Review article
Prognostication — The lost skill of medicine

https://doi.org/10.1016/j.ejim.2007.09.002Get rights and content

Abstract

Making a prognosis is one of the primary functions of the medical profession. At the end of the nineteenth century prognostication took up approximately ten percent of medical textbooks, by 1970 this had fallen to nearly zero. Given medical technology's awesome ability to prolong the process and suffering of dying today's patients need to know their prognosis in order to make choices about their treatment options. Whilst precise predictions of the future are obviously not possible, relatively simple mathematical modelling techniques can make reasonable estimates of likely outcomes for individual patients. The life expectancy of a patient of any age with any illness can be estimated provided the disease-specific mortality of the illness is known. Decision analysis or logistic regression models can then be used to determine the risks and benefits of various treatment options. A patient's prognosis does not just depend on their age and primary diagnosis, but also on the severity of their illness, their functional capacity both prior to and during the illness and the number of co-morbidities also suffered from. Several predictive instruments have been developed to help simplify the prediction of the outcome of individual patients. There are conflicting reports on how these models compare with doctors' intuition — whatever their strengths and weaknesses it is unlikely that they worsen clinical judgement. Therefore, all doctors should become familiar with them and use them appropriately.

Introduction

Diagnosis, treatment and prognosis are the three primary functions of the medical profession. The word physician comes from Greek and implies one who understands the natural human condition, the inescapable fundamental features of which are birth, life and death. Prior to the advent of effective therapies, prognostication was the major task of the medical profession. Indeed, predicting the future was what both priests and doctors were expected to do. Doctors were paid for their opinions, not for their doubts, and patients and relatives looked to the profession not only for reassurance but, if that was not possible, for practical advice on how to deal with their illness and when to settle their affairs. Sections on prognostication, therefore, took up a considerable proportion of medical texts, and a large amount of medical teaching was given over to identifying those features of an individual patient that indicated their prognosis. At the end of the nineteenth century prognostication took up approximately ten percent of medical textbooks, by 1970 this had fallen to nearly zero [1]. As a result, although modern physicians commonly encounter situations that require prognostication they feel “poorly trained to do it, and find the whole process difficult and stressful” [2].

Section snippets

Reasons not to prognosticate

Are there any valid reasons for not making a prognosis? A common unfounded assertion is that patients do not want to know what is likely to happen to them or that the truth might kill them. Even if this argument was ever true it is no longer tenable given modern medical technology's awesome ability to prolong the process and suffering of dying [3], [4]. Many physicians, for example, are not aware of the natural history of dementia [5], [6]. As a result, elderly patients with unrecognised

Correcting life expectancy for disease

Life expectancy estimates based on age alone are averages based on the general population and may not, therefore, accurately predict the life expectancy of individual patients. The Declining Exponential Approximation of Life Expectancy (DEALE) has been proposed as a way of estimating the life expectancy of a patient of any age with any illness provided the disease-specific mortality of the illness is available in the literature [23], [24]. The DEALE assumes that the reciprocal of life

Diagnosis and its shortcomings

The simplistic assumption that a patient's diagnosis and age are the major drivers of prognosis often fails because it does not take into account the other factors that influence mortality. Pneumonia is a good example of the limited value of making a diagnosis. When Sir William Osler wrote his classic textbook in 1901 the only purpose for diagnosing of pneumonia was to prognosticate. At that time pneumonia was the principle cause of death, and recognised by the abrupt onset of fever with

Assessing severity of illness

In the day to day practice of medicine physicians rely on their estimates of severity to make many clinical decisions. For example, decisions to admit patients to hospital, to see them in the emergency room, and to initiate or change therapy are made on the basis of physicians' assessments of how sick patients are. Even though a physician's ability to accurately assess illness severity is critical to the practice of medicine, doctors are never explicitly taught how to judge how sick patients

Assessing functional capacity

For most illness the severity of illness also determines the dependency the patient has on others — a patient with mild disease usually has no little or no dependency nursing staff, whereas those with severe disease usually require constant care and attention. Patients who already have a poor functional capacity will, of course, become even more dependent on others even if they develop a relatively mild additional illness [50]. Several instruments can assess both mental and physical functional

Detecting co-morbidities

Acute medical patients admitted to hospital on average are suffering from four co-morbid conditions and those who die from five — only 10% of patients have one diagnosis, and over 30% of patients have at least six diagnoses on discharge. Few patients, therefore, die from a single diagnosis. Nearly all the clinical presentations encountered in an Irish hospital were encompassed within an average of four combinations of 74 conditions, 34 of which were associated with an increased risk of death

Predictive models

Several predictive instruments have been developed to simplify the prediction of life expectancy of individual patients (Table 3). These models can be used to predict either short term or long term outcome either for patients with known diagnoses, or in certain situations (e.g. intensive care, undergoing surgery, acute medical admissions, elderly care units). Most of these predictive tools rest more on the physical and mental performance of the patient than on laboratory or imaging

Doctors' intuition versus prediction models

How do prognoses made using prediction models compare with those based solely on doctors' intuition? There are conflicting reports, depending on the clinical circumstances. A common problem in the application of prognostic models is that their accuracy degrades when used on a different population to the one from which they were developed. Therefore, not all predictive models can be applied to all populations [65]. The “gut feeling” of surgeons after performing surgery is a good predictor of

Conclusion

Reluctant though they may be to provide one, it is clear that the practice of modern high technology medicine requires physicians to prognosticate as accurately as they can. Most patients want and need a prognosis. Even physicians with limited experience can make reasonably accurate prognoses based on their intuitive assessments of illness severity, functional capacity and the presence or absence of a number of well-recognised co-morbid conditions. There are now several sound techniques and

Learning points

  • •

    Given medical technology's awesome ability to prolong the process and suffering of dying today's patients need to know their prognosis in order to make choices about their treatment options.

  • •

    Even physicians with limited experience can make reasonably accurate prognoses based on their intuitive assessments of illness severity, functional capacity and the presence or absence of a number of well-recognised co-morbid conditions.

  • •

    There are several techniques and predictive models available that can

References (174)

  • T. Killip et al.

    Treatment of myocardial infarction in a coronary care unit: a two year experience with 250 patients

    Am J Cardiol

    (1967)
  • R.M. Norris et al.

    A new coronary prognostic index

    Lancet

    (1969)
  • M. Folstein et al.

    “Mini-Mental State” a practical method for grading the cognitive state of patients for the clinician

    J Psych Res

    (1975)
  • J.A. Yesavage et al.

    Development and validation of a geriatric depression screening scale

    J Psych R

    (1983)
  • J. Kellett et al.

    The diagnoses and co-morbidity in the hospital practice of acute internal medicine

    Eur J Int Med

    (2007)
  • M.E. Charlson et al.

    A new method of classifying prognostic comorbidity in longitudinal studies: development and validation

    J Chron Dis

    (1987)
  • E.J. Simchuk et al.

    Computed tomography severity index is a predictor of outcomes for severe pancreatitis

    Am J Surg

    (2000)
  • T.R. Hucker et al.

    Identifying the sick: can biochemical measurements be used to aid decision making on presentation to the accident and emergency department?

    Br J Anaesth

    (2005)
  • W.R. Smith et al.

    Prognostic judgments and triage decisions for patients with acute congestive heart failure

    Chest

    (May 2002)
  • R.M. Poses et al.

    Physicians' judgments of survival after medical management and mortality risk reduction due to revascularization procedures for patients with coronary artery disease

    Chest

    (Jul 2002)
  • D. Listello et al.

    Unplanned extubation: clinical predictors for re-intubation

    Chest

    (1994)
  • G.A. Tindol et al.

    Unplanned extubations

    Chest

    (1994)
  • J.W. Burnside

    Pre-arrest morbidity index in prediction of survival after cardiopulmonary resuscitation

    Am J Med

    (1990)
  • M. Maltoni et al.

    Successful validation of the Palliative Prognostic Score (P&P Score) in terminally ill cancer patients

    J Pain Symp Manag

    (1999)
  • L.S. Murray et al.

    Does prediction of outcome alter patient management?

    Lancet

    (1993)
  • J.D. Puskas et al.

    Stroke after coronary artery operation: incidence, correlates, outcome and cost

    Ann Thorac Surg

    (2000)
  • N.A. Christakis et al.

    Attitude and self-reported practice regarding prognostication in a national sample of internists

    Arch Intern Med.

    (1998)
  • C. Meredith et al.

    Information needs of cancer patients in west Scotland: cross sectional survey of patients' views

    BMJ

    (1996)
  • A.M. The et al.

    Collusion in doctor-patient communication about imminent death: an ethnographic study

    BMJ

    (2000)
  • C. Wolfson et al.

    A re-evaluation of the duration of survival after the onset of dementia

    N Engl J Med

    (2001)
  • E.B. Larson et al.

    Survival after initial diagnosis of Alzheimer disease

    Ann Intern Med

    (2004)
  • D.E. Meier et al.

    High short-term mortality in hospitalised patients with advanced dementia

    Arch Intern Med

    (2001)
  • R.S. Morrison et al.

    Survival in end-stage dementia following acute illness

    JAMA

    (2000)
  • T.J. Iwashyna

    Critical care use during the course of serious illness

    Am J Respir Crit Care Med.

    (2004)
  • S.P. Keenan et al.

    A retrospective review of a large cohort of patients undergoing the process of withholding or withdrawal of life support

    Crit Care Med.

    (1997)
  • I. Noone et al.

    Telling the truth about cancer: views of elderly patients and their relatives

    Irish Med J.

    (2000)
  • G. Leydon et al.

    Cancer patients' information needs and information seeking behaviour: in depth interview study

    BMJ

    (2000)
  • J.C. Weeks et al.

    Relationship between cancer patients' predictions of prognosis and their treatment preferences

    JAMA

    (1998)
  • T.J. Smith et al.

    Telling the truth about terminal cancer

    JAMA

    (1998)
  • C.P. Gross et al.

    The effects of age and chronic illness on life expectancy after a diagnosis of colorectal cancer: implications for screening

    Ann Intern Med.

    (2006)
  • S. Kajbaf et al.

    Cancer screening and life expectancy of Canadian patients with kidney failure

    Nephrol Dial Transpl

    (2002)
  • L. Balducci et al.

    Management of cancer in the older person: a practical approach

    Oncologist

    (2000)
  • H.C. Sox et al.

    Medical Decision Making

    (1988)

  • Irish Life Tables No. 14

    (2001–2003)
  • R.J. Lifford et al.

    Decision analysis and the implementation of research findings

    BMJ

    (1998)
  • A.S. Detsky et al.

    Primer on medical decision analysis: part 1 — getting started

    Med Decis Making

    (1997)
  • J.C. Wright et al.

    Gains in life expectancy from medical interventions — standardising data on outcomes

    N Engl J Med

    (1998)
  • R.T. Vollmer

    Multivariate statistical analysis for pathologists. Part 1. The logistic model

    Am J Clin Pathol

    (1996)
  • S. Lemeshow et al.

    Logistic regression

  • S.A. Glantz et al.

    Primer of Applied Regression and Analysis of Variance

    (1990)

    • Cited by (31)

    • Death is not the only healthcare outcome important to patients

      2016, European Journal of Internal Medicine

    • Acute hospital medicine - A new sub-speciality or internal medicine re-born?

      2011, European Journal of Internal Medicine

    • Development and validation of a nomogram for prediction of survival and local control in laryngeal carcinoma patients treated with radiotherapy alone: A cohort study based on 994 patients

      2011, Radiotherapy and Oncology

    • "Identifying the hospitalised patient in crisis"-A consensus conference on the afferent limb of Rapid Response Systems

      2010, Resuscitation
      Citation Excerpt :

      Predicting deterioration (prognostication) is the estimation of the risk of deterioration over various time frames and is traditionally based on a history and physical examination, requiring considerable medical knowledge and experience. The teaching and validation of prognostic skills is a neglected area in modern medicine, although predictive tools (usually confined to a single diagnosis) have been developed.64 Detecting deterioration, on the other hand, is the recognition of critical illness at a particular moment and is related to the type, availability, and performance characteristics of the monitor – whether it is a human observer or an electronic detector.

    • Wounds and survival in cancer patients

      2009, European Journal of Cancer
      Citation Excerpt :

      The consequences of inadequate attention to prognostication are considerable and include delayed transitioning of patients to a completely palliative mode of care along with the resultant high rates of late-stage futile interventions.1,3,4 Prognostic models for newly diagnosed cancer patients are generally based on the degree of disease burden as evidenced by clinical, imaging, laboratory, tumour histology and molecular features.1,3–6 An example is the SEER (Surveillance, Epidemiology and End-Results) programme of the National Cancer Institute which publishes survival data from population-based cancer registries.

    • Ability of physiological parameters versus clinical categories to predict mortality on admission to an internal medicine ward

      2009, European Journal of Internal Medicine
      Citation Excerpt :

      There are several possible explanations for this. It is generally accepted that prediction models lose power when used on populations other than that from which they were developed [17]. The most plausible explanation, however, is that our series of patients was very different in outcome as compared to previously published series.

    View all citing articles on Scopus
    View full text
    Copyright © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved.