Abstract
This study investigates risk factors for the development of preeclampsia (PE) in women with twin pregnancies and constructs and validates a column-line diagram prediction model for clinical decision-making. Records of 70 women with PE and 70 women without PE were selected from twin pregnancies who underwent labor and delivery at Huzhou Maternity and Child Health Care Hospital between September 2021 and June 2023. The cohort was then divided into a training set (98 cases) and a validation set (42 cases) in the ratio of 7:3 using a simple random sampling method. Clinical risk factors, blood biochemical indexes, and uterine artery pulsatility index of all pregnant women were collected to assess the risk of PE. The results were presented as odds ratios (OR) with 90% confidence intervals (CI). Least absolute shrinkage and selection operator regression analysis was used to screen the predictors and establish an optimized, multifactorial logistic regression-based columnar graph model. Distinction, calibration, and clinical utility of the columnar plot model were evaluated by using the receiver operator characteristic curve, calibration plot, and decision curve analysis. Age (OR = 13.39, 95% CI = 2.152-157.0, P = .014), prepregnancy body mass index (OR = 5.979, 95% CI = 1.365-34.27, P = .027), mode of conception (OR = 3.498, 95% CI = 1.071-12.79, P = .045), serum homocysteine cysteine level (OR = 2.079, 95% CI = 1.193-4.005, P = .016), serum β-human chorionic gonadotropin level (Log10; OR = 9.984, 95% CI = 1.467-82.77, P = .024), uterine artery pulsatility index (per0.1; OR = 1.347, 95% CI = 1.11-1.7, P = .005) were independent risk factors for PE (P < .05), and the column-line graph prediction model based on the above 6 risk factors had a good discriminatory degree (area under curve value: 0.880, 95% CI = 0.817-0.944 for training set validation, and 0.831, 95% CI = 0.704-0.958 for validation set validation). The calibration curve showed good agreement between the predicted and actual probabilities of the model (P > .05), and the decision curve analysis showed that the model had a high net clinical benefit (threshold probability values: >2.5% for the training set, 18% to 75% for the validation set). The column-line diagram model developed in this study can more accurately predict the risk of developing PE in women with twin pregnancies.