The SIPHER Consortium: Introducing the new UK hub for systems science in public health and health economic research

Framing healthy public policy as a complex systems problem

Non-communicable diseases (NCDs) account for 89% of UK premature deaths at an estimated societal cost of at least £100bn a year¹. Health policy often tries to prevent NCDs and reduce health inequalities by tackling health behaviours (e.g. alcohol and tobacco consumption, exercise, diet). Yet social and structural determinants of health (SDoH) account for a far greater burden of NCD and contribute more to health inequalities than individual behaviours². Thus, the greatest prevention opportunities lie outside the health sector’s control but within reach of sectors such as welfare, housing, and employment³. Three main challenges hinder efforts to realise these opportunities. First, non-health policymakers often lack incentives, resources, expertise and evidence to assess their policies’ health impacts. Second, policy costs and benefits are spread across sectors, making it harder to identify synergies, negotiate trade-offs and mobilise cross-sector political will. Third, each policy area is an ‘open’ and dynamic system, subject to external forces (e.g. the global economy) that can shift rapidly, causing policies to become untenable, ineffective or to have unintended consequences.

The move towards healthy public policy (sometimes known as ‘Health in All Policies’ or ‘Wellbeing in All Policies’) is a key strategic response to the realisation that policies in sectors where health is not the primary concern may be among the most potent drivers of health improvements^4–6. However, the current evidence base on (cost-) effective SDoH-policies is not well-aligned with public health decision-making processes as it lacks robust quantitative evaluations and appraisals of the likely health effects of available policy options and their prioritisation^7,8. This is compounded by healthy public policy being a complex systems problem that involves: (a) many interdependent causal mechanisms, where effect directions, sizes and timings are not well-understood or captured by dominant research methods; (b) actors with diverse values for whom health and wellbeing may just one of many priorities or no priority at all; and (c) policy design and implementation under conditions of deep uncertainty due to evolving geo-political and economic contexts⁹.

Complex systems modelling in policy analysis

Complex systems modelling (CSM) may present a useful way forward. CSM is highly effective for understanding and forecasting the response of physical, natural and social systems to policy change and providing evidence for decision-making under conditions of uncertainty^9,10. CSM can capture hypothesised causal relationships between a multiplicity of diverse factors, using simulation techniques to examine the system-wide consequences of changes made in one or more areas within the system¹¹. CSM offers opportunities for ‘what-if’ simulations, which help identify policy options that are desirable (e.g. cost-effective, high-leverage, acceptable to the public and politicians) and robust (i.e. performing well across a wide range of possible, but uncertain, futures)^9,12. SIPHER will use two different methods: policy-focused microsimulations (individual-level models), which are widely used in parts of UK central government, e.g. the Department for Work and Pensions^13,14 and system dynamics (population-level) models, which have also been used successfully in policy contexts (e.g. housing and energy¹⁵). In public health, CSM approaches have had notable impact on infectious disease policy such as for poliomyelitis¹⁶ and measles¹⁷. Yet, despite a substantial body of literature that describes the systems affecting NCDs and highlights the policy potential of CSM¹⁸, examples of CSM influences on NCD policy remain rare^19,20. Exceptions are an agent-based model that shaped tobacco regulation²¹ and a co-produced system dynamics model that influenced county-level cardiovascular disease policies²². A recent review on CSM in mental health research concludes that participatory or co-production methods which engage with policymakers are beneficial for communication, buy-in and identifying leverage points in the system²³.

Policy partnerships, policy prioritisation and focal outcomes

SIPHER was co-developed with three highly-committed policy partners who represent local, regional and national scales of government:

Exploiting the opportunities offered by new UK localism and devolution settlements, each partner has put in place ambitious plans to develop new working practices that break down departmental silos and align strategies across policy sectors see 24–27 to promote wellbeing, prosperity and tackle above-UK average levels of NCDs and inequalities.

Each policy partner prioritised key policy areas for SIPHER to focus on, accounting for the anticipated scale of health and equity impacts deemed achievable, the buy-in across relevant sectors, and the anticipated utility of SIPHER’s models and outputs. Academic and policy partners then jointly selected four ‘whole-systems’ policy areas that exhibited major aspects of complexity – one shared by all policy partners, and one unique to each (see Box 1).

Reflecting partners’ priorities, SIPHER proposes to focus on increasing healthy life expectancy, reducing multi-morbidity and closing the health gap between advantaged and disadvantaged population groups. SIPHER will also include broader wellbeing outcomes (measured objectively and subjectively) across a number of relevant domains. The choice of wellbeing domains will be informed by systems mapping – a process described later - but aside from health will likely include financial stability, employment, housing conditions and positive social and community relationships. We will value the economic costs and benefits across sectors, enabling cost-effectiveness estimates that use a societal perspective to tie in with policy partners’ dominant approaches e.g.32.

Research questions

Three overarching research questions will guide the consortium:

1.   How can we capitalise on recent advances in complex systems science and multi-criteria optimisation to maximise the health-generating potential of public policy?

2.   How can we design complex systems research processes, models and decision tools to be most useful to academic and policy audiences?

3.   Which pathways and strategies best translate complex systems science evidence into policy?

Theory of change

Figure 1 presents SIPHER’s theory of change, which is mapped onto the UK Prevention Research Partnership Impact and Evaluation Framework. Working with our policy partners, we anticipate our new systems science evidence will initially contribute to policy debates, narratives and agenda-setting in our focus policy areas. As policy opportunities arise, models and tools are then available to answer budget prioritisation and policy design and implementation questions. Specifically, SIPHER will identify which policies maximise benefits across organisational aims, and which policies might lead to important disbenefits in specific sectors. It will also provide ongoing monitoring of policy contexts and effects over time to support policy refinement. In our policy partners’ words, SIPHER evidence will “provide decision makers with the confidence to invest in new approaches”. It is through such evidence-informed investment in SDoH-relevant policies that SIPHER’s NCD and health inequality benefits will be realised. Wider policy networks also benefit: previous research shows that decision models, co-produced with policy-makers to tackle key policy questions, can play a powerful role in advocacy and political process³⁶. Drawing on our experience of policy appraisals and known enablers of evidence use in policy³⁷, we specified the necessary pre-conditions and actions required by academics and partners and created conditions known to facilitate evidence uptake: 1) we have built strong relationships with champions in each organisation which allow us to identify key features in policy debate and ensure our evidence can speak to these; 2) we focus on partners’ policy priorities, and make provision for these to change over time; 3) policy partners are co-investing time and money and will co-own outputs; and 4) we have well-developed plans for the migration of models and decision tools to policy partners via embedded, co-funded analysts. To scale-up our impact, we will: 1) work with knowledge transfer partners to transform decision support practices to routinely consider whether a policy problem would benefit from a systems science approach; and 2) ensure our models and tools transfer between policies and jurisdictions, in recognition that policy makers prefer context-specific results.

Figure 1. Theory of change.

WS1: Understanding policy systems, policy processes and evidence needs

In WS1, we will formally engage with and capture policy developments in each organisation, including external influences on the organisations’ policy environments. Research in this WS will provide evidence on decision contexts, structures, practices, influencers and engagement with both the SIPHER programme and its outputs. Its longitudinal multi-method study of cross-sector policy development across three government levels provides empirical insights to advance theoretical accounts of policy change and evidence use. WS1 will also be a crucial source of information regarding changing policy contexts for the rest of the SIPHER work programme, for example around external events, influential actors and issue framing. WS1 will use a novel combination of qualitative methods including interviews, ethnographic research, documentary analysis, and participatory group model building (system mapping) to understand the evolving policy systems in our partner sites – the first study of its kind to use this range of methods to collaboratively explore, over time, changing decision contexts, stakeholders’ mental models of the factors influencing each policy system, their attitudes, values, and ways of working.

WS2: Synthesis of published evidence

WS2 will develop literature search and review strategies that are suitable for supporting complex systems modelling. Rapid evidence reviews will address a range of consortium concerns including the refinement of hypothesised causal links and system maps, population and spatial distribution of policy impact, state transition estimates, preference-based utilities evidence, and cost-benefit evidence. Given the whole-systems scale, the key is the balance between breadth and depth in the review process. We will adopt a dynamic information-seeking approach³⁸ that accommodates ill-defined and fuzzy review questions, diverse evidence formats and addresses the absence of ‘ideal evidence’ for some information needs by making explicit the judgments and assumptions associated with adopting ‘next best’ evidence. We will use recommended rapid, iterative review methods³⁹ to maximise the breadth, efficiency and relevance of the overall review process whilst achieving greater depth where needed (e.g. areas of particular importance or contention to policy partners, evidence to populate models where available data is limited). Expert knowledge will be sought and quantified using formal elicitation frameworks⁴⁰ where relevant evidence is absent or insufficient.

WS3: Data acquisition, management and generation of synthetic populations

WS3 will build a secure data infrastructure, create detailed simulated populations and develop a system monitoring function to inform adaptive policymaking. The secure data infrastructure will enable ongoing sharing of sensitive, individual-level data, the creation of detailed simulated (yet representative) populations, and the development of a system monitoring functionality to inform adaptive policymaking. Data acquisition and warehousing draws on existing infrastructure, providing secure data transfer, storage and curation. A comprehensive data audit will identify the type, scale and accessibility of data needed for each focus area. Data requirements will be informed by WS1 system mapping and WS4 and WS5 model requirements. Using microsimulation, we will create attribute-rich population datasets (e.g. disaggregated by age, gender, socioeconomic status and by geography), linked with attributes relevant to the social determinants of health. This will be achieved by combining partner data with established datasets (e.g. census, government surveys). These ‘synthetic’ data preserve the distribution of original data (crucial for modelling inequalities) and combine attributes from multiple sources which would not ordinarily be released together.

WS4: Causal system dynamics modelling

WS4 will develop novel system dynamics methodology to robustly estimate the health, social and economic impacts of policies. We will develop causal models for inclusive growth, housing, mental wellbeing and ACEs, and parameterise the models for each partner. Causal modelling captures the dynamics and feedback effects of the causal processes represented on WS1 system maps⁴¹, for example relationships between unemployment, poverty and healthy life expectancy. The causal relationships between variables on the system maps can be represented at three levels reflecting increasing demands on the supporting evidence base: evidence where we only have effect direction; evidence that provides effect magnitude, and evidence that informs the temporal relationship of cause and effect. Causal models will be built at each of these levels (including re-use of existing models where available^15,42), then multi-level methods will be used, allowing the inputs and outputs of models with different types of evidence to be integrated to give projected outputs at each level⁴³, fully accounting for uncertainty. WS4 will generate open-source tools to accelerate future CSM applications.

WS5: Policy microsimulation

WS5 will build dynamic microsimulation models for each policy area to produce estimates of distributional (inequality) impacts needed for policy appraisal, along with a quantification of all relevant uncertainties. We will model the impact of changes in relevant parameters (e.g. employment status) on sub-populations (e.g. by sex, age, socioeconomic status) and at a sub-LA spatial resolution and assesses the evolution over time of effects in the system. The outcome is an open-source dynamic microsimulation tool for future applications.

The CSM methods in WS4 and WS5 are complementary: system dynamics models focus on the relationships between higher-level concepts while microsimulation captures the impacts of environment and policy on the characteristics of individuals and households. For example, WS4 might identify that in one of our jurisdictions, increasing the availability of affordable housing would affect disposable income and health. The WS5 micromodel would then assess changes in individual-level housing, income and health to understand distributional impacts. The combined use of micro and macro models will ensure maximum exploitation of evidence from WS2 and WS3 since each requires different types of data.

WS6: Social valuations of cross-sectoral outcomes and equity implications

WS6 seeks to provide insight into how people value different policy outcomes (e.g. increased income vs health) and different distributions of outcomes (increased total income vs increased income equality), which will allow policy analysts to incorporate information on the relative acceptability of different trade-offs into their policy assessments. In WS6, we will also convert the multiple outcomes that arise from a whole-systems perspective into two common wellbeing measures needed for economic evaluation. Using a combination of online surveys and discussion groups, it estimates statistical relationships between non-health outcomes and a monetary measure of well-being (equivalent income) and an extended quality-adjusted life-years (E-QALY) measure which incorporates non-health outcomes. The resulting information will then be used in WS7 to compare the preference-weighted costs and benefits of action targeted on different policy levers.

WS7: Economic evaluation and decision support

WS7 will use multi-criteria decision analysis methods to develop a decision support tool with visualisations that allows researchers and partners to interactively explore tensions and synergies across the full range of outcomes (captured by WS3-6) under alternative scenarios. The tool will incorporate values and preferences generated in WS6 and identify options that maximise benefits and minimise opportunity costs across policy outcomes. The economic modelling will also estimate the accrual profile of health, wellbeing and non-health outcomes over time. WS7 draws on distributed, robust multi-objective optimization techniques from engineering^1,44 and methods for adaptive policy design which have yet to be used in public health⁹. The latter identify strategies that perform well across different plausible future scenarios, updating projections as uncertainties resolve. Decision tools will help visualise and interactively explore conflicts and synergies between outcomes; see how different sets of preferences (stakeholders’ own or society’s) affect best-buy judgments; and see how economic metrics, e.g. net benefit, map to underpinning tangible outcomes. To encourage uptake, tools will be configured to pre-populate relevant analysis templates and integrated into existing and emerging toolkits available to analysts.

WS8: Evaluation

Guided by our theory of change, WS8 continuously and reflexively evaluates how the consortium’s activities and outputs translate into impact. It assesses how the partner organisations engage with the SIPHER programme and the evidence it generates, and evaluates changes in partner organisations’ approach to cross-sector policy decisions. It takes a multi-perspective approach to evaluation: we will conduct a continuous, mixed-methods, reflexive process evaluation of each consortium activity and of our collective contribution, seeking to gauge SIPHER-attributable changes in partner organisations’ and proactively identify and manage challenges. Using baseline and follow-up interviews and the regular researcher, WS and consortium meetings, we will assess the effectiveness of our collaborative working, use of SIPHER outputs, models and decision-support tools in policy conversations and decisions, volume and value of further research commissioned or funded and any evidence of SIPHER’s work gaining traction in policy organisations and academia beyond the initial consortium. Five years from now, we will commission an external evaluation of SIPHER’s influence on health and wellbeing-related policy decisions.