An Efficient Data Integration Scheme for Synthesizing Information from Multiple Secondary Datasets for the Parameter Inference of the Main Analysis

TL;DR

This paper introduces MinBo, a robust data integration method that leverages secondary datasets to enhance the efficiency of primary analysis in observational studies and clinical trials.

Contribution

The paper presents a novel method called MinBo for integrating secondary data to improve main analysis efficiency, robust to model misspecification.

Findings

MinBo outperforms existing methods in efficiency gains.

Theoretical proofs support MinBo's robustness.

Case study demonstrates practical effectiveness in hypertension risk assessment.

Abstract

Many observational studies and clinical trials collect various secondary outcomes that may be highly correlated with the primary endpoint. These secondary outcomes are often analyzed in secondary analyses separately from the main data analysis. However, these secondary outcomes can be used to improve the estimation precision in the main analysis. We propose a method called Multiple Information Borrowing (MinBo) that borrows information from secondary data (containing secondary outcomes and covariates) to improve the efficiency of the main analysis. The proposed method is robust against model misspecification of the secondary data. Both theoretical and case studies demonstrate that MinBo outperforms existing methods in terms of efficiency gain. We apply MinBo to data from the Atherosclerosis Risk in Communities study to assess risk factors for hypertension.