Data driven dynamic systems modelling in public health

This PhD studentship is offered as part of the SIPHER (Systems science in Public Health and Health Economics Research) consortium. Preventing ill health related to the “social determinants of health” requires well-coordinated policies across many sectors. SIPHER is a major investment by the UK Prevention Research Partnership, which brings together scientists across six universities, three government partners at local, regional and national level, and ten practice partner organisations. SIPHER will deliver novel evidence of the costs and benefits of the complex, interlinked and long-term consequences of policy decisions.

For further details of SIPHER see https://sipher.ac.uk/. The SIPHER project is focused on using a systems science approach to advance the healthy public policy agenda towards achieving reductions in non-communicable diseases and health inequalities. A key driver for this approach is the recognition that policies operate in complex adaptive systems with temporal dynamics and complex inter-dependencies amongst socio-economic factors such as economy, welfare, housing, education and employment with health outcomes. The systems view is captured at multiple levels with systems maps representing the causal dependencies amongst the policies, factors and health outcomes, maro and micro models capturing their temporal evolution and spatial variability to policy interventions which are then aggregated to a multicriteria decision-support system that provides evidential information to policy makers. The focus of this PhD project is on the development of system dynamic models that capture the temporal evolution of the factors and the resulting health outcomes at the population and subgroups of population levels. These models will be evidential and so are data-driven with the information extracted from multiple data sources. The model structures will also be informed by the systems maps that represent the causal dependencies amongst the factors and health outcomes. The models must be capable of not only representing the temporal evolution of the variables of interest, but is also expected to quantify the degree of confidence in any predictions made from the models which are important for the decision-making stage. The model building will be based on the state-of-the-art in system identification, data analytics, time-series and Bayesian dynamic modelling methods. They will include both linearand nonlinear dynamic models fitted with parameter estimation methods such as least squares, maximum likelihood and Bayesian methods. A key challenge, and an opportunity for novel developments in model estimation, is to account for the variability in the quality of the data, both temporally and spatially. The model development can be undertaken in a suitable software platform such as Matlab, R or Python. There is potential for the developed models to be adopted by the SIPHER project policy partners and hence for generating impact from this research.

For an informal conversation about this opportunity, please contact Professor Visakan Kadirkamanathan ([[Email Address Removed]]).

Applications will close on 25 September 2021 and interviews are expected to be held week commencing the 18 October 2021.