Outcome measurement error correction for survival analyses with multiple failure types: application to hearing loss studies

TL;DR

This paper develops statistical methods to correct measurement errors in survival analysis with multiple failure types, specifically applied to hearing loss studies, improving the accuracy of risk factor assessments from self-reported data.

Contribution

It introduces a reweighting strategy for bias correction in survival analysis with multiple failure types due to measurement errors, with theoretical properties and practical application.

Findings

Self-reported hearing outcomes can be corrected for measurement errors.

Adjusted analysis shows tinnitus is positively associated with moderate hearing loss.

Effects are similar across different hearing frequencies.

Abstract

In epidemiological studies, participants' disease status is often collected through self-reported outcomes in place of formal medical tests due to budget constraints. However, self-reported outcomes are often subject to measurement errors, and may lead to biased estimates if used in statistical analyses. In this paper, we propose statistical methods to correct for outcome measurement errors in survival analyses with multiple failure types through a reweighting strategy. We also discuss asymptotic properties of the proposed estimators and derive their asymptotic variances. The work is motivated by Conservation of Hearing Study (CHEARS) which aims to evaluate risk factors for hearing loss in the Nurses' Health Studies II (NHS II). We apply the proposed method to adjust for the measurement errors in self-reported hearing outcomes; the analysis results suggest that tinnitus is positively associated with moderate hearing loss at both low or mid and high sound frequencies, while the effects between different frequencies are similar.