Article Text

Cost-effectiveness of a community-based physical activity programme for adults (Be Active) in the UK: an economic analysis within a natural experiment
  1. Emma J Frew1,
  2. Mobeen Bhatti2,
  3. Khine Win3,
  4. Alice Sitch4,
  5. Anna Lyon5,
  6. Miranda Pallan5,
  7. Peymane Adab5
  1. 1Health Economics Unit, Public Health Building, University of Birmingham, Birmingham, UK
  2. 2NHS Heart of Birmingham Teaching Primary Care Trust, Birmingham, UK
  3. 3NHS Birmingham East & North Primary Care Trust, Birmingham, UK
  4. 4Department of Statistics, Public Health Building, University of Birmingham, Birmingham, UK
  5. 5Public Health Department, Public Health Building, University of Birmingham, Birmingham, UK
  1. Correspondence to Dr Emma J Frew, Health Economics Unit, Public Health Building, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK;e.frew{at}bham.ac.uk.

Abstract

Objective To determine the cost-effectiveness of a physical activity programme (Be Active) aimed at city-dwelling adults living in Birmingham, UK.

Methods Very little is known about the cost-effectiveness of public health programmes to improve city-wide physical activity rates. This paper presents a cost-effectiveness analysis that compares a physical activity intervention (Be Active) with no intervention (usual care) using an economic model to quantify the reduction in disease risk over a lifetime. Metabolic equivalent minutes achieved per week, quality-adjusted life years (QALYs) gained and healthcare costs were all included as the main outcome measures in the model. A cost-benefit analysis was also conducted using ‘willingness-to-pay’ as a measure of value.

Results Under base-case assumptions—that is, assuming that the benefits of increased physical activity are sustained over 5 years, participation in the Be Active programme increased quality-adjusted life expectancy by 0.06 years, at an expected discounted cost of £3552, and thus the cost-effectiveness of Be Active is £400 per QALY. When the start-up costs of the programme are removed from the economic model, the cost-effectiveness is further improved to £16 per QALY. The societal value placed on the Be Active programme was greater than the operation cost therefore the Be Active physical activity intervention results in a net benefit to society.

Conclusions Participation in Be Active appeared to be cost-effective and cost-beneficial. These results support the use of Be Active as part of a public health programme to improve physical activity levels within the Birmingham-wide population.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Physical inactivity is an important and largely avoidable cause of ill health, costing the National Health Service and the UK economy around £8.2 billion per year.1 There is strong and growing evidence that regular physical activity reduces the risk of many chronic diseases, including ischaemic heart disease, hypertension, non-insulin-dependent diabetes mellitus, osteoporosis, various cancers and depression. Regular physical activity also improves general health, enhances quality of life, relieves symptoms of anxiety, improves mood and raises self-esteem. The UK Department of Health recommends that adults should undertake a minimum of 30 minutes of moderate intensity physical activity on at least 5 days per week. However, 65% of men and 76% of women in England do not achieve this,2 and physical activity levels are particularly low in the West Midlands compared to the rest of the country.3

‘Be Active’ is a Birmingham-wide initiative developed in collaboration between the City Council and the NHS Primary Care Trusts, which allows Birmingham city residents to access their local Council-run leisure centres without charge at certain times of the day. During the period covered by this evaluation, registrants of the scheme had free access to fitness gyms (including induction sessions), swimming pools and group fitness classes during off-peak hours (until 17:00) on weekdays and limited hours (after 13:00) on weekends. The aim of Be Active is to increase levels of physical activity among Birmingham residents.

This article reports the results of an economic evaluation of Be Active in the city of Birmingham, UK. Cost-effectiveness is measured using quality-adjusted life years (QALYs) which is commonly referred to as cost-utility analysis. A cost-benefit analysis is also reported using the outcome of ‘willingness to pay (WTP)’.

Methods

Model development

We developed a Markov state-transition model that was adapted from a previously published model that assessed the cost-effectiveness of physical activity interventions in an American context.4 The analysis was conducted from both a healthcare and a wider perspective, and, in all cases, the intervention (Be Active) is compared with a reference scenario where no intervention is available. ‘No intervention’ in this context means that Birmingham city residents have access to the leisure centres at all times of the day but they have to pay to use the facilities. A Markov-type model was chosen as it allows us to incorporate the ‘risk’ of future disease associated with physical inactivity. For the cost-effectiveness analysis (CEA), the model simulated a closed cohort of the Birmingham adult population aged 16–70 years. It was assumed that at the beginning of the model, all members of the cohort were well and did not suffer from any chronic disease. Within the model, physical activity is classified into three health states using the metabolic equivalent (MET) as a unit of measurement that expresses the energy cost of physical activity. One MET is defined as the ‘resting metabolic rate’ and is equivalent to consuming 3.5 mm of oxygen per kilogram of body weight per minute. The three health states were defined as ‘inactive’ (<181 METs per week: equivalent to less than two 30 min of moderate activity per week); ‘active’ (181–449 METs per week: equivalent to two–four 30 min of moderate activity per week) and ‘recommended activity’ (>449 METs per week: equivalent to five or more 30 min of moderate activity per week). The most prevalent chronic diseases linked to physical inactivity are coronary heart disease, stroke, type II diabetes, breast and colorectal cancer, and this is incorporated within the model through the assignment of relative risk within each physical activity health state. At the end of the first cycle of the model, the cohort members are assigned an intervention-specific probability of improving their physical activity levels. With each subsequent year, they then have an annual probability of either remaining in the same physical activity health state or moving to a disease state. Mortality risk was estimated as an adjusted risk based on all-cause mortality combined with the age structure for the population of Birmingham city. Figure 1 shows the design of the model. In total there were three physical activity health states, five disease states and death. We assessed all costs in British pound sterling (GBP) and based them on the 2009/2010 price level or adjusted them accordingly using national indices.5 Costs and benefits are discounted at 3.5% per year.6

Figure 1

Illustration of model design.

Data sources

Physical activity

At the time of the evaluation, Be Active had been rolled out across the city of Birmingham and no comparator population was available, we therefore adopted a pragmatic approach to collect the data to populate the economic model. A ‘before and after’ approach was adopted whereby physical activity levels before registering with the scheme were regarded as a reasonable approximation of physical activity levels with no scheme in place. To gather this information, 19 Council-run leisure centres participating in the Be Active scheme across Birmingham were selected and all individuals who signed up to Be Active at the participating centres within an 8-week period were aged 16 or over (between 21 June and 13 August 2010) were invited to take part in the study. The leisure centres were chosen based on geographical and socioeconomic spread to ensure representativeness of the Birmingham population. Participants were asked to complete a questionnaire at baseline (point of registration) which obtained data on physical activity levels using the Godin Leisure Time Exercise Questionnaire.7 This questionnaire was then repeated at follow-up, 3–4 months after joining the scheme. Physical activity levels were then converted into MET units with the focus being on the change in levels between registration and follow-up. Web appendix 1 provides full details of the study protocol.

Quality of life

As well as gathering information on physical activity levels, health-related quality of life (HRQL) was assessed using the EuroQoL EQ-5D classification system, administered at baseline and follow-up.8 The EQ-5D instrument is a generic instrument for describing and valuing HRQL, the single-index HRQL scores are used to weight survival years and thus generate QALYs. For example, a patient who lives for 10 years and reports a detriment in HRQL of 0.5 on the EQ-5D scale would be attributed (10×0.5)=5 QALYs. HRQL for each of the disease states within the model were derived from a literature review. Full details are available in table 1.

Table 1

Parameters used in the economic model

Costs

The cost of the intervention was determined by dividing the total cost of setting up and running the scheme by the number of Be Active members, to obtain average per-participant annual costs. A detailed breakdown of the cost of the intervention is available in web appendix 2. This cost was then inserted into the model as a triangular distribution to account for the effects of changing levels of participation, for example, over time the number of people actually using the scheme will vary relative to the total number of registered members. We used a changing annual usage rate (ie, proportion of members who would be actively using the facilities over the course of a year) of 50–100%. The annual health service cost associated with each of the disease states were obtained from the literature and incorporated within the model. All of the key model parameters are detailed in table 1.

Disease risk

To account for the differential risk of developing chronic disease by physical activity level, a systematic literature review was conducted. The objective was to assess the impact of physical inactivity upon the future risk of developing coronary heart disease, ischaemic stroke, breast cancer, type II diabetes and colon cancer and to incorporate this risk in the form of annual transition probabilities (movement between the physical activity health states and the disease states) within the model. Web appendix 3 provides full details of the literature review and meta-analyses. Estimates of the disease risk associated with physical activity from this review are summarised in table 2.

Table 2

Estimates of the disease risk associated with physical activity

Cost-effectiveness analysis

The Markov model tracks all the disease cases, and changes in costs and QALYs to allow us to determine the difference in costs and QALYs for a situation with Be Active and a situation without Be Active. Results are expressed as an incremental cost-effectiveness ratio which is the difference in cost (cost (Be Active)−cost (no scheme)) divided by the difference in outcome (outcome (Be Active)−outcome (no scheme)). The base-case model was run for 5 years as it was felt that this time horizon reflected a time period that decision-makers would find useful because of the financial structures within Primary Care Trusts.

Sensitivity analysis

To assess the robustness of the results we carried out three sensitivity analyses or rather scenario analyses. First, we varied the ‘sustainability’ of the intervention by imposing a reduction in physical activity after the first year following the intervention. This is because it is plausible that adults will show an initial motivation to exercise regularly but that this will dissipate with time. We did this by incorporating another ‘disease’ path into the model to allow individuals to move from the ‘recommended activity’ health state into the ‘active’ health state (health state with lower physical activity levels). We assumed that half of the sample who had initially improved their physical activity levels in the first year (eg, moved from ‘active’ to ‘recommended activity’) had dropped back down into the active physical activity health state by the end of year 1. Second, we altered the time-horizon of the model to 10 years and then 2 years to assess the impact upon cost-effectiveness. Third, we removed the ‘start-up’ costs from the intervention.

Willingness-to-pay

We are not aware of any cost-benefit analysis using the contingent valuation methodology in the area of community-wide physical activity programmes. This approach lends itself well to an evaluation of this type as the outcome measure (WTP) allows the participants to include all potential benefits of Be Active (health and non-health related) and it provides a framework for users to reveal their preferences. Average values can then be weighed up against average costs to assess the net benefit of the scheme. A WTP question was incorporated into the baseline and follow-up questionnaire and thus a value from the participants was elicited from an ‘ex-ante’ and an ‘ex-post’ perspective (before and after using the Be Active scheme). The question was in the form of a payment scale format that asked participants to reveal what their maximum out-of-pocket WTP (monthly) would be to have access to the Be Active facilities. Participants were then provided with a vertically arranged list of values that ranged from the top, £0, in increments of £10, to £100 as the maximum amount. An open-ended response was provided for participants who had a monthly WTP greater than £100. The payment scale was the preferred choice of the elicitation method because it has been found to produce the highest response rate (compared to the open-ended design)30 and it does not require an a priori distribution of WTP values (as in the closed-ended design).31 The values of £0 to £100 were chosen as the lower and upper end of reasonable WTP values for a monthly leisure club membership. ‘Monthly’ was chosen as the time frame as it was felt that this reflected a ‘real-life’ scenario of paying monthly premiums for a private leisure club.

Results

Impact upon physical activity

Between June 2010 and August 2010, 797 new adults who registered with Be Active completed the baseline and follow-up physical activity questions. Physical activity levels improved during this time period with an increase of 15% in the ‘recommended activity’ group and a reduction of 9% in the ‘inactive’ group. All physical activity levels are shown in table 3.

Table 3

Physical activity groups

Cost-effectiveness

The total costs, total effectiveness and incremental cost-effectiveness are presented for the Be Active scheme in place versus no scheme. Results are cumulative over a 5-year time frame for the whole Birmingham population, aged 16–70 years, but average per-person values are reported here. Table 4 shows the results for the cost-effectiveness for the base-case analyses and the sensitivity analyses where we explored the impact of the sustainability of the affect, the time horizon (2 and 10 years) and the inclusion of start-up costs. It appears that as we increase the time horizon of the analysis, the cost-effectiveness of Be Active improves as we realise the cost-savings from preventing future disease associated with physical inactivity.

Table 4

Incremental cost-effectiveness ratios in the base-case analysis and sensitivity analysis

Net benefit

Only participants who supplied a WTP value at baseline and follow-up are included in the analysis (n=996). Maximum ex-ante WTP averaged £12 and ex-post WTP averaged £20 (table 5).

Table 5

Descriptive statistics for WTP

This indicates that once individuals had the chance to experience Be Active, the value placed on it (as reflected in the WTP value) has almost doubled. Taking the most conservative assumptions possible, that only 50% of registered Be Active members actively use the facilities, and including the start-up costs of Be Active the per-participant annual Be Active cost is equal to £75. The average maximum ex-ante and ex-post WTP value exceed this amount (ex-ante: £12×12 months=£144 and ex-post: £20×12 months=£240) therefore by applying the decision rules of a cost-benefit analysis, Be Active is showing to produce a positive net benefit.

Discussion

Our analysis indicates that the Be Active programme, with an annual cost of around £4.4 million is likely to be cost-effective. As the duration of the physical activity effects increases (from base-case analysis of 5–10 years), the Be Active programme shows strong dominance and therefore is more effective and less costly than no Be Active programme in place. Several parameters and model assumptions had a strong influence on costs and health gains from Be Active, but even with the most conservative assumptions (inclusion of start-up costs, reduction of physical activity effects and a small timeframe for analysis), Be Active appears to be cost-effective.

Strengths and weaknesses

Be Active is a novel initiative that was piloted through multiagency partnership, with ambitious aims. A pragmatic approach to evaluating a service that already exists within a community was undertaken. One limitation of the study was that the collection of physical data was based on self-report. The assessment of physical activity was by using the modified Godin Leisure Time Exercise Questionnaire, which has been validated against objective measures and shown to provide a reliable and fairly valid estimate of physical activity categories.7 Owing to time constraints, the follow-up period was reduced from 6 to 3 months. Despite this limitation, it is generally thought that attendance at leisure centres reaches a set pattern by 3 months, so greater duration of follow-up would have been unlikely to affect levels of physical activity. We have tested for this effect in the model (by imposing a percentage reduction in physical activity levels after the first year). The model, however, does rest on the underlying assumption that improved physical activity levels will be sustained to realise the long-term health effects.

The conventional application of the QALY outcome measure is usually within the context of assessing the cost-effectiveness of a healthcare treatment that moves a patient from a ‘sick’ to a ‘well’ state. Thus, the starting point of the analysis is a person whose HRQL is impaired. It is debatable whether HRQL is directly affected from being physically inactive (in the short-term at least) and therefore the capacity to benefit from an intervention like Be Active (in terms of conventional HRQL) is limited. Nevertheless, Be Active improved physical activity levels from baseline to follow-up, and the related improvement in terms of HRQL was sufficient to offset the costs of Be Active.

In recognising the potential methodological limitations of the QALY in this setting, we applied a novel WTP approach. WTP has been shown to be sensitive to the choice of an elicitation format with each type of format having its own set of advantages and disadvantages. The payment scale is prone to range bias whereby average WTP values are unduly influenced by the range of values within the scale,32 it was felt, however, that this potential limitation was justified when balanced against the limitations of the alternative elicitation methods. By using WTP, participants had the opportunity to include non-health (and health) related benefits when valuing the programme. Albeit a more exploratory part of this analysis, the method revealed that there is a positive net-societal benefit from having Be Active in place.

‘Natural experiments’ are defined as ‘events, interventions or policies that are not under the control of researchers, but which are amenable to research which uses the variation in exposure that they generate to analyse their impact’.33 We have used a ‘natural experiment’ framework to evaluate Be Active. There is growing support for the use of natural experiments with the 2004 Wanless report,34 the 2007 Foresight report on obesity35 and the Department of Health in its obesity research and surveillance plan, Healthy weight, healthy lives36 all calling for more support for experiments of this type. Recent Medical Research Council guidelines on the use of natural experiments to evaluate population health interventions, state that these experiments are useful when ‘… there is scientific uncertainty about the size or nature of the effects of the intervention but for practical, political or ethical reasons the intervention cannot be introduced as a true experiment …’.33 It was not possible to deliver Be Active as a true experiment because access to leisure centres were not restricted to geographical areas of residence (within the city council boundary) and therefore population groups could not be randomised into clearly defined ‘exposed’ and ‘unexposed’ groups. We adopted a pragmatic study design and estimated impact using a validated questionnaire for physical activity levels and conventional economic tools for the cost-effectiveness. The impact of a public health intervention like Be Active is not short term; this paper is an example of how economic models can be used to model longer-term impacts such as the reduced disease risk associated with being more physically active. It also shows how economic methods such as cost-effectiveness and cost-benefit analysis can be used to model the impact of a public health intervention on population behaviour. By using the conventional QALY measure to capture the outcome of Be Active, this places the intervention within a context where direct comparisons can be drawn between Be Active and other healthcare interventions. It is feasible therefore that on the basis of these results that healthcare decision makers may choose to fund Be Active instead of other competing uses of the healthcare budget.

Using a before and after natural experiment framework creates challenges as with all quasi-experimental designs, natural experiments will never unequivocally determine causation due to the lack of control over the experiment. Individuals may also self-select in ways that are not observed but are correlated with the outcome. Subgroup analysis is limited because of the lack of power in the sample to perform such tests.

Comparison to other studies

To our knowledge, no other published studies of community-based initiatives involving free access to leisure centres, fitness centres or gyms have been published. We identified three UK-based community programmes aimed at increasing physical activity levels37–39 but no published economic evaluations alongside.

Implications and future research

Participation in Be Active resulted in increased physical activity levels. Although we have included the estimated impact of increased physical activity levels on risk of future disease, some of the literature available to estimate these risks was limited. The literature used non-standard classification of physical activity levels and there appeared to be some evidence of heterogeneity across studies, suggesting perhaps that the estimates should not be pooled. Since physical activity will impact upon future risk of disease, future modelling and surveillance studies should deal with these aspects. Furthermore, our model assumes that the Be Active programme will continue unchanged, which is likely in the short term, but political and budgetary factors may vary the design of the programme. If Be Active does increase physical activity levels to a level that offsets the costs, as our model predicts, then it may be possible to extend the coverage of Be Active to perhaps longer time-periods or to cover more leisure centre facilities. Models are a means of consolidating multiple sets of information into one framework, but the conclusions drawn from a model are only as good as the quality of data used to populate it. High-quality surveillance is required to understand the effects of physical activity upon future disease risk and to measure the true-effect of a community-based physical activity programme like Be Active. Models can then be refined as the quality of data improves.

What is already known on this topic

  • There is strong evidence that regular physical activity can reduce the risk of many chronic diseases. A large proportion of adults in the UK do not achieve the Department of Health recommended levels of physical activity and rates are particularly low in the West Midlands.

  • While public health initiatives to promote physical activity have been advocated, the availability of economic evidence to support their use is sparse.

What this study adds

  • We believe this to be the first study to estimate the cost-effectiveness of a public health initiative promoted through Council-run leisure centres to improve population physical activity levels.

  • A physical activity intervention such as Be Active appears to offer good value for money, even with the use of conservative assumptions.

  • Traditional economic tools can be used to evaluate complex community-wide interventions and can help to identify gaps in information that should be a priority for future research.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors EF was responsible for the general coordination of the study and the design and analyses of the economic model. MB and KW were responsible for the systematic review of the literature on physical activity and disease risk. AS was responsible for statistical support. AL, MP and PA were responsible for the analyses of the physical activity data. All authors designed the study, helped to write the manuscript, and read and approved the final version of the manuscript.

  • Funding EF, MP, AL and PA had financial support in the form of a consultancy grant from NHS South Birmingham PCT paid to the University of Birmingham for the submitted work. The authors were independent of the funders and the funders had no role in the economic analyses.

  • Provenance and peer review Not commissioned; externally peer reviewed.