Predicting the end of life in children has received less attention in the literature than the related topic of predicting survival. Recognising that cure is no longer possible, however, is important for several reasons.

Families benefit from the opportunity to make informed choices about the care of their child,1 such as place of end of life care, saying goodbyes and achieving important goals. For professionals, it is an opportunity for a change to potentially achievable goals of treatment such as control of distressing symptoms, or to offer specific interventions such as referral to a children’s hospice or a specialist palliative care team,2 or discussion of advance directives regarding resuscitation.3

Parents of children with cancer value direct and accurate prognostic information expressed in numerical terms wherever possible.4 Families who find the information upsetting nevertheless recognise the importance of prognostic information for decision making and the maintenance of hope.

Parents value direct communication with the physician during end of life care.5 Direct communication with the child is also valued, providing the family consider the child old enough for such communication. Clear information from physicians concerning what to expect in the period immediately before death, communicating with care and sensitivity, and communicating directly with the child are all associated with higher parental ratings of physician care.

Relapse and disease progression are critical junctures for families of children with cancer, as well as for professionals.6 In non-malignant palliative care, also (Hain, personal communication), families retrospectively identify commencement of the dying phase as a key milestone in the disease trajectory. The duration of palliative care itself (table 1) remains poorly delineated7^–10 and is characterised by gradual awareness rather than a distinct shift in treatment goals.

Exactly how and when palliative care should start continues to concern many professionals,11 12 and many still seek a specific point in the illness trajectory to introduce it.13 In cancer, the two greatest barriers to high quality palliative care identified by paediatric oncologists12 are unrealistic expectations of cure by parents (and patients), and family denial. The opportunity for advanced care planning appears beneficial in both adult and paediatric palliative care,3 but uncertainty around suitable timing means that difficult discussions of options regarding end of life care, including withholding or withdrawing treatment, are often delayed until death is imminent.14^–16

The Gold Standards Framework (GSF)17 18 has been designed to identify patients in the primary care setting who are in their last 6–12 months of life, assess their current and future needs and ensure that these needs are met. The GSF was designed for adults with cancer but the concepts are equally relevant to non-oncology diagnoses and paediatrics. The GSF for adults includes guidance on prognostic indicators that may be useful in identifying those in the last 6–12 months of life and early work is now in progress to adapt these for children (Brook, personal communication). Interestingly, although the GSF originated to support adults with cancer in the last 6–12 months of life, it is now recognised that in this group the GSF is relevant from diagnosis.

PREDICTING END OF LIFE

Physicians tend to overestimate prognosis in adults with cancer, with only 61% accurate to within 4 weeks of expected survival.19 However, they are relatively consistent in their overestimates. Accuracy improves with clinical experience but decreases with duration of the doctor–patient relationship.20 Furthermore, the survival estimates that physicians communicate to their patients are also significantly over-optimistic, even compared with their own private “formulated” survival estimates.21

In order to improve prognostic estimates, specific indicators of survival have been identified (table 2) and combined into prognostic scales or scores.22^–25 Unfortunately, there are a number of methodological problems with many of the studies, including lack of consensus over the definitions of key terms such as “palliative” or “terminal”, failure to define fully or to measure proposed prognostic variables, and excessive loss to follow-up. One proposed prognostic score was validated using the dataset from which it was derived.26

A number of prognostic scales have been devised for adults with cancer (table 3). Prognoses are generally more accurate as death approaches.20 Work in adult palliative care suggests that a series of physiological changes occurs in the last 24–48 h of life, characterised by slowing or cessation of normal physiological processes.27 Recognition of this “dying phase” by healthcare professionals is variable, emphasising the importance of the multidisciplinary team approach.28

In paediatrics the evidence is more limited. Paediatricians appear reluctant to document a diagnosis of dying, or discussions about resuscitation, in the child’s case notes until the last 24–48 h of life.29 Furthermore, even at this stage there are significant problems with accuracy. In one series of 17 children transferred home for one-way extubation and end of life care,30 all were expected to die within 48 h. However, although eight (47%) died within 24 h, four (24%) lived for more than 2 weeks (one for a further 18 months). Medical seniority has been associated with more accurate prognostication in paediatric intensive care units (PICUs).31 32 Confidence in prognostication appears to be associated with greater accuracy but certainly does not guarantee it. As in adults, both confidence and accuracy are greater as death approaches.

Rarely, even the fact of death itself may be incorrectly predicted, for example when a diagnosis is revised or when treatment intended to prolong life without cure instead induces long-term remission. It may be that these highly unusual cases disproportionately influence clinicians to be reluctant to predict death until it is not only inevitable but also imminent.

Early work adapting the Gold Standards Framework project to the needs of dying children (Brook, personal communication) has begun to identify factors that may potentially facilitate a diagnosis of dying. The wide range of candidate factors proposed for study underlines the complexity of the task. They include increasing frequency of hospital or PICU admissions, episodes of acute illness without recovery to the child’s usual best level of health, and physiological changes such as decreased oral intake, changes in breathing pattern and decreasing periods of wakefulness and activity.

A proposed prognostic score for children was presented at the recent SIOP 2007 meeting in Mumbai (Brook, personal communication) but has not yet been validated. There are, however, studies on the related topic of predicting survival in critical care. Tools such as the Pediatric Index of Mortality (PIM and PIM2) and the Pediatric Risk of Mortality (PRISM, PRISM III) have been designed to allow population estimates of mortality.33 The measures are not intended to estimate individual mortality, however, and are of limited value in identifying risk before PICU admission. They are also subject to considerable error, and their dependence on invasive tests and monitoring makes them unsuitable for routine use in palliative care.

PHASES: MANAGING UNCERTAINTY

Although clear and accurate prognostic information is valued by families and simplifies management for professionals, arbitrary and spuriously precise estimates are discouraged in palliative medicine.34 35 The problem is not simply one of giving information that may be untrue. Numbers are always memorable; the concept they illustrate less so. In saying “I think your child has around three months to live”, the physician may intend to convey that the child’s life is likely to be months rather than years, but the family will remember the figure itself, and invest it with an authority it cannot possibly merit. A golden rule is to avoid numbers altogether in this context.

If by some chance it is an exact and accurate prediction, it may result in the child and family feeling they are “living on death row” and compromise coping mechanisms. If, on the other hand, their child dies before the expected time, families may feel cheated. Conversely, the expectation that all the suffering and turmoil will be over by a certain time can be an important coping factor. When a child lives unexpectedly through a critical illness, but with the inevitability of another and eventual death, families may experience a complex mixture of relief at their child’s survival, anger that their child has not died, and guilt at the ambivalence of their feelings.

Physicians usually report recognising that there is no realistic chance of cure some time before the child’s parents.36 Decision making in this context is characterised by a lack of clarity, uncertainty and continued hope.37 Where professionals and parents both recognise the child’s prognosis early, discussion of palliative care is easier.38 Families report a difference between intellectual acknowledgement of inevitable death and emotional acceptance,39 which usually follows some time later.

The information needs of adult cancer patients and relatives change and diverge as illness progresses,40 and as communication between them typically becomes less verbally explicit. Patients express ambivalence; they want to be told, but do not want to know. Patients describe living parallel realities: “I know it’s growing again… I still feel invincible, but I know I am going to die. I manage to have these two things in my head at the same time”.

There is a growing consensus that the needs of the child and family for palliative care are best met if palliative care is available from the point of diagnosis16 41 42 and incorporates open and honest communication throughout.39 Parents themselves do not see, for example, cancer-directed therapy and symptom-directed therapy as mutually exclusive alternative approaches.43 They prefer cancer-directed, symptom-directed and supportive care all to be available throughout the child’s illness. Transition to palliative care appears to be most successful when the concept of palliative care, and access to practitioners with skills and expertise in it, is introduced at the point of diagnosis.44 45 Involvement of a children’s hospice in the care of children dying from cancer is often made only at end of life2 but may more appropriately be introduced at relapse. The most frequently given reason for not doing so is ongoing therapy. The decrease in priority for cure-oriented treatment and the increase in the importance of treatment aimed at enhancing quality of life should be seen as a natural transition rather than abandonment.38

CONCLUSION

There is evidence from studies in adults and children that both patients and families value the chance to discuss the end of potentially curative approaches and, later, the approach of end of life. It gives an opportunity to explore concerns, make preparations and express preferences. While families are often upset by the nature of the information they are given, the fact that it is made available in a sensitive, timely and empathic manner strongly influences their sense of comfort with the medical and nursing care they receive.

Professionals are more ambivalent. The identification that a patient is “palliative” clarifies treatment goals, allowing measures aimed at comfort to take priority over those aimed at a cure that is unlikely and permitting planning and anticipation of the child’s palliative needs. At the same time, there is discomfort and uncertainty about the appropriate point in a child’s illness trajectory at which to begin discussing end of life care issues, including those around advanced planning and resuscitation. This is compounded by the diverse trajectories of conditions that can limit life in childhood. Nevertheless, the issues have to be addressed with honesty and confidence if the needs of dying children are to be met.

The diagnosis of dying has two important facets. Firstly, conditions that are potentially or always fatal should be recognised from the outset as life-limiting conditions, and palliative care should be made available as soon as the diagnosis is made. Secondly, clinicians should develop skills that allow prompt and confident recognition that the imminent end of life is likely.

The aim should be both that “curative” approaches can give way gradually and smoothly to “palliative” ones as the illness progresses, and that in the final few days, appropriate arrangements can be made for expert end of life care.