Abstract 4366954: Leveraging Large Language Models to Extract Parity from the Electronic Health Record and Reveal Hidden Cardiovascular Risk Factors: A Retrospective Study of Takotsubo Cardiomyopathy

Abstract

Background: Takotsubo cardiomyopathy (TTCM) is a form of heart failure first described in the 1990s that was believed to be triggered by significant emotional events or stressors - giving it the moniker “broken heart syndrome”. It is now known that both emotional and physiological stressors can induce TTCM in at-risk individuals. Incidence of TTCM is skewed with 80-90% of cases occurring in females over age 50. One small cohort-based study reported an association between parity and risk of TTCM. Hypothesis and Purpose: We propose that parity is a risk factor for TTCM that can be quantified using a data science approach using real world data (RWD) to evaluate long-term biological impact of parity on cardiovascular health. To evaluate the association, we propose to develop a novel methodology to replicably extract parity information from the structured and unstructured data in the electronic health record (EHR). Study Design and Methods: Analysis was conducted using an access-limited, privacy-preserving analytic platform hosting >7.3 million unique records of clinical encounters at a multistate integrated health system. The study cohort was restricted to data from individuals 50-80 years old with ≥1 electrocardiogram in the clinical record. Parity data was determined for 99.8% of records using a replicable, novel method to capture structured and unstructured data. Association between TTCM and parity was evaluated in the cohort. Results: Parity information for 122,769 females was extracted from clinical record data. There were no demographic differences observed in the parity cohort. We observed a 10-20% increased TTCM risk (Figure) among females with nonzero parity values compared to nulliparous controls (n=14,081).To validate our methodology for extracting and analyzing parity data in relation to disease risk, we leveraged the established inverse association between parity and ovarian cancer as a proof-of-concept, which was confirmed in our analysis. Conclusion: Our findings suggest that women with higher parity are at greater risk of developing TTCM. This finding points to potential pregnancy-related contributions to the significant sex-bias in prevalence of TTCM.