The work of an emergency physician is an exercise in multitasking. We see multiple patients in parallel and respond to all the responsibilities that come with this patient care in addition to being aware of new, potentially sick patients. A typical workflow includes signing EKGs of new patients, documenting on a patient encounter, responding to radiology discrepancies from radiologists, answering patient care questions from nurses, discussing patient care plans with patients, consultant physicians, social workers, case managers, and physical therapists, and responding to emergently ill patients. By the end of a ten-hour shift, I will have made approximately 4000 clicks in the electronic health record, had 3 hours of direct patient care time, 1-hour reviewing test results and records, and 1 hour of discussing the care plan with colleagues. The real challenge is not necessarily due to the magnitude of these tasks but that they often occur in parallel with repeated interruptions. A typical Emergency Physician treats as many as many as 12 patients simultaneously but interacted with as many as 100 individuals. We are interrupted on average once every ten minutes with greater than 10% of my tasks being interrupted [Link]. In short, our workflow is interruption-driven with many of these interruptions resulting in a break in the task. We are often working near the maximum of the capacities of our short-term working memories.
In computer science, preemption is defined as being able to stop one task irrespective whether it was finished and switch to another one. This property allows computers to prioritize or weight certain tasks based on algorithms with predefined goals. However, like most things, preemption does not come without a trade-off. The process of switching – context switching – requires processing power as the computer has to take a number of steps in order to start a new task. First of all, the prior task need to be bookmarked in memory so it can be retrieved at a later time, secondly, resources required for the new task has to be determined and then accessed. All of the extra work required to switch task is known as “metawork” and it only minimally advances the state towards any goal. Metawork can be mostly be considered as wasted work. Furthermore, it should not come as any surprise that the effects of switching tasks include both delays and errors and these delays and errors only increase as the amount of switching required per unit of time increases. Similarly, when humans are interrupted mid-task, we undergo the process of context switching. We store the prior task in short-term memory and activate a new set of resources required for the next task. Additionally, similar to computers, these task switching requests do not come without an associated cost. Some studies have suggested that when doctors are interrupted they fail to return to the interrupted task 1 in 5 times [Link]. There is also evidence supporting the negative effects on task performance and perception of stress. Give a person an overwhelming number of trivial things to do, and the important things get lost in the chaos. Furthermore, if you interrupt anyone more than a few times they fall in danger of doing no work at all.
The seemingly seamless running of multiple programs and switching between them is accomplished by this process known as threading. However, the capacity to multitask is limited due to memory constraints and there is an often unpredictable threshold when multiple programs are running and competing for memory and this in turn drastically slows down the response to commands and the system comes to a grinding halt. This state is called thrashing and is also a very recognizable human state and state in the emergency department. These are the moments when the lists of tasks (or the cue of patients) continue to accrue at a more rapid rate than you can clear causing the emergency physician to thrash. The word thrashing is an appropriate adjective because you are mentally flailing about in a seeming state of panic and paralysis without completing any tasks. Eventually, you are forced to step out of this loop and stop doing everything and reset your priorities. As part of emergency medicine training, emergency physicians unofficially become better multitaskers with concomitant decreases in costs related to context switching. However, this improvement is asymptomatic and runs up against biological, cognitive constraints. According to George Miller’s magical number, most humans can store between 5 and 9 items in their working memory and multitasking beyond that is not possible despite training.
Therefore, for emergency physician and well-functioning emergency departments, it is imperative that technological and systems based solutions are in place or developed to filter information. In my opinion, information should be filtered through an information triage system. Consequently, non-urgent information should be presented via indirect channels and more urgent information should be presented via more direct mediums. Part of the filtration process requires understanding the meaning of the information and therefore, improved probabilistic models of decision support could serve as the basis for that filter. For example, intra-EHR filters could display information relevant to the patient presentation or just as importantly not display irrelevant information. Additionally, intra-EHR and other non-verbal communication tools should be utilized to communication non-urgent information. In terms system-wide improvements, information should move up the channel on a need to know basis. For example, all radiology discrepancy calls should not be routed to the Emergency Physician, every “critical” lab values do not need to be received by the physician, and every non-emergent medication order does not need to be directly communicated. This would entail not only having all providers in the department working at the top of their training with the ability to filter information to their training but also having the end decision makers (physicians) willing to relinquish the control associated with lower level decisions.
Internet scholar Clay Shirky stated that “there is no such thing as information overload. There’s only filter failure.” The fast-paced and chaotic culture of the Emergency Department is inherently characterized by information overload in the form of distractions, interruptions, and burdens of multitasking. Furthermore, emergency physicians do not have the technological tools nor operational filters in place to curtail the deleterious effects of information overload. The burdens of context switching are prevalent and not infrequently lead to a state of thrashing. Speaking personally from experience, there are moments in a shift where the burden of task switching far exceed my capacity and cause me to step out the department to reset. Until more effective filters are developed to stymie the effects of information overload, emergency department care will suffer from the decline in performance and delay in task completion associated with information overload.