Improving the validity and usability of decision models: case studies with a focus on physical activity

Background: Health economic evaluation has a crucial role to play in the allocation of scarce societal resources. Economic models used in these evaluations must have a high degree of external validity but must also be usable in order to effectively inform policy. However, there is sometimes a trade-off between the realism of models (external validity) and the ease with which stakeholders can understand and interact with them (usability). This trade-off is particularly relevant in the field of physical activity where modelling is complicated and data availability is limited. The aim of this thesis is to investigate the balance between the external-validity and usability of models used in health economic evaluations of physical activity interventions and develop ways to build models that are more externally valid and usable. Methods: The study begins by identifying limitations in the external-validity and usability of published physical activity models, with a particular focus on models used to inform National Institute for Health and Care Excellence (NICE) guidance. Three case studies of adaptations to improve external validity are provided, with a discussion of their implications for usability. Additionally, ways to improve the usability of models are examined, with methods proposed to make models more accessible, transparent, secure, and efficient to construct and maintain. Results: The results of this thesis demonstrate that models can be improved in terms of both external validity and/or usability. The case studies provided show that methodological developments to physical activity models are feasible given new modelling methods and advancements in computing power, but despite improving external validity may reduce usability. Additionally, this thesis outlines methods by which health economic models can be made more accessible, transparent, secure, and efficient to construct and maintain, thereby improving their usability. Discussion: The overall conclusion of this thesis is that economic evaluation models should be as externally valid and usable as possible. However, a trade-off sometimes exists between the two. With a fixed budget for evaluation, attempts to improve external validity can have an opportunity cost in terms of resources allocated to making models easy to use and understand. The incorporation of methods from computing and data science can help mitigate this trade-off.