Clinical Outcome Prediction
Overview
Clinical outcome prediction estimates what may happen to a patient, from short-term deterioration and readmission to disease progression, treatment response, and survival. Common approaches include tree-based models and survival analysis, temporal transformers and recurrent networks, multimodal fusion of EHRs with imaging or omics, and newer foundation models, federated learning, and causal or digital-twin methods. Evaluation spans settings such as ICU records from MIMIC and eICU, longitudinal Alzheimer’s cohorts such as ADNI, and multi-site cancer and imaging benchmarks, with attention to external validation, calibration, and subgroup performance. A central tension is that complex models can improve prediction, yet missing data, site shifts, label leakage, and weak uncertainty or fairness assessment often limit whether results generalize and support clinical decisions.
Latest papers
266 papers24 papers in the last four weeks, down 8% on the four weeks before. 0.1% of all new papers.