The Prophetic Art of Prognosis

by Nicholas A. Christakis

n numerous occasions, patients have asked me if they would be cured of their cancer, if their pain would ever stop, if they would live until the end of the week, or until their next Christmas, their next anniversary, their next child's graduation from high school. These questions pained me, and not just because they touched on the ineffable sadness of deadly disease or the efforts the dying often make to stay connected to the living. They pained me as well because it was so difficult, yet at the same time so essential, to answer them.

Over the course of my clinical training, I came to regard explicit, precise and compassionate responses to patients' requests for prognosis to be a key part of my role as a physician. I came to see the deliberate assessment of prognosis as absolutely obligatory, even if patients did not happen to ask. Yet I found that I had been poorly trained for this.

Textbooks omitted prognosis, journals avoided it and medical schools ignored it. The whole profession seemed to overlook prognosis. And few of my colleagues shared my conviction that this might be a problem. They even seemed bemused by my interest in this "marginal" topic, as if the proper and scientific role of medicine were only to diagnose and treat disease, not to predict its outcome.

Was prognosis really so unimportant? Were the obstacles to it really insurmountable? When it came to prognostication, physicians appeared to speak hesitantly, softly and ambiguously--if at all. This contrasted markedly with the confidence they exuded when making a diagnosis or prescribing a therapy.

I ultimately concluded that the muffled presence of prognosis had a lot to do with the raging authority of death. Prognostication and death are tightly interwoven. Although there certainly are more routine, less serious incarnations of prognostication, if one asks doctors to free-associate with the word "prognosis," they are apt to say "death." And when physicians are asked to think about the role of prognosis in their practice, the question they imagine--and dread--is "How long do I have to live?" Like death, prognostication seems mysterious, final, powerful and dangerous.

On more than one occasion, I have seen the avoidance of prognostication, or needlessly incorrect prognoses, harm patients. A recent newspaper article entitled "For Cancer Patients, Hope Can Add to Pain" poignantly captured how such harm can come about. It quoted a patient's wife as saying:

The Thursday before my husband died, I thought he was dying. But the doctor was talking about aggressive chemotherapy. I asked if this was palliative, and he said that he still hoped for a cure. I was with him at the time of his death [three days later], but the room was filled with eight other people hanging bags of blood and monitoring vital signs. It was about as horrifying as anything that could have happened. I don't think the [doctors] were trying to mislead us. They thought he might be the one case that would have a positive outcome. [But if I had been told the truth,] we could have spent days with the children, together, not filled with painful regimens in the hospital.

The failure to predict this patient's death--in the sense of not thinking about the prognosis clearly, in the sense of not articulating it and in the sense of encouraging an unduly optimistic expectation--was harmful. Such a failure can contribute to a therapeutic imperative that prevents families from taking steps to prepare for death. The physicians did not want to see that the therapy would not result in the desired outcome. They did not want to predict that the patient was about to die. And they did not want to take action, such as limiting therapy, based on such a prediction. My feeling is that the problem in this case, and countless other cases like it, arises more from errors in prognosis than from errors in therapy. In our rush not to abandon patients therapeutically at the end of life, we abandon them prognostically.

And yet, cogent and compassionate prognostication could decrease the prevalence of bad deaths in our society. Such prognostication is a sensitively delivered and well-calibrated best guess about the patient's future. It requires physicians to be as versed in the art and science of prognosis as they are in diagnosis and therapy, to make strenuous efforts both to learn the state of the art with respect to the prognostic problem presented by the patient and to communicate that knowledge in a way that the patient can comprehend, to the extent that the patient wants this. Moreover, it requires physicians to adopt a broader view of the meaning of hope and to realize that there is much that patients can realistically hope for even if death is imminent and unavoidable.

Such prognostication includes physicians' willingness to spend time talking with patients, assuring them that they will not be abandoned. It entails, finally, the willingness of physicians to act on predictions, despite the risk of error. Such behavior by physicians would reflect the realization that temporizing or self-delusion in prognosis can be as harmful to patients as an incorrect diagnosis or a mistaken treatment. And such behavior would, ultimately, reflect the very real moral aspects of prognostication.

These are not easy things to do, of course, and there is good reason that physicians avoid prognostication, beyond the fact that it is technically difficult and emotionally frightening. Prognostication can hurt patients, and not just when it is inaccurate. Some patients do not wish to be provided with prognostic information. And I am deeply empathetic to the complexity of prognostication. What I am suggesting is an approach that balances the benefits of prognostication against the benefits of avoiding it, an approach that realizes that unfavorable predictions do not mean physicians have nothing to do and patients nothing to hope for, an approach that recognizes that acting on an uncertain prediction may be better than making no prediction at all.

A better understanding of how physicians prognosticate is justified not only by the prospect of enhancing patient care but also by contemporary developments that are increasing the importance of predictions.

The number and percentage of patients with chronic disease--for whom the diagnosis is already known and for whom therapy is often simply the continuation of previously initiated interventions--are increasing. In such cases, because curative therapy is limited and the course of the disease is long, prognostication can become especially prominent. The clinical encounter is focused on the anticipation, avoidance and mitigation of complications of the underlying disease itself or of the treatment. Patients with long-standing diabetes, for example, do not need to be told their diagnosis (which is known to them) or their therapy (for example, insulin--also known to them). Rather, they and their physicians are concerned with such questions as "Will my kidneys fail, and if so, when?" "Will I become blind?" "How long will I be able to care for myself?" "How long do I have to live?" Moreover, in chronic conditions there is more opportunity to revise previously rendered prognoses as, over time, the physician learns more about the patient.

Several developments in health-care delivery and technology also support the increasing importance of prognosis. A key factor is the increasingly bureaucratic structure of American medical practice. More and more, physicians are becoming salaried employees beholden to bureaucratic management or are otherwise losing their economic independence and some of their professional autonomy. External review of physicians' clinical behavior often focuses on actions that are at least implicitly based on their prognostic judgments. For example, administrative oversight plays an increasing role in therapy management, as physicians are being asked, in the context of cost-effectiveness, to predict outcomes among a variety of (more or less) costly treatment options or to estimate the length of the hospital stay necessary for an anticipated medical outcome. Better prognostication, in the sense of a superior ability to foresee the outcome of a patient's illness, can help to optimize the choice of treatment and the timing of hospital admission, thus reducing costs.

A greater focus on prognosis also results from the increasing need to compare the quality of health-care providers, along with a societal interest in rationalizing health-care expenditures by directing resources to those most likely to benefit. Accurate and reliable prognostic assessments are central to identifying, developing and implementing optimal health-care delivery systems in that they help patients and payers determine which systems lead to good or bad outcomes. In the same way, evaluating the cost-effectiveness of different medical therapies relies on access to accurate information about the probabilities of various outcomes, which are, in essence, prognoses. Some interventions may be cost-effective only in subpopulations where the patients are at particularly high or low risk of mortality, and payers may target interventions--conditional, again, on the patient's prognosis--to cases in which they feel the benefit justifies the cost. In each case, the ability to develop and analyze prognoses is essential to asking the right counterfactual questions and getting the right policy answers.

Another factor lending salience to prognosis is the increasing frequency of randomized, controlled clinical trials. The customary role of prognosis in clinical trials has been to ensure that patients of similar illness severity are compared, as groups of patients with an equivalent average prognosis (e.g., for death) are formed by the randomization process. However, rather than waiting for uncommon or temporally distant outcomes to occur in long-term clinical trials, which can be difficult and expensive, investigators are increasingly making use of "intermediate endpoints," that is, findings taken to be predictive of long-term outcomes. A decrease in a lab test is taken to indicate a decreased risk of eventual death, making shorter and less expensive trials possible. The booming industry in clinical trials thus supports increasing interest in the development and use of various prognostic staging systems and clinical markers.

The emergence of certain medical technologies also increases the relevance of prognosis. Such technologies--although often directed at improving diagnosis and therapy--also provide, directly or indirectly, more accurate and earlier prognostic information and therefore foster the rendering of prognoses. One example is obstetrical ultrasound, which may incidentally reveal information about the internal anatomy of a baby that would not otherwise be detected until well after the baby was born. Thus, parents and physicians are made aware of conditions long before the child is able to present with the illness. Such early findings, not yet corresponding to any observed symptom, demand explanation: patients wish to know what the findings will mean, what the future has in store. One pediatrician gives a typical example:

Nowadays, many kids come to our clinic when they're two weeks old because their kidneys were abnormal on Mom's prenatal ultrasound! In the old days when someone was diagnosed with polycystic kidney disease in the neonatal period, they died before they were one year old, but now there are so many much milder cases that only come to our attention because of an ultrasound that happens to have been done for other [obstetrical] reasons. We have no idea what to tell the parents about what to expect....

The parents ask for predictions. They want to know when will the kid need dialysis, when will they have to start thinking about transplant, when will this kidney disease that we see on the ultrasound affect them, because right now their kids are thriving, they're doing well. But at some point they're going to need erythropoeitin shots and growth hormone and vitamin D and all these different things that we see with renal failure. We have no idea when that will be.

The advent of genetic-testing technology provides an important new arena for prognostication. Analysis of a person's genes may reveal relevant medical outcomes years or decades in advance in what are generally termed "presymptomatic carriers." In contrast to other tests applied for diagnostic purposes, these tests have specifically prognostic importance, which is made all the more apparent by the fact that, in most cases, no specific interventions are available to cure the condition so detected, or even to delay its onset. A prototypic example is Huntington's disease, a fatal, degenerative neurological disease that usually begins in the patient's forties. Tests can now reveal with certainty, decades before any symptoms are noted, whether asymptomatic individuals will or will not develop the disease. Because no treatment is available for the condition, this is an example of the phenomenon of the complementary relationship between therapy and prognosis, in that the prognostic significance of these tests eclipses their therapeutic utility.

In other situations, genetic tests are used to develop prognostic information that does have therapeutic implications. For example, some women, on the basis of the prognostic information provided by genetic tests to evaluate their risk for breast cancer, undergo prophylactic mastectomy--even though this prognosis is not certain. As more studies reveal genetic bases for diseases as diverse as emphysema, diabetes, dementia, cognitive disability and alcoholism, the prognostic use of genetic tests will certainly rise. The use of such tests is also likely to rise outside of medical arenas, for example, in law.

The increasingly technological, "postclinical" nature of medical practice fosters the availability of information that is presymptomatic in nature, and therefore inherently prognostic. "Presymptomatic" illness is, indeed, the specifically prognostic analog of "asymptomatic" illness. The notion of asymptomatic or "occult" or "silent" illness is itself interesting in that it posits a phenomenological realm of disease of which the patient has no subjective experience. This realm requires the intercession of an expert, a physician, to be comprehended. Typically, the expert must use technology to approach this realm, as when the physician uses diagnostic tests to adduce the presence of disease even when the patient has no symptoms.

The implications of the notion of presymptomatic disease, however, extend even beyond those of asymptomatic disease. Rather than indications of an already present disease, the expert is said to have discovered indications of a disease that is not yet even present. Some physicians have even begun to call individuals whose genetic tests are positive for a worrisome gene "prepatients." The notion of presymptomatic illness thus represents an even further distancing of the patients' subjective experience of disease from the everyday practice of medicine. Moreover, the term "presymptomatic" suggests an inexorable outcome: the patient will eventually develop symptoms.

The application of new technologies to patient care increases the importance of prognostication in one other way: it creates a whole new class of things about which to prognosticate, namely the complications of the technology. Beneficial new technologies in medicine--from computed tomography to chemotherapy to open heart surgery--have not come without risk. Predicting their consequences is important, and doctors are frequently called upon to explain to patients a potentially confusing array of possible outcomes.

In addition to changes in types of medical problems people face and in the ways physicians confront them, there have been changes in how patients and physicians think about the ethical duties of physicians to their patients. Prognosis is a fundamental, though implicit, basis for many theoretical and practical ethical decisions in medical care, and prognostic uncertainty may complicate such decisions considerably. Ethical decision-making is increasingly finding its way to the bedside. The elaborate informed consent process that patients undergo before having procedures or participating in research, for example, is predicated upon predicting risks and benefits. Prognosis also profoundly affects decisions to initiate, withhold or terminate life support for critically ill newborns and adults, and it figures in the discussions about these decisions that doctors have with patients' families. Prognostication is critical when one must allocate scarce medical resources to those patients for whom they can do the most good.

Finally, it is central to the notion of "futility," a concept usually invoked in situations where death is predicted to be imminent and inevitable. The relatively recent emergence of futility as a theme in bioethics reflects the moral desirability of acknowledging medical limitations and the practical necessity of allocating scarce resources. Futility is based on a prognosis not only that the patient is unlikely to recover spontaneously but also that any intervention will likely be ineffective. As the avoidance of futile treatment has assumed increasing prominence, for reasons of justice, beneficence or economy, prognostication--which is, after all the fundamental and essential basis for a determination of futility--has increased in importance.

Broad changes in American society are influencing the doctor-patient relationship and fostering an increased interest in prognosis. In areas from childbearing to terminal care, patients want information about expected outcomes that they can use to manage their care actively. This is especially true with respect to care at the end of life. Now physicians have the duty to inform their patients about their illness, and patients have a right to know.

In recent years, the American public has become more focused on planning for death, a development reflected in the increasing interest in everything from living wills to physician-assisted suicide. There has been a profusion of books on caring for the terminally ill at home, which include vivid, nontechnical descriptions of what to expect and which document the impact of death on family members. There have been best-selling how-to books on "self-deliverance." And there have been books describing the process of dying, often using detailed and intimate case histories. These latter books typically reflect an attempt to help people find meaning in dying, and they suggest that death is increasingly viewed as a passage that can be actively anticipated and therefore managed. To enact these popular visions of death, however, patients must rely on reasonably accurate prognoses from professional physicians.

Ongoing changes in the nature of illness in contemporary American society, as well as in the way medical problems are being confronted, have substantially increased the relevance of prognostication in clinical care. As these trends converge, the nation's physicians will increasingly be expected to prognosticate--they may be reluctant prophets, but they will be prophets nonetheless.

ABOUT THE AUTHOR | Nicholas A. Christakis

Nicholas A. Christakis has a joint appointment in the departments of medicine and sociology at the University of Chicago. His clinical practice consists of caring for terminally ill patients at home, and his research interests focus on the social outcomes of medical care and the consequences of medical care for morbidity and mortality patterns. Christakis is particularly interested in the role of prognostication in medicine and in the end-of-life care, including the ability of different forms of end-of-life care, such as hospice or intensive care units, to affect mortality. Christakis is the author of Death Foretold: Prophecy and Prognosis in Medical Care (University of Chicago Press, 1999).

COPYRIGHT | This article originally appeared in University of Chicago Magazine, Vol. 92, No. 4, and is adapted from Death Foretold: Prophecy and Prognosis in Medical Care (University of Chicago Press, 1999). Copyright 2000 The University of Chicago.

(c) 2004 The University of Chicago :: Please direct questions or comments to