50 Systematic review looking at how effective clinicians are at prognostication at the end of life – temporal and probabilistic

Background Clinician’s estimation of patient survival can influence decisions regarding treatment, enable patients to make plans, improve quality of life and increase meeting preferred place of care and death. The two survival predictors frequently used are temporal and probabilistic. Temporal-the patient is predicted to live a certain amount of time, probabilistic-the chance of a person surviving to a certain time. Aims Describe the published evidence, relating to the effectiveness and accuracy of clinicians at predicting clinical survival in patients with cancer and non-cancer. Methodology The databases Embase, Cinahl, Medline and Emcare were searched using the terms ‘prognosis’, ‘prognostication’, ‘surprise question’ and ‘advanced care planning’. Duplicates were removed. 127 papers were identified, 40 papers were selected as relevant to clinical effectiveness of prognostication at end of life. Literature was reviewed using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses principles. Results Clinicians are overly optimistic in their estimated survival predictions (studies found by at least 50%). Recent systematic reviews of clinician’s survival predictions suggest that they were correct in only 25% of cancer patients by within a week. This accuracy of predictions increased to 70% when patients were in the last days to weeks of life. Predictions utilizing probabilistic measures over the temporal approach have been found to be significantly more accurate. A large meta-analysis looking at the accuracy of the ‘surprise question’ and outcomes in 22 studies showed that 75% of clinicians accurately predicted if a patient would die within 12 months. Conclusion Clinician’s prognostication predictions are complex and evidence shows that clinicians are often over optimistic in their estimated survival predictions. Clinicians are significantly more accurate in prognosticating survival with use of probabilistic measures than temporal approach.