Hepatic steatosis is a common radiographic "incidental finding" that is overlooked and underreported to patients. The investigators developed a clinical decision support system using machine learning and natural language processing that will prompt reporting to patients and provide ED clinicians risk stratified follow-up care recommendations. Data on both the implementation and effectiveness of our intervention resulting from this trial will inform future use with a goal of ultimately improving diagnostic safety and outcomes for patients with hepatic steatosis.

Metabolic dysfunction-associated steatotic liver disease (MASLD) is a public health crisis affecting one third of the US adult population. Steatotic liver disease is an obesity related disease with serious health and quality of life consequences. Hepatic steatosis is reported on ~10% of written radiology reports. Unfortunately, lack of awareness often leads to hepatic steatosis being overlooked by providers and unreported to patients leading to delays in diagnosis and referral to specialty care.

This project, Improving Diagnostic Safety through STeatosis Identification, Risk stratification, and Referral pathway in the ED (STIRRED), proposes to improve diagnostic safety by ensuring that patients receive timely notification of the new finding and referral to follow up care. The STIRRED CDSS intervention employs natural language processing and machine learning to identify hepatic steatosis in written emergency department (ED) radiology reports and combine it with common clinical data in the electronic health record to create an individualized risk profile and follow up recommendations for patients with suspected MASLD. This recommendation is communicated to the ED clinician during the ED discharge process via an OurPractice Advisory (OPA) and triggering outpatient referral for follow-up care stratified by risk level.

This trial will evaluate STIRRED across a large health system using a type 2 implementation-effectiveness stepped wedge cluster randomized trial across 11 Emergency Departments in a single health system. Effectiveness of providing risk-based care linkage and implementation outcomes will be evaluated.

Inclusion Criteria:

• imaging finding of hepatic steatosis on ED imaging study
• discharged from the ED to home
• first Northwestern Medicine ED visit in the study period with a hepatic steatosis finding

Exclusion Criteria:

• admitted to the hospital
• age < 18
• pre-existing Liver Disease diagnosis (Liver Cancer, HCV, HBV, Cirrhosis, NAFLD/MASLD/NASH/MASH, Alcohol Liver Disease, PSC, PBC and autoimmune hepatitis)

De-identified IPD used in the analysis of the hybrid implementation effectiveness trial will be shared. All published data will be formatted according to the principals of Common Data Elements (CDE) as defined by the NIH. Clinical and survey response data will contain raw item responses as well as any aggregate scores for participants. To facilitate future use and interoperability of our data, we will employ standard processing and documentation protocols adopted by the Inter-university Consortium for Political and Social Research (ICPSR) for data formats and dictionaries as well as for variable names, descriptions, and labels.

Data will be made available through a data repository (still to be determined) following the completion of the grant period and final publication of the trial manuscript.