mean difference between the Booster and Control groups, based on the observed standard deviation from the feasibility stage of 34,465 PAC per week and 102 kcal per day for TEE.

To have a 90% power of detecting a mean difference of 102 kcal in mean TEE per day is between the groups would require 426 participants in total with evaluable data (control = 142; intervention = 254). Similarly, the total required sample size under the above conditions to detect a mean difference of a 34,465 PAC per week is 429. Therefore since the re-estimated sample size, of around 430 participants, is lower than the original estimate of 450 participants the trial will proceed with the original sample size estimate of 450 participants with evaluable data.

Statistical criteria to terminate the trial

There are no statistical criteria for stopping the trial early; as the intervention is considered low risk, there is no DMEC and decisions to stop the trial will be made on safety grounds by the Trial Steering Committee.

Procedure for accounting for missing data

The primary analysis will be an ITT analysis with participants with complete accelerometry data at three months post-randomisation. A sensitivity analysis will be undertaken to impute missing accelerometry data using baseline and follow-up data from the group of patients with valid accelerometry data at three months post-randomisation. As this is an ITT analysis, withdrawals and protocol violations will be analysed in their groups as randomised.

Analysis of primary objective

As the trial is a parallel group RCT data will be reported according to the revised CONSORT statement. 47 The statistical analyses will be performed on an intention-to-treat basis. All statistical exploratory tests will be two-tailed with alpha = 0.05. Baseline demographic variables (age, gender), physical measurements (e.g. weight, height, BMI), and health-related quality of life data (SF-36) will be summarised with appropriate summary statistics, tabulated and assessed for comparability between the treatment groups. For example, categorical variables (e.g. gender, the number and percentage who are male and female will be reported). For continuous variables, e.g. age, depending on the distribution of the data, if it is symmetric, the data will be summarised with a mean and standard deviation; if it has a non-symmetric distribution it will be summarised with a median and inter-quartile range.

The primary aim is to compare the intervention (Full or Mini Booster) versus control treatment (No booster). Secondary aims are to compare the two interventions (Full versus Mini booster). The primary comparison will be between the mean physical activity levels from the Actiheart accelerometer (average Total Energy Expenditure per day in Kcal, from a seven-day assessment) in the two ‘booster’ arms combined compared with the mean physical activity levels in the control arm at 6 months follow-up (3 months post-randomisation). This difference in means, between the intervention and control groups, will be compared using a two independent samples t-test and a 95% confidence interval for estimated mean difference between the groups will also be calculated. In the event of differences between the Booster and Control groups with respect to baseline demographic, physical, and health-related quality of life and accelerometer measurements, multiple regression will be used to adjust the treatment effect for these variables. The ordinary least squares adjusted regression coefficient estimate for the treatment group parameter along with its 95% confidence interval (CI) will then be reported.

The research hypothesis is that the booster interventions will have greater levels of physical activity than the control. The statistical and null hypothesis is that there are no differences between the intervention and control groups at follow up. The alternative hypothesis is that there is a difference in physical activity levels between the intervention and control groups at follow up.

Secondary aims are to compare the effect of the two interventions (Full versus Mini booster). This will be done using the same methods as for the primary endpoint as described above. Interim analyses will not be required. An exploratory sub-group analysis using multiple linear regression, with the primary outcome the mean physical activity levels from the Actiheart accelerometer (average Total Energy Expenditure per day in Kcal, from a seven-day assessment) at 6 months (3 months post-randomisation), will look for an interaction between treatment group (Booster or control) and sub-groups defined by gender, ethnicity and access to community facilities (self reported use versus no use of community facilities).

Analysis of secondary outcomes

Analyses will identify any significant different between groups for each outcome measure, at three months and nine months from randomisation:

1. Objective measures of physical activity from the Actiheart:

   1. – Physical activity counts (PAC) per week;

   2. – Minutes of moderate/vigorous physical activity per day;

   3. – Meeting the current physical activity recommendation of at least 30 min per day (continuous or in bouts of at least 10 min of at least moderate intensity) for at least 5 days a week (yes or no).

2. Physiological measures of fitness (12 minute walk test) and types of physical activity (self report) and change in self-reported physical activity levels

3. Change in health-related quality of life measured by changes in SF-12v2 plus 4 (converted to SF-6D)

4. Health and social care contacts

5. Changes in psychological measures of motivation, intention and stages of change, and self-efficacy

Secondary categorical outcomes such as the proportions maintaining (or increasing) their weekly duration of physical activity in the two ‘booster’ arms combined compared with the proportion in the control arm at 6 months follow-up (3 months post-randomisation), will be compared between the intervention and control groups, using a continuity corrected chi squared test and a 95% confidence interval for estimated differences in proportions will also be calculated. In the event of differences between the groups with respect to baseline demographic, physical, and health-related quality of life measurements, multiple logistic regression will be used to adjust the treatment effect for these variables. The maximum likelihood estimated regression coefficient for the treatment group parameter (odds ratio) along with its 95% confidence interval (CI) will then be reported.

Secondary outcomes such as HRQoL (SF-12v2 plus 4 dimension scores) and distance walked on 12 minute walk test, at six month follow-up, will be assumed to be continuous outcomes. A two independent samples t-test will be used to compare mean outcomes between the Booster and control groups in this parameter. A 95% confidence interval (CI) for the mean difference in this parameter between the groups will also be calculated. In the event of differences between the Booster and Control groups with respect to baseline demographic, physical, and health-related quality of life measurements, multiple regression or analysis of covariance (ANCOVA) will be used to adjust the treatment effect for these variables. The ordinary least squares adjusted regression coefficient estimate for the treatment group parameter along with its 95% confidence interval (CI) will then be reported. Twelve month outcomes will be analysed in a similar way. We shall also compare the effect of the two interventions (Full versus Mini booster) on these secondary outcomes at 3 and 9 months post-randomisation, using the same methods as described above.

Economic analysis

The basic design of the health economic component of the study will be to estimate the incremental cost effectiveness of the mini-booster and full booster interventions compared to no booster. It will include an estimation of the cost effectiveness of the intervention from a NHS perspective in terms of their incremental cost per quality adjusted life year (QALY) and a broader societal assessment of efficiency that includes costs for other Government agencies and productivity (inside and outside the home). It uses similar methods to those used in the successfully completed evaluation of a community exercise programme. 48

There will be two components to the costing. The interventions will be costed, as well as the consequences for the use of health and social services in general. The costs of the booster consultations will be assessed in a micro costing study. The costs will include enrolment of participants, training and time of facilitators, travel and telephone calls. Actual cost data will be collected for consumables and facilitator time will be costed using national grades. Despite being a highly pragmatic trial, there are some features of the programme which are specific to the research study and it will be necessary to adjust for these in order to make the results generalisable. Care will also be taken to compare costs assuming a routine level of throughput, rather than that achieved in the trial. Any research related costs will be excluded.

The consequences for use of health and social services will use resource data collected from participants. Use of primary, secondary, community and social services will be obtained using a self-completed resource questionnaire administered to participants at each assessment at baseline, three months and nine months. Resources will be costed using the best available national estimates. Where appropriate, national unit costs will be used. 49

SF-12v2 plus 4 data will be converted into health state utility values using the SF-6D preference-based algorithm. 50 The area under the curve between assessments will be used to provide an overall estimate of the QALY difference between the intervention arms and the control arm after adjusting for significant baseline variables. 51 Given cost and benefit data will only be collected for nine months, the on-going costs and health benefits will not be discounted, though start-up costs, including training costs, will be annuitised over a five year period. The sensitivity of the results to possible uncertainties in key parameters will be explored by a full sensitivity analysis, including a probabilistic sensitivity analysis.

Ground work and typical day

Using key skills in MI including Open ended questions, Affirmations of change talk, Reflective listening and Summarising (OARS) to elicit client attitudes towards adopting lifestyle change. This opening exchange will also ensure client centredness through empathetic listening and respectfulness. The main outcome is a greater understanding (for both the client and therapist) of the client’s current situation and their relevant health history and envisaging the impact of changing and not changing. TYPICAL DAY: Exploration (guiding and supporting the client) to understand a typical day and what ‘high risk’ scenario’s exist as well as opportunities for change and adapting current behaviours. Key skills used are again OARS.

Assess importance and readiness

A. How important is the change? (e.g. physical activity) (1–10) ( ) Why X Not Y?

B. How ready are you to change? (1–10) ( ) Why X Not Y?

C. How confident are you to maintain the change? (1–10) ( ) Why X Not Y?

This measure is mapped (conceptually) onto the transtheoretical model (TTM) of behaviour change,147 although, unlike most other applications, the MI interview integrates more than stages of change and incorporates ‘decisional balance’, ‘self-efficacy/temptation’ and ‘processes of change’. Many other interventions in health settings wrongly assume that stages of change and TTM are the same thing when stages of change is just one of four dimensions. This measure helps both parties understand the client readiness to change across his or her own selected health behaviour change aspects (i.e. smoking, diet, exercise, alcohol).

Assessing ambivalence

Whenever a client is considering change (from a position of contemplation, for example), ambivalence is a normal consequence. Eliciting all aspects of this position is important and can be done in MI using a four-cell decisional balance grid.

When we think about making changes, most of us don’t really consider all ‘sides’ in a complete way. Instead, we often do what we think we ‘should’ do, avoid doing things we don’t feel like doing, or just feel confused or overwhelmed and give up thinking about it at all.

Thinking through the pros and cons of both changing and not making a change is one way to help us make sure we have fully considered a possible change. This can help us to ‘hang on’ to our plan in times of stress or temptation. Below, write in the reasons that you can think of in each of the boxes.

Summary

Again, the key counselling skill applied here is reflective listening, in which the therapist draws together the main facets of the conversation and provides a reflective summary for the client. This enables the client to ‘listen back’ to the points they have raised themself and provides an opportunity to clarify points, adjust the inference or level of importance or merely have it verbalised from an objective source. This is an important phase in preparation for action planning and will include a recapitulation and asking key questions about subsequent stages.