Bayesian Survival Prediction For Female Breast Cancer Patients

Survival prediction models for cancer patients are important for healthcare researchers to identify the high-medical risk patients,to adequately distribute medical resources and to estimate clinical expenditures.The aim of this study is to build an effective survival prediction model for female breast cancer patients by incorporating the highly significant prognostic marker-estimated LNR(lymph node ratio).Data were obtained from the NCI SEER cancer registry on 4024 patients with infiltrating duct and lobular carcinoma breast cancer diagnosed in 2006-2010 time period.The Bayesian statistical approach was developed to model the LNR and patients' survival simultaneously.The web application is employed for survival prediction for individualized patient.Moreover,various versions of the Cox regression model were examined in order to select the most precise survival prediction model.In particular,the violation of the proportionality assumption led us to stratify the Cox model by variable Age.The work is focused on determining the prognostic value of the LNR and on examination of extended Cox regression models.The estimated LNR was found to be of great significance in this work since the predictive ability of the models with the LNR are higher than the ones without LNR.Among implemented survival model,the most precise is the Cox regression model.The extensions of the Cox regression model did not provide accuracy improvements conceding only with 0.03 differences.