Description of CAS Properties (From Paina and Peters, 2011)
Path dependence is an important phenomenon in the physical and social sciences, and describes that “history matters” by demonstrating how non-reversible processes have similar starting points yet lead to different outcomes, even if they follow the same rules (30). Outcomes are sensitive not only to initial conditions, but also to bifurcations and choices made along the way and single events can have system-wide effects that persist for a long time. Path dependence complicates predictions of the system’s evolution over time and often occurs when there are rapid changes in technologies and heterogeneity in the types of actors involved.
For example, early entry and success in a market often force rivals to cooperate on the question of standards and compatibility of technology, as was the case for national grid voltages, the QWERTY keyboard, railway gauges, or vaccines. Any one of the present-day standards may not be selected today as the most efficient if not for its historical advantages or the transaction costs required to change standards. There is considerable diversity in the codification of health technologies, pharmaceuticals, and services standards across countries due to historical preferences and local regulatory processes and actors. Ball observed that such aggregation around standards are rarely created or enforced by legislation alone or agreed to by an industry, but more frequently involves the uniting of organizations and alliances related to market forces (37).
There are also relevant institutional examples from the health sector. It has been argued that the British National Health Service is a product of a particular cultural legacy that determines the success of reform efforts (38). Similarly, Bloom and Standing (39) argue that you cannot simply copy health reforms from advanced market economies (like creating national health services in post-colonial states, or introducing internal markets in the public sector), and expect them to work in countries that have not had the political processes or institutions (e.g. for health insurance or quality assurance regulation) in place to make them work. Much of the focus on scaling up health interventions in developing countries pay little attention to organizational arrangements needed to support the spread of access to health services.
Feedback loops occur when an output of a process within the system is fed back as an input into the same system (30). Positive feedback increases the rate of change of a factor towards an extreme in one direction (i.e. is self-reinforcing), whereas negative feedback modulates the direction of change (i.e. is balancing). Some feedback mechanisms can also lead to repetitive behaviors (or dead-end loops). In general, feedback loops reinforce common perceptions that the “rich get richer”, and the poor are left behind. In health, feedback loops have been used to describe “vicious circles,” for example between poverty and ill health or malnutrition and infection.
Feedback loops have been used to analyze variations in supply and demand for health care services. For example, studies have described feedback loops between individual and community health. Typically called “neighborhood effects” or “place effects”, these phenomena capture how an individual’s community and environment can affect that individual’s health in both the short-term and the long-term. Most such studies about “neighborhood effects” have been based in developed countries (40,41), though with some applications in developing ones (42).
Studies on provider practice and variation capture the heterogeneity in provider behavior and how clinical practices become reinforced within provider networks. On one hand, these analyses have uncovered practices inconsistent with the state-of-the art evidence-based medicine, which resulted in the provision of ineffective care (43,44). On the other, they also helped to identify variation in provider practice connected to quality improvement and the diffusion of innovation (43).
Scale-free networks are characterized by a structure which is dominated by a few focal points or hubs with an unlimited number of links, following a power law distribution. They are not, in contrast to previous beliefs, comprised of randomly connected actors with a similar number of links to one another (45). One implication is that they have heavy-tailed distributions, so that extreme events happen much more frequently than is expected when one assumes a world of so-called “normal” distribution. Stock markets, the world-wide web, power grids, business alliances, and the human brain are all examples of systems in which scale-free network structures have been identified. All such networks are known to maintain their cohesive structure in spite of breaks in random ties, such that the overall network remains undisturbed despite, for example, multiple daily minor disruptions across the internet or errors in cell mutations (45). Yet the same is not true when networks are faced with coordinated damage on the major hubs they rely on, such as when viruses enter a network through key sites or key cellular structures or organs are damaged. Feedback loops within a network can lead to chain reactions (also called “failure cascades” or “domino effects”) that can lead to epidemic spreading of disease or the sudden collapse of markets (e.g. global financial markets or local health insurance markets) that depend on commonly held perceptions of trust.
Scale-free networks are particularly important in scaling up health services, because they can provide insights into the diffusion of health knowledge, technology, and practices that are central to questions of increasing access to services. For example, networks with scale-free characteristics have been identified as key in the diffusion of knowledge about child health, as well as in overcoming barriers to access to child immunization services in Ghana (46) and in Ethiopia (47). In Kenya (48) understanding networks was shown to be important to how to change norms about unhealthy community practices. Networks are also useful for understanding how and why new health care practices and technologies, such as electronic medical records, are adopted or rejected (49).
Emergent behavior, or the spontaneous creation of order, appears when smaller entities on their own jointly contribute to organized behaviors as a collective, resulting in the whole being greater and more complex than the sum of the parts (30). Emergent behavior can refer to any kind of learning or new pattern that emerges from the complex interactions of a system’s components. The flocking behavior of birds is a common example of how animals organize themselves. Humans self-organize in many ways, particularly in decentralized systems and as a way of establishing social norms, though not always with the most optimal results. The sudden transformation of a peaceful gathering into a violent one without planning is a more obviously harmful behavior. In the health sector, emergent behavior can be seen when informal sector health providers form organizations to protect practices in their trade, or when health workers suddenly organize to go on strike. In environments where central regulation of health systems has been ineffective, self-organization and self-control of a health system is an important means of regulating health services (50). Because emergent behavior may be particularly difficult to predict, plans for scaling up need to monitor and adapt to such events.
Phase transitions are tipping points that occur when radical changes take place in the features of system parameters as they reach certain critical points (30). The transformation from one phase to another has frequently been described in the physical sciences when substances change between gases, liquids, and solids. In nature, they may occur abruptly (e.g. water at its boiling point), or gradually until a critical point is reached (e.g. the loss of magnetization as temperature changes). In a health system, abrupt changes are unusual. However, transitions have often been described as threshold effects. Nonetheless, particularly in the process of scaling up, it is useful to identify the conditions under which both rapid and gradual transitions can occur, whether around the rapid adoption of a policy stalled for years, changes in social norms concerning health behaviors, or new demand for health services.
CAS Pathways in the Health Sector