Diagnostic uncertainty is common when patients are admitted from the ED to the hospital. Multiple studies have identified large discrepancies between patients’ presenting symptoms and final diagnoses, between admission and discharge diagnoses, and between clinical diagnoses and autopsy findings among patients admitted through the ED.6–8 The extent to which these discrepancies represent true diagnostic errors is less clear in the emergency setting than in other hospital settings.
The available measures in the ED (e.g., preliminary diagnoses) do not quantify uncertainty that may be clinically acknowledged. They also do not recognize the imperative for emergency care to focus on clinical stabilization and triage as much as, if not more so, than diagnostic specificity. Conceptually, then, diagnostic uncertainty at ED-to-hospital transitions may be inherent to emergency care or may relate to factors connected with the ED itself, the ED-to-hospital transfer communication, the receiving hospital clinicians, or some combination thereof. This section will review each of these in turn.
ED-Specific Contributors to Diagnostic Challenges and to Diagnostic Errors and Uncertainty
EDs are among the most common settings in which diagnostic error may occur,9–13 for many reasons, including encounter brevity, high patient acuity and volumes, staffing issues, and undifferentiated presentations with fewer available data points.14–16 These challenges produce second-order issues that further complicate the diagnostic process. Such issues include narrative uncertainty (i.e., unclear patient history) from patients who are incapacitated, experiencing impaired mentation or cognition, or lacking relevant cognitive (e.g., medication lists) or sensory aids.17
Further confounding the diagnostic processes are nonlinear emergency care pathways (e.g., departing the ED for imaging or procedures) and the frequent task switching in which clinicians must engage to process these dynamic streams of information.12,18 Finally, some diagnostic investigations may have outstanding results at the time of admission, while others may have yielded “incidentalomas” of unclear urgency or importance; each possibility augments the potential for uncertainty and error.12 Unsurprisingly, then, diagnostic errors in the ED are generally considered multifactorial.19
In addition to being common, diagnostic errors and uncertainty at the time of ED-hospital admission are high risk: approximately 3 in 20 occur in patients who ultimately experience severe harm or death,20 for several reasons.21 First, diagnostic error recognition in the ED is especially difficult: no clear standards for delays have been defined, and clinician perceptions of delays may differ from those of patients and families.15 Furthermore, complexity and acuity correlate with both uncertainty and risk. Thus, those patients ill enough to require hospitalization and to have worse outcomes are also most likely to experience diagnostic uncertainty.
Second, and relatedly, early diagnostic errors can propagate other types of medical errors such as admission decisions (e.g., triage to an inappropriate level of care) or inappropriate medication choices. Finally, admission occurs early in a patient’s hospital course, when patients may be medically unstable or undifferentiated and therefore most vulnerable to cascading errors.12,15
ED-to-Ward Handoff-Specific Contributors to Diagnostic Errors and Uncertainty
Because of these issues and because the language and goals of emergency care (i.e., triage and safe disposition) often differ from clinical goals in other settings (e.g., diagnostic accuracy), ED clinicians may have different cognitive and clinical reasoning processes than other clinicians.16,22 Specifically, while hypothetico-deductive cognitive models are predominant in many clinical settings, Kovacs and Croskerry describe how ED clinicians often appropriately rely on other models. They use hypothesis elimination (i.e., ruling out high-risk “can’t miss,” or “worst case” diagnoses) rather than hypothesis verification (i.e., Bayesian “ruling in” of the most likely diagnosis) for most diagnostic decisions.16
A key implication of this difference is that at hospital admission, these clinicians may have different expectations regarding diagnostic certainty than their handoff receivers.12,23 For example, ED clinicians might feel comfortable—and diagnostically “certain” that immediately life-threatening pathologies are absent—admitting a patient with acute abdominal pain after “ruling out” perforated viscus and ruptured aortic aneurysm.16
In contrast, an admitting hospitalist might ideally view such a patient as undifferentiated and thus having a large degree of diagnostic uncertainty. However, common phenomena such as diagnostic momentum and anchoring bias could instead yield diagnostic error if, for example, the hospitalist were to interpret ED “certainty” as synonymous with a final diagnosis (such as considering an acute abdomen fully “ruled out”).24 Poor communication, both about the diagnostic process and these overall expectations, may yield diagnostic uncertainty, error, and conflict, which can again perpetuate diagnostic error.22,25,26
Receiving Hospital Contributors to Diagnostic Error and Uncertainty
Receiver-specific issues may also influence diagnostic error and uncertainty. As above, diagnostic momentum, anchoring bias, and early closure can result from poor handoff communication, lack of shared mental models, and lack of shared understanding between handoff senders and receivers. Handoff receivers may perpetuate these challenges through passive communication,27 including not asking questions to clarify or check understanding, not seeking help from consultants or other clinicians, and not challenging unclear or potentially incorrect diagnoses when recognized.26
Strategies To Mitigate Diagnostic Errors and Uncertainty at ED-Hospital Transitions
Each of these factors might be targeted for preventing or mitigating diagnostic errors and uncertainty at the ED-hospital interface. In the ED and inpatient settings, computational approaches such as machine learning and natural language processing might in the future be able to analyze written notes and identify potential diagnostic uncertainty.28 Automation and standardization may also add clarity to diagnostic test ordering and interpretation (e.g., standard content and formatting of imaging results).29
Diagnostic feedback to clinicians, such as through autopsies, both standardized and ad hoc, is important for individual diagnostic calibration.15 For example, to promote clinician feedback and reflective practice on diagnostic performance, Ƶdeveloped the Calibrate Dx resource to help clinicians evaluate and calibrate their own diagnostic performance.30 ED and inpatient clinicians and trainees can use this resource to reflect on recently discharged or admitted patients. Both emergency and inpatient settings may also benefit from universal availability of timely expert consultation,15 including radiology double-reads.20
Although these setting-specific interventions may decrease diagnostic uncertainty and errors, these challenges will not be completely resolved without improved communication around diagnostic uncertainty at patient handoffs. Expert consensus recommends that ED-hospital admission handoffs explicitly mention the certainty of the provisional/working diagnosis,12,22 and recent literature indicates that hospital-based clinicians want better skills with which to communicate uncertainty.31,32 Educational products, such as the uncertainty communication checklist, may be useful in teaching these skills.31
Importantly, despite the theoretical benefits of explicitly calling out uncertainty (less diagnostic anchoring, more appropriate and targeted diagnostic testing, and potentially fewer diagnostic delays and errors), patients may perceive clinicians who discuss diagnostic uncertainty as less confident and competent than those who do not.33–35 Thus, thoughtful communication strategies should be used in these situations.
In 2012, Beach, et. al., published best practice recommendations for ED-to-inpatient handoff communication, including style, form, and content.22 These authors suggest synchronous, two-way, closed-loop communication, with the goal of constructing a shared mental model of patient care between ED and admitting clinicians. These recommendations are consistent with best practices from other handoff contexts,36–42 including a focus on clinical judgment and the patient’s clinical trajectory, diagnostic uncertainty, and outstanding tasks.22
However, although synchronous, two-way, closed-loop bedside communication is often considered the preferred gold standard, asynchronous ED-hospital handoffs remain common.26,43–47 Interestingly, in one before and after study, asynchronous voicemail-based signout did not change patient outcomes such as intensive care unit (ICU) transfers or perceived adverse events; this study did not measure diagnostic errors or uncertainty in either period.48
Structured handoff tools have been studied at the ED-hospital interface. Notable examples include DE-PASS:
- Decisive reason for admission,
- Evaluation time,
- Patient summary,
- Acute issues/action list,
- Situation awareness, and
- Signed out to whom.
This tool was associated with improved clinician satisfaction and statistically fewer ICU transfers from the wards in a pre-post study at a tertiary cancer center.49 While the DE-PASS study did not measure diagnostic errors or quantify uncertainty, the similarly designed SBAR-DR tool includes the presence of both a working diagnosis and the certainty around it.50 (SBAR-DR stands for Situation, Background, Assessment, Responsibilities/Risk, Discussion/ Disposition, Readback/Record.) SBAR-DR was perceived positively by clinicians and yielded improved handoff quality after implementation but did not lessen perceived diagnostic uncertainty.
Research on diagnostic uncertainty and error at the ED-hospital interface should move forward in two parallel tracks. First, standardized approaches for handoff communication based on existing tools (e.g., DE-PASS, SBAR-DR) or newly developed tools can be implemented and evaluated now to improve the transfer of information when patients are admitted from the ED to the hospital. Methods from design, human factors, and implementation science can inform the tailoring of these interventions and the assessment of communication quality.51,52
Second, improved operational measurements of diagnostic error and diagnostic uncertainty are needed in order to research these concepts more effectively at the ED-hospital interface. Promising approaches toward this end (e.g., Symptom-Disease Pair Analysis of Diagnostic Error)53 must be accompanied by methods to quantify uncertainty and to contextualize existing outcome measures with counterfactual information when possible.54 Future work might use these measures to examine the extent to which diagnostic errors can be further reduced.
In conclusion, the ED-to-hospital transition is a high-risk time for diagnostic errors. Specific communication strategies focusing on explicitly acknowledging diagnostic uncertainty and creating shared mental models can help debias receiving clinicians, promote broader differential diagnoses, and prevent premature closure.
