Association between survival time and ordinal covariates

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Methods

Background:

• Assessing independence between survival time and covariates is crucial in medical research.
• Existing methods may have limitations with censored data and ordinal covariates.
• Rank correlation offers a non-parametric approach suitable for various data distributions.

Purpose of the Study:

• To propose a rank correlation-based method for testing independence between censored survival time and an ordinal covariate.
• To develop a test statistic that is interpretable within the proportional hazards framework.
• To demonstrate the proposed test as a generalization of established statistical procedures.

Main Methods:

• The proposed method utilizes rank correlation principles to evaluate the relationship between survival outcomes and ordinal predictor variables.
• A test statistic is constructed by summing the differences between concordant and discordant pairs across event times.
• The statistic is shown to be a score statistic derivable from the proportional hazards model.

Main Results:

• The developed rank correlation test effectively assesses independence in the presence of censored survival data and ordinal covariates.
• The test statistic is mathematically linked to the proportional hazards model, enhancing its theoretical foundation.
• The proposed approach encompasses generalizations of Jonckheere's test and the Mantel-Haenszel procedure.

Conclusions:

• The novel rank correlation test provides a valuable tool for survival data analysis, particularly with ordinal covariates.
• This method offers a unified framework that extends existing statistical techniques.
• The application of rank correlation in this context enhances the analysis of time-to-event data.