Year: 2019
Type: conference-abstract
Abstract: To detect the associated factors of FHD based on a big data using Bayesian network, and quantitatively analyse the risk ratio of combined multiple factors. Pregnant women who underwent fetal echocardiography in our centre (N = 16,000) were continuously recruited from July 2013 to July 2018. Twenty-five maternal and fetal factors were obtained. Learning process of Bayesian network was composed of two parts: structure learning and parameter estimation. Several methods were used to learn the structure of Bayesian networks, including K2+T (K2 with MWST initialisation), K2-T (K2 with MWST inverse initialisation) GS+T (GS starting from a MWST-initialised structure) and GES (greedy search in the space of equivalent classes). Best Bayesian network model was chosen according to the greater area under the curve (AUC) after k-Fold Cross-Validation on data. Based on the cause inference of the chosen model, risk ratio (RR) was calculated ranging from single factor to multiple factors compared to the normal group. Among them, 3,312 cases of fetuses had FHD. The single factor analysis showed that RR of the number of births, abortion, or with a spouse who smoked was 1.50, 1.38, or 1.11 respectively. When the number of births was combined with abortion or with a spouse who smoked, we found the RR for FHD was greater (RR = 1.96 or 1.63). By that analogy, when we took the five factors into account, including the number of births, upper respiratory tract infection during early pregnancy, anemia, and mental stress for pregnant women, as well as the history of abortion or with a spouse who smoked, the risk was higher than that of the combined factors less than 5 (RR = 2.62 or 2.28). Other than the factors mentioned above, other factors do not continue to contribute to the risk of FHD. Based on the Bayesian network, we find that the six directly associated factors with FHD. The more factors combined above, the higher risk for FHD. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.
DOI: 10.1002/uog.21305
Institutions: Beijing Anzhen Hospital, Beihang University
Language: English
Subfield: Epidemiology
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W2979034756