The Case

A 9-year-old girl with a past medical history of cerebral palsy, developmental delay, and epilepsy presented to the emergency department (ED) for increased frequency of seizures which lasted about 5 minutes, necessitating a rescue dose of diazepam nasal spray. The patient was stabilized in the ED but developed hypoxic respiratory failure requiring endotracheal intubation, sedation, and mechanical ventilation. The pediatric neurology team was involved in her care starting on the second hospital day. Their initial recommendations involved some changes in her anti-epileptic medications, but on hospital day 7, they recommended testing for acetylcholine receptor antibodies, muscle-specific tyrosine kinase (MuSK) antibody, anti-ganglioside antibodies, creatine phosphokinase, and aldolase due to ongoing symptoms.

Most of these tests had to be sent to an external laboratory and the results returned intermittently over several weeks. On hospital day 13 (after the patient was transferred out of intensive care), the acetylcholine receptor antibody test results returned markedly elevated at 302 nmol/L (normal is <0.5 nmol/L), which is concerning for myasthenia gravis. No hospital notes reflected this positive result, and all subsequent daily progress notes included the phrase, “awaiting lab results for neuromuscular workup”. The patient was discharged on hospital day 18, with no mention of the positive result in the discharge summary. The patient followed up with her primary care pediatrician (PCP) and pediatric neurologist 13 days after hospital discharge. Neither the neuromuscular workup nor the positive acetylcholine receptor antibody test was noted at either visit.

Several days later, the patient returned to the ED for acute respiratory failure and cardiac arrest, requiring pre-hospital cardiopulmonary resuscitation, intubation, and respiratory support. She had a prolonged hospital course and was discharged against medical advice on hospital day 15. After thorough review of her prior laboratory results, her final diagnosis was respiratory failure secondary to myasthenia gravis crisis. She received intravenous immunoglobulin and corticosteroids, with gradual improvement in her acetylcholine receptor antibody levels.

The Commentary

By Zachary Chaffin, MD

Background

Children and youth with special healthcare needs (CYSHCN) represent a vulnerable subset of children with at least one chronic physical, developmental, or behavioral condition, who therefore have more frequent than usual interactions with the healthcare system. They are at increased risk of medical errors during hospital stays, and this risk scales with the number of their chronic conditions.¹ In this case, the delayed diagnosis of myasthenia gravis is illustrative of common challenges faced by providers caring for CYSHCN. The laboratory finding that established this diagnosis was available in the electronic health record (EHR), but it was invisible to multiple teams of providers across multiple phases of care due to issues related to the EHR itself, challenges in clinical reasoning, and hazards in the workflow around transitions of care.

Approaches to Improving Patient Safety

Effective use of electronic health records (EHRs)

One of the strengths of the EHR is its ability to track and flag abnormal laboratory results for review by a provider. This is especially effective for common laboratory studies: most EHRs use displays that require few clicks and clearly denote abnormalities – sometimes even the severity of the abnormality. However, for less common studies that must be sent to an off-site laboratory (like some of the antibody serology studies for the patient in this case), the results may be initially received as hard copies that must be scanned manually into the EHR. In this case, the results are often not flagged as abnormal by the EHR itself; review requires a provider to search for the result and manually open a media file on an appropriate hardware platform. Finding such results is straightforward if a provider is looking for them specifically, but they easily blend into the background during a routine review of new laboratory results.

As a backup, most EHRs have a system of sending alerts directly to providers to notify them of new results. Ideally, these alerts are appropriate, timely, and displayed to a provider with the necessary expertise to interpret them. Most EHR systems route alerts to the ordering provider by default and may require that provider to mark when they review the results. Unfortunately, these alerts do not always have the intended effect: one study in a primary care setting found no association between whether an alert for abnormal result was acknowledged by a provider and subsequent timely follow-up on the result.² Seeing an abnormal result is not the same as digesting it clinically, and busy providers are often reviewing dozens of alerts in one sitting. In an academic health system, the ordering provider is often a trainee, who may be at high risk for cognitive overload and may have rotated to a new service or site by the time a result is available in the EHR. In this case, the patient had been transferred out of the ICU when the diagnostic studies became available, which means that the provider who ordered the studies, and who may have received an alert when the result was available, was no longer directly caring for the patient. Indeed, if this trainee was offsite and relying on a health information exchange platform, it may have been impossible for them to view the scanned result.

The problem of EHR “mis-routing” has been previously identified as a root cause of medical errors and, in one review of IT-related malpractice claims, was identified as a contributing problem in about 10% of all cases reviewed.³ There may be an opportunity for creating a process in the EHR that allows routing of some results for review by both the ordering or attending provider and the recommending specialty service. Such a process may be especially useful for tests that are typically ordered and acted upon by a single specialty service, such as tests for rare neuromuscular disorders.

Much has been written about the use of copy-and-paste in daily progress notes, and without belaboring the point, one review of EHR contributions to diagnostic error found multiple studies detailing how poorly maintained problem lists or copy-pasted notes can hide important details that contribute to diagnostic error or delays.⁴ In this case, the phrase “awaiting lab results for neuromuscular workup” falsely suggested that none of the results were available. This phrase serves the same notational purpose as “continue to monitor”; it provides the author cognitive relief of an accomplished task, but it does not reflect the author’s thought process or actual review of the findings in question. One potential solution is to create a new display workflow in the EHR dedicated to the diagnostic process that could extract recommendations from specialist notes and document a running list of which tests have been sent and which have resulted.

Clinical reasoning

Clinical reasoning, the process of synthesizing raw clinical information and evaluating that information against known illness scripts, often involves problem representation:⁵ a succinct narrative that encapsulates the key points of the patient’s presentation and available data. The subsequent diagnostic plan is only as good as this story that providers tell themselves and each other. If a previously healthy patient has ongoing, unexplained weakness that prolonged her ICU stay and made her difficult to extubate, providers should be vigilant about looking for the results of a “neuromuscular workup.” If the patient is well-known to the service as having a seizure disorder and cerebral palsy and was admitted for increased seizure frequency, then her pre-existing medical problems create a fog that may blind providers to new abnormalities. There is special risk for this type of anchoring bias in children and youth with special healthcare needs, as a problem representation that starts with “cerebral palsy, developmental delay, and epilepsy” admitted for “increased frequency of seizures” focuses attention on a well-recognized pattern. Many children with pre-existing seizure disorders are admitted to intensive care with increased seizure frequency and associated respiratory failure – “Got it, let’s move on,” says our brain. Breaking out of this system 1, automatic thinking⁶ is a significant challenge, and may not always be necessary. Interestingly, in the study of EHR-generated alerts for abnormal laboratory results mentioned above, alerts for results that indicated a new diagnosis were more likely to lack timely follow-up than results consistent with a pre-existing disorder.² Even when the EHR appropriately alerts us that something new and different may be happening with our patient, changing our internal narrative and giving the patient a new diagnosis can be a significant cognitive task. The simple addition of the phrase “with unexplained weakness leading to delayed extubation” points our brain in a different direction and breaks the pattern recognition. Harnessing the power of problem representation can help us continue to think critically about our patients and avoid anchoring and premature closure.

Transitions of care

One of the sobering realities of this case is that the laboratory test result suggesting myasthenia gravis, although available in the EHR, was invisible to multiple teams of providers in the intensive care unit, ward, and two outpatient clinics. It is well-documented that transitions of care are a time of increased risk for medical error, as they are always associated with a handoff. Communication failures are the most common cause of sentinel events, and up to half of communication failures occur during patient handoffs.⁷ In the inpatient setting, hand-offs are both verbal and written, so when this child transitioned from the ICU to the ward, the providers would have spoken verbally and written a transfer summary. Though the laboratory results were not yet available at the time of transfer, that handoff should have included a clear description of the pending studies. At the time of discharge from the hospital, there is usually not a verbal hand-off; instead, outpatient providers rely on the discharge summary to learn the details of the hospital course. Without specific direction to review tests with pending results at the time of discharge, the primary care physician and outpatient neurologist would have only the narrative description of the hospital stay available to them. Furthermore, in the common scenario in which the hospital is in a different medical system from the child’s PCP, lack of EHR interoperability may be a limiting factor, and the PCP may be unable to directly view results of studies sent during the hospital stay. Any loose ends left at the time of discharge, therefore, must be clearly defined in the discharge summary just as in any other patient handoff. One solution, recently adopted by our own institution, is to adjust the standardized discharge summary template so that it begins with “Important Comments to Subsequent Providers” before the description of the hospital stay.

Take Home Points

• Improved EHR routing systems for results of send-out laboratory tests may prevent delays in diagnosis; possibilities include routing to the ordering provider, the current primary provider, the recommending specialty service, and at least one attending physician.
• Diagnostic workups that include multiple send-out laboratory tests are often difficult to follow across phases of care. A new EHR display focused on the diagnostic workup could be designed to track and group laboratory tests based on the differential diagnosis.
• Careful constructions of problem presentations, particularly in children with chronic medical conditions, in combination with standardized handoffs and discharge summaries that prioritize to-do lists and clarify loose ends may help us coordinate optimal follow-up care for patients.

Zack Chaffin, MD
Health Sciences Assistant Clinical Professor
Department of Pediatrics, Critical Care Medicine
UC Davis Health
zrchaffin@ucdavis.edu

References

1. Ahuja N, Zhao W, Xiang H. Medical errors in US pediatric inpatients with chronic conditions. Pediatrics. 2012;130(4). [Available at]
2. Singh H, Thomas EJ, Sittig DF, et al. Notification of abnormal lab test results in an electronic medical record: Do any safety concerns remain? Am J Med. 2010;123(3):238-244. [Free full text]
3. Graber ML, Siegal D, Riah H, et al. Electronic health record-related events in medical malpractice claims. J Patient Saf. 2019;15(2):77-85. [Free full text]
4. Dixit RA, Boxley CL, Samuel S, et al. Electronic health record use issues and diagnostic error: a scoping review and framework. J Patient Saf. 2023;19(1):e25-e30. [Available at]
5. Chang RW, Bordage G, Connell KJ. The importance of early problem representation during case presentations. Acad Med. 1998;73(10 Suppl):S109-S111. [Free full text]
6. Tay SW, Ryan P, Ryan CA. Systems 1 and 2 thinking processes and cognitive reflection testing in medical students. Can Med Educ J. 2016;7(2):e97-e103. [Free full text]
7. Blazin LJ, Sitthi-Amorn J, Hoffman JM, et al. Improving patient handoffs and transitions through adaptation and implementation of I-PASS across multiple handoff settings. Pediatr Qual Saf. 2020;5(4):E323. [Free full text]