Abstract: Objective To devise and verify a screening model for nonalcoholic fatty liver disease(NAFLD). Methods From the “Shandong Multi-center Longitudinal Cohort for Health Management”, 8,993 health examination clients without excessive drinking habits were randomly selected and divided into 2 groups, internal group(80% subjects, for derivation and internal assessment), and external group(20% subjects, for external assessment). Multivariate stepwise logistic regression was used to build a screening model, and prediction power of the model was assessed. Results Multivariate analysis demonstrated that gender, body mass index(BMI), hypertension, dyslipidemia, aspartate transaminase to alanine aminotransferase ratio(AST/ALT)and fasting blood glucose(FBG)were involved in the model and fatty liver index(FLI)could be constructed. The discriminatory power of the model was tested with the area under the receiver-operating characteristic curve(AUC),(0.859, 95%CI, 0.851 2-0.867 in the internal group, and 0.853, 95%CI, 0.835-0.869 in the external group). When FLI≤1.25 was used to exclude the possibility of NAFLD, the negative predictive value was 93.1% and 93.3% respectively for the internal and external group. When FLI≥2.25 was used to detect NAFLD, the positive predictive value was 74.6% and 72.7% respectively. Conclusion FLI is an effective screening tool for the identification of high-risk subjects of NAFLD.

Key words: Nonalcoholic fatty liver disease, Health examination clients, Fatty liver index, Screening