������Ƶ

Theoretical or conceptual models are used in research to organize complex and abstract concepts and describe relationships among them so that those relationships can be tested, which facilitates advancement in a field.³⁴ Several conceptual models in the literature provide guidance on how to reduce diagnostic error. NASEM developed a conceptual model to describe the diagnostic process,⁵ which was further refined for the ED context.³⁵

The Safer Dx framework provides a comprehensive overview of how to reduce diagnostic error specifically.³⁶ Gleason and colleagues provided the first conceptual model that illustrates the critical roles nurses can play in the diagnostic process and all the touch points in the NASEM process map of diagnosis where nurses are involved.¹³

A major shortcoming of all these models, however, is that because of differences in workflow, many actions nurses take are not in sync with those of their physician colleagues. Therefore, no standard mechanism exists for bringing the nursing perspective to a physician’s attention. Using an approach grounded in distributed cognition could address this critical flaw in diagnostic processing to help prevent and catch diagnostic errors. Figure 2 illustrates this point.

In the current model of diagnosis (left side of the model), horizontal arrows represent individual workflows and activities of physicians and nurses before and after seeing the patient. The nurse usually sees the patient first and gathers information, integrates and interprets that information, and frequently comes up with a working diagnosis. There is no expectation of interacting with the physician, although the circles overlap because sometimes the physician and nurse see the patient at the same time. In general, though, the physician sees the patient later and goes through the same process as the nurse while also ordering diagnostic tests and consulting with others as needed.

The current approach is in contrast with our proposed physician-nurse dyad model of diagnosis (right side of the model) that uses a distributed cognition framework to insert deliberate interactions between physician and nurse before, during, and after patient contact, as indicated by the slanted arrows. While overlap in patient interactions still occurs, the physician and nurse make a deliberate effort to see the patient together, which standardizes the message the patient receives and decreases the need for clarification of the plan later in the patient’s stay.³⁷ Our dyad model is a direct application of studies on how to optimize joint decision making; the best decisions come from allowing individuals to first think on their own, before collaborating to arrive at a final answer.³⁸

Figure 2. Application of Distributed Cognition to Model of Diagnosis

Research studies will need to evaluate and compare the best ways to achieve effective integration of physician and nursing input. One possible intervention would be to reconfigure the physical organization of work to bring physicians and nurses physically together at specific stages of the patient’s journey through the ED, as depicted in Figure 2.

Such an approach would serve multiple purposes. First, it would foster the teamwork NASEM recommends, providing opportunities to discuss a patient’s presentation and how each member of the physician-nurse dyad views it. Without scheduled interactions, physician and nurse workflows are asynchronous³⁹ and disconnected, resembling parallel play.^40,41

The potential value of teamwork of this sort was demonstrated in a cluster randomized crossover trial conducted in six EDs, where physicians cross-checked their diagnosis with another physician by presenting a case and receiving feedback.⁴² The study reported a significant reduction in adverse events and near-misses,⁴² suggesting that distributed cognition played a role by having the physician summarize information for the peer cross-checker or by having the cross-checker’s fresh “outsider” perspective bring new insights to the discussion. Although in this study cross-checking was done between physicians, a similar strategy could be trialed between physicians and nurses.

Second, bringing clinicians of various types together more than once reflects the temporal nature of both the diagnostic process and a patient’s journey through the ED and may help prevent the anchoring bias that can occur early in a patient’s presentation. Nurses can be just as susceptible to anchoring bias as physicians. For example, the triage process can be anchoring when a nurse elicits a chief complaint that is really secondary or incorrectly triages a patient to a lower priority category. Multiple interactions may be useful in mitigating bias effects.

Finally, many aspects of distributed cognition overlap with concepts inherent in “Safety-II” science, such as teamwork and feedback mentioned above, providing yet another avenue for intervention development.⁴³

Another area that would benefit from additional research involves optimizing information flow, specifically by improving communication. The explicit role of communication in diagnostic processes, as well as the impact of suboptimal communication on patient harm, has been well documented but solutions to the problem are sparse.^44-46 For example, communication breakdowns can occur at multiple levels and at varied points during the patient’s diagnostic journey in the ED. Patients arriving via ambulance often have no previous relationship with the transporting clinicians, who can only base their medical decisions on what is being told to them or what they observe.²

The EHR is also becoming increasingly important for information flow. For example, although they have been available for years, the implementation of communication devices that accept messages has led to a huge increase in the use of secure chat messaging between providers and nurses in the EHR. But given differences in perspective between providers and nurses, we know little about whether secure chat messaging via the EHR is an effective communication medium.

Once in the ED, patients, nurses, and physicians all engage in framing.⁴⁷ Furthermore, patients may not be able to accurately communicate their symptoms or history to clinicians for many reasons (e.g., low health literacy, mistrust, language barriers, cognitive impairment due to illness). Not enough research attention has been paid to differences in communication that can contribute to the framing biases or context errors that lead to diagnostic error. As described above, differences in perspective contribute to the framing effect, but instead of minimizing those differences, a distributed cognition approach would seek to bridge them, recognizing the value in bringing all perspectives to bear on a specific problem.

Several strategies can be used to bridge differences in perspective and improve communication, such as:

• Interprofessional education in the health professions that could provide formal training in communication between nurses and physicians and develop a shared, common language for diagnosis.⁴⁸
• A system in place where nursing input is actively solicited in the diagnostic process, possibly during an ED huddle.⁴⁹
• A requirement that physicians communicate their plan to patients in the presence of nurses.⁴⁹ Although this strategy might be difficult to implement when the ED is very busy, the patient would derive benefit from hearing the physician and nurse perspective at the same time and possibly allow more meaningful dialogue among them.

Research into artifacts and how they can be used to facilitate diagnosis through a distributed cognition lens is a final area of research that would benefit from further development. A tragic example of the need for research in this area was provided several years ago by a patient who had been traveling in Africa and presented to an ED in Dallas, Texas, with signs and symptoms of Ebola.⁵⁰

Although the nurse documented the patient’s travel history in the nursing notes in the EHR, the nurse did not otherwise communicate this information to the treating physician. The patient was discharged from the ED only to return 2 days later, much sicker. Known as “Patient Zero” because they were the first patient diagnosed with travel-associated Ebola in the United States, the patient was admitted but died within a few days.

This case revealed numerous system failures that contributed to the initial misdiagnosis and needless exposure of others to the Ebola virus.⁵⁰ Ineffective communication between the nurse and the physician was the dominant issue, and a central problem in this regard was the failure of the EHR to serve as an effective artifact for communication between them. From a distributed cognition perspective, the EHR was not then—and still is not—configured to optimize information sharing between members of a physician-nurse dyad, because they either view or have access to different screens, and information is not shared.

Other types of artifacts include schedules, white boards, and worksheets, all of which contribute to distributed cognition but only if they are shared. These artifacts mediate collective work but need to be shared as a way to maintain an overview of the total activity.⁹ Schedules that include both nurse and physician activities and worksheets that can be used by both physicians and nurses are just two ways artifacts can be used to promote distributed cognition. Table 1 summarizes our recommendations, organized by distributed cognition concepts, and provides associated rationales.

Table 1. Recommendations To Improve Interprofessional Diagnosis in the ED

Anecdotally, in our experience, a physician-nurse dyad approach is best for bringing the nursing perspective to bear so that it contributes to diagnosis. However, this hypothesis has not been tested, and a physician-nurse dyad approach may have negative aspects:

• First, a dyadic approach takes time to establish, which is always in short supply in the ED.
• Second, with so much nursing staff turnover (especially post-COVID), dyad stability may weaken.
• Third, in some cases, a dyadic approach may introduce biases, including “group think” effects if the same physician and nurse work together frequently.
• Finally, dyad combinations are not fixed and may change frequently during an ED shift, with one physician needing to establish a dyad with several nurses at the same time and vice versa because of differences in shift length or patient assignment.