S2429 Beyond FIB-4: A Multivariable Risk Prediction Model for Advanced Fibrosis in Type 2 Diabetes Mellitus Using Outpatient Data

Introduction: FIB-4 is a valuable noninvasive score for identifying patients at risk for hepatic fibrosis, but it is often underutilized or unavailable at the point of care. We sought to develop a multivariable model using routine outpatient clinical data to predict elevated FIB-4 in patients with Type 2 Diabetes Mellitus (T2DM), offering an adjunctive tool to trigger MASLD care pathways. The aim of the study was to develop and evaluate a predictive model for identifying patients with FIB-4 >1.3 using demographic and laboratory data. Methods: A retrospective cohort of T2DM patients seen in outpatient care in 2021 was analyzed. Predictor variables included age, sex, body mass index, A1c, race, and provider type. The outcome was a calculated FIB-4 >1.3. A logistic regression model with L2 regularization was trained and validated on a 75/25 split. Model performance was assessed using AUC and confusion matrix. Results: Among 1,143 patients: The final model demonstrated an AUC of 0.79. Top predictors included race (Nepalese, Somali, Hispanic, Nigerian), with NP/PAs less likely to care for patients with elevated FIB-4. In the test set, the model had: sensitivity 41.5%, specificity 94.2%, accuracy 78%. Conclusion: We present a novel predictive model for advanced fibrosis risk in patients with T2DM using standard outpatient data. With strong discriminatory performance, this model may serve as a surrogate flag to prompt further MASLD workup or specialist referral—especially when full lab-based scores like FIB-4 are unavailable. Integration into EMR systems can help close the gap in MASLD recognition.