Research units A-Z

Policies can affect the health of the public in ways that are not straightforward to understand.

To give an example, an employment policy to reduce underemployment may have beneficial health effects if it generates higher incomes (e.g. increasing people’s ability to afford healthy food and good quality housing) and productivity (e.g. generating more government money through income tax that is available to spend on healthcare).  

It may also have detrimental effects, for example increasing stress, limiting time for healthy activities and social relationships, or by making it more difficult for those with existing health problems or caring responsibilities to get employment. If the negative health effects outweigh the positive effects, then a less healthy population may be able to work less, or less effectively, thereby leading to the opposite of the intended policy outcome.

Often there are multiple pathways linking a policy with health and these pathways are not independent from one another (e.g. higher incomes allow people to balance out some of the negatives, such as enjoying holidays in their leisure time).

To add to the complexity, the various influences can work to different time scales, with some effects being immediate and other effects only occurring after a delay (e.g. no longer being underemployed may increase your income immediately but health problems may take a while to show and then get worse as time goes on).

To know the full influence of a policy on health, we need to understand how the effects along the different pathways combine. The “system” in systems dynamic modelling is a set of equations capturing of all the different outcomes and pathways we are interested in, with the numbers in the equations representing the direction of an effect, the strength of an effect, and its timing.

Workstrand 4 is developing a systems dynamics modelling methodology to estimate the health, social and economic impacts of different policies, taking into account the effects of multiple, dependent pathways, and how effects accumulate over time.

An important aspect in modelling is the question of “causality” – that means that one thing causes another rather than things just cooccurring by chance or as a result of something else. In the above example we would need to find out whether it is really the underemployment change that is responsible for health changes, and to what degree health changes also cause a reduction in underemployment. Analysis of causal effects and their directions is part of our systems dynamics modelling methodology.

Workstrand 4 has four major objectives:

• To develop the set of equations that describe the causal pathways for each topic of interest and for each policy partner.

• To implement the equations as a computer model in a way that can be run by Workstrand 7 to simulate the impacts of different policies. The computer models can be run to estimate what happens over time to all the factors in the system when a hypothetical event happens (such as a policy implementation or an external event like an economic recession).  Since some of these events will actually happen and Workstrand 3 is able to capture evidence of the effects, we can test how well the model was able to correctly estimate what happened. We are using this learning to progressively improve the models, working closely with policy partners.

• To collaborate with Workstrand 2 and 3, together with topic experts, to identify evidence for the numbers in those equations (and account for the uncertainty about what the ‘real’ numbers should be).  Where there is lots of evidence, we are able to generate detailed equations that capture the flow of cause and effect over time in detail; where there is not much evidence, we generate much simpler equations.  We are developing methods that allow models of different ‘granularities’ (i.e. different levels of detail) to be used together. This is important because the evidence relating to the different pathways is patchy, so different pathways will have equations at different levels of detail, and we cannot combine all the pathways together unless we can handle those different granularities at the same time.

• To develop insights and recommendations for how to do this kind of whole-systems modelling outside of the original SIPHER topics and/or policy partners. So far, the effects we represent in the models have been for the whole population. Later, in order to account for inequalities between different groups of people, we will also build models that capture subgroup effects. These subgroups may be in the form of geographical regions, income groups, men and women, or people of different ages for example. Here, we will be able to check our Workstrand 4 model predictions against the micro modelling predictions (Workstrand 5), which use individual-level representations of the population. These predictions of effects at the individual-level can be added to the same groups so that we can compare what the two models say about how young women would likely be affected by a policy. These modelling methods are very different, so if they arrive at similar conclusions, this can give our policy partners greater confidence in the usefulness of the predictions for policy design.

Workstrand 4 is building working computer models for each topic area and for each policy partner that can be refined over time to give the best possible estimates of policy impact, over time, across the system.

Learning from this work will inform best practice guidance for how to do whole-systems public policy modelling across local, regional, and national government.