Teams evaluating the effectiveness and cost-effectiveness of a set of interventions could identify which theories were applied (if any) when interventions were developed, if these are reported. The interventions and their influences could then be added to a behavioural systems map (see Method 4). This information can be presented to and discussed with policy makers and other relevant stakeholders to help select interventions for evaluation. Behavioural theory could be qualitatively utilised to inform long term assumptions within both cohort and individual level models. Where intervention studies present theory of change diagrams (identifying an intervention’s impact, outcomes, outputs, activities, and inputs and describing why interventions will create which outcomes) these could be used to inform long term assumptions.
Most existing theories and intervention studies focus upon behaviour change rather than behaviour maintenance (i.e. sustained behaviour over more than 6 months) (Kwasnicka et al., 2016). The influences on behaviour change should not be assumed to be the same as those affecting behaviour maintenance. For the prevention of many non-communicable diseases, the goal is to maintain healthier behaviour over the longer term, to impact long term health outcomes. Different interventions are likely to work in different ways, and in different populations, and hence some interventions may be more likely to assist behavioural maintenance rather than others. Most studies report only the effectiveness of the interventions over 6 or 12 months. There are a small number of intervention studies which are designed to capture behaviour maintenance outcomes, as well as some meta-analyses of these studies, mainly for physical activity maintenance (Howlett et al., 2019; Madigan et al., 2021; Van der Heijden et al., 2018; Flore et al., 2022; Murray et al., 2017). A literature search for relevant behaviour maintenance studies could be undertaken. It may be possible to apply these studies to inform model predictions; however, outcomes are still relatively short term (generally, a maximum of three years of data), and only average effects are reported. These data could be used directly, or to inform elicitation, or as a calibration target. Information specialist input would be required to design these literature searches.