The assessment of patient life-expectancy: how accurate are urologists and oncologists?

Objective: To assess the degree of accuracy, precision and consistency with which consultant urologists, oncologists and junior doctors predict a patient's 10-year life-expectancy.

Subjects And Methods: Eighteen doctors of varying seniority independently examined 70 patient case scenarios containing detailed medical histories; 13 of these cases were duplicate scenarios. Bland-Altman analyses were used to compare doctors' estimates of the probability of each hypothetical patient surviving 10 years with that calculated using actuarial methods. Intra- and interdoctor reliability were also assessed.

Results: Compared with actuarial estimates, doctors underestimated the 10-year survival probability by an overall mean of 10.8% (95% confidence interval, 10.1-11.5%). The 18 individual doctors ranged from a mean underestimation of 33.2% to a mean overestimation of 3.9%. Variation around these means was considerable for each doctor, the standard deviations being 14.5-20.9%. Inter-doctor reliability was 0.58, while overall intra-doctor reliability was 0.74, but for individual doctors was 0.31-0.94. Junior doctors were less accurate in their predictions than the senior doctors. Five doctors tended to overestimate where life-expectancy was poor and underestimate where it was good.

Conclusions: Doctors were poor at predicting 10-year survival, tending to underestimate when compared with actuarial estimates. There was also substantial variability both within and between doctors. The inaccuracy, imprecision and inconsistency amongst the doctors in assessing patient life-expectancy is an important finding and has significant implications for managing patients. Many patients may be denied treatment after a pessimistic assessment of life-expectancy and (less commonly) some may inappropriately be offered treatment after an optimistic assessment. The particular inaccuracy in junior doctors compared with their senior colleagues also highlights the need for training. The development of a tool to assist in both training and clinical practice has the potential to improve doctors' decision-making and patient care.