A prediction model for moderate to severe periventricular-intraventricular hemorrhage in very/extremely preterm infants
10.3760/cma.j.issn.2096-2932.2023.12.003
- VernacularTitle:极/超早产儿中重度脑室周围-脑室内出血预测模型的研究
- Author:
Xi LIU
1
;
Dan QIAO
;
Fei BEI
Author Information
1. 国家儿童医学中心，上海交通大学医学院附属上海儿童医学中心新生儿科，上海 200127
- Keywords:
Periventricular-intraventricular hemorrhage;
Preterm infant;
Prediction model
- From:Chinese Journal of Neonatology
2023;38(12):715-720
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To develop and validate an early prediction model for moderate and severe periventricular-intraventricular hemorrhage (M/S PIVH) in very/extremely preterm infants (V/EPIs).Methods:From January 1, 2017 to December 31, 2021, preterm infants with gestational age (GA) <32 w admitted to the Neonatal Intensive Care Unit of our hospital within 24 h after birth were enrolled. The infants were assigned into no-or-mild (N/M) PIVH group and M/S PIVH group according to postnatal cranial ultrasound within 2 w after birth. Clinical data including pregnancy and perinatal history, complete blood counts (CBC) and blood gas analysis (BGA) within 24 h after birth were retrospectively collected. The univariate analysis, stepwise regression analysis and multivariate logistic regression analysis were used to determine possible predictive risk factors and to establish an early prediction model. The receiver operating characteristic (ROC) curve was used to evaluate the efficacy of the model. The Hosmer-Lemesshow test was introduced to make calibrations of the model. Bootstrap method was used for internal validation.Results:Among 512 preterm infants, 460 (89.8%) were in N/M PIVH group and 52 (10.2%) in M/S PIVH group. Cesarean section ( OR=0.323, 95% CI 0.155-0.669, P<0.001), GA ( OR=0.789, 95% CI 0.633-0.979, P<0.001), use of vasoactive drugs ( OR=2.487,95% CI 1.152-5.184, P=0.008), mean corpuscular hemoglobin concentration (MCHC) ( OR=0.956,95% CI 0.930-0.981, P<0.001) and maximum lactate level ( OR=1.246, 95% CI 1.075-1.440, P=0.004) were independent risk factors of M/S PIVH in the early stage (sensitivity=73.1%, specificity=81.2%, AUC=0.818). The Hosmer-Lemesshow test showed that the model correlated well with the actual incidence of M/S PIVH ( χ2=2.394, P=0.302). The Bootstrap internal validation showed that the model had a realistic estimate of its performances (AUC=0.801). Conclusions:An early prediction model for M/S PIVH can be established based on pregnancy/perinatal history and clinical data within 24 h after birth. The model is helpful for prognosis evaluation and clinical decision-making.
