Structure Generates Behavior: A Blueprint for Whole-System Change
Complexity theory and whole-system change programmes in healthcare
“Whole system change” is underlying much rhetoric of healthcare policy. Yet, when performance lags, the instinctive solution called for by the frontline is almost always for more: more beds, more doctors, more nurses.
It is a natural reaction. When you see a crowded Emergency Department (ED) / Accident and Emergency (A&E), the problem looks like a capacity shortage. But in complex systems, adding resources often fails to solve the problem. As systems theory teaches us, it is structure that generates behavior, not just the value of the variables within it.
In our research with Professor James Barlow into the Unscheduled Care Collaborative Programme (UCCP) in NHS Scotland, we observed a powerful example of how changing the structure of a system—rather than just pumping in resources—can trigger massive improvements. But we also saw how even the best structural fixes hit a “glass ceiling” when they fail to account for the wider world outside.
Here is what we learned about leverage points, black boxes, and the counterintuitive reality of changing large social systems.
The Counterintuitive Fix: Fixing the Outflow to Clear the Front
The Scottish programme was driven by a single, high-stakes target: a 4-hour maximum wait in Accident & Emergency (A&E).
Intuitively, if you have a bottleneck at the front door (A&E), you might think the solution is to add more resources at the front door—more doctors, more beds, more emergency nurses or a bigger waiting room. But the programme took a counterintuitive approach. It recognized that the leverage points for fixing A&E were not actually in A&E.
Yes, large improvements were gained by using big data on attendances to mobilize the appropriate staff at different days of the year. For example, anticipate three severe burns on Saturday afternoon first week in May at an specific hospital because the weather will be fair, a football game is on, and people will cook barbecues.
But, the national programme guided local teams in realizing that their A&E was crowded not because they worked too slowly, but because they couldn’t move patients out into the wards. The blockage to address was occurring in the “outflow”.
By shifting the focus to patient flows—specifically looking at how patients moved through minor injuries, medical admissions, and surgical admissions—hospitals began to dissolve the organizational silos between their departments. They didn’t need a massive influx of new budget; they needed to redesign the relationships between the emergency room, the labs, and the wards.
This approach based on distributed innovations and changes for structural adaptation was cumulatively successful. By streamlining the outflow at the back of the hospital, they released the pressure upstream at the front. It proved that you can achieve “whole system” change within a hospital if people update their mental model to collectively identify a few high-leverage control points and actually change the structure of their mutual interdependencies to rigorously manage the flow between them.
Hitting the Wall: The “Black Box” Effect
However, even the most streamlined hospital exists within a larger ecosystem. And this is where the structural approach encountered a hard limit.
While acute care staff successfully “connected the dots” inside the building, they eventually reached the hospital exit. To further improve patient flows, they needed to influence factors outside the control of acute care sub-system: social care packages, nursing home admissions, and GP availability.
Here, they encountered what we call the “Black Box” effect.
To the hospital managers, the outside world (primary and social care) became a dissipative black box—a massive, opaque system whose momentum they couldn’t influence. They understood the interdependencies—for example, waiting for a social care package to be agreed upon for discharging a frail elderly patient, thus freeing up a bed in the ward, in turn allowing to move a patient out of A&E—but they lacked the leverage to change them.
Conversely, for the stakeholders inside that black box (GPs, social workers), the “A&E Target” seemed irrelevant. Because the goal was framed entirely around a hospital metric, they asked, “What does this have to do with us?”. They perceived it solely as an acute care target. The incentives were misaligned, and the collaboration stopped at the hospital walls.
The Sting in the Tail: Compensating Feedback
This brings us to another dangerous trap in complex systems: compensating feedback.
Because the acute hospitals were remarkably successful at restructuring their patient flows, A&E performance improved dramatically. Waiting times dropped. The service became faster and more reliable.
But in a complex social system, success can breed its own problems. As A&E became more efficient, the public noticed. Patients realized that going to A&E was now faster and more convenient than waiting for a GP appointment.
The result? Attendance increased. In some areas, workload jumped by 12-20%, threatening to erode the very efficiency gains the hospitals had worked so hard to achieve.
This is a classic compensating feedback. It demonstrates why you cannot treat a hospital in isolation. If you optimize one node in the network without addressing the wider structure, the system will re-balance itself in ways you didn’t predict.
The Way Forward: A Portfolio of Leverage Points
So, what is the lesson for policymakers and leaders? It isn’t that targets don’t work—the A&E target worked brilliantly to drive internal structural change, throughout the acute care ‘sub-system’. The lesson is that we need to extend that structural thinking further out.
To avoid the “Black Box” effect and manage compensating feedback, our findings showed that we need to design a portfolio of targets that operates at three levels: system behavior targets (the “why”), subsystem performance targets (the “what”), and interdependency targets (the “how”). We cannot rely on a single metric to do all the heavy lifting.
While the theory of complex systems is appealing for understanding the behaviour of health care systems, the way it is usually used misses the important notion of scale. Currently numerous uncoordinated, parallel performance targets are introduced. This may well achieve ‘whole system’ changes but in reality this is only at a ‘whole subsystem’ level. It is essential that a clearer differentiation between targets for system behaviour and targets as performance monitoring is made. This would lead to a smaller number of strategic targets which take into account the interactions between subsystems and potentially limit the second-order impacts of competing objectives.
Conclusion
The Scottish experience offers a hopeful message. It proves that we can fix defaillant systems without always needing a massive influx of new resources. By finding the right leverage points and redesigning the structure of our workflows, we can achieve remarkable results. This is one of the core message of complex system theory: structure generates behavior. So, instead of changing the values of variable s (the ‘more’ approach), identify a few high leverage points leading to structural adaptation.
We just need to ensure our vision is wide enough to see the whole picture. When we design policy that respects the scale of the system—connecting the local control room to the wider world—we turn the “Black Box” into a glass house.
For those interested in the full details of my research, you can access the paper here: https://journals.sagepub.com/doi/abs/10.1258/jhsrp.2009.009097
