ABSTRACT Venous thromboembolism (VTE), consisting of pulmonary embolism and deep vein thrombosis, is a common and consequential public health problem affecting up to 600,000 adults in the United States annually. VTE requires timely detection and treatment, but the VTE diagnosis workflow in ambulatory care is fraught with challenges, including delays, inaccuracies, and misdirection, influenced by multiple factors including nonspecific symptoms and a lack of systematic measurement and quality improvement tools. VTE is associated with health disparities (highest in black and African American populations) and missed or delayed VTE diagnosis can have serious consequences for patient and healthcare cost outcomes, resulting in increased risk of morbidity, mortality, and prolonged hospital stays. These issues may be further compounded by variation in the types of ambulatory care practices (primary care versus urgent care) and the geographical locations and socio-economic status where patients seek care. Our team developed an electronic clinical quality measure (eCQM) that uses structured and unstructured EHR data to measure Diagnostic Delay of Venous Thromboembolism (DOVE) in primary care settings that was recently endorsed by the Partnership for Quality Measurement. Using this eCQM, the rate of delayed VTE diagnosis in urban, metro, and rural primary care practices across three large healthcare systems using different EHR systems was found to be consistently over 70%, suggesting that this is an important type of diagnostic error (DE) with likely negative impacts on patient outcomes. Building on our preliminary work, we propose to leverage EHR data and stakeholder expertise to gain an understanding of VTE DE risk factors, disparities and outcomes. We will develop artificial intelligence (AI) and statistical learning tools to identify, factorize, and address vulnerabilities in a range of VTE DE workflows including delayed and missed VTE diagnosis in both ambulatory and urgent care settings. This study brings together a strong interdisciplinary team of experts in primary care, VTE diagnosis, informatics, data science (natural language processing and machine learning). The advanced data-driven eCQM will be refined using high-dimensional EHR data (structured and unstructured) to quantfy timely, delayed and missed VTE diagnoses. To increase generalizability of the results, we will use multiple data sources from 220 primary care/urgent care practices and clinics associated with 13 hospitals from urban, metro and rural clinics in the Northeastern and Southern United States (2 different EHR vendors). If successful, this approach will substantially improve our understanding of DEs and related risk factors, VTE DE related costs and build a foundation for improving VTE diagnostic accuracy and precision and diminish disparities in healthcare outcomes.