Sign in Subscribe
Machine learning

Machine Learning Model Accurately Forecasts ICU Mortality in Lung Cancer Patients with Sepsis

Strategic Insights

— 2 min read

Research conducted at the Department of Endocrinology and Metabolism, Huai'an Hospital Affiliated to Xuzhou Medical University and Huai'an Second People's Hospital, Huai'an, Jiangsu, People's Republic of China, has established a machine learning (ML) model to predict ICU mortality in patients with sepsis combined lung cancer. The model, based on a combination of 13 clinical variables, demonstrated excellent accuracy and reliability, facilitating personalized prognostic forecasts for lung cancer patients with sepsis. The study utilized the Medical Information Mart for Intensive Care IV (MIMIC IV) database and included a total of 1347 patients, with 1096 patients used for training and 251 patients used for external validation.

Key Takeaways:

  • Researchers from the Department of Endocrinology and Metabolism, Huai'an Hospital Affiliated to Xuzhou Medical University and Huai'an Second People's Hospital, Huai'an, Jiangsu, People's Republic of China, developed a machine learning (ML) model to predict ICU mortality in patients with sepsis combined lung cancer.
  • The model, based on 13 clinical variables, achieved an area under the curve (AUC) of 0.931 in the training cohort, 0.698 in the internal validation cohort, and 0.794 in the external validation cohort.
  • The Oxford Acute Severity of Illness Score (OASIS) was identified as the greatest influence on ICU mortality according to SHAP (SHapley Additive exPlanations) interpretation.
  • The research concluded that the ML models demonstrate excellent accuracy and reliability, facilitating more rigorous personalized prognostic forecasts for lung cancer patients combined sepsis.
  • A total of 1347 patients were included in the study, with 1096 patients from the MIMIC IV database and 251 patients from an external validation set.
  • The study demonstrated the potential of machine learning algorithms in predicting ICU mortality in patients with sepsis combined lung cancer.

Statistics:

  • 1347 patients were included in the study (1096 from the MIMIC IV database and 251 from an external validation set).
  • The area under the curve (AUC) for the training cohort was 0.931 (0.921, 0.945).
  • The AUC for the internal validation cohort was 0.698 (0.673, 0.724).
  • The AUC for the external validation cohort was 0.794 (0.725, 0.879).
  • The CatBoost model was identified as the prime prediction model with the highest AUC in the training, internal validation, and external validation cohorts.

Sources:

  • Personalized ICU mortality assessment by interpretable machine learning algorithms in patients with sepsis combined lung cancer: a population-based study and an external validation cohort. Frontiers in Oncology, 2025;15:1661212.
  • Researchers from Department of Endocrinology and Metabolism Discuss Findings in Lung Cancer (Personalized ICU mortality assessment by interpretable machine learning algorithms in patients with sepsis combined lung cancer: a population-based ...). Cancer Weekly. October 28, 2025; p 744.

Read more