INTRODUCTION — Prognosis is the science of estimating the likelihood of an outcome (eg, death, disability) due to a medical condition (eg, cancer, heart failure, late-life disability). Prognosis may address many outcomes that patients care deeply about, such as the likelihood of losing the ability to care for oneself independently, of treatment leading to a cure, or of developing a debilitating condition such as dementia. This topic will focus primarily on communicating prognosis for life expectancy or survival.
Nicholas Christakis has described the “ellipsis of prognosis” in medical teaching [1,2]. In the time of Osler, textbooks focused equally on diagnosis, treatment, and prognosis. Prognosis now makes up a scant one to two lines per chapter of most medical textbooks, and medical students and residents receive little training in how to estimate or communicate prognosis. Additionally, patterns of current practice suggest that prognosis is not routinely incorporated into clinical decision-making.
The growing palliative care movement is refocusing attention on the importance of prognosis in decision-making for seriously ill patients [3,4]. Prognosis reaches paramount importance at the end of life, when the risks and burdens of treatments and their alternatives must clearly be weighed in view of the patient’s life expectancy and personal goals.
This topic will discuss the science of estimating prognosis and focus on how best to communicate this information to patients and family members. Indicators of prognosis for specific medical conditions are available in disease-specific topics in UpToDate. A more general discussion about discussing serious news is provided separately. (See "Discussing serious news".)
THE IMPORTANCE OF PROGNOSIS IN CLINICAL PRACTICE — The risks and benefits of any test or treatment must be considered in light of the patient’s expected prognosis. Thus, understanding and communicating prognosis is an important element of all clinical decision-making. Determination of eligibility for hospice, where clinicians must certify a prognosis of six months or less should the disease run its usual course, is the paradigmatic example of the need to be able to determine prognosis. The importance of understanding prognosis for multiple other clinical decisions is shown in a table, organized by shortest prognosis to longest (table 1). The table includes only those decisions where a guideline, professional organization, or recognized experts have recommended a specific prognostic cutoff; these represent only a subset of decisions that are routinely made in light of prognosis.
While prognosis is clearly vital to many end-of-life decisions, clinical guidelines rarely provide specific timeframes to inform prognosis and influence clinical decisions. Such common decisions include initiation of methadone for pain without monitoring QTc interval , discontinuation of anticoagulants in patients receiving end-of-life care , and deactivation of implantable cardioverter-defibrillators . (See "Cancer pain management with opioids: Optimizing analgesia", section on 'Methadone' and "Palliative care: The last hours and days of life", section on 'Eliminating non-essential medications'.)
Patient-important factors — Patients frequently cite life choices rather than medical decisions for wanting to know prognosis. In one study involving 60 Black, Hispanic, Chinese, and White disabled older American adults, 75 percent stated they would want to know if their doctor thought they had less than one year to live . Reasons given included having time to put their spiritual affairs in order, to say “goodbye,” and to prepare financially.
RATIONALE FOR DISCUSSING PROGNOSIS — Western medicine has slowly transitioned from a paternalistic model to a model based on informed shared decision-making. The doctrine of informed consent obligates clinicians to provide patients with information, such as prognosis, that is critical to their decision to accept or refuse treatments. Going beyond informed consent, the shared decision-making model promotes a bidirectional exchange of information so that patients and clinicians can deliberate collaboratively about the best courses of action [9,10].
Prognostic information is often integral to the health decisions and life choices patients are making in their last years of life . Patients who have a good understanding of their prognosis tend to make different decisions than those who are not aware, including less use of aggressive life-sustaining treatments at the end of life [12,13].
Too often, however, patients are making decisions with little prognostic awareness. As an example, in a study of 1193 patients with metastatic lung and colorectal cancer, 69 percent of those with lung cancer and 81 percent of those with colorectal cancer did not report understanding that chemotherapy was not at all likely to cure their cancer . Inaccurate perception of treatment intent has also been noted among women with advanced breast cancer .
Clinicians may be reluctant to discuss prognostic information because they are concerned it will negatively affect the patient-doctor relationship or the patient’s psychological state [16,17]. Clinicians frequently articulate the concern that discussing prognosis will “take away hope.” Indeed, in the study of patients with metastatic lung and colorectal cancer, understanding that chemotherapy was not curative was associated with lower patient ratings of clinician satisfaction . Furthermore, in a study of surrogates of chronically ventilated intensive care unit (ICU) patients, the addition of a structured informational session about prognosis by non-ICU (palliative care) clinicians did not improve anxiety or depression among family members compared with a control condition, and exploratory analysis suggested an increase in symptoms of posttraumatic stress disorder . (See "Palliative care: Issues in the intensive care unit in adults".)
In other studies, however, prognostic disclosure has not harmed the patient-clinician relationship, and instead it has led to greater patient satisfaction, lower anxiety and depression, less likelihood of using alternative therapies, a more realistic expectation of life expectancy, and possibly a stronger therapeutic alliance [19-23]. As an example, in a prospective longitudinal cohort study of 265 adult patients with advanced cancer who visited 38 oncologists in either community- or hospital-based cancer clinics in Western New York or Northern California, the prognostic discussion was assessed qualitatively by coding a transcribed audio-recorded visit using the Prognostic and Treatment Choices (PTCC) scale (which measures the extent to which oncologists engaged patients regarding prognosis and treatment options ) . Patients were surveyed at baseline, at two to seven days, and at three months after the clinic visit using validated tools (the Human Connection  and the Perceived Efficacy in Patient-Physician Interactions  scales) to assess the strength of their perceived therapeutic alliance with their clinician. Prognostic discussions were not associated with a temporal decline in the strength of the patient-clinician relationship. In fact, a one-unit increase in the PTCC during the audio-recorded visit was associated with a small but statistically significant improvement in patient ratings of the therapeutic alliance both at two to seven days and at three months after the visit.
An important point is that patient satisfaction may be the wrong metric for measuring the outcomes of these conversations, however. A potentially more appropriate measure, as yet untested, might be to ask patients and families/loved ones later in their illness trajectory if having the discussion about prognosis helped them to prepare for the end of life. Evidence is mounting that inadequate preparation for death leads to poor bereavement outcomes, including depression, a sense of regret, and complicated grief [13,27,28]. Thus, the evidence suggests that prognosis discussions are likely to lead to not worse, but better psychological outcomes for patients and their families or other loved ones.
THE SCIENCE OF ESTIMATING PROGNOSIS — A major practical barrier to discussing prognosis is the lack of certainty in prognostic information. With increasing availability of data from national surveys and clinical trials, the science of estimating prognosis has made remarkable progress. Nonetheless, some argue that available prognostic tools lack the accuracy, generalizability, and usability for routine clinical use.
Available prognostic tools for use in clinical decision-making for older adults are available through ePrognosis by the University of California, San Francisco. Prognostic calculators for patients with specific conditions are available in UpToDate as follows:
●Advanced cancer (see "Survival estimates in advanced terminal cancer")
●Advanced dementia (see "Care of patients with advanced dementia")
●Advanced liver disease (see "Model for End-stage Liver Disease (MELD)")
●Coronary heart disease (see "Risk stratification after acute ST-elevation myocardial infarction" and "Risk stratification after non-ST elevation acute coronary syndrome")
●Intensive care unit (ICU) mortality (see "Predictive scoring systems in the intensive care unit")
●Pancreatitis (see "Predicting the severity of acute pancreatitis")
Prognostic tools are more accurate at predicting short-term than long-term prognosis. In addition, prognostic tools that include functional status of the patient (eg, ability to carry out activities of daily living) tend to perform better than those that rely only on diagnostic predictors (eg, history of cancer) and demographic factors (eg, age, sex).
In addition to the calculators above, two systematic reviews may provide guidance for identifying patients with end-stage illness [29,30]. The systematic reviews, based on a literature search of over 1000 articles, have identified patient presentations associated with median survival of six months or less for patients with cancer  or six noncancer illnesses (heart failure, chronic obstructive pulmonary disease, dementia, geriatric failure to thrive, cirrhosis, and end-stage renal failure) . These sample presentations may identify patients in whom aggressive treatment is unlikely to prolong survival and for whom hospice services may be appropriate. Guidelines for determining eligibility for the hospice benefit from the United States Medicare program (which generally requires that the clinician estimate that survival is six months or less) are also available for several diseases (table 2). (See "Hospice: Philosophy of care and appropriate utilization in the United States".)
Selecting a prognostic tool and assessing limitations — Several factors must be weighed in the decision to use a given prognostic tool in clinical decision-making, including:
●The extent to which a given tool was tested in a population of patients that resembles the clinical setting at hand
●The extent to which the index has been tested across geographic regions and time
●Its accuracy when applied to patients of varying illness severity
For patients with a terminal condition such as advanced cancer or dementia, a disease-specific prognostic tool’s performance will be superior to the performance of a non-disease-specific tool. However, many older adults have a high mortality risk from a complex mixture of multiple chronic conditions, physical disability, and cognitive impairment. For such patients, non-disease-specific prognostic tools may be more accurate. Examples of these are available through ePrognosis by the University of California, San Francisco.
One critical metric for evaluating the accuracy of a prognostic tool is the c-statistic, also referred to as the Area Under the Receiver Operating Characteristic Curve (AUC-ROC). This is a measure of how well a given prognostic tool separates subjects who lived from subjects who died. By chance alone, a tool would sort such patients correctly 50 percent of the time (c-statistic 0.50), whereas a perfect tool would sort patients accurately 100 percent of the time (c-statistic 1.00). Most prognostic tools have c-statistics in the range of 0.60 to 0.80, with only a few performing better than 0.80 . These values are comparable to other widely used prognostic tools, such as the Framingham risk index (c-statistic 0.63 to 0.83)  and the Thrombolysis in Myocardial Infarction (TIMI) risk score (c-statistic 0.65) .
Clinician errors in prognostication — Although the information from a prognostic tool is not perfectly accurate, clinicians’ own empiric estimates of prognosis are likely to be far less so, and studies suggest that clinicians consistently overestimate survival .
In addition, two studies that looked at factors that predict clinician error in estimating survival did not report associations between accuracy and the clinician and patient characteristics examined (eg, clinician or patient sex) [34,35]. However, in one of these studies, two factors predicted clinician errors :
●Experienced clinicians were less likely to make prognostic errors than less experienced clinicians.
●Perhaps surprisingly, longer durations of patient-clinician relationships were associated with greater error in predicting survival and a more optimistic prognosis.
Multiple other potential reasons that clinicians err in prognostication are plausible, though they have yet to be fully assessed, likely because they are more difficult to measure. These include clinician biases such as:
●Prior negative experience, or the so-called “chagrin factor” (eg, telling another patient she would die shortly, only to be wrong and she lives for years) 
●Over-reliance on the importance of one particular test (eg, serum albumin levels)
●Enthusiasm for a new treatment (eg, a new chemotherapeutic agent for a patient with advanced cancer) 
Computed prognostic models, by contrast, do not suffer from these documented and potential clinician biases, which makes them more objective and potentially more accurate. In general, the combination of information from prognostic tools and clinical judgment result in a better prognostic estimate than either does when used alone [38-40].
ASSESSING PATIENT AND/OR FAMILY READINESS TO DISCUSS PROGNOSIS — Studies of patients with serious illness demonstrate that a majority of adults would want their clinician to discuss prognostic information. In a national survey of 1272 older adults presented with a range of hypothetical scenarios, preference to discuss prognosis with a doctor increased with decreasing life expectancy: 11.4 percent at 20-year life expectancy, 55.8 percent at two-year life expectancy, and 83.5 percent at one-month life expectancy. In a study of 60 White, Chinese, Hispanic, and Black American disabled older adults, 75 percent would want their clinician to talk to them about prognosis if they had less than one year to live; results were similar across ethnic groups . In another study of 214 community-dwelling adults with disability and advanced cancer, chronic obstructive pulmonary disease, or congestive heart failure, over half of subjects wanted to discuss prognosis with their doctor . Variation by age and gender was not found in these studies, though other studies in patients with cancer suggest that older patients and men are more reluctant to discuss prognosis [42,43].
In some cultures, there is a higher prevalence of wanting to avoid prognostic information or to prefer that family be told the information [44-47]. Clinicians should be aware that variation within cultures is likely as great as variation between cultures, however, and avoid stereotyping based on characteristics such as age, sex, or race/ethnicity . (See "Discussing serious news", section on 'Patients’ preferences when receiving serious news'.)
When a patient’s family expresses a strong preference against sharing prognostic information with the patient, clinicians should first explore the family’s concerns about sharing the information, as family members may have important information. The clinician should explain that the decision about whether to receive prognostic information belongs to the patient. The family should be invited to join the discussion when the clinician asks the patient whether they want this information. The clinician should explain in advance that, if the patient requests prognostic information, it will be shared, and if the patient prefers not to receive the information, the clinician will not provide it. Usually, this approach is acceptable to the family, who then hears directly from the patient about their information preferences, reducing conflict and disagreement.
Patients who prefer that prognosis not be discussed with their health providers cite several reasons for this preference: concern that discussing prognosis is too emotionally difficult, that the information would not be useful, or that they may not believe that doctors can estimate prognosis accurately .
Clinicians should offer to discuss prognostic information but respect the preferences of patients who are not willing or prepared to discuss prognosis. For example, clinicians can say:
●“In order to make this decision (or ‘to plan for the future’), I think it might be helpful if we talked about how much time you might have left to live. What are your thoughts on discussing this information?”
If the need to discuss prognosis is critical to decision-making, but patients are reluctant to discuss the information, clinicians should obtain an understanding of the reasons for reluctance. Understanding these reasons may provide important clues about how to approach this difficult topic in a way that is acceptable to the patient . For example, clinicians might say:
●“Please help me to understand why you would prefer not to discuss this information, so that I have a better sense of how to approach these difficult topics with you in the future.”
In some cases, when the decision is particularly urgent, prognosis might be brought into the conversation hypothetically, to obtain a sense of the patient’s values and goals. For example, the clinician might say:
●“If you knew that time was short for you, what would be important to you in making this decision?”
Alternatively, patients may cede decision-making to a caregiver who is more willing to discuss prognosis.
When to discuss prognosis — Ideally, prognosis discussions should take place when the patient is not acutely ill and is therefore better able to process and understand information and is not overwhelmed by anxiety due to a serious medical event. Unfortunately, these discussions too often are initiated after acute deteriorations or progression of disease . Patients may be overwhelmed with anxiety and there is evidence to suggest that patients tend to opt for highly aggressive therapies when they are anxious or experiencing strong emotions [51,52]. A study of 1231 patients with stage IV lung or colorectal cancer found that patients who have end-of-life discussions with their clinician prior to the last month of life were less likely to receive highly aggressive care, such as hospitalizations and chemotherapy .
How to discuss prognosis — As with any discussion in which clinicians need to share serious news with patients, clinicians need to prepare in advance for the discussion . This includes making sure the discussion will take place in a private setting without disturbances such as frequent pager calls. Patients should have the support of a caregiver, if desired, to provide support and help process and interpret the prognostic information.
Good communication skills are essential. A year 2017 guideline from the American Society of Clinical Oncology (ASCO) provides some recommendations about how to optimize communications about goals of care, treatment options, and prognosis (table 3) . The Serious Illness Conversation Guide provides a practical framework to guide conversations in patients with advanced cancer and other serious illnesses (table 4). In a trial of 278 outpatients with advanced cancer and 91 oncologists, use of the Serious Illness Conversation Guide resulted in earlier and higher-quality communication about serious illness and reduced patient anxiety and depression, though it had no impact on receipt of goal concordant care or peacefulness at the end of life [23,56].
The following section is a general guide to a stepwise strategy for discussing prognosis; it is expected that clinicians will adapt this approach to the case at hand and their own communication style. A more general discussion of this type of model or framework to deliver bad, sad, or serious news is presented elsewhere. (See "Discussing serious news", section on 'Approaches to breaking serious news: existing models and key elements'.)
Stepwise strategy for discussion
●Step one: Place the prognosis discussion in context. A discussion of prognosis rarely occurs in a vacuum and is almost always tied to treatment decisions. Clinician should explain that an understanding of prognosis can be helpful, or even critical, to making informed health care decisions and life choices.
●Step two: Ask patients and families/loved ones about their own understanding of prognosis. This can be accomplished with a question such as:
•“So that I can understand where we are starting from, I wonder if you have some general sense of how much time you might have, and how you came up with that information?”
This step is often overlooked. However, it is critically important to have a sense of the patient and family’s understanding of prognosis before revealing the prognosis. Clinicians can gauge this by asking themselves, “Will the news that the patient has only a few weeks left to live be a huge surprise to him, or is he fully aware?”
Despite this, studies of surrogates of critically ill patients suggest that families/loved ones rarely rely on clinicians alone for prognostic information. In one study of 179 surrogates of critically ill patients, less than 2 percent relied exclusively on clinicians for prognostic information . Other sources of information used by patients and families/loved ones include their own observations about the trajectory of illness, the patient’s will to live, their sense of optimism, and faith. Being aware of the understanding of the patient and family about prognosis, and the reasons for their perception, is a critical starting point for clinicians, who may later wonder why their prognostic estimates were “not believed.”
●Step three: Deliver the prognostic information. Clinicians should use language that emphasizes the range of possible outcomes. Ranges are important because of the inherent uncertainty in any prognostic information. For example, one might give a prognosis for life expectancy in hours to days, days to weeks, weeks to months, months on one hand, or months to a year at most. When dealing with a choice between treatment options, such as chemotherapy for advanced cancer versus hospice, presenting the best and worst case for each treatment can help clarify the range of possibilities . A clinician could say, for example:
•“With chemotherapy, in the best case you would have few side effects and might live for several months, perhaps as long as a year. In the worst case, the chemotherapy might cause you to feel sick and spend most of your time in bed or the hospital, and complications from the chemotherapy would cause you to die sooner, perhaps as soon as one month. With hospice, in the best case you might live for six months and those would be months where you are alert and your symptoms are well controlled. In the worst case, your symptoms from the cancer would be hard to control and the cancer might progress rapidly, and time could be as short as two to three months.”
Clinicians can follow best- and worst-case scenarios by making estimate of the most likely case. In the previous example, the clinician might conclude by saying:
•“In the most likely case with chemotherapy, you would have some side effects that are manageable, and it would help you live a month or two longer than you might with hospice, maybe six months or so. With hospice, the most likely case is that you would avoid the symptoms and complications from the chemotherapy, and you might have three or four symptom-free months with your family.”
Clinicians can acknowledge that dealing with uncertainty may be difficult psychologically for patients and families/loved ones, as in:
•“I wish the information were more certain. I understand that you would like more certain information. But that certainty does not exist. How can I help you to cope with that uncertainty?”
Some patients may find it helpful to refocus on the present, rather than dwell on the uncertainty of the future. When time is likely to be short, helping patients focus on personal goals can be an antidote to strong feelings that are often generated by an awareness of limited time.
●Step four: Respond to emotion. Attending to the affective response to prognostic information is of critical importance. Clinicians need to be sensitive to patient and family distress at hearing the prognosis. This can be accomplished by providing supportive statements and questions. However, it is important to avoid providing false reassurance to the patient in response to intense distress. One useful mnemonic for attending to affect are the NURSE statements and questions, as described in a table (table 5) .
●Step five: Agree on a plan for the next steps. This may be a follow-up appointment to discuss treatment options in light of the prognostic information, additional testing that may help clarify the prognosis, or a supportive visit to provide additional time to process the emotional content of the prognosis and its implications. For many patients, “structure is the antidote to despair,” and they rely on a sense of a plan for the future to help cope with unwanted news. For example:
•“At this time, I suggest we focus on getting you enrolled in hospice now so that we can get you out of the hospital and let you be with your family.”
●Prognostic information is a critical component of informed shared decision-making. The risks and benefits of any test or treatment must be considered in light of the patient’s expected prognosis. Prognosis has bearing both on medical decisions as well as other factors important to patients, including addressing spiritual, financial, and interpersonal issues. (See 'The importance of prognosis in clinical practice' above.)
●Patients who have a good understanding of their prognosis tend to make different medical decisions than those who are not aware. A clinician’s fear of discussing prognosis because it may provoke negative feelings is not an adequate reason for avoiding discussions of unwelcome news. Evidence is mounting that inadequate preparation for death leads to poor family bereavement outcomes, including depression, a sense of regret, and complicated grief. (See 'Rationale for discussing prognosis' above.)
●Although prognostic information from a prognostic tool is not perfectly accurate, clinicians’ own empiric estimates of prognosis are likely to be far less accurate. The combination of information from prognostic tools and clinical judgment results in a better prognostic estimate than either alone. (See 'The science of estimating prognosis' above.)
●A majority of patients with serious illness would want their clinician to discuss prognostic information. Such information should be offered to patients and never forced upon them. (See 'Assessing patient and/or family readiness to discuss prognosis' above.)
●Discussions of prognosis should proceed in a stepwise manner that reflects the science of how patients and families/loved ones perceive and respond to prognostic information. Above all, prognosis discussions must proceed with an intense sensitivity to the emotional reactions of patients and families/loved ones. (See 'Discussing prognosis' above.)