A Nomogram Model for Predicting Early Hyperglycemia in Premature Infants

Abstract

Background: Hyperglycemia in preterm infants is likely to lead to severe complications and higher mortality. Timely identification of hyperglycemia in preterm infants is vital for the prognosis of patients. We developed and validated predictive models for hyperglycemia in preterm infants < 32 weeks of gestational age to aid in the early detection of these patients.

Methods: A retrospective analysis was performed on 460 premature infants to examine the association of various clinical variables with hyperglycemia. We collected data from June 1, 2021, to May 31, 2023. clinical and demographic parameters were analyzed using univariable and multivariable logistic regression analysis (backward method). We constructed a nomogram to assess the risk of hyperglycemia. The model's accuracy was validated using bootstrap resampling (n=500), and the POC curve was used for discrimination analysis to calibrate function and value. Calibration was evaluated via a calibration curve. The model's clinical utility was evaluated through decision curve analysis

Results: Of the 29 potential predictors analyzed in 460 premature infants, the incidence of hyperglycemia was 24.1%. Multivariable logistic regression analysis identified birth weight, invasive ventilation, and Intraventricular hemorrhage as independent risk factors for premature infants with hyperglycemia. The resulting nomogram accurately predicted hyperglycemia risk with an area under the curve of 0.735(95%CI: 0.685-0.786). The bootstrap-validated area under the curve remained at 0.735(95%CI: 0.687-0.785). This model exhibited excellent calibration and demonstrated greater predictive efficacy and clinical utility for hyperglycemia.

Conclusion: We have developed a prediction nomogram of hyperglycemia that can assist clinical treatment decision-making.