Addressing Bias from Missing Data in EHR Based Studies of CVD

Project Summary
This NHLBI K01 application supports the career development of Dr. Nrupen A. Bhavsar, PhD, an Assistant Professor
of Medicine at the Duke University School of Medicine. Dr. Bhavsar is a chronic disease epidemiologist who has
performed multidisciplinary studies in the epidemiology of CVD, chronic kidney disease and cancer. He is passionate
about pursuing a career in clinical research at the intersection of epidemiology, informatics, and biostatistics. At the
end of the award period, Dr. Bhavsar will be an independent investigator applying knowledge gained through this K01
to develop large scale EHR-based population studies that identify individuals at high risk for cardiovascular disease
(CVD) events. Through training in data linkage, machine learning, and causal inference, he will apply missing data
methods to conduct rigorous non-randomized studies to improve health. The topical areas of the proposed training
and research are diabetes and incident CVD events in the application of data linkage, machine learning, and causal
inference. Career development aim: Obtain transdisciplinary competencies within informatics, biostatistics, and
population sciences to investigate methodological challenges inherent in the use of multi-health system EHR data for
clinical research. The training approach will leverage didactic and experiential training complemented by analyses of
data derived from a multi-health system, multi-state research collaborative. Study population: Patients who received
care at one of the North Carolina or South Carolina “Carolinas Collaborative” institutions (Duke University Medical
Center, UNC-CH Health System, Wake Forest Baptist Health Center, and 9 additional health systems collaborating
within the Health Sciences of South Carolina institutions). Specific aims: This proposal will identify approaches to
account for missing data when patients seek care across multiple health systems but EHR data is only available from
a single health system. Estimating the systemic bias introduced by missing data for single institution studies and
identifying methods for accounting for missing data biases may improve the ability of EHR data to be used for clinical
research. Anticipated results: Through this NHLBI K01 Research Scientist Career Development Award, Dr. Bhavsar
will acquire essential training and research experience to develop large scale EHR-based population studies in CVD.