Accuracy of clinical prediction of survival in a palliative care unit

Abstract

Background: Clinical prediction of survival (CPS) by clinicians is generally overestimated. No study has been done in Hong Kong addressing the accuracy of survival prediction in advanced cancer patients. The objective of this study was to examine CPS by a local palliative care physician in a palliative care unit.

Patients and methods: This was a prospective study in a local palliative care unit. After the enrolment of advanced cancer patients into the palliative care service, the physician-in-charge, experienced in palliative care, estimated the survival of the patients in weeks or months as a continuous variable. These were compared with the actual survival time (AS), accuracy being defined as observed survival = predicted survival ± 33%. CPS was also re-categorized into three groups (≤ 4 weeks, 5–12 weeks, ≥ 13 weeks) for the analysis.

Results: Of the 167 patients studied, 103 were men and 64 were women (mean age 69.3 years; SD 11.7 years). The median overall survival was 76 days (interquartile range 30–160 days). The median actual survivals among the three different clinical predicted groups were 12, 63 and 170 days, respectively. The differences were statistically significant (P < 0.0001). CPS correlated strongly with AS (r = 0.67; P < 0.0001). The difference between the median CPS and median AS was 6 days in a pessimistic direction and their ratio was 0.92. CPS was accurate in 53 patients (31.7%), overly optimistic in 70 patients (41.9%) and overly pessimistic in 44 patients (26.3%).

Conclusions: In this study, CPS was highly correlated with AS. Physicians with experience in palliative care may discriminate well the different prognostic groups, yet a significant proportion of CPS was still overestimated. Maybe over-optimism of prognostication is a cultural bias.

Keywords:

• CLINICAL PREDICTION OF SURVIVAL
• ADVANCED CANCER
• PALLIATIVE CARE