Approach To Prognostication

Approach to Prognostication – 18 Interesting Facts

Prognosis is the likelihood of a patient developing a particular outcome over a specific amount of time.¹ Physicians can prognosticate about life expectancy, functional status, quality of life, and the ability to live independently²
Despite perceived barriers, research shows that most patients want prognostic information and that sharing it does not diminish hope and may improve anxiety, depression, and the patient-doctor relationship
Prognostic information is important in decision-making for both clinicians and patients
Prognostication is guided by several principles⁹⁴⁰
It is an iterative process
Prognostic accuracy may change as a disease progresses
Prognostic accuracy varies with the patient population and prognostic timeframe
Uncertainty is inherent to prognostication
Patients and surrogates do not solely rely on a physician’s prognosis in their interpretation of prognosis
Surprise question (ie, “Would I be surprised if this patient died within the next year?”) is useful to screen patients who may benefit from a serious illness conversation or a discussion about prognosis
Estimate prognosis using a combination of clinical gestalt, disease-specific models, and functional tables
Discuss prognosis estimates with treating specialists. Prognostic alignment is a unified prognostic narrative developed among multiple teams (or specialists) caring for the patient¹⁹
Communicate the prognosis using patient-centered language in a serious illness discussion with the patient, family, and relevant treating practitioners
Clarify what prognostic information is helpful for a patient and ask permission before sharing prognostic information
When communicating an approximate life expectancy, emphasize uncertainty inherent to prognostication by describing timeframes as a range (eg, days to weeks, weeks to months, months to years)
For patients who do not want information about possible life expectancy, use best-case/worst-case/most likely scenario or pictorial depictions to describe clinical trajectories for their primary diagnosis
Interpret the prognosis and make recommendations about clinical decisions that account for the patient’s expect prognosis. Explain the recommendations³³
Multiple clinical tools have been validated in palliative care settings to provide prognostic information for patients in the last phase of life

Basic Information

Background Information

The term “prognosis” refers to the likelihood of a patient developing a particular outcome over a specific amount of time.¹ Physicians can prognosticate about life expectancy, functional status, quality of life, and the ability to live independently²
Historically, prognostication was 1 of the 3 cardinal skills of competent physicians, alongside diagnosis and treatment³
- As medical advancement led to development of treatments for conditions that were previously fatal, medical education allotted less attention to prognosis
- Prognostication has garnered renewed interest as the burden of chronic incurable diseases has increased. Additionally, the creation of the hospice benefit brought about the need for physicians to prognosticate an expected survival period of 6 months or less to identify eligible patients
Prognostic information is essential to clinical decision-making for both practitioners and patients
Most patients want to know prognosis.⁴ However, misinformation and misunderstanding of prognosis are prevalent among patients⁵⁶⁷
In general, clinicians are overoptimistic when estimating prognosis and communicate an overly optimistic prognosis than their own estimate. Additionally, patients generally interpret prognosis more optimistically than what is communicated by clinicians⁸

Terminology

Prognosis is the likelihood of a patient developing a particular outcome over a specific amount of time¹
Prognostic accuracy can be difficult to define and measure. 2 key aspects of prognostic accuracy are discrimination and calibration⁹
- Discrimination is how effectively a prognostic tool distinguishes between patients who will survive and those who died within a specific time frame
  - A concordance statistic is commonly used to assess discrimination, with values ranging from 0.5 to 1. In binary models (death versus survival), it is measured as the area under the receiver operating characteristic curve
    - A value of 0.5 indicates no better discrimination than chance, while a value close to 1 represents excellent differentiation
- Calibration evaluates how closely a prognostic model’s prediction of survival probability aligns with the actual outcomes observed on a population level
  - For example, if a model predicts a 70% survival probability for a group of patients, calibration determines whether approximately 70% of those patients actually survive
- Discrimination focuses on the model’s ability to differentiate between individual outcomes. Calibration assesses the agreement between predicted probabilities and actual outcomes, typically across a group of patients
- Additionally, values such as sensitivity, specificity, positive predictive value, and negative predictive value can be useful in evaluating prognostic accuracy

Barriers to Prognostication

Discussing prognosis is challenging for physicians because of:
- Emotional content of prognostic disclosures
- Considerable uncertainty and challenges inherent to prognostication¹⁰
Research suggests that physicians are reluctant to provide prognostic information
- Many wait to be asked before disclosing prognostic information and most consciously present more optimistic estimates than they believe to be accurate¹¹
- Physicians commonly use strategies such as realism, optimism, and avoidance in disclosing prognosis¹²
Many barriers exist to discussing prognosis
- Patient-related barriers
  - Stress and emotion may affect how patients process complex prognostic information¹³
  - Cultural norms and communication preference vary among patients²
  - Independent coping strategies may influence how patients prefer prognosis to be communicated¹⁴
- Family-related barriers
  - Families occasionally request that physicians withhold information from patients. Often, the reasons for this are cultural (eg, that disclosing is impolite in some cultures or considered bad luck)¹⁵
  - A recent large multicenter survey of health care workers (nurses, resident physicians, and attending physicians) identified that family members’ difficulty accepting a poor prognosis and disagreement among family members were 2 of the most important barriers to discussing goals of care¹⁶
- Clinician-related barriers
  - Lack of time
  - Lack of confidence, support, or training to initiate and guide the conversation¹¹
  - Concerns that prognostic information will cause depression or take away hope¹⁷
  - Concerns about the inherent uncertainty of prognostication¹¹
  - Difficulties in prognostication for nonmalignant disease¹⁸
  - Concerns that prognostic errors will impact the patient’s confidence in the clinician¹¹
  - Perceptions that prognostic models have poor sensitivity and specificity¹⁸
- System-related barriers
  - US health care system’s default is to provide life-prolonging and life-sustaining treatment unless a patient actively declines it. Physicians are likely to spend more time talking about treatment options than prognostication
  - Hospitalized patients are often cared for by multiple team members. Teams may fail in communicating with each other, which results in prognostic misalignment. Communication is often mediated over short messages (pages or texts), may be mediated by junior team members, and may fail to convey prognostic information or leave space for further discussion¹⁹
  - There is ambiguity about which clinician should be responsible for prognostication discussions (eg, oncologist versus primary care practitioner versus palliative care practitioner)
Despite perceived barriers, research shows that most patients want prognostic information and that sharing it does not diminish hope, cause depression, or impact the physician-patient relationship
- Studies suggest that receiving honest prognostic disclosure does not diminish hope.²⁰ Multiple studies found that patients who received more frequent prognostic disclosure were more likely to report communication-related hope²¹²²²³
- An overt majority (93%) of surrogate decision-makers reported that avoiding discussions about prognosis is an unacceptable way to maintain hope¹⁰
- A study of prognostic disclosures in patients with advanced cancer did not find an association between prognostic disclosure and harm to the patient’s emotional well-being or relationship with the physician⁶
- A study of prognostic disclosures in patients with advanced liver disease found that patients who wished they had more prognostic information had higher rates of anxiety and depression and a higher symptom burden compared with patients who felt they had adequate prognostic information²³
Failure to share prognostic information with patients impedes their ability to make treatment decisions in line with their values. Patients who do not receive prognostic information may spend more time in the hospital, undergo more aggressive treatments, and lose time with their families¹⁷

Ethical Principles Guiding Prognostication Conversations

Underlying Ethical Principles

Clinicians balance transparency with sensitivity in disclosing prognostic information. Balance between providing honest information and maintaining hope is guided by the ethical principles of autonomy, beneficence, and nonmaleficence
- Multiple studies suggest that when patients are given truthful prognostic information, their hope is maintained²⁰²¹
- Clearly documenting an estimated prognosis (especially for specialty treating physicians) can assist other medical practitioners in understanding the clinical trajectory of diseases outside of their primary specialty. As clinics and hospitals move towards open-access notes, clinicians should use caution when documenting a prognosis that has not been disclosed to the patient
Clinicians interpret prognosis to offer treatments that are beneficial and not harmful at a certain phase in the illness. Principle of nonmaleficence encourages clinicians to avoid unnecessary suffering and withhold nonbeneficial interventions
Clinicians use patient-centered language when communicating prognosis. In respect for the ethical principle of autonomy, clinicians share prognostic information after asking the patient’s permission
- Patients may exhibit cultural differences in the desire to know various details about their disease
  - Clinicians often assume that individuals want to know as much as possible about the disease process and trajectory in support of the ethical principle of patient autonomy²⁴
  - Some patients believe that discussing death may quicken the dying process and do not want direct communication about this topic²⁴
  - A large, stratified survey found that Korean American and Mexican American people were significantly less likely than European American and Black people to believe that the patient should be told about terminal prognosis²⁵
- Patients of different cultural backgrounds may have different preferences about group versus individual approaches to decision-making²⁶
  - A large survey found that Korean American and Mexican American people were more likely to believe that the family, not the patient, should make decisions about life-sustaining treatment²⁵
- If the patient does not want prognostic information or would prefer that family members make medical decisions on their behalf, the clinician should respect their preferences. The patient has a right to be informed and make medical decisions, but not an obligation²⁷
Shared decision-making is a better model for sharing prognostic information than the doctrine of informed consent³
- Principle of informed consent was developed for research ethics, not clinical care, and typically refers to a 1-directional transmission of information from physician to patient. Presenting realistic prognostic information bluntly, without structuring the conversation, can be perceived as uncaring¹²
- Disclosing and discussing prognosis should be a 2-way exchange between patient and physician over time

Patient Preferences/Autonomy

Principle of protecting patient autonomy is central to discussing prognosis and its implications:
- Respect the patient’s right to make informed decisions about their care and ensure they have access to all relevant information
- Encourage the patient to express values, goals, and treatment preferences
- Assess the patient’s understanding of their illness, prognosis, and potential outcomes of different treatment options
- Clarify that patients have the right to accept or decline specific interventions
Patients do not lose the right of self-determination when they lose capacity. Several tools, such as an advance directive, living will, or designation of a surrogate decision-maker, are mechanisms through which the patient’s voice can continue to direct their care²⁸
Determine whether patients have designated a medical power of attorney. If they have not, ask who would serve as their surrogate decision-maker should they become incapacitated. Encourage patients to discuss their goals and preferences with their surrogate decision-maker. Examples of questions patients should discuss with their surrogate include:²⁹
- What is most important to you in life?
- How would you think about the tradeoffs between living as long as possible and avoiding prolonged disability? Would you prefer treatments that might extend your life, even if it involves risks of significant disability, or would you rather avoid anything that reduces your independence?
- Would you rather die at home or be in the hospital during the final days of your life?
- Are there any spiritual, religious, or cultural values that you consider when making decisions about your care?
- Have you already made decisions about the kind of medical care you would or wouldn’t want, like in a living will, or any specific instructions about life-sustaining treatment?
Asking a patient to appoint a surrogate decision-maker and to discuss their goals with them is an important task that can be done at any stage of illness and can provide opportunities to elicit a patient’s thoughts about end-of-life care, without the pressure of discussing specific decisions³⁰
After disclosing prognosis to the patient (or their surrogate), use shared decision-making to inform a treatment plan³¹
- Shared decision-making is a collaborative process that allows patients, or their surrogates, and clinicians to make health care decisions together, taking into account the best scientific evidence available and the patient’s values, goals, and preferences³¹
- Patients and surrogates may choose from various treatment options recommended by the treatment team
- Patients and surrogates may allow clinicians to guide decision-making, accounting for the patient’s or surrogate’s preferences
- When deciding between value-laden choices, evidence suggests that 75% to 85% of surrogates prefer to share responsibility for decision-making with clinicians or make decisions after considering clinicians’ recommendations. Some patients or surrogates (5%-15%) may wish to defer decision-making to clinicians³¹

Role of the Surrogate Decision-Maker

A durable power of attorney for health care is a legal document that grants a particular person the ability to make health care decisions on another’s behalf if they can no longer make health care decisions for themselves²⁹
A surrogate’s role is to apply these standards in the following order:²⁸
- Implement any wishes or instructions explicitly expressed by the patient
- Use substituted judgment and decide in accordance with the patient’s preferences and values
- Act in the patient’s best interest
Clinicians should identify whether the patient has designated a health care proxy or medical power of attorney for medical decisions
- If the patient has the capacity to make their own decisions, they are the decision-maker
- If the patient does not have the capacity to make their own decisions, the surrogate is the decision-maker in discussions
If a patient has not previously appointed a surrogate decision-maker, a default decision-maker is specified by state laws using the patient’s next of kin
- A typical default order involves a spouse, then adult children, then parents, then siblings, then other kin
- If no kin are available, in some states, an adult friend can serve as default surrogate
Clinicians guide the surrogate to use substituted judgment (what the patient would want) rather than their own preferences
Clinicians involve the medical decision-maker (either the patient or surrogate) in ongoing discussions that include prognostic information. Prognosis informs a patient’s (or surrogate’s) decisions about electing to proceed with certain types of treatment, as well as making personal, familial, financial, and legal decisions

Guide for Medical Practitioners

Rationale for Prognostication

Prognostication plays a crucial role in guiding clinical decision-making for physicians to recommend appropriate treatments⁹
- Risk-benefit ratio for many interventions increases as estimated length of survival decreases⁹
  - 1 meta-analysis found that 33% to 38% of patients received nonbeneficial treatments near the end of life³²
  - Accurate prognostication of life expectancy can assist physicians in interpreting and educating patients about the risks and benefits of possible treatments
  - Patients with limited life expectancy are unlikely to live long enough to benefit from certain tests or treatments³³
- Anticipated life expectancy can impact short-term clinical decisions (eg, how to approach current care, cancer screening, surgical versus nonsurgical management) and long-term decisions (ie, anticipating care needs and transitioning to a nursing home or discussing advance care planning)³³
- Prognosis may also impact access to certain health benefits. For example, hospice is provided to those with a prognosis of 6 months or less²
Prognostication is crucial in guiding patient decision-making. Prognostic information allows patients and family members to prepare for what is to come and to make decisions in line with their values³⁴
- Examples of life choices that hinge on prognostic information:³
  - Moving in with adult children in anticipation of the loss of independence
  - Undertaking legal measures to allow a caregiver or surrogate to control finances
  - Repairing relationships
  - Receiving last rites or other religious rituals to prepare oneself for death
  - Arranging personal affairs
  - Ensuring meaningful time with friends and family and saying goodbye to loved ones before death³⁵
- Expected quality of life also impacts patients’ decisions to accept or forgo certain treatments
  - In 1 study, 74.4% of patients said they would forgo treatment if the treatment burden was low but the probability of severe functional impairment was high. 88.8% of patients would forgo treatment if probability of severe cognitive impairment was high³⁶
- Accurate prognostic information may allow family members or surrogates to better support the patient and each other¹⁰
Patients want prognostic information³⁷
- 1 study of patients with advanced cancer found that although 71% of patients want information about their life expectancy, only 17% recall having a prognostic conversation with their physician⁶
- 1 survey found that about 80% of patients wanted a qualitative prognosis (ie, information about whether they will die from their disease) while only about 50% wanted a quantitative prognosis (estimated life expectancy)³⁸
- A systematic review of patients with cancer found that many patients want information about the chance of a cure, the extent of disease spread, their life expectancy, and the impact of cancer on their quality of life. Many patients indicated they wanted to be asked first if they would like to receive prognostic information³⁸
- Another study found that even when patients request prognostic information, physicians reported communicating accurate information only 37% of the time and otherwise provided no estimate, an intentional overestimate, or an intentional underestimate³⁹
Prognostication provides an opportunity to correct misconceptions
- Patients overestimate their life expectancy
  - In 1 cohort of patients with advanced cancer (n = 590), only 50% of patients were willing to estimate their life expectancy. In those who did so, 45% of patients overestimated their prognosis by 2 years or more⁶
- Patients overestimate their chance of cure
  - In 1 large study (n = 1193) of patients with metastatic lung and colorectal cancer, 74% incorrectly believed chemotherapy was a curative treatment⁵
- 1 systematic review found that even after specific interventions aimed to improve prognostic understanding, prognostic misconception was common (31%-95%). Further research is necessary to better understand which interventions most effectively correct prognostic misconceptions and whether single-time interventions or repeated interventions are necessary⁷

Principles of Prognostication

Prognostication is an iterative process, not a single event in the clinical journey⁹
- A patient’s prognosis may evolve based on their response to treatment, development of an acute complication, or competing comorbidities
- Prognosis must be revisited over the course of disease
Prognostic modeling may change throughout stages of disease⁹
- For example, in early-stage cancer, prognosis may be influenced by cancer-specific characteristics such as tumor stage, histologic grade, and mutation status. In advanced disease, patient-related factors such as performance status, dyspnea, and cachexia may drive prognosis
Prognostic accuracy is impacted by the definition of accuracy, patient population, and prognostic timeframe⁹
Uncertainty is inherent to prognostication⁹
- When communicating prognostic information to patients, clinicians should highlight the inherent uncertainty. Using ranges (eg, weeks to months) for survival predictions can help underscore the uncertainty in prognostication
Patients and surrogates do not solely rely on a physician’s prognosis in their interpretation of prognosis
- In 1 study of surrogate decision-makers, less than 2% reported that their beliefs about prognosis relied exclusively on information from physicians⁴⁰
- Contributing factors to the patient’s or surrogate’s belief about prognosis include perceptions of the patient’s strength of character, perceptions about the patient’s will to live, the patient’s unique history of illness and survival, the surrogate’s own observations of the patient, the surrogate’s belief that being present for their loved one may improve the prognosis, and the surrogate’s intuition and faith⁸⁴⁰

Approach to Prognostication

Use the surprise question to screen patients who may benefit from a serious illness conversation or a discussion about prognosis: “Would I be surprised if this patient died within the next year?”
- If the answer is “no,” initiate a serious illness discussion with the patient with the input of their relevant consulting practitioners
- The surprise question, although based on gestalt, has been shown to have modest ability to predict survival
  - 1 study of emergency medicine physicians treating patients aged 65 years and older found that the surprise question sorted the patient who lived from the patient who died 67% and 72% of the time on univariate and multivariate analyses, respectively⁴¹
  - 1 meta-analysis estimated that the surprise question has a sensitivity of 71.4%, specificity of 74%, a high negative predictive value (greater than 85%), and a low positive predictive value (12%-47% with varying mortality rates).⁴² Another meta-analysis found that pooled accuracy of the surprise question was 74%⁴³
- For patients with serious illness, the benefits of these discussions include promoting patient autonomy and patient-centered care, identifying care in line with the patient’s values, avoiding unwanted care, and conferring psychological and emotional support to patients and families facing complex decisions⁴⁴
Clinicians may communicate prognosis in multiple settings. For example, discussions about prognostication are relevant to a planned serious illness discussion (ie, multidisciplinary family meeting) for a critically ill hospitalized patient, a clinic visit in a specialty clinic, or a primary care clinic visit to discuss advance care planning
- Conversations may be planned (if initiated by the clinician) or unplanned (if initiated by the patient)
- In acute settings, communication of prognosis may occur in a planned family meeting (such as a serious illness discussion) but may also occur in 1-on-1 conversations with the patient
  - The hospital setting may allow a physician to build rapport over consecutive daily visits
  - Acute care settings may signal a health event that prompts either patients or clinicians to reevaluate prognosis
- In outpatient settings, communication of prognosis may be initiated by either the patient or physician
  - Most patients want their physician to initiate conversations about prognosis³⁰
  - Clinician-initiated conversations allow a clinician to adequately prepare and involve other treating specialists, if necessary
Estimate prognosis using a combination of clinical gestalt, disease-specific models, and functional tables
- If the patient has a dominant life-limiting illness, use a disease-specific prognostic calculator or research data²
- If the patient does not have a dominant life-limiting illness, use a general prognostic calculator. Calculators that incorporate age, functional status, and clinical location may offer improved accuracy²
- Tailor the prognostic estimation using information specific to the clinical scenario, such as comorbid conditions, disease- or treatment-related complications, frailty, recent or recurrent hospitalizations, and other contributing social factors²
Discuss prognosis estimates with treating specialists. Prognostic alignment is a unified prognostic narrative developed among multiple teams (or specialists) caring for the patient¹⁹
Communicate the prognosis using patient-centered language in a serious illness discussion with the patient, family, and relevant treating practitioners. Tailor communication to the patient’s readiness
Interpret the prognosis and make recommendations about clinical decisions that account for the patient’s expected prognosis. Explain the recommendations³³
Formulate a treatment plan using a shared decision-making model
Emphasize that ongoing discussions about prognosis as disease progresses can ensure that patients receive care aligned with their values

Estimating Prognosis

Clinicians may estimate prognosis subjectively (ie, using clinical gestalt) or objectively (ie, using disease-specific prediction models)⁹
Determine whether a dominant life-limiting illness is present²
- If present, use a disease-specific prognostic calculator or disease-specific research to assist in estimating life expectancy
- If absent, use a general prognostic calculator (eg, https://www.eprognosis.ucsf.edu⁴⁵). Calculators that incorporate age, functional status, and clinical location may offer improved accuracy
Tailor the prognostic estimate using patient-specific information, such as comorbid conditions, disease- or treatment-related complications, frailty, recent hospitalizations, or social factors²
- Consider clinical context and illness trajectory (eg, acute versus chronic, reversible versus progressive)
- If using population-level statistics (eg, median survival), identify potential biases in the data and consider its timeframe. Recent advances in treatment may alter patient outcomes
- Adjust prognostic predictions based on age, comorbidities, functional status, and treatment preferences
- Consider psychosocial determinants of health, including support systems and access to care
- Collaborate with specialists for their input
Table 1 describes benefits and drawbacks to different models and tools used to estimate prognosis. Typically, prognosis is estimated using disease-specific or functional status data, then adjusted with clinician gestalt given patient-specific factors

Table

Table 1. Tools and models to approach prognostication with benefits and drawbacks.

Clinical tool	Description or examples	Benefits	Drawbacks
Physician estimates (ie, clinical gestalt)	Clinician’s intuitive assessment of overall health and prognosis based on observation and clinical experience	Rapid and accessible; requires no additional testingAllows clinician to synthesize their clinical experience into actionable insightsMay identify subtle cues not captured by structured tools or laboratory resultsIf clinician has a longitudinal relationship with a patient, clinical gestalt may involve dynamic assessments	Not standardized; may vary among practitionersProne to bias (typically with physicians overestimating prognosis)Less accurate for clinicians with less experience
Performance status scales	Tools that prognosticate are typically based on functional status. Examples include the Karnofsky Performance Scale, the Palliative Performance Scale, and the Eastern Cooperative Oncology Group Performance Status Scale	Typically easy to use in clinical settings and require minimal training and equipmentCan be applied across various diseases and settingsStrongly associated with survival, quality of life, and treatment toleranceCan be combined with other prognostic indicators for greater accuracy	Interrater variability can affect reliabilityDoes not account for reversible factors affecting performance and may be less accurate in certain contexts (eg, hospital setting for acute illness)
Disease-specific prognostic tools	Tools that assist with prognostication tailored to a specific disease process (eg, Model for End-Stage Liver Disease, Seattle Heart Failure Model)	Evidence-based: typically developed with robust clinical data and validated in disease-specific populationsAllow physician to prognosticate accurately about a single diagnosisMay be used in research as benchmarks for comparing interventions and outcomes in clinical trials	Individual patients may have both protective features and harmful risk factors not captured by prognostic predictors³Do not account for patients with multiple comorbidities; may oversimplify complex clinical situationsMay become outdated as new treatments, diagnostics, and disease understanding evolve
Biomarkers	Measurable biological indicators that offer objective data to predict a disease trajectory, treatment response, or survival. Examples include oncologic biomarkers (eg, prostate-specific antigen, cancer antigen 125) that indicate cancer progression and its response to treatment and brain natriuretic peptide, which predicts outcomes in heart failure, or other inflammatory markers	Provides measurable and reproducible information and is standardized across clinical populationsMay allow for early detection of disease progression earlier than clinical symptomsCan be used in dynamic monitoring to track disease status over time or indicate response to treatment	Can be context dependent and may be affected by comorbidities or the clinical settingSome biomarkers (eg, C-reactive protein) are not disease specific and may reflect multiple conditionsRequires some clinician interpretation; thus prone to subjectivity
Actuarial data	Statistical data used to prognosticate based on historical data derived from large representative samples. Examples include life tables and the Surveillance, Epidemiology, and End Results Program in patients with cancer	Assist in understanding disease trends and prognosis at the population levelFreely available to researchers and clinicians	May not account for changes in medical technology or new treatmentsDo not account for individual protective or harmful risk factors
Dynamic assessments	Tracking clinical trends over time, such as weight loss, recurrent hospitalizations, or functional decline	Account for real-time changesUseful for refining prognosis as new data emerge	Require consistent monitoring and documentationMay be disrupted if care is fragmented with different practitioners or locations

Clinical Judgment

Clinical judgment is clinician gestalt, which includes their experience, pattern recognition, and intuition. Physicians use measurable factors, such as laboratory results and diagnoses, and often use the rate of patient deterioration in their estimates¹⁸
Clinical experience is associated with improved prognostication abilities⁴⁶
Longer doctor-patient relationship is associated with less prognostic accuracy⁴⁶
Clinicians may predict survival using 3 general forms: the temporal approach (How long will this patient live?); surprise question (Would I be surprised if this patient died within a [specific timeframe]?); and the probabilistic approach (What is the probability of survival of this patient in [specific time frame]?)⁴⁷
- Studies suggest that when making temporal estimates (ie, how long will this patient live?), physicians tend to overestimate prognosis.⁴⁸ 1 prospective cohort study of terminally ill patients found that physicians’ prognostications about life expectancy were overoptimistic 63% of the time, typically overestimating survival by factor of 5⁴⁶
- Studies suggest that clinician estimates perform better with a probabilistic approach than a temporal approach⁴⁹⁵⁰
- Surprise question predicts survival relatively well
- Limiting the timeframe of prediction may allow clinicians to predict survival more accurately⁴⁷

Prognostic Models and Tools

Clinicians may use different prognostic models, tools, or calculators to assist in prognostication
- Disease-specific models
  - Offer mortality predictions for patients with specific diseases
  - Examples include:
    - BODE index (BMI, Obstruction, Dyspnea, Exercise) for patients with COPD (chronic obstructive pulmonary disease)
    - Seattle Heart Failure Model for patients with heart failure
    - ADEPT score for patients with dementia
- Life tables and actuarial data
  - Life tables are a statistical tool used to summarize a population’s mortality and survival patterns over time. Life tables provide a way to estimate life expectancy based on age, sex, and other demographic factors
  - Life tables of the general population describe how life expectancy decreases with age. This information is often intuitively incorporated into a clinician’s gestalt when estimating prognosis
  - Disease-specific life tables provide information about survival for patients with a specific disease process (eg, a 5-year survival curve for patients with prostate cancer)
- Performance status models
  - Multiple studies indicate that functional status (at various times in the clinical trajectory) are key predictors of clinical outcomes
    - In a prospective cohort of hospitalized patients, a functional axis (involving impairment in instrumental activities of daily living, a Mini Mental State Examination score lower than 20, and a short-form Geriatric Depression Scale score of 7 or higher) predicted 90-day and 2-year mortality after hospitalization⁵¹
    - A large retrospective cohort study (n = 80,020) found that diminished functional status at discharge was an independent predictor of long-term care facility admission and death at 180 days from discharge⁵²
  - General (non–disease-specific) calculators often use patient information and performance status including ability to complete specific activities of daily living or instrumental activities of daily living in mortality predictions
    - Most commonly used functional tools are the Karnofsky Performance Scale, the Palliative Performance Scale, and the Eastern Cooperative Oncology Group Performance Status Scale⁵³
    - Palliative Performance Scale (https://www.npcrc.org/files/news/palliative_performance_scale_ppsv2.pdf)
      - Scored in 10% increments from 0% (death) to 100% (fully ambulatory)
      - Focuses on ambulation, activity level, evidence of disease, self-care, oral intake, and level of consciousness
      - Commonly used in palliative care to predict survival and guide advance care planning
    - Karnofsky Performance Scale (https://www.npcrc.org/files/news/karnofsky_performance_scale.pdf)
      - Scale ranging from 0 to 100 in 10-point increments, with 100 indicating full functional ability and 0 indicating death
      - Scores greater than 70% suggest a patient can care for themselves independently, while scores lower than 50% suggest significant disability
      - Primarily used to assess functional status and predict prognosis in patients with cancer
    - Eastern Cooperative Oncology Group Performance Status Scale (https://www.ecog-acrin.org/resources/ecog-performance-status)
      - Ranges from 0 (fully active) to 5 (death), with intermediate scores reflecting increasing levels of disability
      - Widely used in oncology to assess the impact of disease on daily living and determine eligibility for cancer-directed treatment
Typically, prognosis is estimated using disease-specific or functional status data, then adjusted with clinician gestalt given patient-specific factors
Tables 2 and 3 detail select prognostic calculator resources for clinicians
- Prognostic calculators are evaluated by discrimination and calibration
  - Discrimination is a prognostic test’s ability to differentiate between individuals who do and do not have the predicted outcome
  - Calibration evaluates how closely a prognostic model’s prediction of survival probability aligns with the actual outcomes observed on a population level
  - Discrimination and calibration of a test may vary as the cohort population changes (eg, a heart failure mortality predictor may have different discrimination in populations with heart failure with preserved ejection fraction versus with reduced ejection fraction)

Table

Table 2. Disease-specific prognostic tools.

Serious illness	Tool	Population and inputs	Outcome	Accuracy	Online calculator
Heart failure
	Seattle Heart Failure Model (outpatient setting)	Population: adults with heart failureInputs include patient characteristics, ejection fraction, NYHA class, laboratory values, and treatment with common heart failure medications	1- and 5-year mortality	Discrimination of 72.9%⁵⁴Discrimination may vary with specific cohort (eg, may be more accurate in patients with heart failure with reduced ejection fraction than in those with preserved ejection fraction)⁵⁵	https://www.mdcalc.com/calc/3808/seattle-heart-failure-model⁵⁶
	MAGGIC Heart Failure Risk Score	Population: adults with heart failureInputs include age, patient characteristics, ejection fraction, NYHA class, BMI, and treatment with common heart failure medications	1- and 3-year mortality	Discrimination of 74% for 3-year mortality⁵⁷⁵⁸	https://www.mdcalc.com/calc/3803/maggic-risk-calculator-heart-failure⁵⁹
COPD
	BODE index	Population: patients with COPDInputs include FEV₁, 6-minute walk distance, BMI, and dyspnea scale score	4-year survival	Discrimination of 74%⁶⁰	https://www.mdcalc.com/calc/3916/bode-index-copd-survival⁶¹
Dementia
	Deardoff mortality risk for community-dwelling older adults with dementia⁶²⁶³	Population: adults aged 65 years and older with dementia living in the communityInputs include age, BMI, sex, instrumental ADL impairments, and presence and absence of several comorbidities	1-, 2-, 5-, and 10-year all-cause mortality	Discrimination of 73%, 73%, 75%, and 84% at 1, 2, 5, and 10 years respectively⁶²	https://www.eprognosis.ucsf.edu/dementia.php⁶³
	Deardoff model to predict need for nursing home level of care in community-dwelling adults with dementia⁶⁴⁶⁵	Population: adults aged 65 years and older with probable dementia living in the communityInputs include age, sex, ADL impairments, and instrumental ADL impairments	2-, 5-, and 10-year risk of needing nursing home care*	Discrimination of 66%-72% if completed by the proxy and 64% based on patient’s self report⁶⁴	https://www.eprognosis.ucsf.edu/nhloc.php⁶⁵
	ADEPT Score⁶⁶	Population: adults aged 65 years and older with probable dementia living in a nursing homeInputs include age, sex, ADL impairments, presence of sufficient oral intake, BMI, and presence of pressure ulcers or heart failure	6-month survival	Discrimination of 67%⁶⁶	https://www.eprognosis.ucsf.edu/mitchell.php⁶⁷
Cirrhosis
	MELD-Na score	Population: adults with chronic liver disease on liver transplant waiting listInputs include serum creatinine, serum total bilirubin, and serum sodium levels; INR; and whether patient has recently received dialysisNew prognostic model, titled MELD 3.0, better accounts for female sex and serum albumin level⁶⁸	3-month survival	Discrimination of 86%-88%⁶⁸⁶⁹	https://www.mdcalc.com/calc/10437/model-end-stage-liver-disease-meld⁷⁰
End-stage kidney disease
	Cohen Prognostic Model	Population: patients receiving hemodialysisInputs include age, serum albumin level, presence or absence of dementia and peripheral vascular disease, and answer to surprise question†	6-, 12-, and 18-month mortality	Discrimination of 71%-72% at all 3 time points⁷¹	https://www.qxmd.com/calculate/calculator_135/6-month-mortality-on-hd⁷²

Caption: ADEPT, Advanced Dementia Prognostic Tool; ADL, activities of daily living; BODE, BMI, Obstruction, Dyspnea, Exercise; COPD, chronic obstructive pulmonary disease; MAGGIC, Meta-Analysis Global Group in Chronic Heart Failure; MELD, Model for End-Stage Liver Disease; MELD-Na, Model for End-Stage Liver Disease with sodium level; NYHA, New York Heart Association.

*Outcome of nursing home level of care was defined as 1 of the following: 3 or more ADL dependencies, 2 or more ADL dependencies and proxy report that patient wanders or cannot be left alone, or eating dependency (eg, needing help cutting up food).

†Whether physician would be surprised if patient died within 1 year.

Table

Table 3. General prognostic calculators.

Setting	Prognostic model	Population and inputs	Outcome	Accuracy	Online calculator
Hospital
	Walter Index⁷³	Population: hospitalized adults aged 70 years and olderInputs include functional status on discharge, select comorbidities, creatinine level, and albumin level	Death within 1 year (all-cause mortality)	Discrimination of 79%⁷³	https://www.eprognosis.ucsf.edu/walter.php⁷⁴
	Palliative Performance Scale⁷⁵⁷⁶	Population: patients who received a palliative medicine consultation at an academic medical centerInputs are driven by functional status	1- and 6-month mortality, median survival	Recent research suggests that prognostic estimates associated with this scale are longer (2.3- to 11.7-fold) than prior estimates⁷⁶In inpatient setting, has a discrimination of 74%⁷⁶In outpatient setting, has a discrimination of 67%⁷⁶	https://www.eprognosis.ucsf.edu/pps.php?p=palliative⁷⁷
Nursing home
	Porock Index (also called the 6-Month Minimum Dataset Mortality Risk Index)	Population: nursing home residentsInputs include recent admission to nursing home; weight loss; presence of renal failure, heart failure, dehydration, or cognitive impairment; and functional status	6-month mortality	Discrimination of 75%⁷⁸	https://www.eprognosis.ucsf.edu/porock.php⁷⁹
Community and/or clinic
	Suemoto 10-year International Mortality Index⁸⁰	Population: adults living in the community aged 60 years and older in low-, high-, and middle-income countriesInputs include age, sex, select medical history, select functional status questions, and self-reported health	10-year all-cause mortality	Discrimination of 76%⁸⁰	https://www.eprognosis.ucsf.edu/suemoto.php⁸¹
	Alex Lee Comprehensive Prognostic Tool for Mortality and Disability⁸²	Population: adults aged 70 years and olderInputs include age, select medical history (strokes, heart failure, diabetes, high blood pressure, lung disease, and cancer) and questions about functional status	5-, 10-, and 14-year risk of mortality and incident ADL disability	Discrimination of 72% for correctly predicting patients who died and those who lived⁸³Discrimination of 63% for correctly predicting patients who developed ADL disability and those who did not⁸³	https://www.eprognosis.ucsf.edu/alexlee.php⁸³
	Combined Lee Schonberg 4- to 14-Year Mortality Index⁸⁴	Population: adults aged 50 years and olderInputs include age, select medical history (strokes, heart failure, diabetes, high blood pressure, lung disease, and cancer) and questions about functional status	4-, 5-, 10-, and 15-year mortality	Lee index has a discrimination of 82% for 4-year mortality⁸⁵Schonberg index has a discrimination of 75%, 73%, and 72% for 5-, 10-, and 14-year mortality respectively⁸⁴⁸⁶	https://www.eprognosis.ucsf.edu/leeschonberg.php⁸⁴

Caption: ADL, activities of daily living.

Interpreting Prognosis

Interpreting prognosis involves using the prognosis to aid clinical decision-making and to make relevant treatment recommendations
- Risk-benefit ratio for many interventions increases as estimated length of survival decreases⁹
  - Accurate prognostication of life expectancy can assist physicians in interpreting and educating patients about risks and benefits of possible treatments. Patients with limited life expectancy are unlikely to live long enough to benefit from certain tests or treatments³³
  - Table 4 provides examples of clinical decisions impacted by patient prognosis
Asking patients about their goals can help a clinician translate medical information into what it means to the patient
- “What are you expecting as your cancer progresses? What are you hoping for?”
- “Given what we discussed about your prognosis, what are you hoping for now?”
Translate the prognosis, focusing on what quality of life may be possible for the patient’s illness trajectory
- Some goals (such as curing advanced cancer) may be out of reach. Consider what goals are possible and how the health care system may support them (eg, spending time at home, getting stronger with physical therapy, living until a loved one’s wedding)
Make a recommendation for treatment in line with the patient’s values
- Review medications, treatments, and procedures in light of the estimated prognosis. Consider the time to benefit from each intervention and make a medical recommendation about each treatment
- For patients whose prognosis is less than 6 months who would like to maximize time at home, discuss hospice care and comfort care

Table

Table 4. Examples of clinical decisions impacted by patient prognosis.

Prognosis	Diagnosis or condition	Clinical decision and rationale
< 4 to 6 weeks	Depression at end of life	Selective serotonin reuptake inhibitors take weeks to reduce depressive symptoms; patients with limited life expectancy may not live long enough to benefit from initiating themMethylphenidate acts within days⁸⁷
< 3 months	Spinal metastases	Time to benefit from spinal cord decompression is approximately 3 months, suggesting that those with a shorter prognosis may be considered for conservative management⁸⁸
< 1-3 years	Hypertension	1 meta-analysis of antihypertensive control in patients aged 60 years or older determined that the time to benefit from intensive blood pressure treatment was on average 9 months, 19.1 months, or 34.4 months respectively to avoid 1 MACE (1 MACE per 500 patients, 1 MACE per 200 patients, or 1 MACE per 100 patients respectively)⁸⁹
< 2-3 years	Hyperlipidemia	Expert consensus recommends against lipid-lowering medications in patients with limited life expectancy⁹⁰1 meta-analysis concluded that statins may help prevent a first MACE in patients aged 50-70 years if life expectancy is at least 2.5 years⁹¹
< 10 years	Colon cancer screening	Expert consensus recommends stopping screening for colorectal cancer in asymptomatic average-risk adults with life expectancy of 10 years or less⁹²

Caption: MACE, major adverse cardiac event.

Communicating Prognosis

Physicians may communicate prognosis in multiple settings (eg, a planned serious illness discussion for a critically ill hospitalized patient, a clinic visit in a specialty clinic such as oncology or cardiology); this may occur in both planned and unplanned conversations
Consider initiating a conversation about prognosis in any patient with a chronic illness, diminished functional status, or recurrent hospitalizations
- Most patients want their physician to initiate the conversation
- Initiating the conversation allows the practitioner to plan in advance (with input of relevant specialists), if necessary
- Some conversations occur organically when a patient asks, “What is my prognosis?”
  - When this occurs, physicians should ensure they are prepared to communicate prognosis
  - If not, a clinician should explain that they would like to share prognostic information at a future visit (eg, “That’s an important question and I think we should talk about it. I’d like to check with your cardiologist and make sure I’m understanding fully. Can we schedule a visit to talk about this in the next 2 weeks?”)
Prepare in advance to discuss prognosis using principles from serious illness discussions
- Establish the right frame of mind and plan before seeing the patient and family. A serious illness discussion can be thought of as a medical procedure and is more successful with thorough planning
- Discuss the prognosis with relevant specialists
- Arrange a time that works for the patient, family, surrogate decision-maker (if applicable), and key members of the health care team
- Meet with relevant specialists beforehand to ensure that all members of the treating team are aligned⁹³
  - Discuss prognosis, treatment options, and possible clinical trajectories. Attempt to achieve prognostic alignment. If differing opinions exist, they should be discussed before the meeting with the family
  - Discuss or determine a “headliner:” 1 to 2 sentences that summarize the key information to communicate. A “headliner” often involves information about the prognosis and how it will impact clinical decisions
- Ask the family about their preferred setting in advance of prognostic conversations. For example, some patients prefer to meet in the ICU room with their loved one while others prefer a conference room
- Create a comfortable setting for the discussion, ensuring a quiet, private setting free of interruptions
Use patient-centered language without jargon or euphemisms
- Calibrate for your own biases. Avoid being intentionally vague, optimistic, or extreme³
- Avoid being blunt or giving more information than the patient desires³
- Ask permission before disclosing prognostic information
- Clarify what kind of information patients would find helpful
- Communicate prognosis clearly and directly, using empathetic language
  - Frame the conversation with both hopefulness and reality¹⁴
    - “We are hoping for the best and preparing for the rest”
    - “We hope for the best and prepare for the unexpected”
  - For most patients, use ranges when providing estimates of life expectancy: “hours to days,” “days to weeks,” “weeks to short months,” “long months to short years”
    - 1 large study showed that patients receiving explicit prognostic information preferred when a range of survival time was added¹⁴
  - Consider using best-case/worst-case/most likely scenarios when explaining life expectancy or providing information about quality of life¹⁴
- Anticipate that discussing prognosis may be difficult for some patients. However, do not assume patients will be uncomfortable with the topic³³
- Respond to emotion. The NURSE acronym (naming, understanding, respecting, supporting, exploring) (Table 5) can be useful in responding to challenging emotions
- Respect patient autonomy, including the need to follow alternative paths/treatments or to elect for a treatment that you would not choose for yourself
- Explore and facilitate realistic goals as appropriate³
- Table 6 describes a conversation guide intended to provide examples of language that may be helpful when discussing prognosis
Use a framework with which you are comfortable to help guide the conversation. Accepted and well-known frameworks for serious illness discussions include the SPIKES protocol; Ask, Tell, Ask model; REMAP acronym; and ALIGN framework
- SPIKES protocol⁹⁴
  - Setting up the interview: mental rehearsal, arranging for privacy, involving the family and/or power of attorney, managing time constraints and interruptions
  - Perception: ask the patient and family open-ended questions about how they perceive the medical situation
  - Invitation: ask permission to discuss test results, diagnoses, prognoses, or other information about the patient’s health
  - Knowledge: share knowledge and information with the patient and family
  - Emotion and empathy: address the patient’s emotions with empathy
  - Summarize and strategize: if the family is ready, discuss a treatment plan. Otherwise, plan to revisit this conversation at a future date
- Ask, Tell, Ask⁹⁵
  - Ask: investigate the patient’s and/or family’s understanding of their health and build a relationship by listening
  - Tell: explain diagnosis, prognosis, or concerns about clinical trajectory in straightforward, patient-centered language
  - Ask: ask about the patient’s understanding, thoughts, and feelings about the news
  - See Table 6 for an example of a practical conversation guide using the Ask, Tell, Ask conversation tool. While setting is not included in the Ask, Tell, Ask framework, it is essential for every serious illness discussion and/or communicating prognosis
- REMAP tool⁹⁶
  - Reframe the situation
  - Expect emotion and empathize
  - Map what is most important
  - Align with the patient’s values by repeating what they said is most important
  - Plan medical treatments that match patient values
- ALIGN framework⁹⁷
  - Approach: consider the appropriate team member to relay prognostic information. Ensure the setting is appropriate and that the medical team agrees with the prognostic estimate
  - Learn: elicit the patient’s or family’s communication preferences. Assess their baseline understanding and goals
  - Inform: acknowledge uncertainty and provide balanced information. Discuss the impact of the illness on day-to-day life
  - Give support: offer support in real time (eg, social workers for social support or chaplains for spiritual support). Clinicians should offer support in response to emotional reactions
  - Next steps: connect with peer support, caregiving support, or other resources to help families understand prognosis

Table

Table 5. Examples of NURSE statements.

Component	Examples
Naming	“It sounds like you’re feeling frustrated right now””It seems like you’re feeling overwhelmed””It sounds like you’re feeling scared about what is ahead””It sounds like you’re feeling uncertain about what comes next””You seem disappointed with the way things have unfolded”
Understanding	“I can’t imagine what it would be like to hear this news””This must feel like a lot to take in all at once””I get how this feels so unfair””I know this wasn’t the news you were hoping for””I can see why you feel upset; you have really been through a lot”
Respecting	Acknowledge the effort by patients and families and express your appreciation or respect:”It’s easy to see how much you care about and advocate for your father””Your family has really rallied around your mother in this time of need. You have been incredible supporters and advocates””I really admire how thoughtfully you consider the next steps””I can see your commitment to doing everything in your power to maintain as much independence as possible””I’m so impressed with your strength and perseverance as you’ve battled the ups and downs of your cancer””You’re asking all the right questions”
Supporting	“I will keep checking in and will keep you updated as we navigate the next steps””I’m not going anywhere; I’ll be here to support you and your family however I can”
Exploring	“Tell me more about what’s running through your mind””Can you tell me more about what’s worrying you most right now?””What else are you feeling that you may not have yet shared?””Tell me more about what getting cancer treatment means to you””Tell me more about what hoping for a miracle means to you”

Caption: NURSE, naming, understanding, respecting, supporting, exploring.

Table

Table 6. A practical conversation guide using the SPIKES protocol and Ask, Tell, Ask tools.*

Component	Conversation tips
Setting	Prepare for the conversation in advanceIf necessary, discuss disease-specific prognosis and available treatment pathways with relevant specialistsThink of a “headliner:” a 1- to 2-sentence summary of the patient’s health, including your worries as a medical practitionerInclude the right peopleWho would the patient like present for support?Identify the appropriate surrogate medical decision-maker if the patient does not have capacityInclude the right members of the health care team, if availableAllot ample time for discussion of prognosticationFor patients who cannot participate in discussions (eg, intubated, confused):Ensure the patient is clean, comfortable, and turnedAttend to patient needs such as suctioning, cleaning, or medicating before the conversationAsk the patient and/or family about the preferred location for the discussionDo they want to be at the bedside in the ICU with their loved one? Would they prefer a quiet conference room?If the patient is alert but confused (eg, dementia), would they rather include the patient in the conversation or discuss it separately?Minimize interruptions (eg, phones, pagers)Locate chairs so that participants can sit comfortably and converse at eye levelIn a large multidisciplinary conversation:Start each conversation by introducing everyone in the roomIt is often helpful to start by stating an overall purpose”The purpose of this meeting is to get on the same page about ____ (eg, your heart failure, your cancer, your father’s injuries) and to discuss what may lie ahead. We hope for the best and we prepare for the rest”
Ask	Ask about the patient’s or family’s knowledge of their illness or prognosis. Examples include:”What have you been hearing from the doctors?””What have you been hearing from the cancer doctors about your cancer?””What do you know about what lies ahead with your illness and treatment?””What did you hear in the last medical update?”When possible, do not use the phrasing “What is your understanding of your medical illness?” Occasionally, patients take offense that the speaker is implying they lack understandingAsk for permission”Would it be OK if I shared with you what I’ve been hearing from the doctors?””Would it be OK if I shared with you what I’m worried about?”Ask for specific permission to share prognostic information”Would you like me to talk about the prognosis for this type of cancer? What kind of information do you want me to cover?””If doctors had an idea about how long your life expectancy may be, would you want that information?””Some patients tell me they want as much information as possible, including estimates of how long they may live. Other patients tell me they don’t want this kind of information. What kind of information do you want about the future?”If the patient wants information about life expectancy, clarify what kind of information they would like. If the conversation is initiated by the patient asking, “What is my prognosis?,” this would be a good place to start”What kind of information do you want about the future?””There are a few ways I can give you information about prognosis, and I’d like to know which is the best way for you. I could share an approximate life expectancy for someone with cancer at a similar stage. I could talk about the best- and worst-case scenarios. Sometimes people are thinking about a certain event in the future, like they are hoping to live until their son’s graduation. Which ones of these would be most helpful?”¹²If patients do not want information about life expectancy, ask permission to share information about quality of life. Many patients, surrogates, and families deeply care about the likelihood of returning to independent function, which in some scenarios may be more certain than estimates of survival time”Would you want to know about what to expect with your quality of life moving forward?””Would it be OK if I shared what life may look like now that Dad has a tracheostomy and percutaneous endoscopic gastrostomy tube?”
Tell	Always acknowledge or underscore the uncertainty inherent in prognostication”I wish I could predict exactly what will happen, but I can’t. I can share what I think is very likely to happen with your disease”Some families appreciate comparing prognostication to weather prediction:”Prognosis is a bit like weather forecasts. For example, a hurricane is often forecasted with a ‘cone of uncertainty,’ which emphasizes the range of pathways that the storm may take. Similarly, when we prognosticate about the future, there is a range of possibilities. Prognosis tends to become more accurate as time goes on”Given inherent uncertainty, use ranges when discussing prognosisBased on my clinical experience, I would expect someone with a diagnosis like yours to have weeks to months to live”Use “hours to days,” “days to weeks,” “weeks to short months,” “many months to short years,” or “many years” when discussing prognosisIf using median survival data, highlight the average survival and offer information about the best- and worst-case scenarios to highlight inherent uncertaintyIf a patient has found median survival information on their own, emphasize that “the median is not the message.” Some patients live longer than the median, while others live for less time⁹⁸“Considering an average patient with your diagnosis, I think the life expectancy is approximately 2 years, but it may vary from 1 to 4 years in the average patient. This is an estimate based on the average and does not tell us what will happen to you exactly”¹⁴Use a mix of positive and negative language to frame the information³⁷Using numeric statements to convey prognosis is not more effective than qualitative statements. If using numeric statements, use whole numbers rather than fractions (eg, “roughly 20 out of 100 patients with this disease will be alive at 5 years”)³¹⁹⁸Use “best-case/worst-case/most likely scenario,” which may be particularly useful when highlighting changes in quality of life or functional status”I expect someone with metastatic colon cancer to live for months to short years. In a bad case scenario, a few months. In a good case scenario, a few years. If you look at the average, the average is around 2 years”Use a hope/prepare statement that mirrors the sentiment “we hope for the best and prepare for the worst””We will do our best to ensure that you have a better-than-average outcome. On the other hand, if you do progress faster than average, I think it is a good idea to prepare yourself for the unexpected”¹⁴For some patients, using third-person references may be best”I would expect a patient with an infection and organ failure, like what your dad is experiencing, to die in the ICU””Given my experience, I would estimate that the life expectancy of someone with your type of cancer, which has spread to other parts of the body and stopped responding to treatment, would be in the range of weeks to months”Other phrases that may be useful, particularly if the patient does not want a specific life expectancy estimate but the medical teams hope to convey the seriousness of illnessLast phase of life: “I know you’ve been managing your liver disease for a long time. I worry we are in the last phase of life”Different phase of life: “I know Mom has bounced back from her heart failure before, but with her new kidney failure, I really worry we are in a different phase of life”Time is short: “I’m worried time may be short”Offer a recommendation that considers your prognostic estimate. Here, you can offer your interpretation of the prognosis and what it means for the treatment. Many patients need time between the disclosure of prognosis and further discussion of next steps”Is it OK if I share a recommendation?””Seeing how sick Mom is in the ICU, I would recommend that anyone who wants to visit her prioritize doing so””Given what you’ve shared is most important to you – wanting to be with family and avoid suffering – I want to talk about how best to care for you in the final phase of life. We have reached a place in your illness where adverse effects of treatment may outweigh benefit. I think we should shift our focus to ensuring that you are comfortable and that we are optimizing time with your family and friends. I want to talk about how hospice may help us best achieve this”
Ask	Check for understanding”Can you tell me what you’re taking away from our conversation today?””Can you share with me what you will tell your spouse about our conversation?”Ask “What questions do you have?” (this phrasing is preferred over “Do you have questions?”)Ask about follow-up and next steps. If possible, provide a timeframe for follow-up on these discussions. In critically ill patients, a follow-up may occur within hours. In the clinic setting, you may want to indicate this is an ongoing conversation that may change as treatment or disease progresses”Would it be OK if I give you and your family time to process what we’ve talked about and check in with you this afternoon?””Would it be OK if I give you time to think over some of the things we discussed and we can address them next time I see you in clinic?””I know these conversations can be tough. I’d like to check in with you when I see you next month in clinic”

Caption: SPIKES, setting up, perception, invitation, knowledge, emotion, summarize.

*The intent is not for a clinician to use all of the communication techniques within 1 serious illness discussion, but rather to present options for language that may be adapted to each clinical circumstance.

Discussing Disease Trajectories

Figure 1. Visual representations of disease trajectories for common incurable chronic conditions.* – *T = 0 represents time at or just before diagnosis.Adapted from Lunney JR et al. Patterns of functional decline at the end of life. JAMA. 2003;289(18):2387-2392.

Disease trajectories offer another framework to discuss prognostic information and inform patients about what to expect during the course of their illness
- A large observational study described distinct illness trajectories at the end of life for the common diagnoses of dementia, frailty, cancer, and organ failure (eg, heart failure, COPD)⁹⁹
- Subsequent research has shown that disease trajectory may not differ by diagnostic category but may be impacted by disease severity, demographics, emotional and social well-being, presence and degree of frailty, and frequency of hospitalization¹⁰⁰¹⁰¹¹⁰²
Discussing illness trajectories may still benefit patients and can be done early in illness. These discussions may be useful on diagnosis of the disease, in a clinic setting to provide education and encourage advance care planning discussions, or in an acute hospital setting to show an advanced stage of illness
Patients who are visual learners may respond well to a clinician’s drawing a graph and discussing how their disease may progress over time. Figure 1 shows visual representations of illness trajectories for common incurable chronic conditions
- Dementia and/or frailty
  - Cognitive and physical capabilities may decrease as disease progresses
  - Some patients with dementia experience low functional status and significant impairment for a long time, as indicated by a possible tail on the right side of the graph
  - Clinicians can educate patients and caregivers on the trajectory of dementia, emphasizing the importance of advance care planning while the patient has improved cognitive function and the need to plan or prepare for a prolonged period where the patient may have many care needs
- Advanced cancer
  - In early illness, many patients with cancer have stable or little impairment of functional status. Patients are likely relatively well (ie, coming back and forth to outpatient chemotherapy appointments)
  - Many patients with cancer experience a steep decline in function in the months preceding their death, which may take patients and families by surprise
  - If a patient has had a rapid deterioration and is currently in the hospital setting, it can be helpful to draw a graph of disease trajectory (see Figure 1), describe the typical clinical course, place an X on the drawing, and say, “I am worried we are here in Mom’s illness”
- Organ failure (eg, congestive heart failure, COPD, decompensated cirrhosis)
  - Patients may experience decreases in functional status that correlate with acute exacerbations (and possible hospitalizations)
  - After these periods of acute decline, functional status may improve (as the result of treatment or hospitalization), but often not to the level before the acute exacerbation
  - Frequency of exacerbations and hospitalizations may increase as a patient nears the end of life
  - Many patients with COPD, heart failure, and cirrhosis do not realize that these diseases are chronic and cannot be cured completely. It may be helpful to draw a graph of disease trajectory (see Figure 1), describe the typical clinical course, place an X on the drawing, and say, “I think we are here in your illness. I wish we had a cure. Unfortunately, I would expect, over time, for your heart failure to give you more and more trouble. You may find yourself in the hospital more and more”

Responding to Emotion

Avoid responding to feelings with facts. Aim to respond with empathy
NURSE acronym is frequently used as a framework to respond to difficult emotions, convey empathy, and provide emotional support
- Components include naming the emotion, understanding the emotion, conveying respect for the patient’s efforts and emotions, offering support, and exploring emotions further
See Table 5 for examples of NURSE statements that can be adapted for use

Resources for Clinicians

Table 7 describes resources that may assist clinicians in estimating and communicating prognosis

Table

Table 7. Selected resources to assist in prognostication and communicating prognosis to patients.

Resource	Description	Link
ePrognosis	Comprehensive resource that estimates mortality for patients across 3 settings (hospitalized, living in a nursing home, dwelling in the community)Links to other disease-specific prognostic calculatorsHas videos to assist clinicians in communicating prognosis to patients	https://www.eprognosis.ucsf.edu¹⁰³
Predict Survival	Calculator to estimate prognosis in patients with advanced cancerPools data from multiple common prognostic predictors and suggests a timeframe to communicate to patients and familiesIntended for patients with expected survival of < 6 months	https://www.predictsurvival.com¹⁰⁴
MDCalc	Online medical reference with many point of care clinical decision support toolsIncludes many disease-specific calculators	https://www.mdcalc.com¹⁰⁵
VitalTalk	Resources and courses to strengthen clinicians’ communication skills that can be used to communicate prognosis	https://www.vitaltalk.org/¹⁰⁶

Special Populations

Patients in Palliative Care

Prognostication in the last phase of life is essential for anticipating symptom management needs and for providing anticipatory guidance to patients and families¹⁰⁷
Multiple prognostic tools exist for calculating life expectancy in palliative care settings¹⁰⁸
Table 8 describes characteristics of commonly used prognostic tools validated in palliative care settings
Palliative Performance Scale (Table 9) is a functional scale with high interrater accuracy
- Widely used for patients with serious illnesses and is the most commonly used tool on ePrognosis (https://www.eprognosis.ucsf.edu)⁴⁵⁷⁶
- Recent research shows that patients survive significantly longer (2.3- to 11.7-fold) than predicted by older estimates commonly used by clinicians⁷⁶
  - Many prior studies validated the Palliative Performance Scale in inpatient hospital or hospice settings, although it is now widely used in both inpatient and outpatient settings
  - This scale is commonly used by palliative care specialists, who may see patients earlier in their course of illness
- More accurate in the inpatient setting than in the outpatient setting⁷⁶
- Average prognosis predicted by the PPS is affected by the care setting (inpatient or outpatient) and the type of illness (cancer versus noncancer)⁷⁶

Table

Table 8. Examples of prognostic tools validated in palliative care settings.

Prognostic tool	Validated populations	Type of prediction	Factors included in score	Comments
Palliative Prognostic Score	Mixed advanced disease: cancer and noncancer	Probability of surviving 30 days; score assigns patients to 1 of 3 groups with 30%, 30%-70%, or > 70% probability of survival	Symptoms of dyspnea and anorexia Functional status Clinician prediction of survival Laboratory results: WBC count and lymphocyte percentage	Hybrid assessment method that combines clinicians’ survival estimates with clinical features and blood results
Palliative Performance Scale	Mixed advanced disease: cancer and noncancer	Each decreasing level (deciles from 100%- 0%) is associated with shorter survival; a study has derived median survival in days for levels 10%-70%	Functional status based on ambulation, activity, evidence of disease, ability to perform self-care, intake (food and fluid), and consciousness level	Does not rely on blood results or clinician predictions of survival. Not specifically developed as a prognostic tool and may therefore be missing some key prognostic variables
Prognosis in Palliative Care Study A score	Advanced incurable cancer	Provides a probability of surviving days (0-14 days), weeks (15-56 days), or months (> 56 days)	Clinical information on diagnosis Sites of metastases Presence or absence of key symptoms Cognitive status Functional status	Does not rely on blood results or clinician predictions of survival
Prognosis in Palliative Care Study B score	Advanced incurable cancer	Provides a probability of surviving days (0-14 days), weeks (15-56 days), or months (> 56 days)	Similar factors as for Prognosis in Palliative Care Study A score but with addition of blood results	Does not rely on clinician predictions of survival. In 1 study, was found to be better than a doctor’s or nurse’s survival prediction
Palliative Prognostic Index	Advanced incurable cancer	Probability of surviving < 3 weeks or < 6 weeks	Performance score Oral intake Clinical signs of edema and delirium Symptoms of dyspnea	Does not rely on blood results or clinician predictions of survival

Caption: Adapted from Chu C et al. Prognostication in palliative care. Clin Med. 2019;19(4):306-310, Table 2.

Table

Table 9. Palliative Performance Scale.

Percentage	Ambulation	Activity and evidence of disease	Self-care	Intake	Level of consciousness
100	Full	Normal No disease	Full	Normal	Full
90	Full	Normal Some disease	Full	Normal	Full
80	Full	Normal with effort Some disease	Full	Normal or reduced	Full
70	Reduced	Cannot do normal job/work Some disease	Full	Normal or reduced	Full
60	Reduced	Cannot do hobbies or housework Significant disease	Occasional assistance needed	Normal or reduced	Full or confusion
50	Mainly sit or lie	Cannot do any work Extensive disease	Considerable assistance needed	Normal or reduced	Full or confusion
40	Mainly in bed	Cannot do any work Extensive disease	Mainly assistance	Normal or reduced	Full or drowsy or confusion
30	Always in bed	Cannot do any work Extensive disease	Total care	Reduced	Full or drowsy or confusion
20	Always in bed	Cannot do any work Extensive disease	Total care	Minimal	Full or drowsy or confusion
10	Always in bed	Cannot do any work Extensive disease	Total care	Mouth care only	Drowsy or coma
0	Death	N/A	N/A	N/A	N/A

Caption: N/A, not applicable.

Data from Victoria Hospice Society. Palliative Performance Scale. Accessed February 25, 2025. http://www.npcrc.org/files/news/palliative_performance_scale_ppsv2.pdf

Patients Who Are Imminently Dying

Frequently evaluate for signs that suggest patients are entering the actively dying phase (ie, prognosis of hours to days) and provide anticipatory guidance to families about what to expect
1 large study found that clinicians are fairly accurate in prognosticating patients with less than 24 hours or 1 to 7 days of expected survival (approximately 80% and 65%, respectively). Clinicians became less accurate in prognosticating in survival categories beyond 1 week, with a tendency to overestimate prognosis¹⁰⁹
The following physical signs can be useful in prognostication in the final hours or days of life. These signs were studied in a large prospective study of patients with advanced cancer and showed high specificity but low sensitivity. These signs suggest higher likelihood of death within 3 days:¹¹⁰¹¹¹
- Cheyne-Stokes breathing (LR [likelihood ratio] of death within 3 days: 12.4)
- Death rattle (LR 9)
- Pulselessness of the radial artery (LR 15.6)
- Nonreactive pupils (LR 16.7)
- Respiration with mandibular movement (LR 10)
- Decreased urine output (less than 100 mL over 12 hours [LR 15.2])
- Drooping of the nasolabial fold (LR 8.3)
- Hyperextension of the neck (LR 7.3)
- Grunting of vocal cords (LR 11.8)
- Inability to close eyelids (LR 13.6)
- Decreased response to verbal stimuli (LR 8.3)
- Decreased response to visual stimuli (LR 6.7)
- Presence of upper gastrointestinal bleed (LR 10.3)
- Peripheral cyanosis (LR 5.7)
- Periods of apnea (LR 4.5)
In the same study, the following clinical signs had higher sensitivity but lower specificity in predicting impending death. These signs are very common in the last 3 days of life but are less predictive of death:¹¹⁰
- Palliative Performance Scale score less than 20% (bedbound, total care, with minimal intake; see Table 8) (LR 3.5)
- Agitation, based on the Richmond Agitation Sedation Scale score of -2 or lower (LR 4.9)
- Dysphagia to liquids (LR 1.9)
Pulselessness of the radial artery typically predicts death within 12 hours¹⁰⁷
Combination of low functional status (bedbound, total care, with minimal intake) and drooping of the nasolabial fold may predict death within 3 days. In a large prospective study, in patients with both of these signs, 3-day mortality rate was 94%¹¹²

References

1.Martin EJ et al. Prognostication in serious illness. Med Clin North Am. 2020;104(3):391-403.

Approach to Prognostication – 18 Interesting Facts

Basic Information

Background Information

Terminology

Barriers to Prognostication

Ethical Principles Guiding Prognostication Conversations

Underlying Ethical Principles

Patient Preferences/Autonomy

Role of the Surrogate Decision-Maker

Guide for Medical Practitioners

Rationale for Prognostication

Principles of Prognostication

Approach to Prognostication

Estimating Prognosis

Clinical Judgment

Prognostic Models and Tools

Interpreting Prognosis

Communicating Prognosis

Discussing Disease Trajectories

Responding to Emotion

Resources for Clinicians

Special Populations

Patients in Palliative Care

Patients Who Are Imminently Dying

References

Related Posts

Related Posts:

Recent Post

features

Categories

Weekly Newsletter