Predicting disease risks by matching quantiles estimation for censored data.

Area of Science:

• Biostatistics
• Survival Analysis
• Predictive Modeling

Background:

• Accurate prediction of survival outcomes is crucial in time-to-event data analysis.
• Traditional semiparametric regression models for survival analysis often struggle with prediction accuracy due to challenges in correctly specifying covariate-survival associations, especially in heterogeneous patient populations or complex models.

Purpose of the Study:

• To propose a novel, flexible disease-risk prediction approach for time-to-event data analysis.
• To develop a method that bypasses the need for a priori model specification and directly links covariate distributions to survival time quantiles.

Main Methods:

• A disease-risk prediction approach is proposed by matching an optimal combination of covariates with the survival time in terms of distribution quantiles.
• The redistribution-of-mass technique is employed to effectively handle censored data.
• Theoretical properties of the proposed method are established.

Main Results:

• The proposed method demonstrates ease of implementation and flexibility, operating without assuming an a priori model.
• Simulation studies and a real data example confirm the practical utility and accuracy of the disease-risk prediction approach.
• The method provides an alternative to traditional approaches that rely on hazard function or estimating equation estimation.

Conclusions:

• The developed disease-risk prediction method offers a robust and flexible alternative for survival analysis.
• This approach enhances prediction accuracy by directly modeling the relationship between covariate distributions and survival time quantiles.
• The method is suitable for complex scenarios and heterogeneous patient populations where traditional models may falter.