A nomogram developed using machine learning methods and with an assessment of clinical value for predicting chronic kidney disease in children.

Background. Prevention of chronic kidney disease (CKD) is based not only on the determination of etiologic and pathogenetic features of the disease in the patient, but also on the use of information tools that help the doctor to identify the disease in the child and timely refer to a nephrologist. Aim of the study: to develop a graphical tool to predict the risk of developing chronic kidney disease in children. Materials and methods of the study: data were obtained from a single-center ospective cohort study (2011–2022) of children with CKD stages 1–4 aged 0 to 18 years. In the group of children with CKD (n=128) girls 88 (68.8%), boys 40 (31.2%). In the group of children without CKD (n=30), girls 18 (60.0%), boys 12 (40.0%). The median age was 6 (5.0–8.0) years. Kidney disease in patients lasted 2 (1.0–3.0) years. Children in the identified groups did not differ statistically significantly by sex and age. Results: Important precursors of chronic kidney disease in children were: urinary protein loss (0.24 [0.07; 0.42], vs. 0.0 [0.0; 0.0], n=30, p<0.001; hematuria (2.0 [0.0; 4. 0], vs. 0.0 [0.0; 0.0], p<0.001); asthenic physique (76.6% vs. 10.0% p<0.001; OR29.4 [8.33; 103.7]); aggravated heredity for nephrologic or metabolic pathology in the first and second generations (77.3% vs. 0%, p<0.001; OR205.75 [12.21; 3467.52]; anemia during pregnancy (70.3% vs. 10.0%, p<0.001; OR21.32 [6.1; 74.54]). A model of excellent quality (ROCAUC> 90%) was built and a nomogram was developed based on it. Conclusion. The nomogram is intended for prediction or screening of CKD in children and for use in primary health care.