Harnessing Big Data to Identify Effective Peripheral Artery Disease Treatments in Chronic Kidney Disease

PROJECT SUMMARY / ABSTRACT
Peripheral artery disease (PAD), characterized by diseased arteries to the limbs, affects 200 million people
worldwide and 9 million people in the U.S. Chronic kidney disease (CKD) affects 20 million people in the U.S.
and confers a markedly higher risk for PAD. Yet patients with CKD are less likely to have revascularization
procedures and are more likely to undergo lower extremity amputation than patients without CKD. In addition
to a high prevalence of traditional risk factors such as hypertension and diabetes mellitus, patients with CKD
have other unique risk factors such as chronic inflammation or uremia, which in turn can lead to more
aggressive PAD at a younger age. Therefore, patients with CKD need dedicated study. Our overarching goal is
to help close these evidence gaps and address these limitations by harnessing the power of Optum
Clinformatics Data Mart, which includes over 7 billion claims records on over 83 million unique lives from all 50
states spanning 2005-2019. Our secondary goal is to facilitate future PAD studies using real-world data by
leveraging the power of natural language processing to improve our ability to accurately and automatically
ascertain PAD from large electronic health record databases. Our innovative algorithm will be of particular
importance among subgroups where clinical trial evidence is limited, such as in advanced CKD. Our proposal
has the Specific Aims. Aim 1: To evaluate lower extremity revascularization in patients with non-dialysis-
requiring CKD. We hypothesize that patients with CKD undergoing surgical versus endovascular
revascularization will have longer initial hospitalization, but fewer subsequent major adverse limb events. AIM
2: To evaluate antiplatelet and anticoagulant medications after lower extremity revascularization in patients
with non-dialysis-requiring CKD. We hypothesize that real-world patients with CKD treated with antiplatelet
medications or direct oral anticoagulants after lower extremity revascularization will have higher rates of
bleeding but lower rates of major adverse limb events. AIM 3: To develop an algorithm that accurately and
automatically ascertains PAD from electronic health record databases. We hypothesize that a natural language
processing-approach applied to diagnostic vascular testing reports will have better test performance (i.e.
sensitivity, specificity, positive and negative predictive values) for identifying PAD than a traditional approach
that uses administrative billing codes. Manual chart review will serve as the gold standard.

Grant Number: 5R01HL151351-05
NIH Institute/Center: NIH
Principal Investigator: Tara Chang