Constructing a predictive model for the risk of periventricular leukomalacia in very low birth weight premature infants based on Lasso-Logistic regression analysis

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objective To analyze the predictors of periventricular leukomalacia (PVL) in very low birth weight premature infants based on Lasso-Logistic regression and construct a nomogram prediction model. Methods From March 2022 to May 2024, 248 very low birth weight premature infants born and hospitalized in our hospital were marked as the modeling group. Another 106 very low birth weight premature infants hospitalized from June 2024 to March 2025 were selected as the verification group. Lasso Logistic regression was used to screen for the predictors of PVL in very low birth weight premature infants. The nomogram risk prediction model was constructed. Hosmer-Lemeshow test and ROC analysis were used to evaluate the predictive value of the model. The decision curve analysis (DCA) was used to evaluate the clinical application value of the model. Results The total incidence of PVL in 354 premature infants was 23.73%. Lasso and logistic regression results showed that Apgar score at 1 min after birth (OR = 0.645), C-reactive protein (CRP) (OR = 5.687), neonatal respiratory distress syndrome (NRDS) (OR = 5.412), sepsis (OR = 6.752), patent ductus arteriosus (OR = 9.428), and chorioamnionitis (OR = 8.342) were the predictors of PVL in very low birth weight premature infants ( P < 0.05). ROC analysis and Hosmer-Lemeshow test suggested that the model had high predictive consistency and discrimination, while the DCA curve suggested that this model had high clinical application value. Conclusion The risk of PVL in very low birth weight premature infants is closely related to Apgar score at 1 min after birth, CRP, NRDS, sepsis, patent ductus arteriosus, and chorioamnionitis. The model established based on these six factors has good predictive performance.