The duration of the terminal period of cancer allows us to determine its prevalence, which is necessary to plan palliative care services. Clinical prediction of survival influences access to palliative care and the healthcare approach to be adopted. The objective of this study was to determine the duration of the terminal period, the prognostic ability of healthcare professionals to predict this terminal period and the factors that can improve the prognostic accuracy. In the island of Mallorca, Spain, we followed 200 cancer patients at the inception of the terminal period. Twenty-one symptoms, quality of life, prognosis and duration of survival were measured. Using a Cox regression model, a predictive survival model was built. Median duration was 59 days; 95% confidence interval (CI)=49-69 days, mean=99 days. The oncologists were accurate in their predictions (+/-1/3 duration) in 25.7% of cases, the nurses in 21.5% of cases and the family physicians in 21.7% of cases. Errors of overestimation occurred 2.86-4.14 times more frequently than underestimation. In the final model, in addition to clinical prognosis (P=0.0094), asthenia (P=0.0257) and the Hebrew Rehabilitation Centre for Aged Quality of Life (HRCA-QL) Index (P=0.0002) were shown to be independent predictors of survival. In this study, the estimated duration of the terminal period was greater than that reported in a series of palliative care programmes, and survival was overestimated. Oncologists could estimate prognosis more accurately if they also take into account asthenia and HRCA-QL Index.