Analysis of survival data having time-dependent covariates

Area of Science:

• Biostatistics
• Machine Learning in Medicine
• Survival Analysis

Background:

• Cox's proportional hazards model is standard for censored survival data analysis.
• Accurate survival function estimation and prediction are crucial for patient management.
• Limitations exist in current models for dynamic short-term survival prediction.

Purpose of the Study:

• To propose a novel neural network model incorporating bootstrapping for survival data analysis.
• To enhance the estimation of survival functions.
• To improve short-term survival prediction during disease progression.

Main Methods:

• Development of a neural network model.
• Integration of bootstrapping for model optimization (hidden units) and goodness-of-fit testing.
• Application to a long-term primary biliary cirrhosis (PBC) patient cohort.

Main Results:

• The proposed bootstrapping neural network model effectively estimates survival functions.
• Accurate prediction of short-term survival is achieved at various time points.
• The model demonstrates utility in analyzing complex survival data.

Conclusions:

• The bootstrapping neural network offers a powerful alternative to traditional models for survival data.
• This approach improves prognostic accuracy and dynamic survival prediction.
• The method shows promise for clinical applications, particularly in chronic diseases like PBC.