Abstract 4353854: An Electronic Health Record Multimodal Data Integration Platform for Comprehensive Analysis of Single Ventricle Physiology

BACKGROUND: Multimodal Electronic Health Record (EHR) data for congenital heart disease (CHD) is challenging to integrate and requires subspecialty expertise across several medical domains. Single ventricle physiology (SVP), a CHD subtype, exemplifies this challenge due to its heterogeneous traits and variable outcomes. HYPOTHESIS: An integrative, multimodal EHR platform informed by subspecialty domain knowledge can improve SVP outcome analysis by enabling intuitive visualizations, survival analysis, disease trajectory tracking, and predictive modeling—thereby reducing the complexity and manual integration effort. METHODS: Using an in-house prototype, we analyzed 25,000 CHD patients from UCLA Health. We identified 732 SVP patients using a phenotype algorithm with adjudication. International Classification of Diseases (ICD) and CPT codes were collected, and the CHD patient landscape was visualized using UMAP. Kaplan-Meier curves compared survival across SVP subtypes over three years. Additional analyses included outcomes, including arrhythmias, heart failure, stroke, liver disease, protein-losing enteropathy, and surgical procedures (Norwood, Glenn, Fontan, heart transplantation). RESULTS: A prototype platform was developed to study SVP outcomes. Figure 2A showed 732 SVP patients among 25,000 CHD cases, and Figure 2B highlighted the top 15 ICD codes. Four SVP subtypes were summarized in Figure 2C , while Figure 2D displayed patient encounters from 2013–2023. Survival analysis by birth decade ( Figure 2E ) identified 167 patients with complete records, 138 with at least 36 months of follow-up. Fontan and Double Inlet Left Ventricle patients showed higher survival rate ( Figures 2F–2G ). Disease trajectory analysis ( Figure 2H ) revealed reduced arrhythmias and heart failure severity in the first year, underscoring the importance of early intervention. DISCUSSION: This study demonstrated that an integrative, multimodal EHR-based platform can streamline the analysis of SVP patient outcomes, enabling visualization, survival analysis, disease trajectory tracking, and predictive modeling. This approach reduces the domain-specific complexity and highlights the potential of data-driven tools to assist physicians with integration of multimodal data to enhance patient care delivery. CONCLUSION: An EHR-based platform can enhance SVP outcome analysis by integrating multimodal data and alleviating complexity. Future work will incorporate imaging data and extend to other rare diseases.