Learning from tragedy: the Julia Berg story

Introduction

Medical errors have been implicated as the third leading cause of death in the US [1], and diagnostic errors account for many of these [2], [3]. Diagnostic errors result from complex interactions between health care systems, providers and patients and are almost always caused by multiple factors. Previous studies have shown that most diagnostic errors result from a combination of systems and cognitive factors, often resulting in patient harm.

One of the most important responses when diagnostic errors occur is to formally analyze cases to determine why the error occurred and implement effective feedback for providers and systems to avoid similar errors in the future. This is the key to developing expertise for providers and is a nearly universal desire of patients and their families who have been harmed by diagnostic error. Here we share the story of Julia Berg, the diagnostic error that contributed to her death, and how providers, learners and systems continue to learn from Julia’s story.

Case presentation

The parent’s perspective (Dan Berg and Welcome Jerde)

We were visiting friends west of Minneapolis when our daughter Hannah called about her sister’s nosebleed. Hannah was 12; Julia was 15. The normal remedies hadn’t worked, and there was a bit of panic in her voice. We coached her through some alternative strategies and began the 40-minute drive home.

That evening, the nosebleed was under control, but her mother, Welcome, took Julia to urgent care. This wasn’t a normal nosebleed and she hadn’t felt well all week, with fatigue, a sore throat, and sometimes a low-grade fever.

After a 90-minute wait, the physician on duty determined from a urine sample that Julia had a kidney infection. As they prepared to leave the office with a prescription for an antibiotic, the doctor asked Welcome, “Does Julia look normal to you?” “Well, more or less, yes.” “Huh”, he said. “My first thought was that she might have leukemia”.

That was a Sunday. By Tuesday, Julia wasn’t getting better so Welcome took her to our pediatrician’s office. The antibiotic was changed, and we agreed to stop back on Thursday for a final checkup before a long-planned vacation.

By Thursday, Julia said she felt better and was eager to put this behind her and hit the road. As she rose from the exam table the pediatrician noticed a wince. Probing, it seemed her right side was still pretty tender.

The doctor thought we should go to Minneapolis Children’s Hospital for a CT scan, just to make sure there wasn’t something else going on. Following the scan, the doctor said that Julia didn’t have a kidney problem. Great news! But...she had a gall bladder infection. Gall bladder? In a fit, slender 15-year-old? Unusual, but not unheard of, we were told. We were persuaded that Julia should be admitted for further assessment and possible surgery.

Over the next 5 days, and a parade of attendings, residents and specialists, we learned that Julia’s platelets were low and that her liver tests were abnormal. We heard from one doctor that she didn’t appear to have any gall stones; but, according to another, she had lots, but they were small. More than once, someone looked at her chart, and said “weird”. She was scheduled for surgery. Then she wasn’t. There were a series of tests, but all were taken with the assumption that the gall bladder needed to be removed.

Her platelets stabilized, and surgery was finally scheduled for Wednesday morning. But, as a final check on Tuesday night, she was transported to the adjacent adult hospital for an ERCP, a special type of endoscopy to visualize the bile drainage system. The gastroenterologist sat down with us following the procedure. He was puzzled. There didn’t appear to be any obstructions or stones. He had performed a sphincterotomy, just in case. The results weren’t as expected, he wasn’t sure of the diagnosis, and he wasn’t sure Julia should have surgery. He said he would have a conversation with the surgeon in the morning to discuss his observations. Julia went back to her room and we went home. Our sleep was fitful; but, we reminded ourselves that we were at a great hospital with great physicians.

We were back at the hospital at 6:30 a.m., uncertain whether surgery would happen. By late morning it was scheduled, but it wasn’t until we were outside the operating room, that the surgeon confirmed that he had spoken with the gastroenterologist. They agreed that he would proceed with the cholecystectomy and do a liver biopsy to rule out other contributing factors.

Julia was a bit lightheaded as she was prepared for surgery. We met with the anesthesiologist. Consent forms were signed. We kissed our daughter and assured her that everything would be okay.

After 90 minutes the surgeon called us from the waiting room. He said the procedure had been successful and without complications. They had removed the gall bladder laparoscopically, as planned, and taken the liver biopsy. The resident surgeon would complete the procedure and let us know when Julia was in post-op. After another half hour, the resident reported in as promised. Everything looked fine, although Julia had received a unit of blood during surgery, and her hemoglobin was unusually low. Additional blood and fresh-frozen plasma had been ordered as a precaution. They would let us know when she was ready to return to her room.

Eventually Julia was the only one left in post-op, so we could enter and visit. She was talking with the nurse about the movie that she had watched that morning. Apart from a dark stool post-surgery, everything seemed under control. The post-op nurse spoke with the charge nurse on Julia’s non-surgical floor and was assured that she would monitor Julia’s status. The expected blood and FFP would be forwarded to her room on the sixth floor. We accompanied her to her room, where she could rest and recover. Dan went home to pick up Hannah for a post-surgery visit. It was 3 pm.

Welcome stayed with Julia and assisted the nurse in turning Julia as she passed another bloody stool. Once the bed was cleaned, there was another bloody stool. Julia was in pain and asked for more pain medication. Welcome noticed that when mentioning the pain, Julia pointed to her chest, not the surgical site. Welcome repeatedly asked when the blood and FFP would arrive. She was told it would be there by 4:00. At about 3:40, Dan checked in by phone. Welcome sensed that things were going wrong asked him to come back immediately.

Welcome noted that on the monitors Julia’s blood pressure was dropping and heart rate increasing. The nurse dismissed her concern and continued to focus on cleaning up the bed. Welcome continued to watch the monitor in dismay as Julia’s BP and heart rate continued to spin out of control. At Julia’s side she attempted to comfort her...“it will be okay, Honey.” Then, in her mother’s arms, Julia’s eyes rolled back. The code team descended and tried but failed to save Julia. Ninety minutes later, the surgeon, who had left the hospital after completing the surgery and was called back, delivered the news to Welcome and Dan.

Perspective of the hospital: Philip Kibort

There are certain events in a clinician’s life that are never forgotten. The death of one’s patient seldom doesn’t cause some amount of angst, self-reflection, questioning of competency and raises moral as well as ethical questions. In the management of healthcare, tragic events are also major emotional episodes that are seldom forgotten.

Julia’s death was a watershed tragedy for our hospital and a significant, if not quote “the event” that changed me personally and professionally, as a clinician and as an administrator.

In 1999 my Children’s Hospital began its journey to wanting to become one of the safest if not the safest children’s hospitals in the United States or North America. Our CEO after being told to withhold information from a family whose son died of an abnormal and poor diagnosis met with the family and basically didn’t tell them the truth. He came out of that meeting vowing never to do that again and to become a company and hospital that would be fully transparent from that point on. He went on to hire a new chief operating officer who had great expertise in patient safety and it was she who changed our culture. One of the main tenants of the organization’s cultural change was to be fully transparent with all families whenever a medical error occurred.

One of my responsibilities as chief medical officer was to be the person from administration who would meet with the families to apologize when an error that we created caused harm to their child. It was the most difficult part of my job, but one that was the right thing to do. Our commitment as an organization was always to tell the family right away what we knew.

Julia’s death impacted me as a physician and as Chief Medical Officer. It was the first time in my 30-year career in medicine that I realized and understood that a cognitive “bias” caused the death of a patient under my responsibility. It was immediately after her unfortunate death that I became interested and committed to personally learn and understand the impact of cognitive biases on diagnostic errors. At the same time as CMO it became my goal and commitment to begin changing the culture of both our medical and clinical staff to understand how our cognitive biases are potentially hurting our pediatric patients. In addition, because of the Bergs’ willingness to work with us at Children’s Minnesota we were able to use Julia’s story multiple times over the years as a teaching opportunity for all our staff and trainees.

Julia’s death changed me drastically as a physician and as an administrator. Julia’s parents, because of their unbelievable resiliency, kindness and compassion, allowed themselves, myself, and our Children’s Hospital to heal together after her death.

Discussion

This case involved a fatal diagnostic error: a previously-healthy 15-year-old girl was diagnosed with cholangitis and suspected biliary obstruction, whereas the autopsy-confirmed diagnosis was something completely different – active EBV infection (mononucleosis).

Although other causative factors almost certainly contributed, the most notable factor in this case is that none of the providers seemed to have noticed or considered the importance of the patient’s lymphocytosis. Atypical lymphocytes are the hallmark of mononucleosis and the disease owes its name to this key manifestation [6]. None of the medicine, surgical or gastroenterology notes commented on either the leukocytosis, the marked lymphocytosis, or the atypical cells. On 3 days, there are notes in the laboratory results section documenting that the elevated white blood cell count was called up to the ward and reported to nursing staff. A question remains as to whether informing nursing staff by the laboratory personnel is sufficient to meet the requirement of informing a responsible care provider of a result. Nursing staff, especially those who are on general surgery floors, may not have been conversant with abnormal hematology results. They may have focused only on the elevated white blood count, a common enough finding in the setting of infectious conditions including cholecystitis. Whether laboratory staff highlighted the more unusual and critical finding of lymphocytosis and in particular the presence of atypical lymphocytes is not known. There is no documentation as to whether the physicians were ever informed, either by the laboratory staff or nursing staff, or if they were, their thoughts on this. Another potential causative factor may be the quality of laboratory report itself. Traditional laboratory reports list numeric values (e.g. hemoglobin 12.2) or short descriptors (80% lymphocytes) with control values. Interpretive results are limited to a small number of circumstances. An interpretive result, where the report said, “atypical lymphocytosis – this finding is commonly seen in the setting of viral infections, most commonly infectious mononucleosis”, may have triggered a different response from the patient’s care providers.

Mononucleosis classically presents with sore throat, fevers and a profound sense of fatigue, especially in teenagers and young adults. All of these symptoms were noted in the ambulatory care visits but were downplayed or lost completely when Julia was admitted to the hospital; the inpatient diagnostic efforts were re-directed to understand her cholestasis and direct hyperbilirubinemia. Acalculous cholecystitis is a rare presentation of Epstein-Barr infection, and though now well documented in the medical literature [7], [8], [9], [10], [11], [12], [13], [14], was not considered by any of the clinicians ante-mortem. Further, the initial reading of the ultrasound mentioned that the gallbladder abnormalities could have been secondary to another, possibly hepatic process.

Based on the combined findings of fever, right-upper quadrant pain and tenderness, and markedly abnormal bilirubin and LFTs, Julia was thought to have obstructive biliary disease requiring surgical intervention. Surgical intervention would have been clearly indicated if stones had been documented and the patient was clinically deteriorating, but neither of these were present. It is possible, given the surgeon’s note, that they mistakenly believed Julia did have gallstones, but one then wonders why this was not aired at the conference with the gastroenterologist pre-operatively.

Trying to understand the diagnostic error

When analyzing diagnostic errors, it is fundamental to use a standard approach that seeks to delve systematically into the error’s causative factors and minimizes bias. In other areas of patient safety, this has been accomplished through the use of a “fishbone” diagram that categorizes causative factors that, in combination, lead to the error as an outcome. This approach has been successfully adapted to analyze diagnostic errors [27]. Figure 1 uses a modified fishbone to designate some of the causative factors for the diagnostic error in Julia’s case. The diagram illustrates the many different potential sources of error in this case and also the complexity of the diagnostic process.

Figure 1:

A fishbone diagram illustrating factors that might have contributed to the diagnostic error, missing the diagnosis of EB virus infection.

The National Academy of Medicine described diagnosis as a process and most diagnostic errors involve some breakdown in one or more steps of this process [2], [28]. The key failing in this case lies squarely in failing to appreciate salient laboratory data; Julia’s atypical lymphocytosis was the clue to unlocking the correct diagnosis. Closer attention to the history might also have triggered consideration of a viral illness, given the complaints of sore throat, fever and prominent fatigue. Further, it appears that the providers focused entirely on the biliary tract without stepping back to ask, “What causes fever, sore throat, lymphocytosis and thrombocytopenia in a teenager?”

Cognitive shortcomings can be identified retrospectively in most cases of diagnostic error [29]. Problems that could have conceivably contributed to the diagnostic error in the case of Julia Berg included:

1. Framing and context errors – from the moment of admission, the suggested diagnosis was cholecystitis, most likely due to gallstones. Framing the case as being a “GI” problem may have discouraged reconsidering it in an infectious disease context.

2. Diagnostic momentum – findings that could have led to a different diagnosis were discounted, if not completely ignored (not volitionally).

3. Premature closure – because it explained many features of the case, the diagnosis of acalculous cholecystitis was plausible, and no other entities were seriously mentioned. An assigned diagnosis tends to stop further thinking. Was a comprehensive differential diagnosis constructed at any point? In a study by Singh and colleagues of diagnostic error cases, there was no documentation of the differential diagnosis 80% of the time [30]. Similarly, premature closure may have contributed to a curtailed autopsy. Because the “proximate” cause of death (in pathology, that which in natural and continuous sequence led to the series of events that eventually resulted in the patient’s demise) – EBV infection – was discovered early, efforts to identify the “immediate” cause of death, the sudden cardiovascular collapse, were not considered, explorations that often requires detailed biochemical, hematologic and microscopic assessments.

Similarly, system factors can be identified in most cases of diagnostic error, and there were several such factors to consider in this case:

1. Residency supervision: there are only a few notes by the medical, GI and surgical attendings. Would closer supervision have made a difference?

2. Decision support: many effective and easy-to-use web-based tools are available to assist with differential diagnosis. Were these available?

3. Test result communication: the ward staff were notified of Julia’s highly abnormal white cell count on at least three occasions. Were the results communicated to the appropriate decision maker with respect to diagnostic responsibility? Is it reasonable to deliver results and receive results of some urgency by non-diagnostic care providers? Were the results ever communicated and acknowledged by the physician staff?

4. Laboratory involvement: was the laboratory pathologist consulted about the meaning of the “abnormal lab finding”? Was anyone in the lab curious about such extreme lymphocytosis in a non-cancer patient? Should the lab report have included a note that said “atypical lymphocytes in a teenager is most commonly associated with EBV infection”?

5. Multidisciplinary huddle/discussion: multidisciplinary discussions (tumor boards) are common for cancer diagnosis, where specialty expertise needs to come together in the context of the whole. A parallel process, allowing communication with pathologists, hematologists, surgeons, the ward team and the family may have helped paint a complete picture.

The case also prominently brings to attention how the family and caregivers can contribute to patient safety. Julia’s mother noticed Julia’s progressive deterioration in the recovery room, but her observations were dismissed by the staff. It is impossible to know if a more receptive response would have made a difference. Acknowledging the voice of the patient and the family is now thought to be one of the most powerful interventions available to improve safety, by creating a new safety net to catch deterioration and errors. Accordingly, many hospitals now encourage patients and families to participate in “rapid response” programs [31], and the efforts of the Children’s Hospital to promote patient engagement are exemplary.

In summary, the death of Julia Berg illustrates the multifactorial origins of diagnostic error and identifies many opportunities for improving the diagnostic process going forward. The case provides lessons for providers and healthcare systems. This feedback is vital, so that when errors occur, we can promise families that “We are doing everything humanly possible to learn from this case so this never happens again”.