Semiparametric bayes' proportional odds models for current status data with underreporting.

Area of Science:

• Biostatistics
• Epidemiology
• Statistical Modeling

Background:

• Current status data present unique challenges in event time analysis due to interval censoring.
• Existing semiparametric Bayesian models often struggle with this data type.
• Understanding fibroid onset timing is crucial for women's health research.

Purpose of the Study:

• To propose a novel semiparametric Bayesian proportional odds model for current status data.
• To nonparametrically estimate the baseline event time distribution using adaptive monotone splines.
• To incorporate risk factors into the model's mean structure and handle potential data underreporting.

Main Methods:

• Developed a semiparametric Bayesian proportional odds model.
• Utilized adaptive monotone splines within a logistic regression framework for nonparametric baseline estimation.
• Employed an efficient Gibbs sampler for model implementation.
• Extended the model to account for systematic underreporting in data subsets.

Main Results:

• The proposed model provides a straightforward and efficient implementation using a Gibbs sampler.
• Successfully applied the methods to an epidemiologic study of uterine fibroids.
• Demonstrated the model's capability to handle interval-censored data and risk factor analysis.

Conclusions:

• The new Bayesian model offers a robust and implementable solution for analyzing current status data.
• This approach advances the statistical methods available for epidemiological studies, particularly those involving interval-censored outcomes like fibroid onset.
• The model's flexibility in handling underreporting enhances its applicability in real-world research.