Related Experiment Video
Updated: Jul 12, 2026
Characterization of Electrode Materials for Lithium Ion and Sodium Ion Batteries Using Synchrotron Radiation Techniques
Published on: November 11, 2013
Using artificial intelligence to identify characteristics associated with clinical and economic outcomes in MASH
Kamal Kant Mangla1, Semiu O Gbadamosi2, Daniel Semeniuta3
1Novo Nordisk Service Centre India Pvt. Ltd., Bengaluru, KA 560066, India.
Background:
Heterogeneity in the disease characteristics of metabolic dysfunction-associated steatohepatitis (MASH) complicates efforts to identify individuals with high unmet need.
Objective:
We used artificial intelligence (AI) phenotyping to identify factors associated with rapid fibrosis progression, long-term clinical outcomes, and high healthcare costs.
Design:
In this retrospective cohort study (January 1, 2013-September 1, 2022), patients with a MASH diagnosis, a Fibrosis-4 index score within 90 days of first MASH diagnosis, and no other liver diseases were identified in a United States electronic medical records and claims dataset.
Methods:
Using machine learning, characteristics, including diagnoses, procedures, and medications, were grouped into phenotypic signals, which were iteratively refined and evaluated for strength of association with each outcome. Differences between subgroups in the proportions of patients with specific phenotypic signals were calculated.
Results:
In total, 14,707 patients were included in at least one analysis. Rapid fibrosis progression (n = 1795) was associated with a history of anemia, thrombocytopenia, and cardiovascular (CV) diagnoses. Long-term liver- and CV-related outcomes (n = 13,880) were associated with chronic diseases, kidney diagnoses, cardiac diagnoses, medications, lab tests, and injuries. High healthcare costs (n = 10,133) were associated with heart failure procedure codes, cardiac diagnoses and testing, hospitalization procedure codes and kidney diagnoses, and gastrointestinal diagnoses and abdominal imaging.
Conclusion:
This study demonstrates the application of AI phenotyping to multidimensional real-world data. This approach could support new methods for proactively identifying patients in clinical practice who require closer monitoring and management.
