Skip to content

Innovation

Definition

We propose 2nd-order approaches as those most aligned with systems engineering in healthcare (Clarkson 2018; RAE, 2017).

In common with 1st-order approaches, systems are treated as ontological entities (i.e. a collection of components and dependencies that exists “in the real world”) at a defined level of abstraction. 2nd-order approaches also make use of and deliver formal models but differ from 1st-order approaches by allowing

  • specification of a system with broad and imperfectly specified boundaries
  • more tolerance for uncertainty and “unknowns”
  • less emphasis on analytical tractability and instead, numerical approximation and simulation methods are common; for example discrete event simulation models (Vázquez-Serrano, 2021), system dynamics (Sterman et al. 2015) or when detailed enumeration of a system’s components is difficult, statistical process control (Benneyan, Lloyd, and Plsek 2003).

Qualitatively, 1st- and 2nd-order approaches differ because:

  • 1st-order models usually assume a mechanistic (often deterministic, law-like) relationship between system inputs and outputs
  • feedback -- either negative (damping) or positive (amplifying) -- is rarely explicitly modelled in 1st-order systems to maintain the tractability of the models
  • 2nd-order models allow for sophisticated behaviour-based descriptions of components (e.g. modelling clinicians as agents, with beliefs, goals and repetoires of actions having consequences external to the agent)
  • 2nd-order models allow for complex dependencies (e.g. the behaviour of one component or actor has effects on others that may not be immediately or easily predicted by having a complete description of the components/actors)

Both attempt to both describe systems and enable predictions or synthetic experiments to test counter-factuals (e.g. "if we increase the number of beds, does service throughput change?") that are impractical or costly to test practically without being constrained by some modelling.

Importantly, because the system components are considered at a higher level of abstraction than those typifying 1st-order approaches, 2nd-order models are often large, require time-based simulation and experimentation and sacrifice the detail and tractability that make 1st order modelling attractive. We propose 2nd-order approaches are suitable for system innovation (e.g. improving care pathways/services by remodelling) and revision (i.e. “designing out” bugs in healthcare processes such as inefficiencies and bottlenecks).

Examples

An unusual (but instructive model) uses cellular automata to simulate infectious disease propogation (Zhou, 2019) that produce results compatible with classical compartmental models. Cellular automata (CA) are essentially "grids" of simple components (cells, agents, components), each defined by homogenous rules that govern each cell's behaviour and resulting state. The rules governing cell states (e.g. alive or dead) depends -- sometimes stochastically -- on the states of it's neighbours. In classical two-dimensional CA, the grid arrangement implies a spatial relationships relevant to the system being modelled -- so for example, cells can only be influenced by their immediate neighbours (e.g. the 8 neighbours in a grid of square cells). By increasing the sophistication of the rules governing the behaviour of a cell (agent, component) and relaxing the strict homogenous grid-like structure leads to models such as agent-based methods.

The healthcare digital twin (Masison, 2021; Shen, 2024; Katsoulakis, 2024) represents one contemporary framework that shares modelling approaches with 2nd-order systems, but often, in an attempt to model a single component or agent (patient) as an aggregate of complex sub-systems.

An overview of systems approaches in health care can be found in Komashie et al's (2021) systematic review.

An overview of 2nd-order techniques employed in healthcare settings is provided in (Pitt, 2016).

For a systematic review of different modelling approaches used for the specific healthcare problem of improving patient flow in emergency departments, see (Mohiuddin, 2017).

References and Further Reading