Anticoagulant Treatment Effects Assessment in Antiphospholipid Syndrome Patients Using Targeted Learning and the Oracle Health EHR Data

<p>Antiphospholipid syndrome (APS) is a chronic autoimmune condition characterized by an increased risk of both arterial and venous blood clots, leading to significant health complications and even death. While warfarin has traditionally been the go-to anticoagulant for most patients with APS, many clinicians are now turning to direct oral anticoagulants (DOACs). However, there's still a lot of uncertainty about how effective these newer medications are compared to warfarin, especially for patients who are considered high-risk. The randomized studies we have are somewhat limited and might not reflect the diverse experiences of patients in the real world. </p> <p>In my dissertation, I used extensive electronic health record (EHR) data to explore the impact of different anticoagulant options on mortality rates in adults diagnosed with APS. I analyzed de-identified EHR data from Oracle Health, creating a scenario similar to a clinical trial where patients started either warfarin or a DOAC at a specific point in their treatment. To address potential biases that could have influenced the results, I employed advanced statistical methods, including targeted maximum likelihood estimation (TMLE) with Super Learner techniques. My primary focus was on the difference in one-year all-cause mortality risk, but I also looked at long-term outcomes up to five years and examined differences across various patient subgroups. </p> <p>The results indicated that starting DOAC therapy was linked to a higher estimated risk of death within the first year compared to warfarin, even after adjusting for various confounding factors. This difference was especially notable in the earlier follow-up period and less pronounced as VI </p> <p>time went on, with only minor differences observed after three to five years. Interestingly, the analysis showed that patients who had experienced an arterial clot before had a significantly higher risk, reinforcing the notion that certain APS patients are at greater risk. </p> <p>Overall, these findings bolster the idea that warfarin should still be the preferred treatment for APS, particularly in the early stages and among patients who are at higher risk. Beyond this specific conclusion, my work highlights how we can use sophisticated methods to analyze complex EHR data, ultimately providing valuable insights in circumstances where traditional clinical trials may not be feasible. Complementary descriptive analyses using the IQVIA national claims database document a 40 percent rise in APS incidence between 2019 and 2024, pronounced geographic and socioeconomic disparities in diagnosis, and a 50- to 60-fold drug cost differential favoring warfarin over DOACs — findings that contextualize the clinical results and strengthen the policy case for warfarin as the preferred anticoagulant in this population. </p>

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computational and Data Sciences

Year dc:date.available

2026

Author dc:creator

• Odabas, Omer J

Contributors dc:contributor

• Cyril Rakovski
• Adrian Vajiac
• Ehsan Yaghmaei

• Causal Inference
• Machine Learning
• Superlearner
• Targeted Learning
• TMLE
• Biostatistics
• Data Science
• Longitudinal Data Analysis and Time Series
• Statistical Models
• Survival Analysis

Repository record dc:identifier

https://digitalcommons.chapman.edu/cads_dissertations/52

OAI identifier oai:identifier

oai:digitalcommons.chapman.edu:cads_dissertations-1053