When the Unknown Is Unknowable: Confronting Diagnostic Uncertainty

Nathalia had an unremarkable pregnancy. Her daughter Evangelina was born at 35 weeks via “crash” cesarean section secondary to repeated late decelerations. She was transferred to a level 4 NICU shortly after birth because of hypovolemic shock caused by multiple hemorrhagic lesions throughout the body.

At birth, Evangelina was found to have numerous large, dark red-purple hemorrhagic masses on her head, mouth, right axilla, and other sites. The cranial mass was bleeding profusely from iatrogenic laceration. She developed hypovolemic shock and severe metabolic acidosis, requiring extensive volume and pressor support. Evangelina underwent two resuscitation “code” events before being transferred. On arrival, Evangelina was encephalopathic, anuric, and had an initial blood pH of 6.61. Therapeutic hypothermia could not be offered because of ongoing massive bleeding. She received inhaled nitric oxide for severe persistent pulmonary hypertension.

Based on the clinical presentation at admission, the differential diagnosis for Evangelia’s condition included multifocal lymphangioendotheliomatosis, diffuse neonatal hemangiomatosis, and congenital vascular malformation of unknown etiology. Further diagnostic testing was not possible because Evangelina’s clinical instability precluded magnetic resonance imaging (MRI) or angiography, and disseminated intravascular coagulation prevented a biopsy. Despite consultation with more than 10 subspecialists, the definitive diagnosis remained unknown.

The attending physician was conflicted about how to tell Nathalia and Daniel, Evangelina’s parents, that he did not think their infant daughter would live when he did not know the cause of her severe condition. How would you advise the attending to proceed?

A newborn infant, not even one week of age, is grievously ill and close to death. The underlying cause of the illness, though, remains unknown. The infant’s physician, tasked with conveying to the parents the consensus opinion of all the consultants involved that their child cannot survive, is frozen, unable to move forward for want of a unifying diagnosis.

Although the particularities of this case are striking, this general scenario of physicians feeling that they cannot render a prognosis for want of a unified underlying diagnosis is not uncommon. Underneath this feeling of being stuck, there often seems to be a deeper feeling, namely the fear of potentially being wrong, of offering a fatal prognosis, and having subsequent events show that death might not have been, or was not, inevitable.

How can we approach this scenario to better understand and address various ethical concerns that tend to arise when a patient is in a grim clinical situation, yet the underlying diagnosis is uncertain? A combination of clinical and epidemiologic reasoning, joined by clear communication, although not a panacea, can help us to move forward.

We can start by thinking of how any piece of information affects the prognosis of a patient similar to how we think of any test result providing a piece of information about the likelihood that a patient has a particular diagnosis. As with any other medical test, prognostic information has test characteristics such as true- and false-positive and true- and false-negative results. Viewed in this framework, the physicians are worried of issuing a false-positive prognosis of dying and the subsequent implications (such as ceasing life-support measures), whereas those around them may be more worried about the implications of continuing care (such as continuing suffering for the patient) being conducted under what they take to be the false-negative prognosis that death is not inevitable.

Next, we should examine the assumption that a diagnosis is a key piece of prognostic information, and that diagnostic uncertainty inevitably results in prognostic uncertainty. Certainly, sometimes a diagnosis carries a clear and precise prognosis, ranging from full recovery to death. More often, however, the given diagnosis has a broad phenotypic range, with some patients affected quite severely while others are so mildly, which in turn is associated with a broad prognostic range. Given this fuzzy correlation between many diagnoses and their prognosis, one should not assume either that a diagnosis implies a given prognosis nor that the absence of the diagnosis makes the act of prognostication impossible, especially if there are other ways of formulating a prognosis but do not rely on a diagnosis.

On what basis can a prognosis be made if not based on a diagnosis? One basis focuses on the nature of the injuries that the patient’s body has sustained and the likelihood and course of recovery from the sustained injuries. Another basis assesses the degree of physiologic derangement and the likelihood of ongoing injury. A third basis would be the availability, or the lack thereof, of countermeasures to offset any physiologic derangements or organ failure.

Let’s pause and consider an analogy to make these points clearer. Suppose that in the middle of the night, a villain entered a house and caused injury to an occupant and then slipped away and was never apprehended. When the injured occupant is brought to the hospital for care, a physician could still make a reasonably accurate prognosis for the patient based on the nature of the injuries and the consequences thus far, even though never knowing who or what caused the injuries. A single uncomplicated fractured bone with no other injuries is likely to heal favorably, whereas a blow to the head resulting in a skull fracture that then resulted in intracranial bleeding and raised intracranial pressure has a much less favorable prognosis. Stated more abstractly in almost algebraic format, if we know that X caused Y, which in turn then caused A, B, and C, even though we do not know what X is, if we clearly know Y, we can proceed with our prognosis from that point forward in the causal chain.

Returning to the case, our infant sadly had ample findings to suggest a fatal outcome. The infant had widely disseminated injury throughout her body. There was evidence of severe physiologic compromise and likely ongoing injury as evidenced by the initial blood gas pH, pulmonary hypertension, encephalopathy, and anuria. Last, the severity of her tenuous condition precluded her from several potential forms of life-sustaining therapy, and no countermeasure exists to tackle the ongoing hemorrhagic process.

Given this confluence of clinical reasons for a fatal prognosis, the absence of the underlying diagnosis has much less potential sway over the accuracy of this prognosis. Why? Recall that test characteristics such as false positives and negatives depend on the prevalence of the underlying condition in the population being tested. In this case, the prevalence of mortality of patients like that in this patient, with similar extensive injuries and physiologic derangements, is extraordinarily high. In such a circumstance, although discovering the underlying diagnosis would provide some additional information, given the lack of tight correlation between diagnosis and prognosis discussed previously, this additional information is unlikely to substantially change the overall prognosis.

Last, we should consider how to communicate not only our bottom-line prognosis but also the reasoning that led us to it to the parents in an appropriate manner. There are many ways to do so. I can imagine, though, what I might offer to the parents, Daniel and Nathalia. “Let me start by saying how much I wish that your daughter was not ill.” I would probably pause here before continuing. “I do not know what started the chain of events that led her to be so ill. What I do know is that 1 of those events was massive bleeding, which then led to what we call shock, where parts of the body did not get enough blood and the oxygen that the blood carries, and that this resulted in injury to the brain and to the kidneys as well as to the heart, which you know stopped twice. I also know that all this injury is continuing.” Another pause. “She is too sick to receive some of the treatments that could keep her body going longer, although none of the treatments would stop the process that started this terrible chain of events.” Another pause. “I’m afraid that she simply cannot survive her extensive injuries.”

I am a neonatologist and a researcher in clinical ethics. I am also the mother of three children, one of whom was born at 24 weeks and another having special needs during his first years of life. While I read the case, I vividly remembered emotions I felt as the stressed mother of sick babies, lost despite all her professional knowledge.⁶  Examining the case as a neonatologist, I determine that Evangelina is certainly dying. The attending physician was conflicted about how to communicate with the parents without a clear diagnosis.

Meeting Daniel and Nathalia in such a situation is no longer to discuss treatment options. The physician’s role becomes one of supporting the parents during the death of their baby and helping them “rewrite their story.” There is evidence on how to communicate with families in these tragic moments.^7,8  For example, using the baby’s name, meeting the parents with a limited number of individuals, sitting on a chair. During the whole conversation, I would make a conscious effort to listen to parents, even leaving intentional silences—intentional because in moments of discomfort, it is easy to speak constantly.⁷  I would tell them unambiguously Evangelina is dying (“We have done everything we can to help and are unable to keep Evangelina alive. She will die soon.”), avoiding euphemisms (such as being with the angels, meeting her maker, or being in a better place). Parents need to hear there is nothing they could have done to prevent what is happening to their baby. I would avoid describing everything we will not do (extracorporeal membrane oxygenation, MRI), nor would I ask for permission to withhold cardiopulmonary resuscitation or any other treatments that would not work. For dying babies, cardiopulmonary resuscitation is a physiologically futile intervention that would not be offered nor discussed in many NICUs around the world.

I would place the emphasis on everything we will do. Parents need to know we have experience caring for dying babies, that we will be there with them. Providing physical proximity between Daniel, Nathalia, and Evangelina is important.⁷  I would promise them that we will do our best to ensure that she will not feel pain during the rest of her life. I would offer to speak directly to their religious representative and offer them spiritual support from our team. The practical processes should be discussed; for example that they can take Evangelina in their arms if they wish, that we can take the tube out or keep her on the respirator, that either option is acceptable, and that different families decide differently and for what reasons. Knowledge about what happens to babies’ bodies after death is also important as well as memory-making. Many physicians are unaware or misinformed about these details,⁷  but we must be ready to accurately answer the family’s questions.

Many parents are in such shock and so lost that they ask for recommendations; in this case, offering a plan is essential. Naming, acknowledging, and normalizing the emotions parents feel (anger, rage, sadness, despair, emptiness) often helps families. Even if not evidence-based, in some circumstances when anger and injustice fill the room, I have used curse words and told parents that “life is a bitch,” that what was happening to their baby is unfair, and that life makes no sense. Later, many of the parents whom I keep in touch with told me this had helped them feel they were not going mad and that nice words sometimes are not enough. I leave my e-mail address with families after the death, send them a bereavement card, and stay in touch with them. Families have taught me a great deal over the years. They help me become a better person and physician.

Uncertainty about the diagnosis of a dying child is a lesser but important problem. Parents of babies in the NICU often experience powerful emotions of inadequacy and guilt. Clinicians must be sensitive to the unintended harm they can cause, such as questioning the mother about her lifestyle and health. With genetic diseases, either of the parents may feel guilt because they “carried bad genes.”⁹  In this case, and in the vast majority of cases in neonatology, we know that the parents carry no personal responsibility for the infant’s illness. Parents should be told so explicitly.

Clinicians may also benefit from a diagnosis. There has been a recent explosion of precision tests (such as whole-genome sequencing, MRI, near-infrared spectroscopy). Doctors are less willing to make redirection of care decisions without a clear diagnosis, even when it is obvious that the clinical situation is extremely reserved. We have become dependent on precision tests and yearn for unambiguous answers. Lacking a diagnosis may make clinicians uneasy and may feel like a failure, especially in the face of death. For some parents also, seeking a diagnosis is perhaps a way of coping. Knowing when to stop the diagnostic odyssey and speak about life and death is an important skill physicians should master.

This search for certainty is illustrated by the recent ability to rapidly sequence an individual’s entire genome.¹⁰  Genomic testing may reveal a diagnosis, but often does not reveal a curative treatment. The results may simply bolster the case for redirection to comfort care, or may, as is more common, be nondiagnostic or of uncertain significance. In such cases, waiting for the results may simply delay what is already the clearly preferable therapeutic option for an infant. The decision to withdraw life-sustaining interventions for infants who are dying, despite not having an explicatory diagnosis, may be emotionally challenging for clinicians. But waiting for any further test result for Evangelina was not necessary and was not in her interest.

A diagnosis may not help the child, but may have some utility for some parents, such as for future pregnancies or to find “a reason.” But a child’s catastrophic illness or death can never be truly “understood” by parents even with precision testing and a “clear diagnosis.”¹⁰  In this case, an autopsy may also reveal a diagnosis. We should focus on other kinds of interventions, such as grief support, or interventions that promote resilience and coping. Such anticipatory discussions will give parents a realistic understanding of the likely impact of a diagnosis.

There are no life-saving interventions we can do for Evangelina but we can ensure that she dies pain-free. There is much we can do for her parents, ensure that they can play a parental role, have the opportunity to express their love for their baby, and write a story about her brief life that will be meaningful for their futures.

Despite continuous volume resuscitation and maximal pressor support over the next few days, Evangelina remained anuric. She was denied candidacy for hemodialysis and extracorporeal membrane oxygenation because of her poor prognosis. The attending physician expressed to her parents that he believed Evangelina would die even though he did not know the underlying cause. He was open about the uncertain diagnosis and acknowledged the epistemic ambiguity inherent in his recommendation to withdraw life-sustaining treatment and disclosed that this was based on his observations relying on experience.

Nathalia and Daniel agreed with transitioning Evangelina to comfort care on day of life 5. They elected to have her extubated, and held her in a private room, with a morphine infusion for comfort. She died 37 minutes after removal from the ventilator, in her mother’s arms. They consented to a postmortem biopsy and autopsy, which confirmed the diagnosis of congenital rhabdoid tumor with placental spread and multiorgan metastasis inclusive of the central nervous system.

Nathalia and Daniel were appreciative of the care given to Evangelina and the support they received while she was admitted. They were grateful for the time they had with her and believed that sharing her story and their experience would bring meaning to Evangelina’s life.

Despite the advances in contemporary medicine, including whole-exome and genome sequencing, providers cannot always make a definitive diagnosis. As Dr Feudtner argues, this does not mean that the prognosis cannot be known with a reasonable degree of certainty. It is important for providers to be able to manage epistemic uncertainty and its associated angst given the potential harms to patients of trying to reduce this uncertainty to make themselves feel better. Drs Feudtner and Janvier provide very helpful suggestions, including examples, about how to discuss uncertainty and a child’s imminent death with the child’s parents and support them in such tragic circumstances.

MRI

magnetic resonance image