Interval-censored data with misclassification: a Bayesian approach.

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Survival data often involve silent events, leading to interval censoring and classification errors from imperfect diagnostic tests.
• Accurate estimation of event occurrence time distribution is crucial but challenging with such data limitations.

Purpose of the Study:

• To develop and evaluate a statistical methodology that improves parameter estimates in parametric proportional hazard models for interval-censored survival data with classification errors.
• To incorporate validation subsets and Bayesian inference to address uncertainties in test sensitivity and specificity.

Main Methods:

• Incorporation of validation subsets into a parametric proportional hazard model.
• Application of Bayesian inference, specifically a Gibbs sampling procedure, for model analysis.
• Evaluation through simulation studies and analysis of real-world HIV acquisition data.

Main Results:

• The proposed model with validation subsets significantly improves parameter estimates, showing lower bias and standard deviation.
• Bayesian inference combined with validation data compensates for unknown test sensitivity and specificity.
• The methodology demonstrates superior performance compared to models lacking validation subsets or assuming perfect test accuracy.

Conclusions:

• The novel methodology effectively enhances the analysis of complex survival data, particularly when dealing with interval censoring and classification errors.
• Validation subsets and Bayesian methods offer a robust solution for improving accuracy in survival time distribution estimation.
• The approach is validated by simulation studies and real-world application in HIV research.