The effectiveness of public health interventions is dependent upon human behaviour. Behaviour is complex and multifaceted, and shaped by many influences which change over time. In some health economic models of public health interventions, only biological risk factors for disease, such as BMI, are incorporated (Breeze et al., 2017), without including the contributing behaviours such as eating and physical activity. In others, behavioural risk factors (e.g., smoking) are included, but the influences upon these behaviours, such as those related to capability (e.g. knowledge/behavioural regulation), opportunity (e.g. environmental context, social influences) and motivation (e.g. beliefs, emotion, reinforcement), are not explicitly considered (Virtanen et al., 2017). The impact of the intervention is generally estimated based on a single study or a meta-analysis of the effectiveness of an intervention (NICE, 2023) and effectiveness evidence is often limited to 6 – 12 months follow up (Willinger et al., 2021; Hubbard et al al., 2016; Skinner et al., 2020). There is a dearth of evidence about behavioural maintenance resulting from interventions and there are no standard approaches for estimating the impacts of the intervention beyond the end of study follow up. Assumptions range from maintaining the effectiveness over an individual’s lifetime, to reverting to the outcomes of the comparator either immediately or gradually over some time period (Bates et al., 2020; Candio et al, 2022). These assumptions are usually based on little theory or evidence, and generally do not vary by individual characteristics or intervention type.
Figure 1 shows some potential alternative modelling assumptions beyond study follow up for a hypothetical intervention which reduces BMI (either by increasing physical activity or improving diet). The cost-effectiveness results are based on the average differences between usual care and the interventions, which may be dramatically different depending on which assumptions are chosen (Bates et al., 2020). For simplicity, this hypothetical example shows only usual care and one intervention, with three alternative modelling assumptions for the cohort. However, there are often multiple interventions to compare, and it may not be appropriate to make the same extrapolation assumptions for each intervention or each individual, given that there are many factors that will affect behaviours over time.
Figure 1: Illustration of the importance of assumptions beyond study follow up
During model development, assumptions about the effectiveness of the interventions are often treated as an ‘add on’ to the main modelling of the current system. This may be because health economic modelling traditions were developed for the evaluation of clinical interventions, where effectiveness evidence tends to come from randomised controlled trials and the reasons why an intervention works or does not work will not affect policy decisions. Yet, for public health economic evaluation, the benefits of developing a more detailed individual-level simulation model are likely thwarted by the basic assumptions about the intervention effectiveness over the long-term. It is not advisable to attempt to predict future outcomes without any understanding of the mechanism of action – i.e., the processes through which an intervention affects behaviour, such as by increasing motivation (Carey et al., 2019). This is particularly important when policy makers are comparing the cost-effectiveness of several alternative interventions which are made up of different behaviour change techniques, some of which are more likely to lead to maintenance of a new behaviour than others (Howlett et al., 2019).
In addition, interventions may affect different behaviours (e.g., eating or physical activity) or different elements of a behaviour domain (e.g., eating fruit and vegetables or salt intake). Intervention strategies may include taxation policies, environmental changes, service provision, or education. Each of these intervention strategies has different mechanisms of action, with some depending more upon changing individual factors, and others on changing the external environment. If the same extrapolation assumptions are made for all interventions this could lead to inappropriate comparisons between the interventions. Moreover, different individuals may be more likely to maintain a behaviour than others, according to some environmental influences, biological factors or psychological attributes, and subsequent inequalities are likely to be important to decision makers. Thus, the underlying evidence, the choice of assumptions, and appropriate representation of uncertainty related to these aspects, is fundamental for informing policy decisions.
Research undertaken within other disciplines including psychology, sociology, complexity science and behavioural economics can help to inform assumptions about behaviour beyond intervention study follow up. Many theories and frameworks have been developed to explain human behaviour. These link a set of biological, psychological, social and/or environmental factors to behaviour, offering a bio-psycho-social understanding of behaviour. Such factors can be thought of as potential intervention targets for change and mechanisms of action, such that the effectiveness of interventions for changing and maintaining behaviour will depend on which influences are targeted, and to what extent changes in these influences (e.g., knowledge), and the behaviour change techniques (e.g., providing information) and intervention strategies (e.g., education) used impact on behaviour. However, there is no one accepted behavioural theory, and there are multiple mechanisms of action and behaviour change techniques (Johnston et al., 2021). Michie (Michie et al., 2014) collated 83 behavioural theories which could inform the development of behavioural interventions. Within this review, the three theories for which the most published papers (more than 50%) were identified were: (i) the Transtheoretical Model of Behaviour Change (which includes progression and feedback loops through the stages precontemplation, contemplation, preparation, action and maintenance, allowing for relapse) (Prochaska et al., 1997); (ii) The Theory of Planned Behaviour (which links attitude, subjective norms, perceived behavioural control and intentions to behaviour) (Ajzen, 1991); and (iii) Social Cognitive Theory (which links the interaction between the individual, environment and behaviour with behavioural capability, observational learning, reinforcement, expectations and self-efficacy) (Bandura, 2001).
Recent literature encourages the use of behavioural theories to inform public health intervention development to understand what works for whom in which contexts (Michie et al., 2018). However, many existing studies do not report their theoretical basis (Prestwich et al, 2014) and a narrative synthesis of nine systematic reviews found no difference in the effectiveness of interventions that were theory-based versus non-theory-based (Dalgetty et al., 2019). The study authors suggested that this could be due to limitations with the theories used, some of which have had calls to be retired (Sniehotta et al., 2014; West, 2005) or issues with fidelity and the way in which they were applied.
The Behaviour Change Wheel developed by Michie et al. (Michie et al., 2011) has become an important framework for developing interventions, because it provides a coherent and comprehensive approach. It was developed based upon a systematic review of the literature and subsequently tested for reliability. The Behaviour Change Wheel includes comprehensive sources of behaviour at the hub, with the full range of interventions and policies set out around the wheel. There are then tables to facilitate a systematic selection of interventions and policies according to the behaviour change components. The sources of behaviour at the hub consist of the COM-B model of behaviour, where Behaviour can be explained by a combination of Capability, Opportunity and Motivation. Capability includes physical (e.g., skills/ strength) and psychological (e.g., knowledge, memory, attention and decision making, behavioural regulation) factors; Motivation includes reflective (e.g. beliefs, intentions, identity) and automatic (e.g. emotion, reinforcement) factors; and Opportunity includes physical (e.g. environmental context and resources) and social (e.g. norms, culture) influences.
Any theory that is used to develop the interventions is generally ignored when they are evaluated within a health economic model. This is partly because health economic modellers tend to use methods developed for the evaluation of clinical interventions, and partly due to limitations of existing behavioural theories, including data limitations to support them and the fidelity with which they are implemented in the interventions. To predict the long-term effectiveness of interventions, it is important to understand the precise content and context of the interventions (Skivington et al., 2021). Combining behavioural theory with modelling, as has been done in other fields such as natural resource management (Schluter et al., 2017), could help to understand the longer-term impacts of a range of interventions with different mechanisms of action upon individuals with different attributes.