Dynamic risk score modeling for multiple longitudinal risk factors and survival

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Informatics

Background:

• Prognostic modeling for disease risk and survival often involves longitudinal data.
• Incorporating multiple longitudinal risk factors and dependent censoring presents analytical challenges.
• Personalized decision-making requires dynamic risk assessment tools.

Purpose of the Study:

• To propose a dynamic risk score modeling framework for multiple longitudinal risk factors and survival.
• To address the challenge of dependent censoring in prognostic modeling.
• To develop a parsimonious model accommodating numerous risk factors.

Main Methods:

• A dynamic risk score modeling framework was developed for competing risks.
• The model accommodates multiple longitudinal risk factors with few random effects.
• Dependent censoring was explicitly handled based on post-baseline clinical progression.

Main Results:

• The proposed method demonstrated satisfactory performance in extensive simulation studies.
• The model was successfully applied to a pediatric acute liver failure registry study.
• It modeled death using trajectories of multiple clinical and biochemical markers.

Conclusions:

• The developed framework provides a parsimonious approach to modeling disease risk with multiple longitudinal factors.
• The model generates an easily calculable longitudinal risk score for disease monitoring.
• This tool can enhance personalized medical decision-making and patient care.