Patients waiting in emergency department

JC: To batch or not to batch? Managing flow in the ED

This JC differs from our usual fare in that it does not focus on a clinical topic; it is not even published in a medical journal. It relates to urgent care flow, which is topical given rampant crowding across urgent care services globally.

I enjoy learning about systems, and have written about it here and here for St.Emlyn’s. Much of what we do in urgent care is underpinned by processes and systems. The interconnectedness of things (staff, stuff, systems and space) is what allows us to navigate safe, quality and efficient urgent care. This paper’s premise, and in particular the questions it raises about urgent care flow, has been cycling around in my head since I first read it.

This paper studies a common ED practice called batching. Batching is the practice where related tasks are grouped, irrespective whether tasks are for different patients: signing on to see multiple patients, and then batching their administration (notes, referrals, etc) for instance, is common practice for some ED clinicians. Most believe that batching speeds up their productivity. But is there any real evidence of this?

This paper answers the question by evaluating how batching of a decision to admit, impacts on clinician productivity and boarding (the process where a patient is held in the ED after a decision to admit was made). As always please read the full paper and make up your own mind before considering a change in practice.

What kind of paper is this?

This is a retrospective observational (cohort) study. Observational studies tell us how things are. There is usually no interference with the subject matter, and no control or treatment groups. As I have already mentioned, this paper is not from a medical journal. It was published in Operations Research which is a journal that focusses on the science of decision-making.

The layout is somewhat different: it follows the Introduction, Methods, Results, Discussion format we are used to, but with substantially more detail: there is a whole page for hypothesis development, five pages describing the various variables included and a post-hoc analysis following the planned results section. There are no numerical citations and references are listed alphabetically at the end of the paper in the APA format.

The reason it looks different is because it isn’t strictly healthcare research. Sure enough it is set in a healthcare environment, but the focus is on the queue, and how a specific process, like batching, affects it. According to Operations Research, it reports on design, analysis, organisational and implementation sciences. And these are reported with substantially more technical information than healthcare research. Thankfully we can still learn from our system engineering colleagues, as long as we are prepared to read substantially longer papers.

What exactly did they study?

The paper describes ED processing as a two-stage queuing system: new patients first queue for care from an ED clinician, and if admitted join a second queue for an inpatient bed. The second queue is wholly dependent on the first queue, including delays that may be unintended – such that may, or may not, arise from batching.

The authors collected 25 months of patient admission and clinician shift data from a single ED (see below). The authors then looked at how admission data behaved in relation to clinicians’ shifts, to answer the following three research questions:

  1. Does batching of the decision to admit increase towards the end of a shift?
  2. Does batching of the decision to admit negatively impact on boarding?
  3. Does batching of the decision to admit negatively impact clinician productivity?

Tell me about the setting

Data collection took place in the ED of a 567-bed academic medical centre, that included a level 1 trauma centre. Authors state that it is one of the top 10 busiest EDs in the US. This particular ED consists of multiple sections, called pods. Pods are staffed by at least one clinician along with three to six nurses. During periods of lower attendance, a single clinician may take responsibility for more than one pod.

Clinicians pick patients up and then assess, diagnose and treat them. Finally, the clinician discharges, transfers or admits the patient. There are no direct financial incentives for clinicians to take on new patients. We are told that clinicians feel responsible for their pods and that this is an important driver for the number of patients seen.

There is however a disincentive to taking a handover from another clinician who was unable to finish with their patients. The new clinician will not be credited for the patient. Therefore, clinicians are reluctant to sign up for new patients toward the end of their shifts. It is reported that this is an important driver for batching.

What did they find?

Batching was more commonly observed at the end of shifts. Batching also increased individual clinician productivity. But despite increased productivity, batching paradoxically worsened boarding. This was in part due to increased batching occurring at the end of a clinician’s shifts, and amplified by several overlapping shifts.

In other words, any positive, individual productivity effect was more than offset by a resulting negative, compounding system effect. The posthoc analysis suggested that without batching, boarding could be reduced by as much as 15%.

How does this translate to the NHS?

This is a US study and as such cannot be generalised for UK EDs. That said, I am fairly certain that UK clinicians are as reluctant to sign up for new patients toward the end of their shifts as their US counterparts. And batching most certainly occurs for various tasks in the UK, including admission decision towards the end of a shift.

Batching presents itself in other ways as well. In the UK (as well as other health systems), performance metrics such as the four-hour standard forces activity early on in the two-stage queue; promoting early decision making for the first part of the queue. We’ve known for a while already that most admission activity happens around the four-hour mark (more recently around the 12-hour mark, but that is just too depressing to unpack here). Given standard patient arrival patterns across the day, it becomes really easy to predict when admission batching (and subsequent boarding) is most likely to occur.

More batching occurs in the second part of the queue, where site or bed managers use a series of escalation triggers to open up inpatient bed space. More recently a third queue was added, ahead of the first queue, for patients to enter the ED (specifically ambulances). Frankly it is all a bit of a mess, and you won’t be blamed for wondering if anything matters anymore.

Proactive vs reactive flow

Looking a bit deeper however (and this is the bit that keeps coming back to me): this study pits a continuous, proactive flow model against the more traditional, reactive flow model for admissions from ED. Interestingly, there is an NHS Trust in the SouthWest of England, with previously reported severe ED crowding, that introduced such a continous flow admission model: a pre-determined number of patients are admitted from ED, on the hour, every hour, with the onus on the receiving ward (admin, nurses and clinicians) to accommodate incoming patients.

The Trust reports reduced ED crowding and boarding in ED. Anecdotally this has driven innovation both in ED and on the ward for safe ward to accommodate a more continuous, proactive flow model. Batching still happens; but the difference is that it is now standardised across the day, by the hour (a fixed variable), rather than patient arrival patterns (a random variable).

Many questions remain for us to ponder over. Although batching practice may result in downstream admission queue disruption, it also speeds up care in the non-admitted cohort – so how do we struck a balance? Is a two-stage queuing process inevitable, or is it simply a social construct due to a particular school of thought, tradition or hierarchy (how we have always done things)? Would a system that supports continuous, proactive flow nudge batching of admitted patients towards a better balance? I’m fairly sure if we ask ten different people we will get ten different answers, none necessarily right or wrong.

Striking a balance between the trade-off involving the first and the second stages of the queue may present the best way to approach urgent care flow. The newly proposed admission within one hour of being clinically ready to proceed metric may just be the type of metric that could drive improvement. But that topic is an entire blogpost on its own for another day.

At the very least this JC should have you thinking more objectively about your batching practice. This paper introduces a fascinating and disruptive concept worth considering in future decision making research. I value your thoughts and ideas, please comment below.

vb

Stevan

References

  1. Feizi A, Carson A, Jaeker JB, Baker WE. To batch or not to batch? impact of admission batching on emergency department boarding time and physician productivity. Operations Research. 2022 Aug 12. https://doi.org/10.1287/opre.2022.2335

Cite this article as: Stevan Bruijns, "JC: To batch or not to batch? Managing flow in the ED," in St.Emlyn's, October 1, 2022, https://www.stemlynsblog.org/jc-to-batch-or-not-to-batch-managing-flow-in-the-ed/.

Thanks so much for following. Viva la #FOAMed

Scroll to Top