A mathematical model to predict the probability of a successful pregnancy

Abstract

Aim: To develop a scoring system for the prediction of a successful pregnancy. Methods: Data were collected prospectively from women diagnosed with pregnancy from January 1, 2015, to December 31, 2018. Pregnant days, hormone levels, and gestational sac diameters were recorded. Relationships among the pregnancy days, hormones, and gestational sac were analyzed by Spearman correlation analysis. A scoring system was established and stratified by the 5th, 50th, and 95th percentile of hormone levels and gestational sac diameters on different pregnancy days. Pregnancy outcomes were predicted by the scores using quadratic polynomial regression analyses. A portable desktop analyzer was developed and the performance was evaluated by receiver operating characteristic (ROC) curve. Results: In 273 successful pregnancy cases, the length of gestational days was significantly correlated to beta-human chorionic gonadotropin (β-hCG) (r = 0.74, p < 0.001) and E 2 (r = 0.79, p < 0.001) levels, and the size of the gestational sac (r = 0.88, p < 0.001). Meanwhile, the size of gestational sac was positively correlated with β-hCG (r = 0.93, p < 0.001) and E 2 (r = 0.55, p < 0.001). For 273 delivery and 103 miscarriage cases included in this study, our scoring-based prediction model rendered an area under the ROC curve (AUC) of 0.86 with the sensitivity of 78.31% and the specificity of 80.83%. Conclusions: We successfully developed a scoring-based analyzer to evaluate the viability of embryos at different gestation stages and to predict the probability of a successful delivery, which would provide a reference for clinicians in postpregnancy management.