Development and validation of a nomogram for predicting multiple-organ dysfunction syndrome in low-birth-weight infants with early-onset sepsis

Abstract Background Early-onset neonatal sepsis is among the most common causes of multiple organ dysfunction syndrome (MODS) in newborns. The incidence of neonatal MODS is 14.6%, and the mortality rate is 75.4%. Low-birth-weight infants with early-onset sepsis face an even higher risk of death when complicated by MODS. Therefore, developing a predictive model to assess the likelihood of MODS in low-birth-weight infants with early-onset sepsis (EOS) has become a key focus in clinical research. Methods Three hundred and seventy low-birth-weight infants diagnosed with EOS from January 2019 to January 2024 were included in this study. The clinical features and pregnancy details of newborns in the MODS group and the non-MODS group were compared. Univariable and multivariable analyses were sequentially performed to identify independent risk factors. A nomogram was developed using multivariable binary logistic regression analysis. The validity of the model was evaluated through discrimination and calibration performance. The model was internally validated using the bootstrapping technique. Results A total of 370 newborns were included in the analysis. Among the 370 newborns, 74 (20%) were in the MODS group. The study revealed that gestational hypothyroidism ( OR = 4.009, P = 0.048, 95% CI 1.010–15.910), glycaemic lability ( OR = 7.707, P < 0.001, 95% CI 3.019–19.671), high-sensitivity troponin T (hs-TnT) ( OR = 1.014, P < 0.001, 95% CI 1.009–1.019), blood lactate (Lac) levels ( OR = 1.375, P < 0.001, 95% CI 1.196–1.581), and a platelet count (PLT) < 100 × 10 9 /L ( OR = 3.263, P = 0.005, 95% CI 1.416–7.517) were independent risk factors for MODS among low-birth-weight infants with EOS. The model had a discriminatory power of 93.7% (95% CI 0.903–0.971). The model retained excellent internal discrimination ability. The sensitivity, specificity, positive predictive value, and negative predictive value of the model were 86.5%, 91.2%, 71.1%, and 96.4%, respectively. Conclusion In this study, a risk scoring model was developed using gestational hypothyroidism, glycaemic lability, hs-TnT levels, blood lactate levels, and platelet counts < 100 × 10 9 /L to predict MODS in low-birth-weight infants with EOS, the results of which were sufficiently predictive and calibrated.