Quantile regression for nonignorable missing data with its application of analyzing electronic medical records

Area of Science:

• Biostatistics
• Health Informatics
• Biomedical Research

Background:

• Electronic Medical Records (EMRs) are increasingly used for biomedical research.
• Quantile regression offers insights into EMR data heterogeneity.
• Nonignorable missing data in EMRs hinders accurate analysis and discovery.

Purpose of the Study:

• To propose a novel method for estimating covariate effects in quantile regression with nonignorable missing responses.
• To address the challenge of missing data in EMRs for robust biomedical research.

Main Methods:

• Developed a method imposing no parametric distribution specifications.
• Utilized implicit distributions from quantile regression models.
• Established consistency and asymptotic normality of the proposed estimator.

Main Results:

• The proposed method accurately estimates covariate effects with nonignorable missing data.
• Demonstrated consistency and asymptotic normality of the estimator.
• Validated performance through numerical studies and real-world EMR data analysis.

Conclusions:

• The novel method effectively handles nonignorable missing data in EMRs for quantile regression.
• Provides a robust approach for biomedical discoveries using EMR data.
• Offers an efficient algorithm and bootstrap method for statistical inference.