Types of diagnostic errors in neurological emergencies in the emergency department

Introduction

Diagnostic errors have been estimated to occur in 0.6%–12% of emergency department (ED) patients [1, 2]. Failure to make a diagnosis can result in patient harm and is a leading source of medical malpractice claims [3, 4]. Conversely, excessive workups and over-diagnosis may also lead to patient harm and inefficient resource utilization. Neurological emergencies may pose particular diagnostic challenges, with one study reporting that misdiagnosis or diagnostic uncertainty occurs in over one-third of ED patients presenting with neurological problems [5].

Although the reasons for this are complex, they can be organized into three categories: knowledge gaps, cognitive errors, and systems-based errors [6, 7]. Patients with neurological emergencies frequently present with “atypical” symptoms that lack the “classic triads” on which traditional medical education focuses [7–14]. Lack of understanding of the limitations of radiologic testing further complicates the matter. The resultant knowledge gaps may result in misdiagnosis. Cognitive strategies and heuristics, which have been previously described in the literature, are often essential for efficient decision making in the busy ED but when they fail, cognitive errors occur [15, 16]. Systems-based factors such as resource limitations, discordance between consulting and emergency physicians’ impressions and radiology interpretation errors also contribute. The prevalence of these types of errors as it applies to the diagnosis of neurological emergencies is unknown.

The goal of study was to identify the types of errors that lead to missed diagnosis and delays in diagnosis of patients with neurological emergencies by reviewing ED quality assurance (QA) records. As a pilot study, we also aim to determine the feasibility of using our institution’s extensive QA database as a means of retrospectively analyzing causes of misdiagnosis and error with the hope of conducting future, larger scale studies using this methodology.

Materials and methods

Results

During the study period, 1168 cases reviewed by the QA Committee had both documented medical record numbers and case descriptions in the QA meeting minutes. Of these, 42 cases were found to include a neurological misdiagnosis that met inclusion criteria. We excluded 13 cases in which no error was found. Twenty-nine cases were determined to have had a diagnostic error in the ED management of the patient (Figure 1). Two of these cases had two diagnostic errors made over multiple ED visits for the same presentation. The patient characteristics and missed diagnoses are shown in Table 1.

Figure 1

Process for identification of cases meeting inclusion criteria and determination of error type.

Knowledge gaps were found to be the most common type of error leading to misdiagnosis. The distribution of error types and subtypes is shown in Table 2. Examples of cases and associated errors are shown in Table 3. Two cases included two errors (knowledge gap, systems-based issue and knowledge gap, cognitive error) over multiple ED visits and these are included in the calculated proportions. There were no cases in which all three reviewers disagreed on the type of error. In five cases, however, one reviewer disagreed with the other two. The Fleiss coefficient for inter-rater reliability was 0.89 (95% confidence interval 0.85 to 0.93).

The types of failed cognitive heuristics and strategies included diagnostic anchoring, diagnostic momentum, premature closure, base rate neglect, and search satisfying. The types of systems-based issues were radiology resident misreads and delay in obtaining imaging study. Radiology resident misread was the most common systems-based issue, accounting for 87.5% of all systems-based issues (Table 4).

In regards to the time of year when the errors were made, there was no statistically significant difference in error rates between the first and second halves of the academic year or among quartiles. In 31.0% of cases with diagnostic errors, a neurology or neurosurgery consultation had been obtained in the ED. Data regarding the use of a consult in the ED was not available for two cases (6.9%).

Discussion

Although we cannot make any conclusions about the frequency of misdiagnoses of patients with neurological problems, we have identified some of the reasons behind misdiagnosis, some of which suggest certain solutions. The most common causes of error in our study were knowledge gaps, which result from education deficits about the conditions that were misdiagnosed. Focusing on improving physician education about neurological emergencies may decrease this type of error. A review of malpractice claims found that residents were involved in 56% of the cases and that in 75% of these cases, trainees had the highest contributory rate of any personnel involved [4]. Because the residents and attending emergency physicians are jointly involved in the decision-making process, our data do not allow us to assign these errors to one group or the other. Nevertheless, improving education about neurological emergencies at the resident level could help reduce these knowledge gaps down the road. Efforts to maximize resident supervision (which is in part a systems-based issue) may also improve quality of care.

More generally, a survey of emergency medicine program directors revealed that only 17% of residency programs require a rotation in neurology and 15% required a rotation in neurosurgery. Program directors indicated that the primary method of educating EM residents on neurological emergencies was through didactics [18]. Residency programs should consider expanding clinical rotations or improving didactic lectures on such topics in order to decrease knowledge gaps at the resident level. The contribution of knowledge gaps among attending emergency physicians is unknown and more difficult to study.

Cognitive errors accounted for 29% of misdiagnoses in this study. While emergency physicians will always use some sort of cognitive heuristics with their clinical acumen to make judgments, there may be strategies to avoid those that lead to error. Croskerry describes cognitive de-biasing through metacognition, “the process by which we reflect upon, and have the option of regulating what we are thinking” [16]. Reflection on one’s thought process can often be difficult in a busy ED, as emergency physicians are faced with constant distractions, variable acuity, and the need for rapid decision-making [15, 16]. Sign out at shift change may provide one instance where metacognition and reevaluation can occur; encouraging the oncoming physician to question a plan that does not sound right may decrease these errors. Perhaps a more realistic tool to avoid cognitive errors is the “diagnostic time-out” [19]. Patient safety has been shown to improve in surgical and procedural suites because of such time outs [20]. The ED team can implement a similar process by taking a “diagnostic time-out” to review the data and consider an expanded differential diagnosis. Such brief exercises may improve diagnostic accuracy and have been suggested for coma patients of unclear etiology [21]. Finally, acknowledging one’s own mistakes encourages self-reflection and allows clinicians to teach others about the error [19].

Systems-based errors accounted for 29% of misdiagnosis in our study, mostly due to radiology resident misreads. At our institution, radiology residents release preliminary reports on all ED imaging except for neurological magnetic resonance imaging (MRI) studies and emergency physicians base their decision making on these initial reads. For patients presenting on overnight shifts, patient dispositions are usually made prior to the attending radiologist report being available. This practice is similar to many other academic centers [22–24]. Of note, during the study period, our institutional policy changed from one that allowed radiology residents to release preliminary neurological MRI reports for decision making to our current policy given the large number of misreads. A report from our institution found the sensitivity for on call residents’ detection of aneurysms on CT angiography studies of patients with question subarachnoid hemorrhage was as low as 62% [25]. Studies from other institutions have found discrepancy rates of resident and attending physicians’ interpretations of neurological imaging to range from 8.4% to 13.5% [23, 24]. One study found the miss rate of radiology residents to be 5.2% per case on CT angiography studies of the head and neck [26]. Neuroimaging is the modality of choice for diagnosing strokes, aneurysms, and arterial dissections and these are the diagnoses that accounted for the majority of misses in our study. Emergency physicians should consider requiring real-time attending radiologist reads on all ED neuroimaging in order to decrease this type of error.

The two most common misses in our study had to do with diagnosis of stroke, particularly cerebellar stroke, and symptomatic aneurysms. This is consistent with existing literature that describes the high miss rate of these diagnoses and the difficulty clinicians face when making these diagnoses [7, 27–29]. All cases of missed cerebellar strokes in our study were due to cognitive errors or knowledge gaps, which further supports the need for increased education on this diagnosis. While radiology misreads accounted for several cases of missed cerebrovascular pathology, it is also important for clinicians to understand the limitations of even the most advanced neuroimaging tests in making these diagnoses, particularly in the early hours after an event [27–30].

Conclusions

In this study, knowledge gaps, cognitive errors, and systems-based errors all contributed to the misdiagnosis of neurological emergencies. The two most common mistakes that suggest specific areas for improvement are missed diagnosis of cerebellar stroke and erroneous radiology resident interpretations of neurological imaging. Understanding these types of errors can enable emergency physicians to develop possible solutions to avoid them in the future.