Significance. Improving the provision of medical care to women during pregnancy, childbirth, and the postpartum period is a priority for the development of health care. The aim of the study was to develop mathematical models to predict adverse outcomes of pregnancy and childbirth, and compare their results with other scales assessing the risk of adverse outcomes. Methods. This paper analyzes the data of the Krasnoyarsk regional perinatal monitoring about 250 122 completed cases of pregnancy in the period from 2014 to 2017. To build the prediction models used the following mathematical model: logistic regression, multilayer artificial neural network, classification tree and discriminant equation. Data on the presence of risk factors in pregnant women were used as input parameters of mathematical models, and analyzed outcomes of pregnancy and childbirth were used as output parameters. Results. The highest values of prediction accuracy of all studied outcomes were obtained using classification trees. Prediction of almost all outcomes, with the exception of the prediction of late births, resulted in a prediction accuracy of more than 90%. When using the classification tree to predict the delayed deliveries were obtained precision equal 77,2 [77,1; 77,4]%, sensitivity: 92,0 [91,9; 92,1]%, specificity is 62.5 [62,4; of 62.6]%. Discussion. Among the two techniques (V. E. Radzinsky and order No. 572) fewer cases of underestimation of perinatal risk allows to use the evaluation criteria perinatal risk, established by order No. 572. However, when these two techniques are used together, fewer cases of underestimation of perinatal risk can be achieved. This indicates the feasibility of including the technique of V. E. Radzinsky in remote perinatal monitoring in the Krasnoyarsk region. Application of the obtained classification trees in combination with the V. E. method. Radzinsky and the order No. 572 allows to reduce to a greater extent the number of cases of underestimation of perinatal risks. Conclusions. In the course of the study, mathematical models were built that allow predicting the occurrence of various adverse outcomes of pregnancy and childbirth on the basis of data on the presence of various risk factors in pregnant women. The highest accuracy rates were obtained with the use of classification trees, which indicates their preferred use for the task of assessing perinatal risk in comparison with other models.