Predicting survival time for critically ill patients with heart failure using conformalized survival analysis

Area of Science:

• Critical Care Medicine
• Biostatistics
• Health Informatics

Background:

• Heart failure (HF) presents a major public health concern, particularly for critically ill patients in intensive care units (ICUs).
• Accurate prediction of survival outcomes with calibrated uncertainty is crucial for effective treatment strategies in ICUs.

Purpose of the Study:

• To introduce and evaluate conformalized survival analysis (CSA) as a novel method for survival time prediction in critically ill HF patients.
• To assess CSA's capability in providing statistically rigorous uncertainty quantification for survival predictions.

Main Methods:

• Developed and applied conformalized survival analysis (CSA) to predict survival times.
• Utilized the large-scale, real-world MIMIC-IV database for analysis.
• Compared CSA's performance against traditional parametric models like Cox and Accelerated Failure Time (AFT) models.

Main Results:

• CSA demonstrated effective calibrated uncertainty quantification for survival predictions in critically ill HF patients.
• CSA provided a statistically rigorous lower bound for each predicted survival time.
• CSA outperformed traditional parametric models in delivering reliable uncertainty estimates.

Conclusions:

• Conformalized survival analysis (CSA) is a promising novel method for improving prognostic accuracy in critical care settings.
• CSA's ability to provide calibrated uncertainty enhances decision-making for critically ill heart failure patients.
• This approach offers a more precise and reliable tool for prognosis where accurate predictions are vital.