Lower extremity peripheral artery disease (PAD), affects an estimated 8 million people in the United States. Nearly 50% of patients with PAD die within a 10-year period. A significantly larger proportion experience disability and poor quality of life due to leg pain, and poor mobility. PAD is also a leading cause of amputation, often in patients with advanced disease who develop critical limb ischemia. Risk of myocardial infarction and stroke is also increased several-fold. While risk factors for PAD, which include smoking, diabetes, hypertension and dyslipidemia are common in veterans, contemporary data regarding the epidemiology, treatment patterns and clinical outcomes in PAD patients remain sparse. Lack of a reliable method of identifying PAD in administrative data is a major impediment to high-quality PAD outcomes research. The overall objective of this pilot application is to develop and validate an automated algorithm to identify patients with PAD in administrative data with a high degree of accuracy, which would facilitate the development of a nationwide PAD registry in the Veterans Health Administration (VHA). Given the high prevalence of PAD risk factors, and its associated morbidity and mortality, such a registry could be a powerful resource to study the epidemiology of PAD, examine intensity of risk factor modification, compare the effectiveness of different treatments, and identify opportunities for improving care. To achieve our goals, we will extract data on key variables from reports of the ankle brachial index (ABI) - a routinely performed diagnostic test in patients with suspected PAD. The ABI is a ratio of Doppler recorded systolic blood pressure in the lower and upper extremity, and a value of < 0.90, is considered abnormal. An ABI value < 0.90 has excellent sensitivity (79%) and specificity (96%) for diagnosing PAD, which is further enhanced with incorporation of toe-brachial index (ratio of toe and brachial pressures), and vessel non- compressibility. In addition, ABI values correlate strongly with severity of PAD, risk of future limb-related and cardiovascular events, and are used for monitoring response to treatment. Currently, the ABI test results reside within dedicated test reports in the VHA?s electronic health record as semi-structured text, and therefore not available for research purposes. We propose to update and implement an existing natural language processing (NLP) system to extract information on the above key variables from the ABI test reports, evaluate its performance on national VHA data, and determine the diagnostic accuracy of an automated algorithm based on NLP-extracted values to identify PAD. Our central hypothesis is that such an algorithm will achieve a high positive predictive value (PPV) for detecting PAD within the VHA. This application will leverage key strengths of the VHA - our nation?s largest integrated health system with a common EHR across sites. The project is a collaborative effort between investigators at the Iowa City, Nashville and Salt Lake City with operational support from VA Informatics and Computing Infrastructure (VINCI) and the VA Office of Specialty Care. To test the above hypothesis, we propose the following specific aims:
Aim 1. Enhance and implement an existing NLP system to extract values for ABI, TBI and presence of non- compressible arteries from ABI test reports using national VHA data Aim 2. Determine the diagnostic accuracy of a structured algorithm based on NLP-extracted values of ABI, TBI and non-compressibility to detect PAD at the patient-level
Peripheral artery disease (PAD), also referred to as blockage in leg arteries affects more than 8 million people in the United States. PAD causes significant morbidity due to leg pain and impaired mobility, and is a leading cause of leg amputation. Among patients who are affected with PAD, risk of heart attack and stroke is also markedly increased. Although a number of treatments for PAD are available, our understanding of how well these treatments are applied to patients in the real world, and how they impact health and quality of life remain limited. We propose to develop a novel method of building a nationwide database of veterans with PAD. Such a database will allow our research team to observe a large number of PAD patients over time with regards to their health outcomes and experience with PAD treatment, and enable us to identify opportunities for improving care for veterans with this debilitating condition.