Title: Improving the statistical rigor of healthcare data integration.

Abstract: The growth of big health data provides remarkable opportunities for integrating different sources of datasets and facilitating healthcare decisions on prevention, diagnostics, and treatments. However, scientific studies, including health sciences, have faced unprecedented challenges due to the failure to replicate influential studies. This talk will discuss how to improve the scientific rigor of using meta-analysis to integrate evidence. Specifically, I developed a novel statistical measure that quantifies replicability and identifies non-replicable studies in a meta-analysis. I will present examples to improve the rigor of three real meta-analyses that study 1) the postoperative complications after cardiac surgery and 2) susceptibility to COVID-19 infection among children and adults. Understanding whether a discrepancy in a meta-analysis arises from non-replicability may give researchers insight to guide further investigation of sources of the discrepancy, polish future study designs, and improve health care decisions.