Comparison with other studies
This study can be placed within the currently limited economic evaluation literature on population-level promotion of PA. In particular, the economic evaluation conducted to assess the cost-effectiveness of the BeActive programme11. This represents the main comparison study. LLGA mirrored the BeActive intervention modality, except that LLGA was offered only in City Council leisure centres located in the most deprived areas of the city. This afforded an opportunity to test the cost-effectiveness of providing universal access to free off-peak leisure centre-based sessions in another similar setting. For BeActive, base-case cost-effectiveness estimates were not dissimilar from those reported here, with an estimated £400 incremental cost per QALY gained. This finding supports the hypothesis that this type of population-level intervention represents good value for money also in the short term, and even when the offer is proportionate to attract hard to reach groups. By contrast with this study, BeActive appeared to be cost-effective even under the most conservative assumptions, though no further details were reported. Another comparable study simulated the implementation of a primary care-based universal intervention and found a 64.7% probability of the intervention being cost-effectiveness at a WTP threshold of £30,000.14 One possible explanation for this difference in results is that, in that study, utility gains were accumulated only as a function of reduction in disease incidence and no utility gains were assigned from transitions to higher PA levels. Nevertheless, although some of the economic evaluation methods used in the present analysis were aligned with those studies (e.g. perspective, short time horizon), differences in the structures and parameters of the economic models limited the ability to directly compare our findings.
Strengths and limitations
To the best of our knowledge, this represents the first cost-utility analysis of a proportionate universal programme to promote free off-peak leisure centre-based exercise in the general population. The programme is relatively easy to incorporate into currently operating public leisure centres (off-peak sessions), and therefore this intervention has the potential to be replicated in other comparable settings (i.e. local City Councils in the UK). As a result, this makes the evidence generated by this analysis particularly important for decision-makers that may be interested in evaluating the impact of implementing this type of intervention in the future.
The study is however subject to a number of limitations. In particular were the lack of experimental design, a non-research led data collection and handling process and restrictions imposed in terms of further data collection on residents/participants. This meant making the validity of effectiveness results depend on the plausibility of a parallel trend assumption, representativeness of the sample of participants providing full outcome data, as well as on untested measures of PA behaviour change which in turn relied on self-report. Previous similar studies share these limitations that cannot be overcome retrospectively and are likely to characterise large-scale programmes. Furthermore, while a sub-group analysis was conducted to account for heterogeneous effects, one of the objective of public health decision-makers is to reduce existing health inequality, which, due to resource constraints, was not possible to ascertain within this study.
Application of the QALY as the consequence considered in the evaluation restricted the evaluative space accordingly, therefore excluding non-health effects potentially generated by the intervention (e.g. increased work-related productivity20). However, in line with previous models11 15 21, the decision-analytic model used for economic evaluation of LLGA was designed to accumulate utility gains/losses as a result of changes in PA state.
A de novo decision-analytic model was developed building on previous models, by incorporating a continuous-time structure which allowed for testing the assumption related to the sustainability of behaviour change over time. Nonetheless, this analysis still relied upon other structural assumptions relating to a fully elastic dose-response relationship between changes in PA and health, compensatory or synergistic effects potentially occurred on the path to health improvement (e.g. changes in dietary habits), increased health expenditure from extended life expectancy, and adverse events (e.g. injuries) which were not formally taken into account. Nevertheless, unlike previous models, negative intervention effects were captured informally by allowing the four PA states to move freely between one another.
Further, these results, like those presented in previous similar studies, rely on a set of structural assumptions which have not been verified yet and have the potential to impact identification of the optimal intervention. In particular, although the decision-analytic model used for this economic evaluation allows for “natural” transitions between PA states to be captured, due to lack of relevant data, PA states were assumed to be stable over time in absence of the intervention. However, this may not always be necessarily the case, especially in the short term 22 and during sensitive life phases (e.g. retirement 23). Furthermore, since the effects of changes in PA on chronic disease are likely to vary between conditions and depend on personal characteristics, as well as on their magnitude/persistence, population-level monitoring studies should deal with these aspects.
In addition, the impact of an intervention like LLGA is likely to vary not only between individuals and over time, but also on whose economic perspective is taken. In this and previous studies11 14, costs and benefits (QALYs) falling on the health care sector only were considered. However, results are likely to change when a local public health agency viewpoint is taken. As the body administering and hosting the intervention, the opportunity cost by the Local Authority may not coincide with the budget expenditure. Potential spill-overs from increased numbers of paying members or reductions in member retention due to the intervention might have occurred.