Background <p>Distinguishing diabetes diagnosis is fundamental to ensuring proper management of individuals with diabetes, but has been challenging, especially in newly diagnosed diabetes onset with ketosis or ketoacidosis in pediatrics.</p> Methods <p>A retrospective analysis was conducted on medical records from 2017/1/1 to 2020/4/30 in pediatrics with new-onset diabetes accompanied with ketosis or ketoacidosis. Data was collected at diabetes onset and two years after discharge. Persons with diabetes were classified as type 1 or 2 diabetes (T1DM; T2DM) based on the person’s medication and final diagnosis. The best diagnostic cut-off point was determined using receiver operating characteristic curves (ROCs) between T1DM and T2DM.</p> Results <p>Among 153 children and adolescent with diabetes, 78 (51.0%) were diagnosed as T1DM and 75 (49.0%) were diagnosed as T2DM after two years of follow-up. There were significant differences in sex, age, family history, BMI, systolic and diastolic blood pressure, lipids, uric acid (UA), C-peptide, combined fatty liver ratio and any islet auto-antibody-positive ratio at the time of onset (<i>P</i> &lt; 0.05). Key discriminators identified by ROC analysis included fatty liver, SBP, BMI, and C-peptide levels (fasting, 1-h, and 2-h), with AUCs of 0.79, 0.83, 0.92, 0.94, 0.96, and 0.95 and optimal cut-offs value of 110.5 mmHg, 20.95&#xa0;kg/m², 0.47 nmol/L (fasting), 0.98 nmol/L (1-h), and 2.03 nmol/L (2-h), respectively.</p> Conclusions <p>Overall, the most sensitive and specific clinical and biochemical criteria for the diagnostic classification of newly diagnosed diabetes onset with ketosis or ketoacidosis in pediatrics should consider C-peptide, BMI, SBP and fatty liver at the time of onset, which have effective diagnostic values.</p> Clinical trial number <p>Not applicable.</p>