Abstract
Background: General practitioners are probably asked regularly about the prognosis for patients with cancer. There is no readily available source of information on the accuracy of their estimates.
Objective: To discover studies that have reported general practitioners estimates of prognosis.
Methods: A search for studies of estimates of prognosis in terminal cancer by general practitioners.
Results: Three papers were discovered, reporting on 836 patients. In only a quarter of patients was prognosis accurately estimated by general practitioners, even within broad limits. Hospital doctors and nurses performed similarly. We suggest considerable scope for further studies in general practice, which should not require sophisticated procedures.
Conclusion: Evidence of GP's limited performance in cancer prognostication should push us to do better.
KEY MESSAGE(S):
Assessing estimates of cancer prognosis by general practitioners has rarely been studied
GPs are probably as inaccurate as hospital doctors. There is considerable scope for further studies of prognostication by GPs which could be done in any country, and with basic resources
Introduction
Doctors’ use their experience and judgment to estimate prognosis when patients with serious illness want to know what the future holds. This may be especially difficult in general practice where the whole range of terminally ill patients is often cared for, but not frequently enough for GPs to become familiar with each type of cancer (Citation1). GPs will often know patients for years before a cancer presents, and follow them up through investigation and treatment until they become terminally ill and care focuses upon the relief of symptoms or other distress, and on death itself. Does knowing the patient help to enhance their judgment of prognosis? So far the limited literature on this subject shows that clinicians (GPs included) tend to overestimate survival so predictive models have been developed in an attempt to improve accuracy (Citation2,Citation3).
In many countries, there are plans to increase the proportion of people dying at home. In England, the Department of Health published an end of life strategy in 2008 to improve the quality of care by giving more support to generalists, along with research to inform service development (Citation4). An important gap in applying generalist end of life care is the lack of prognostic indicators, which in turn can hinder access to appropriate further services (Citation5). The aim was to discover studies that might enable GPs to understand, and if possible improve, their ability to give a prognosis to patients with cancer.
Method
A search of Medline from 1950 and Cinahl from 1981 was undertaken to identify studies of estimates of prognosis in terminal cancer by GPs and nurses working in primary care using the search terms (general practice OR family practice OR primary health care OR family physician) AND (prognosis OR forecasting OR survival analysis OR time factors) AND (neoplasms OR terminally ill OR palliative care).
Many papers were readily rejected as not being concerned about the end of life care, and/or not having numerical results. The abstracts of possibly relevant studies that included numerical estimates of prognosis were retrieved. Appropriate papers with some mention of primary care were read independently in full by both authors. We also searched the references of both relevant studies and systematic reviews of survival predictions in cancer (Citation2,Citation3).
Results
The initial search identified 67 possibly relevant papers. However, only three papers (Citation6–8) reported estimates of prognosis by general practitioners (see summary ).
Table I. Studies of prognosis in terminal cancer by general practitioners and others.
The three studies reported upon 836 patients, most of who lived less than three months after their prognosis was estimated. It would be inappropriate to pool results because a third of patients in the largest study did not have cancer; and ‘accurate’ estimation of survival varied widely, i.e. between 66% and 133%; and 50% and 200%, of that actually achieved. General practitioners were shown to estimate prognosis similar to other doctors or nurses. Unfortunately, none of the studies could help GPs improve their ability to give a prognosis to patients with terminal cancer.
Discussion
Limitations of the study
Each general practitioner only contributed a small number of estimates of survival, i.e. between 1.1 (Citation7) and 1.2 (Citation8), or not stated (Citation6). Patients had all been referred to a hospital or hospice—there are no results available for patients managed by general practitioners alone. GPs also estimated prognosis on referral, whereas hospital doctors and nurses did so later, on admission (Citation6).
Comparison with existing literature
Pooled prognostic performance by all doctors and nurses has previously obscured the ability of general practitioners in this area. The first major systematic review of survival predictions in cancer concluded that survival is typically 30% shorter than predicted, and correct to within one week in just 25% of cases (Citation2). A substantial attempt to systematically estimate survival, the Clinical Prediction of Survival (CPS), was more than twice as likely to be over-optimistic than over-pessimistic, and often over-estimated by three to five times the length of actual survival (Citation3). Unfortunately, the CPS remains heavily influenced by the operator's experience and is poorly reproducible (Citation9).
Implications for clinical practice
Greatly over-estimating prognosis is common and substantially underestimating it is unusual so we must be cautious. For general practitioners too, ‘even the best prognostication will be dramatically inaccurate for a significant number of patients’ (Citation3).
End of life care is probably biased towards cancer and we also need better information on terminal heart failure, neurological conditions, etcetera (Citation5). Furthermore, we need to understand better what patients and their families want to know. It may be that a ‘80% chance of living more than three months’ is far more useful than nothing: perhaps we have been trying too hard to be unrealistically precise? How much is our behaviour interpreted and misinterpreted; which GPs are best at end-of-life care; how could we all learn from them?
Access to appropriate further services for patients may thus be hindered if GPs do not realize that death is likely to be imminent (Citation5). Syringe drivers, night sitters, contact with hospices may be delayed and distant relatives may not be informed, possibly missing the chance to visit. Understanding our professional bias towards optimism and the risk of this could make us more willing to be proactive. The second author (JH) was relieved recently to discover a daughter had arrived from New Zealand a few days after he had predicted an elderly patient was dying, and hours before she did so.
The quality of health care provided is an important factor in allowing patients to die at home but it is not known whether an accurate prognosis from a GP can help this (Citation10). These three studies are preliminary ones and we need to know much more about how precise GPs can be in giving prognoses, and in what conditions their estimates are most accurate. In time we will also need to know whether the CPS and other instruments, almost inevitably developed in secondary care, could perform well in primary care.
We could involve many patients in different countries, and comparisons could start soon. Such studies need simple data collection forms, but to be worthwhile they would require high levels of participation by GPs, with few patients lost to follow-up. General practice should be an ideal place to study prognostication.
Hippocrates told us, ‘[If a physician] is able to tell his patients ... what is going to happen ... he will increase his reputation … and people will have no qualms in putting themselves under his care.’ Hippocrates, Prognosis 1. Quoted in Miles (Citation11).
Conclusion
This evidence of GP's limited performance in cancer prognostication should push us to do better.
Acknowledgements
The authors thank librarians Katherine Dumenil and Sarah Thomas for help with the searches.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
References
- Holden J, O’Donnell J, Brindley J, Miles L. Analysis of 1263 deaths in four general practices. Br J Gen Pract. 1998;48: 1409–12.
- Glare P, Virik K, Jones M, Hudson M, Eychmuller S, Simes J, Christakis N. A systematic review of physicians’ survival predictions in terminally ill cancer patients. Br Med J. 2003;327: 195–8.
- Maltoni M, Caraceni A, Brunelli C, Broeckaert B, Christakis N, Eychmueller S, . Prognostic factors in advanced cancer patients: Evidence-based clinical recommendations—a study by the Steering Committee of the European Association for Palliative Care. J Clin Oncol. 2005;23:6240–8.
- Department of Health. End of life care strategy: Promoting high quality care for all adults at the end of life. London: Department of Health; 2008.
- Shipman C, Gysels M, White P, Worth A, Murray SA, Barclay S, . Improving generalist end of life care: National consultation with practitioners, commissioners, academics, and service users. Br Med J. 2008;337:848–51.
- Parkes CM. Accuracy of predictions of survival in later stages of cancer. Br Med J. 1972;264:29–31.
- Llobera J, Esteva M, Rifa J, Benito E, Terrasa J, Rojas C, . Terminal cancer: Duration and prediction of survival time. Eur J Cancer 2000;36:2036–43.
- Christakis NA, Lamont EB. Extent and determinants of error in doctors’ prognoses in terminally ill patients: Prospective cohort study. Br Med J. 2000;320:469–73.
- Pirovano M, Maltoni M, Nanni O, Marinari M, Indelli M, Zaninetta G, . A new palliative prognostic score: A first step for the staging of terminally ill cancer patients. J Pain Symptom Manage 1999;17:231–9.
- Gomes B, Higginson IJ. Factors influencing death at home in terminally ill patients with cancer: Systematic review. Br Med J. 2006;332:515–8.
- Miles SH. Hippocrates and informed consent. Lancet 2009;374: 1322–3.