Aim — a comprehensive assessment of social, obstetric, gynecological and somatic factors with the development and validation of a model for predicting premature birth in high-risk pregnant women. Material and methods. A cohort study with a mixed cohort was performed. Retrospectively, a comprehensive assessment of social factors, bad habits, somatic diseases, obstetric and gynecological history, course and outcomes of pregnancy, the condition of newborns in 1246 women (106 patients with perinatal losses in the 28—34 weeks of pregnancy, 1039 — with preterm delivery and live birth, 101 — with urgent delivery and live birth) with the subsequent construction of a prognostic model using regression analysis with optimal scaling and prospective validation among 100 women. Results. Based on a retrospective analysis, a prognostic model has been developed that includes a score for the following factors: a history of preterm birth, irregular monitoring and non-compliance with recommendations during pregnancy, a history of pelvic inflammatory diseases, smoking, obesity, the onset of sexual activity under 16 years of age, cardiovascular and endocrine diseases, intellectual labor. With a total score of risk factors of 25 or more, the model had a sensitivity of 73%, specificity of 71%, the area under the ROC curve (AUC) — 0.76; p<0.001. In order to validate the model, the course and outcomes of pregnancy in 100 women were prospectively analyzed: 50 patients in whom the pregnancy ended in premature birth (4th group), and 50 patients in whom the pregnancy ended in timely birth (5th group). When assessing the risk of preterm birth, a total score of ≥25 and higher was observed in 22 (44%) of 50 women in the 4th group and only 5 (10%) of 50 women in the 5th group. This model had sufficient diagnostic significance (sensitivity 81.4%, specificity 61.6%, predictive value of a positive result 44%, predictive value of a negative result 90%, positive likelihood ratio of 2.2 (1.5—3.0), negative likelihood ratio of 0.3 (0.13—0.68)). Conclusion. A model for predicting preterm delivery and adverse perinatal outcomes using commonly available characteristics of pregnant women is proposed and validated.