Модель прогнозу перебігу та результату вагітності

Abstract

The significance of predicting preterm births is determined by their consistently high frequency, high perinatal mortality of premature babies, significant labor and economic costs associated with their care, and the risk of child disability. Based on the analysis of modern literature, we established that most of the predictive methods used in modern obstetrics require additional examinations, financial costs and a certain amount of time, which reduces their effectiveness in cases of the threat of preterm birth. The purpose of the work was to develop and test prediction models of the outcome of pregnancy according to the indicators of laboratory and instrumental studies. Material and methods. The most informative indicators of the health state of 83 pregnant women (41 of them was with a favorable pregnancy outcome and 42 women were with preterm birth and / or antenatal fetal death) were selected to develop prediction models of the outcome of pregnancy (urgent or preterm birth). The models were tested using indicators of 17 pregnant women (11 with urgent birth and 6 with preterm birth). In all women, laboratory and instrumental indicators were determined in accordance with the existing standard of management of pregnancy. Fuzzy logic was used to develop prediction models of the outcome of pregnancy. Results and discussion. Using statistical criteria showed that the outcome of pregnancy in women without signs of threat of preterm birth was associated with the values of clinical (erythrocyte sedimentation rate) and biochemical (content of aspartate aminotransferase and bilirubin) blood tests, biochemical screening (content of estriol, plasma protein-A, associated with pregnancy), indicators of ultrasound (blood flow velocity in the middle cerebral artery and right uterine artery). In women with signs of threat of preterm birth, informative indicators were the content of estriol and bilirubin in the blood, as well as the value of erythrocyte sedimentation rate. Model-A, built on the basis of indicators of biochemical screening, had an overall accuracy of 84%, 86% of sensitivity, and 82% of specificity, which indicated a fairly high probability of identifying pregnant women, who were predicted both urgent and preterm birth from the entire sample. Model C, built on the basis of indicators of Doppler measurements, showed an overall accuracy of 96.4%, 100% of sensitivity and 94% of specificity, which allowed using it to predict the outcome of pregnancy. Model-AC, which uses the results of models A and C, allows classifying women by pregnancy without errors, i.e. the accuracy, sensitivity and specificity of such a combined model was 100%. Model-D for predicting preterm birth in pregnant women with signs of threat of abortion and/or preterm birth was synthesized according to the indicators of clinical and biochemical blood tests, taking into account the concentration of estriol. Testing of the model-D showed an overall accuracy of 96.3%, 100% of sensitivity, and 86% of specificity, which indicates a sufficiently high probability of identifying pregnant women, who were predicted both urgent and preterm birth. Conclusion. Using the developed prediction models will allow diagnosing and predicting preterm birth in each pregnant woman individually at an early (preclinical) stage, which contributes to the timely implementation of precaution measures to prevent the development of preterm birth