Severe sepsis mortality prediction with logistic regression over latent factors

摘要: Sepsis is one of the main causes death for non-coronary ICU (Intensive Care Unit) patients and has become 10th most common cause in western societies. This a transversal condition affecting immunocompromised patients, critically ill post-surgery with AIDS, elderly. In countries, septic account as much 25% bed utilization pathology affects 1-2% all hospitalizations. Its mortality rates range from 12.8% sepsis to 45.7% shock. The prediction caused by is, therefore, relevant research challenge medical viewpoint. clinical indicators currently use this type have been criticized their poor prognostic significance. study, we redescribe through latent model-based feature extraction, using factor analysis. These extracted are then applied sepsis. reported results show that proposed method improves on obtained current standard predictor, which based APACHE II score.