Estimating cost-effectiveness of diet and physical activity interventions for the primary prevention of non-communicable disease at the Local Authorit

Preventable disease is responsible for approximately 40% of the UK's disease burden, largely in the form of non-communicable diseases such as heart disease and stroke. Causes of these diseases, such as obesity, can be prevented, providing an opportunity to avert disability and deaths. It is useful to know how much disease is likely to be prevented for each pound of spending. To work out this value for money, complicated sums called models are used. At present, Local Authorities have no models that account for the characteristics of the local area, such as people's ages, weights or exercise levels. This acts as a translational gap, preventing the consistent application of the best interventions. This DPhil will develop a model for chronic disease prevention that can compare value for money across Local Authorities for approaches to improving diet and physical activity. This will not only help local decision-makers with funding decisions but also allow the local impacts of national public health policy to be estimated.

Modelling works by estimating the effect of each step that leads from an intervention to a disease outcome. At present, models produce an estimate for England as a whole, without being able to estimate how this might vary for different areas. For example, if a local council spends money on gym memberships for diabetic patients, modelling could estimate the increase in activity those individuals do in the real world, the weight they lose and improvement in diabetes control; all based on published clinical research and standard health economic methods. More affluent areas tend to have better responses to health promotion, as well as very different patterns of diseases and risks (eg, having lower obesity rates), so an approach that is effective and good value for money in one area may not be in another. Local area estimates will also allow impacts on health inequalities to be estimated.

The modelling method is called proportionate multistate lifetable modelling, which simulates a population moving through time, aging and developing disease. From there, NHS costs can be estimated. By doing this twice in parallel, the difference between a baseline and an intervention can be estimated. To make the modelling process flexible for local outputs, two additional steps will be added. Firstly, the input risk factor profiles need to reflect local areas, and secondly the results will be standardised to local area age and sex composition. Local risk factor profiles are estimated as locally-collected data are poor. This process uses regression modelling to estimate an area's risk factor profile based on demographics and national risk factor data. Standardisation will be performed based on census data.

Two scenarios will be modelled. The first will estimate local implications of the calorie reduction plan from Public Health England's report Childhood Obesity: A Plan for Action, Chapter 2 (2018), which has major spillover health implications for adults. The second will estimate the local impacts of recent salt reduction in the UK. This second scenario will compare progress under the scrapped industry Responsibility Deal (2011) against the trend under the previous Food Standards Agency's Salt Reduction Strategy from its initiation in 2000 to 2011. Validation and sensitivity analyses will follow. Validation will be performed against international examples of chronic disease scenario models. Deterministic sensitivity analyses will be based on the uncertainty intervals produced in the risk factor profile estimates and probabilistic sensitivity analyses will be based uncertainty around the Relative Risks through Monte Carlo Analysis.

This proposal involves MRC Strategic Skill Priority areas in quantitative skills (applied statistics, computer modelling) and interdisciplinary skills (epidemiology, clinical application, health economics, policy analysis).