Title: Complexities in the analysis of electronic health records: An illustrative case of a chronic kidney disease study
Authors: Guofen Yan - University of Virginia (United States) [presenting]
Abstract: While randomization is generally considered the gold standard approach to address clinical questions such as efficacy of an experimental treatment, some questions may only be addressed using observational data, such as the relationship of race and outcomes. Chronic kidney disease (CKD) is a major public health problem in the US and worldwide, with the current estimated prevalence of 10-12\% in the US. Many questions remain unanswered, from simple facts like the distribution of onset-age to more involved issues such as the differential survival rates among racial and ethnic groups. Utilizing the U.S. veteran electronic health record (EHR) database for more than 8.9 million veterans -- the largest EHR database among U.S. health care systems, we started to address many important questions. We present our project and share our experience, with emphasis on the study design, modelling and analysis, and accompanying challenges such as how to identify the onset time of CKD and issues relate to irregular and informative visits and confounding.