The COVID trap: pediatric diagnostic errors in a pandemic world

Case 1

A previously healthy 6-year-old male presented to the Emergency Department (ED) of a large children’s hospital located in the mid-Atlantic region of the United States in May 2020 with approximately 3 days of fever, bilateral lower extremity pain with inability to walk, rash, diarrhea, and vomiting. At presentation, the patient appeared irritable but alert. Vital signs were notable for temperature of 38.6 °C, mild tachycardia, and elevated blood pressure for age. His exam was notable for dry, cracked lips, erythematous tongue, injection of nasal mucosa, erythrodermic rash of groin and left arm, and tenderness to palpation of the large muscle groups in the lower extremities. Documentation intermittently describes migratory joint pain with palpation and range of motion; however, documentation of the joint exam was missing or less detailed in different clinician notes during a similar time period. Initial laboratory studies revealed a normal white blood cell count with 88% neutrophils on differential, mild anemia, and mild thrombocytopenia. Inflammatory markers were high with erythrocyte sedimentation rate of 49 mm/h (reference range: 0–20 mm/h) and C-reactive protein of 52 mg/dL (reference range: 0.0–0.9 mg/dL). He was also noted to have mild hyponatremia and mildly elevated liver enzymes. The laboratory-developed nasopharyngeal multiplex reverse-transcription polymerase chain reaction (PCR) for respiratory viruses was negative and SARS-CoV-2 nasopharyngeal PCR was negative. He was started on vancomycin, ceftriaxone, and clindamycin. He was also given one dose of doxycycline.

The patient was initially admitted to the general pediatrics floor, but due to concerns of hyponatremia and need for ongoing clinical assessment, he was rapidly transferred to the pediatric intensive care unit (ICU). He did not have any hemodynamic instability. Due to suspicion of MIS-C as a diagnosis, Rheumatology was the first consult team to be called. Immunomodulatory therapies (intravenous immunoglobulin and steroids) were discussed, but not initiated. Notably, the initial differential diagnosis documented in the ED was broad and included atypical Kawasaki, MIS-C, streptococcal or staphylococcal toxic shock syndrome, bacterial sepsis, and tick-borne illnesses. However, with each transfer (from ED to general floor to ICU) the documentation increasingly focused on MIS-C as a working diagnosis.

Around 12 h after admission, bilateral knee radiographs were obtained that revealed unilateral right knee effusion. Orthopedics was consulted and performed an arthrocentesis with synovial fluid analysis revealing 125,139 white blood cells with 73% neutrophils. An ultrasound of the bilateral hips was obtained and revealed unilateral left hip effusion. He was taken for operative washout of the left hip and right knee joints. The Infectious Diseases team was consulted at about 16 h after admission. Blood cultures and synovial cultures from left hip and right knee grew Streptococcus pyogenes and he was diagnosed with invasive group A streptococcal disease presenting with multiple septic joints and likely toxin-mediated rash without shock.

Case commentary

This case presents a series of systems factors and cognitive factors (depicted in Figure 1) that contributed to delayed diagnosis of invasive group A streptococcal disease with septic arthritis. Of the systems factors noted, we see potential for improvement in communication systems to buffer the inherent pitfalls of telephone consultation and multiple transfers including potential loss of critical case information, reliance on communicated physical exam findings rather than primary exam, and misunderstanding of key findings.

Figure 1:

Fishbone diagram analysis of cognitive and systems factors contributing to diagnostic error.

This case highlights important cognitive factors contributing to delayed diagnosis including availability bias, premature closure, anchoring, diagnostic momentum, and confirmation bias [13]. This patient presented to the hospital at a time when MIS-C, a new clinical entity, was being reported; several children at this hospital had already been treated for the syndrome. Although MIS-C was still a rare diagnosis and a diagnosis of exclusion, the heightened interest in this emerging disease likely contributed to a tendency to recall more easily this diagnosis, set the stage for anchoring bias, and led to premature closure on this as the primary working diagnosis. Furthermore, many features consistent with severe infections such as group A streptococcal disease (hyponatremia, fever, rash, gastrointestinal symptoms, myalgias, and elevated laboratory markers of inflammation) were similar to features of MIS-C allowing for a confirmation bias effect and less emphasis on less consistent features (such as specific joint pain, erythrodermic rash). Anchoring on MIS-C also led to less focus on the initial chief complaint, refusal to walk in the setting of fever, which in pediatrics should trigger concern or septic arthritis and requires prompt time-sensitive surgical management.

The particular power of availability bias over clinician differential diagnosis building is palpable in the narrative of this case. The context of the pandemic appeared crucial to driving the diagnostic momentum seen in this case and leading to premature closure on MIS-C. While, ultimately, the patient received appropriate antimicrobials and operative debridement of the affected joints, the time lapse still highlights areas of improvement in our diagnostic process, particularly as related to the pandemic.

Case 2

A 6-day-old full-term female infant born via spontaneous vaginal delivery was admitted to the pediatric ICU in June 2020 for respiratory and fulminant liver failure. The initial working diagnosis was acute COVID-19 vs. MIS-C. Prenatal history was notable for maternal group B streptococcus positive status. Two months prior to delivery, the mother had been diagnosed with COVID-19 by SARS-CoV-2 PCR in the setting of self-limited respiratory symptoms. The mother was noted to have a positive SARS-CoV-2 PCR without symptoms on screening at time of admission to the labor and delivery unit. After delivery, the neonate was tested for SARS-CoV-2 via PCR and was found to be negative. She was discharged from the newborn nursery on day of life (DOL) 2 without complications.

On DOL 5, the baby’s parents noted that she had decreased oral intake and activity for which they sought medical care. At the time of presentation to a local ED, the baby was noted to be ill-appearing with progressive apnea requiring intubation. Blood cultures were obtained and she was treated empirically with ampicillin, gentamicin, and acyclovir. Initial laboratory findings were notable for severe acidosis (pH <6.9), normal white blood cell count with 22% premature neutrophils on differential count, elevated liver enzymes with aspartate transaminase of 2,638 U/L (reference range: 30–140 U/L) and alanine aminotransferase of 318 U/L (reference range: 5–45 U/L), and quickly developed signs of disseminated intravascular coagulation. She was transferred to a large children’s hospital for further care.

Upon arrival at the children’s hospital, additional laboratory studies revealed markedly elevated ferritin (>200,000 ng/mL; reference range: 99.6–717.0 ng/mL) and continued impaired renal function and liver failure. Due to concern for SARS-CoV-2 infection with secondary hemophagocytic lymphohistiocytosis triggered by the anti-viral inflammatory process, the hospital’s specialized multidisciplinary consultation team for immune dysregulation was consulted. As a result of this consultation and concern for hemophagocytic lymphohistiocytosis, dexamethasone was given on DOL 6. Later that day, herpes simplex virus (HSV) PCR from serum was positive and a diagnosis of disseminated neonatal HSV disease was made. Further steroid administration was held and the neonate was continued on IV acyclovir.

Of note, concerns for neonatal COVID-19 continued; a total of four SARS-CoV-2 PCRs were collected from both nasopharyngeal and tracheal aspirate sources from DOL 2 through DOL 6, all of which were negative. Despite these test results, she remained on expanded, COVID-19 transmission-based precautions for about 24 h.

Case commentary

In this case, systems factors contributed to diagnostic delay and erroneous decision to administer steroids, which are contraindicated in disseminated neonatal HSV disease. Specific opportunities were identified for improved communication, as the patient required multiple transfers of care (both between and within hospitals) and telephone (rather than in-person) subspecialty consultation. Local SARS-CoV-2 disease activity likely contributed to availability bias of a SARS-CoV-2 related diagnosis. In particular, we highlight the focus on positive maternal SARS-CoV-2 testing, prior maternal mild COVID-19 disease, documentation of high suspicion for COVID-19 in transfer notes of the neonate, and repetitive SARS-CoV-2 PCR testing of the neonate despite negative results. These factors originate from new processes of routine testing for SARS-CoV-2, the existence of the COVID-19 pandemic, and the emergence of MIS-C that does require immunomodulatory therapy (in contrast to neonatal HSV disease) and led to the cognitive factors of availability bias, diagnostic momentum, leading to premature closure, and continued confirmation bias [13].

This case builds on the availability bias seen in case 1 and particularly highlights the interplay of communication systems (including transfer notes and telephone consultations) with cognitive factors of diagnostic momentum and confirmation bias (Figure 1). The mutual reinforcements of systems and cognitive factors in this case suggests opportunities for systems level interventions to counteract cognitive bias in the diagnostic process. While this case occurred during the COVID-19 pandemic, these interactions are not unique to this moment.

Case 3

A 3-year-old male presented to a community ED in the mid-Atlantic United States in April 2020 with septic shock after 10 days of pale appearance, nosebleeds, fever, and altered mental status with hallucinations. Prior to ED presentation there had been one telephone encounter (two days prior) and one ambulatory clinic visit (one day prior), both of which resulted in a working diagnosis of a viral syndrome. Notably, this clinic visit was identified to be a “drive-through” visit with minimal provider contact. The patient’s parents also expressed hesitation in seeking medical care due to fear of exposure to COVID-19 in the healthcare setting. During the ED visit and ICU admission at the community hospital, he was diagnosed with acute B-cell lymphoblastic leukemia with severe pancytopenia (white blood cell count <2 K/uL, hemoglobin <3 g/dL, and platelets <5 K/uL) and septic shock secondary to group G streptococcal bloodstream infection with subsequent hyperkalemic cardiac arrest. He was transferred to a tertiary care children’s hospital two days after initial presentation for continued management of multiorgan system failure. Several days later, due to the severity of neurologic injury in the setting of cardiac arrest, the decision was made to withdraw life-sustaining measures.

Case commentary

This case presents a striking array of systems factors, primarily related to seeking care in the setting of a pandemic and with active shelter-in-place orders (Figure 1). In the diagnostic error model presented by Gandhi and Singh, the avoidance of seeking healthcare due to the pandemic and fear of exposure to COVID-19, could be classified as a variant of “acute collateral” [8]. These concerns of COVID-19 exposure may have contributed to this child’s delayed diagnosis of acute leukemia, which in turn likely resulted in a more severe disease presentation with septic shock, severe tumor lysis syndrome, and death. The ultimate outcome highlights downstream effects of necessary public health measures, shifts in alternative modes of care (i.e. phone and telehealth appointments), and patient and family perceptions of the healthcare field and media communication surrounding safety and the pandemic.

The cognitive factors at play in the outpatient visits with this case emphasize the typical cognitive biases that can happen even in the absence of a pandemic including availability bias and premature closure on a diagnosis that does not fit all the presenting symptoms [13]. Additionally, there may be an added contributor of “wellness bias,” particularly in the case of a previously healthy child in the ambulatory pediatric clinic setting that may prevent consideration of more serious etiologies for presenting symptoms [14, 15].