What’s Next? Cognitive Task Analysis of Emergency Physicians’ Experience in Multi-Patient Environments

Concurrent management of multiple ill patients is an important skill in emergency medicine, especially given increasing emergency department (ED) patient volumes. In this environment, rapid task prioritization is a critical skill. Regularly, emergency physicians are asked to concurrently manage multiple patients at once at any given point in their shifts, and often have to make time-sensitive decisions around the priorities across multiple patients. The art and science of teaching the critical skill of task prioritization is not well described in the literature. Few studies have explored the cognition of physicians in multi-patient scenarios, and even fewer have examined how this affects their clinical decision-making. We conducted a three-part, mixed-methods cognitive task analysis of attending and resident physicians’ thinking about efficiency and task prioritization in multi-patient environments. The three components of this study included a critical incident interview, a cognitive task (prioritizing patients on a simulated tracker board), and a think aloud experiment of that same cognitive task. This study was completed at multiple teaching hospitals associated with a major Canadian academic institution between March 2014 and September 2015. Ten attending physicians and ten residents engaged in all three parts of our study. In the first part they were asked via a critical incident interview to describe difficult prioritization scenarios, as well as the teaching and learning environments that result in the learning of this skillset. In the second part, participants engaged in simulated prioritization exercises using a novel simulated tracker board interface. Participants were asked to view and interact with a tracker board with various simulated patients, and then prioritize these patients. Participants were asked to describe which patients they would see first, which they would see soon, for which they would initiate orders or tests, and which they deem could wait. Times to completion and interactions with this interface were recorded. We observed the effects on time to completion and task load as measured by a modified version of the NASA Task Load Index (modified NASA-TLX). Finally, the participants were asked to think aloud while completing the prioritization exercise. This part allowed us to complete a modified protocol analysis and generate a new conceptual framework, which explains how physicians engage in task-prioritization processes within these multi-patient environments. For the first part, there were three main themes that emerged from our interviews in our participant’s descriptions of how they taught or learned the skill of task prioritization: 1) formal didactic teaching, 2) observation, and 3) in situ instruction (i.e. on-the-job teaching, informal coaching in the ED). Only one formal teaching strategy was named, and only by a single participant (i.e., formal teaching around the Canadian Triage Acuity Score). The bulk of teaching and learning strategies were more akin to coaching. They tended to be found within the in situ category (e.g., collaborative problem solving; informal conversation with staff, i.e. think aloud, “running the board”, walk-around rounds). A minority of strategies included observation by learners (e.g., residents watching staff perform their duties) or by explicit role-modeling by attendings (e.g., faculty members asking residents to follow them around and observe how the job is done). For the prioritization exercises, we manipulated tracker boards to vary along three factors we anticipated may affect the degree of agreement across different participants’ prioritization decisions and their time to completion: number of patients with similar acuity, number of patients with similar presentations, and number of extraneous patients (i.e. patients already cared for by other physicians). None of the manipulated factors discernably affected novices differentially when compared to experts. There were no specific trends in expert vs. novice agreement within the various conditions as measured by the intraclass correlation statistics for the various tracker boards. There were significant main effects of our three experimental conditions within these simulated tracker boards on the participants’ completion time for scenarios: Increasing the number of patients with similar presentation led to longer time to completion (F(2,17)=35.6, p<0.001; means = 20 seconds with 0 patients with similar presentations, 5 seconds with 2 patients with similar presentations; 20 seconds with 4 patients with similar presentations). Increasing the number of patients with similar acuity led to a decreased time to completion with two similar acuity patients, but then a much higher mean time to completion with four similar acuity patients (F(2,17)= 18.8, p<0.001, quadratic relationship). Increasing the number of extraneous patients led to increased time to completion of the prioritization task ( F(2,17)=11.2, p=0.001, linear relationship). The experimental design only permitted examining two-way interactions while holding the third factor constant at zero each time. The think aloud processes revealed a unified, overall process used by almost all participants. The cognitive task of patient prioritization consisted of three components (Figure 1): 1) viewing the entire board to determine an overall strategy, 2) creating an archetype (a functional ED-context based illness script) from patient-care information available in an initial chart (i.e. vitals, brief clinical history), and 3) creating a relativistic prioritization list. Using a mixed methods study, we generated a cognitive analysis of how physicians perceive multi-patient environments and engage in rapid task-prioritization. This will inform development of didactic and clinical educational materials.