Genetic detection for the diagnosis of maturity-onset diabetes of the young (MODY) in China has low sensitivity and specificity. Better gene detection is urgently needed to distinguish testing subjects. We proposed to use numerous and weighted clinical traits as key indicators for reasonable genetic testing to predict the probability of MODY in the Chinese population. We created a prediction model based on data from 306 patients, including 140 MODY patients, 84 type 1 diabetes (T1D) patients and 82 type 2 diabetes (T2D) patients. This model was evaluated by receiver operating characteristic curves. Compared to T1D patients, MODY patients had higher C-peptide levels and negative antibodies, and most MODY patients had a family history of diabetes. Different from T2D, MODY was characterized by lower BMI and younger diagnostic age. A clinical prediction model was established to define the comprehensive probability of MODY by a weighted consolidation of the most distinguishing features, and the model showed excellent discrimination (AUCs of 0.916 in MODY vs. T1D and 0.942 in MODY vs. T2D). Further, high-sensitivity C-reactive protein (hsCRP), HbA1c, Δ2h-glucose and triglyceride were used as indicators for GCK-MODY, while triglyceride, hsCRP and hepatocellular adenoma were used as indicators for HNF1A-MODY. We developed a practical prediction model that could predict the probability of MODY and provides information to identify GCK-MODY and HNF1A-MODY. These results provide an advanced and more reasonable process to identify the most appropriate patients for genetic testing. Copyright © 2021. Published by Elsevier Inc.