Systems change for the social determinants of health

In social determinants of health research, what is sometimes referred to as ‘upstream’ change – that is change within government and policy – is seen as a more powerful and effective intervention point for addressing the social determinants of health than ‘downstream’ measures which target communities or individuals [2, 27, 28]. This is particularly so when action centres on the ‘determinants of health’, rather than health itself [29, 30]. Upstream action is increasingly emphasised over and above programs or interventions aimed at individuals or community groups [27, 30, 31]. This is because the social determinants of health are now understood to be affected by the organisation of material and social resources amongst the members of societies, which is best addressed through government action [30].

Interestingly, the systems frameworks developed by Meadows [4] and Johnston et al. [3] do not map neatly onto the upstream-downstream dichotomy which now dominates much discussion on population health interventions for the social determinants of health. Rather than the level at which an intervention in made, systems frameworks draw attention to the way we intervene. That is, how we intervene in a system can be much more important than where we intervene; interventions made within government can still fail to take hold thereby generating few positive outcomes [29].

For example, recommendations that called for joined-up action between different policy actors and between different levels within service delivery systems (i.e., a linking of different parts of government, or government and other sectors – sometimes referred to as ‘whole of government approaches’ or ‘horizontal government’ [29]) were amongst the most common recommendations, particularly in later reports. The EU Report on Health Inequalities, for example, called for ‘all governmental levels to liaise and cooperate with other sectoral policies and invest smartly in specific health inequality measures’. This constitutes joined-up action in the sense that different parts of government need to connect with and work closely with other departments (see Table 3 for further examples). An exception to the popularity of recommendations for joined-up action is the UK Marmot Review into Health Inequalities, which did not include explicit recommendations for joined-up action. However, this reflects a limitation of our data extraction methods – the UK Marmot Review is premised on the notion that joined-up action is required to deliver on all recommendations contained in the review. This is reflected throughout the report and in the implementation and measurement plans.

While joined-up government/whole of government approaches are seen as a powerful intervention point, they sit towards the lower (less effective) end in Meadow’s scale. This type of action was coded as changes to actor network structures (10a) and information [6]. Physical system structures (built or otherwise) are tricky to change. In the case of actor networks, the adaptive nature of systems can come into play to mitigate outcomes that could be precipitated by such changes. That is, the self-organizing properties of systems means that they can quickly adapt to changes made at low intervention points, causing these changes to ‘wash out’ and have little effect. Recent research on the type of joined-up government/whole of government changes suggested by these recommendations indicates that more often than not these ‘upstream’ interventions do wash out and the system returns to the status quo [29]. The collection and reporting of information, implicit in joined-up efforts, can be an effective leverage point but only under particular circumstances.

Collection and reporting of information on its own is not enough to generate substantive change. To be effective, information flows must be restored to the right place in the system and in a compelling form [4]. Meadows uses the analogy of a pilot, who receives information on the state of the aircraft and is positioned to act swiftly on this information. Moreover, if a pilot does not act he/she will immediately feel the repercussions of this failure to act. Hence, the type of data collection systems that are recommended across the various reports coded (see Table 3) need to be integrated into the system in such a way as to force decision-makers to act. Viewing data collection and reporting through a systems lens can therefore make the difference between a relatively weak action in terms of systems change and a very powerful one.

In coding to Meadow’s full twelve leverage points, we found several powerful but underutilised leverage points. Few recommendations argued for changes to rules in the system. Rules define the boundaries, or scope of the system. When dealing with inequalities in the social determinants of health, rules become critically important. A simple example of this is how much wealth we allow individuals to accumulate. If this is unlimited, disparities are free to widen. If we cap the amount of wealth any individual can posses, we stop growth at the top end of the social gradient. As Meadows contends, “If you want to understand the deepest malfunctions of systems, pay attention to the rules and to who has power over them” [4]. In our example, these rules are taxes that favour the wealthy. It is worth noting that, while powerful, these types of changes to system rules can be socially and politically difficult to achieve. This is likely to be particularly so in countries with countries that operate under state regimes that favour individualism and fewer government funded services and support (i.e., liberal versus social democratic regimes) [31‐33].

No recommendations were coded as a the power to add to, change, or evolve system structure (leverage point 4 in Meadow’s framework). Systems are naturally self-organising, where complex behaviour emerges from relatively simple building blocks or rules (for example, DNA). Using this self-organising nature to one’s advantage can be a powerful leverage point. A focus on the self-organising tendency of systems can generate highly sophisticated and nuanced approaches to change.

Social determinants of health advocates frequently call for changes to policy. Yet, this is commonly done in broad terms, such as ‘We recommend policies to improve the quality of jobs, and reduce psychosocial work hazards’ (Acheson Report) or ‘Public policy—both national and global—should change to take into account the evidence on social determinants of health and interventions and policies that will address them’ [34]. However, understanding the self-organising properties of systems, and the role of feedback loops in enabling this self-organisation, means we begin to think more carefully about the type of policy changes we recommend. The literature on systems science and public policy argues for ‘adaptive’ or ‘learning’ policies. A dynamic, self-adjusting feedback system cannot be governed by a static, unbending policy. In fact, static policies often fail to produce their intended effect as the dynamic system shifts around them. The Australian government’s taxation increase on ready-to-drink spirits-based alcoholic beverages (referred to as alcopops) sought to decrease harmful drinking by, particularly, young women. The targeted increase saw consumption of other drinks rise [35] and no change in alcohol-related violence [36]. Most relevant is the observation by Doran and Digiusto [37] that ‘it is impossible to know how much of the [consumption] changes were due to the tax, to the ‘global financial crisis’, to adaptive marketing by the alcohol industry, to the Government’s national binge drinking strategy, to mass media coverage of these issues or to other factors.’

Learning, or adaptive, policies change depending on the state of the system [4, 38, 39] . For example, an adaptive education policy would make the proportion of government funding for private schools contingent upon the performance of public schools. When public schools perform well, private schools receive more funding. When they perform poorly, government funds for private schools decreases. This type of adaptive policy changes as the system changes, but also uses the self-organising principles of the system to achieve a particular outcome (i.e., more equality in school outcomes between public and private systems) [38]. Here, the rules and incentives are bent towards favourable action in terms of achieving the goal of reducing the inequalities which stem from tiered education systems. These built in policy adjustments can speed up the process of responding to emergent conditions within the system [39, 40].

Finally, it is worth noting the limitations of this study. The research only considered a subset of all SDOH reports – concentrating on the UK context in the main. Reports from other countries, such as Brazil and other parts of Latin America where action on the SDOH has occurred, could yield different results and would be a worthwhile area of future investigation. These different contexts may require, or potentially enable, different types of action to be taken. It is also worth noting that what is contained in the recommendations of the reports analysed is not necessarily representative of the aspirations of the field of SDOH research as a whole. The reports are produced within particular political contexts which constrain the types of recommendations that can be made. All of the Reports we analysed display a tendency towards centrist policies endorsing neither neoliberal, market-based solutions nor highly socialised market-opposed interventions. These constraints may have some effect on the content of the recommendations but they need not effect the types of leverage points targeted. There is no reason why the right or left of politics would be more likely to target the rules of the system or its goals. Were these constraints lessened, therefore, our analysis of ‘how’ we intervene upstream would remain relevant.