Prevalent cohort studies and unobserved heterogeneity.

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• Survival analysis methods are established for observed covariates.
• Unobserved covariates pose challenges, causing attenuation in prospective studies.
• Prevalent cohort studies face additional selection biases.

Purpose of the Study:

• To survey recent developments in survival analysis for prevalent cohorts with unobserved covariates.
• To discuss the consequences for hazard rate and accelerated failure time models.
• To address issues of unobserved heterogeneity in survival data.

Main Methods:

• Review of established survival analysis techniques.
• Discussion of hazard rate models with frailties.
• Analysis of prevalent cohort studies with left-truncated data.

Main Results:

• Unobserved covariates lead to attenuation in regression parameter estimates.
• Prevalent cohorts experience differential selection effects.
• Simulation results illustrate the impact of unobserved heterogeneity.

Conclusions:

• Unobserved heterogeneity significantly impacts survival analysis in prevalent cohorts.
• Careful consideration of models is needed when covariates are unobserved.
• Findings are relevant for applications like time-to-pregnancy studies.