Mathematical modeling of the risk of antenatal fetal death based on retrospective analysis

Current issue in obstetrics is the early prediction of perinatal pathology arises by determining the degree of risk, especially when it comes to reproductive losses.The objective: to develop a mathematical model for predicting antenatal fetal death (AFD) based on retrospective analysis.Materials and methods. Logistic regression model was constructed by analyzing medical cards for pregnancy monitoring of 46 women with AFD and 50 women with live births who were 18–40 years old and had singleton pregnancy. Variables that demonstrated statistical significance in the univariate logistic model were included in a multivariate stepwise model.Results. Analysis of extragenital pathology allowed us to isolate a number of clinical factors associated with an increased risk of AFD. Thyroid diseases tended to increase the risk of AFD by 2.25 times; cardiovascular pathology – by 2.37 times. The predictive value of sexually transmitted infections was 2.38. According to the results of LASSO regression (Least Absolute Shrin­kage and Selection Operator), the most informative predictors were identified: age > 35 years, body mass index ≥ 30 kg/m2, thyroid pathology, vaginal microbiocenosis disorders, spontaneous abortions in history, placental dysfunction, D-dimer.Conclusions. The identification of the main prognostic markers and the mathematical analysis of the prediction of perinatal pathology in AFD allowed us to assess the predictive power of some anamnestic indicators, which contributed to the creation of a seven-factor prognostic model of AFD.