在基于电子健康记录（EHR）的队列中探索代谢功能障碍相关脂肪性肝病（MASLD）的多层次危险因素

Background: MASLD is a growing risk factor for liver cancer. Few studies have examined multilevel risk factors contributing to MASLD disparities, particularly the impact of neighborhood attributes. Elucidating multilevel risk factors can help identify groups relevant for targeted interventions to mitigate liver disease, including liver cancer. Methods: Our EHR-based cohort included adults with at least one in-person visit between 2000-2017 to Kaiser Permanente Hawai'i. From the EHR, we extracted data on patient sociodemographic characteristics; diagnostic codes; medication prescriptions; and lab results. Patient observations were appended to census tract-level neighborhood SES (nSES). The outcome was incident MASLD, defined using ICD-codes with follow-up until 12/31/2017. We used multivariable Cox proportional hazards regression to explore factors associated with incident MASLD; models were adjusted for baseline year and decile of number of encounters. Model 1 included sociodemographic variables (i.e., sex, race/ethnicity, English proficiency); in Model 2 we retained sociodemographic variables which were multivariably statistically significant and added clinical characteristics (i.e., body mass index [BMI], smoking history, hypertension, diabetes, and hyperlipidemia); and in Model 3 we retained clinical characteristics which were multivariably statistically significant and added nSES. Results: Of 439,709 patients, 8,339 had incident MASLD during 18,350 person-years of follow up. In Model 1, males had higher risk of developing MASLD than females. Compared to non-Hispanic White individuals, those who identified as Chinese, Japanese, Korean, multiracial, or Hispanic had higher risk (HR range 1.08-1.83), and as Native Hawaiian, Pacific Islander, or Black had lower risk of MASLD (HR range 0.54-0.78). Limited English proficiency was associated with increased MASLD risk (HR= 1.52, 95% CI 1.30-1.78). In model 2, overweight and obese BMI, ever smoking, elevated triglycerides, low HDL, and diabetes were associated with higher MASLD risk (HR range 1.02-1.99). In Model 3, nSES was not statistically significantly associated with MASLD risk when added to sociodemographic and clinical factors. Significant factors in prior models remained independently associated with MASLD risk, except for smoking. Discussion: Our study provided critical disaggregated data on MASLD risk for ethnic groups that are typically understudied. We found that among this diverse cohort, disparities in MASLD incidence persist, despite adjusting for other sociodemographic, clinical, and neighborhood SES. Future efforts to reduce MASLD burden may consider focusing on these ethnic groups at higher risk. Furthermore, future studies should explore the role of other neighborhood social and built attributes such as food environment and access to parks.