Emergency Department Escalation in Theory and Practice

Emergency Department Escalation in Theory and Practice

This week we’re taking a look at some research around how Emergency Departments in hospitals actually perform escalation procdues as they become overwhelemed, as opposed to how it is specified in policy. This paper was written by Jonathan Back, Alastair J. Ross, Myanna D. Duncan, Peter Jaye, Katherine Henderson, and Janet E. Anderson.

Emergency departments can be useful places for us to learn from in software because they are places where there is some amount known about expected or predicted usage, but demand can spike at any time and simply turning people a way is rarely an option. This mirrors much of the world we live and work in, if our systems are taking heavy traffic or are subject to some other disturbance, we can’t say “well, we’ll just redirect them to some place else.”

This study took place in the UK at a large hospital with a sizable emergency department. Here the department is subject to regulation by the National Health Service, that requires them to enact “escalation policies” as certain measures are hit. When certain thresholds are crossed the department is required to take certain actions. Though these policies exist at a number of emergency departments, there has been little to no research on how well they work in practice, whether they do in fact help maintain access to emergency services or not.

This is important becuase a gap between what a policy specifies and what adaptions actually take place will always exist so long as there is complexity, expertise, and judgement involved. This isn’t a bad thing.

But if policy or guideance specifies one thing, and something completely different is happening, that’s not ideal. Especially if as in this case there are penalties associated with not meeting certain metrics or following certain policy.

Also, by examining what happens in practice vs in policy, learning can take place to shape policy and help spread useful adaptations. I’ve talked about this idea a bit before.

If it is not commonly understood that such a gap between work as imagined and as done exist, it can lead to a pressure to follow policies exactly as written. This can in turn lead to “secret” policies being created, those that reflect the real way of getting things done that perhaps only those with certain experience will know. This can stand in the way of creating a good environment for resilience to emerge.

In order to understand what was happening at the large hospital that they researched, they spent time initially just observing so that they’d have context as well as analyzing multiple hospital’s policies at a high level. Ultimately they spend 16 months observing the “Majors” area of the emergency department, where the more serious cases go, as this is also where the concerns and discussions of escalation tended to take place.

They shadowed staff and when they weren’t otherwise busy they interviewed them using open ended questions to get more information so that they wouldn’t take routine things for granted. They chose times and days that were expected to be the most busy but also included some other periods to be able to compare and contrast. The researches explain that they:

“used resilience engineering theory to focus on how workers decided on actions, what actions were undertaken, and the effect of those actions.”


From the policy documents that they reviewed, 3 overall goals emerged:

  1. Increasing capacity
    • This allows more patients to be treated by doing things like moving patients to less busy areas of the department or using cubicles.
  2. Reducing demand
    • This helps the ED keep up by doing things like moving patients to other departments or diverting incoming patients to other hospitals.
      • The authors don’t say much about how effective strategies like diverting patients are, but in my experience in the US this is a very difficult thing to do that doesn’t always work.
  3. Increasing efficiency
    • Actions for this goal included changing processes to allow decision making to happen sooner and avoid unneeded investigations and moving staff around to areas they might be needed (but not necessarily normally be).

Some of the policies at hospitals were more prescriptive, some more vague, but goal 3, especially the action of moving staff to where they were most needed was a prominent feature.

Difficulties in practice

Though the department is subject to several rules, that patients need to be seen and admitted or discharged within four hours for example, in practice things didn’t always work out this way.

The hospital had an Emergency Medical Unit, which acted as a short stay ward where patients could go that needed additional diagnostics. This essentially bypassed the 4 hour requirement for those patients. I see this as an adaptation in the face of the rules that may be well intentioned but may also not take into account the situation for those actually doing the work of patient care.

In some cases the 4 hour throughput rule didn’t align with the physicians goals. Especially during busy periods, some physicians in charge would not assist with patient flow management when asked and instead focus on patients with greatest clinical need. This also is not a bad thing from a patient perspective, that physicians are focused on patient care instead of targets. This can be seen as one of the ways in which adaptive systems fail, working at cross purposes.

When things were busy, communication would slow down considerable and silos would begin to form. Senior clinicians, being overburdened with patient care, would leave the whoever was acting as the flow coordinator on their own, which would in turn slow down patient flow. Sometimes this would cause the flow coordinator to enter a sort of “firefighting mode,” where they would have to go track down needed resources such as referrals or missing staff themselves.

During busy periods, there are meetings every 2 hours to help determine when to enact formal escalation procedures. Representatives from different areas of the ED attend.

Through “board rounds,” a different meeting it physicians it was sometimes determined that though the department may be in such a state as to warrant the execution of the procedures by policy that they were more burdensome than helpful given the situation and were deferred. For example in one case a board round determined nearly 1/3 of the patients in the ED were ready to leave, so escalation was not needed.

Department policy though encourages escalations even if the staff don’t believe it will be of benefit. Further, choosing not to escalate could result in a reprimand from service managers, especially when treatment time limits or other targets were not met.

There was also high variability in how coordinating roles were performed, depending on who was staffed in each role. This also could impact patient transfers. In some cases, physicians in charge were able to monitor the department and more junior physicians, asking them to provide information on patients so that decisions could be made quickly on which patients to relocate and to where. Conversely, when physicians didn’t take an overseer type approach, silos tended develop. A similar pattern was noticed in the nurse in charge role as well, so it’s not confined to just physicians. When either role wasn’t acting as an overseer and instead was hands-on assessing patients, they were unable to work with the coordinator, slowing down patient flow.

Additionally, it was difficult to gain information about the journey of patients through the emergency department. There was no device or artifact available (like a white board) to the staff to keep an eye on what was going on. There was a computer system, but it seems to have only been accessible from limited places. This matches my experience in some emergency rooms, where everything is on the computer, but there can be lots of space between where patients are and the next computer, even if large monitors or TVs are used.

I think in software, even though we’re almost always in front of some networked machine we can create similar issues. I’ve worked with many teams where there is some existent monitoring that may provide a view into what is happening, but it lives somewhere “out there,” possibly unavailable or unknown to the people who need it.

There were times that escalation could have been performed, but was not due to other factors:

“During busy periods of the year, it was historically deemed important not to call escalations too frequently because it might result in a degraded response (“crying wolf”)”

Despite all of this many successful adaptations were performed and we can presume that is very often the case, given the continued functioning of the emergency department. But what helped those adaptations in the face of policy that sometimes hindered it?

“Successful adaptations required the ability to maintain an awareness of the state of the wider hospital system so that possible admission pathways were identified in anticipation of needs, and batch referrals to specialty services could be planned to reduce waiting times.”

“When processes were adjusted, successfully expedited patient journeys were reliant on getting the delicate balance between continuing with clinical work and interrupting work flow to perform planning activities right. Being sensitive to these pressures highlighted the need for awareness based on experience and expertise to avoid initiating unnecessary escalations based on codified indicators alone.”

If you’ve been a reader for long you’ve probably head me bang this drum already, but here it goes: expertise and experience were a key part of what made adaptations successful, not rote policies. Guidelines can be helpful, but ultimately what made the difference was expertise.

This is something I run into a lot at almost every level of a technical organization, a confounding of the use of guidelines or tools with policy to be followed exactly. Sometimes this can appear in requiring a certain level of detail in runbooks or runbooks for all sorts of things. I think runbooks can be helpful, but they’re most often helpful for those who already have a somewhat matching level of expertise or awareness. You cannot runbook someone into being an expert at how the system operates. You can however provide opportunities to learn, for them to explore, and to observe what helps and hinders adaptation and adjust accordingly.


  • Though many polices are very prescriptive, none capture all the nuance or were sufficient on their own, adaptation was always required.
  • " At times the escalation policy was triggered without any expectation that it would help. Such policies serve multiple purposes, including satisfying hospital reporting requirements and documenting pressure on the system."
    • Similar to my previous discussion about the role of error, these policies may be organziationally useful, even if they’re not useful to sharp end in preserving service or capacity.
  • " Much insight can be gained organizationally from the extent to which clinicians have developed their own adaptive strategies to manage patient flow. "
  • “For successful intervention it is critical to capture what adaptations are made, rather than assuming it is the policies that enabled the workload to be managed well or poorly.”
  • Creating penalties for not following underspecified procedures (and they’re always underspecified) can create an environment where there are “secret” policies, the real way of getting things done.
  • Ultimately, successful adaptations were often due to experience and expertise in how the department or hospital worked as a whole.
    • No amount of rote policy following can create this.
← The Risks of Autonomy: Doyle’s Catch
Toward General Principles for Resilience Engineering →

Subscribe to Resilience Roundup

Subscribe to the newsletter.