Background: Metabolic associated steatotic liver disease (MASLD) is highly prevalent in minority populations and commonly is undiagnosed due to barriers to care. As opposed to longitudinal and preventative encounters with primary care providers, patients from these populations are more likely to have interactions with the health care system in safety net environments, such as emergency departments (ED). Valuable information related to MASLD diagnosis and risk stratification (e.g., incidental hepatic steatosis on imaging, Fibrosis-4 score [FIB-4]) is obtained during these visits but is not utilized due to scattered data in the health record and not being relevant in the urgent setting. We hypothesized that data science and artificial intelligence (AI) techniques can be harnessed to automate detection of at-risk patients in a complex clinical setting where they would otherwise be missed. Utilizing a data science approach, we aimed to identify and phenotype ED patients at risk for MASLD.

Methods: ED patients from August 2016 to November 2023 between the ages of 18-65 years-old were identified. We created an automated algorithm to calculate the FIB-4 score and ED patients with a FIB-4 of >1.3 were included in analysis. We excluded individuals with prior history of liver disease except MASLD. We used ED diagnosis codes to identify the most common diagnoses. We defined patients with MASLD as having a MASLD diagnosis either prior to their index ED encounter or within 6 months after. We used natural language processing (NLP) on radiology reports to identify incidental findings of hepatic steatosis.

Results: Of the 65,116 ED patients with FIB4> 1.3, 35% (22,818) were Black patients and 7.4% (n = 4844) were Hispanic patients. The median FIB-4 score was 2.35 and median age was 55-years-old. The top 3 ED diagnoses were: Chest Pain 5.3%, COVID-19 1.7%, and Syncope 1.3%. Overall, 1.3% (841) of patients had a diagnosis of MASLD. Of the 18,734 (29%) of patients with abdominal imaging performed during the ED encounter, 21% (3,915) had hepatic steatosis. Of patients with hepatic steatosis, 3.4% (134) had a MASLD diagnosis, of which 57.5% were White patients, 24.6% were Black patients, and 5.2% were Hispanic patients (Figure).

Conclusions: Our results show that there are many individuals with elevated FIB-4 values and hepatic steatosis without a MASLD diagnosis, which represents a potential opportunity to identify undiagnosed patients and connect them to appropriate care. Our findings highlight the need for validation of non-invasive scores in the unique ED population (i.e., false elevations in labs due to acute conditions). Further this work exemplifies using AI-based clinical decision support tools for detection and risk stratification of MASLD with a primary goal to promote equitable care.

Kimia Ghobadi

Appearances