9595 Background: Determining an accurate prognosis for terminally ill cancer patients is one of the biggest challenges that confronts a physician. Correct predictions can be done in only 20–40% of all cases. Although the current prognostic scales are helpful, they have significant limitations. Our objective consists of determining the potential indicators that influence the survival of these patients and develop and validate a new predictive model. Methods: A prospective, multicentric and observational study was conducted in 880 terminally ill cancer patients. At first, 40 clinical, demographic and laboratory variables were recorded in 406 patients. A forward stepwise regression method was applied for the multivariate survival analysis. Hence, a predictive model was constructed. Subsequent validation was performed in 474 patients. Results: Median age was 66.4 years (range 18–95). The median overall survival was 21 days in the first 406 patients studied and 19 days in the validation group. A prognostic model with 9 variables was constructed (age, ECOG, the amount of time between initial diagnosis up to being considered terminal phase, nauseas, anorexia, cognitive impairment, lymphocytes, LDH and albumin). Afterwards, to simplify the model, 4 variables that were considered more objective and with greater Odds ratio were selected and assigned one point per each prognostically poor category. We obtained a survival model that discriminates 3 prognostic categories: Good prognoses (score 0) with a median survival of 95 days (44–146), intermediate prognoses (score 1–2) with a median survival of 33 days (26.8–39.2) and bad prognoses (score 3–4) with a median survival of 15 days (11.1–18.9). In the validation group, median survival times were 60 (47.1–72.8), 27 (22.8–31.1) and 11 days (9.2–12.7) respectively. Conclusions: We propose a predictive score model that is objective and easy to use to help in accurately predicting life expectancy in terminally ill cancer patients. Its effectiveness has been validated in a group of independent centers. No significant financial relationships to disclose.