Introduction: Gastrointestinal cancers survival studies are important in evaluating efficacy of treatments and probable prognostic. Objective: To correlate clinical and socio-epidemiological variables with the survival of patients diagnosed with cancer of the gastrointestinal tract in a hospital in a countryside city of Minas Gerais. Methods: Retrospective, cross-sectional, clinical study carried out through the analysis of 426 medical records of patients diagnosed with gastrointestinal tract cancer from 2010 to 2014. Patients were selected by checking Hospital Cancer Records. The conditional probability of survival was calculated using the Kaplan-Meier method. Results: Most patients were male (62.29%), 70 to 95 years (29.39%), fair-skinned (65.80%),low education level (54.39%), from the Unified Health System (78.29%) with no history of tobacco (44.59%) or alcoholic consumption (60.00%). The most affected site was the large intestine (35.14%), and the predominant stage was stage III (53.28%). The overall survival was 16.5%. Conclusion: Variables relevant to survival were cancer site, educational level, history of tabagism, origin of cancer, clinical staging and basis for diagnosis. Esophagus cancer presented the lowest survival time, with a median of 248 days.