Prediction of neonatal sepsis in Hakka population: PCT time interval analysis and nomogram model incorporating perinatal factors and disease characteristics.

OBJECTIVE

This study aimed to develop a predictive model for neonatal sepsis by analyzing procalcitonin (PCT) levels measured across different age-specific time intervals in the Hakka population, while integrating maternal perinatal factors and neonatal disease characteristics.

METHODS

A retrospective analysis was conducted on 13,884 neonates, encompassing both sepsis and non-sepsis cases, to systematically evaluate the predictive capability of PCT across different time intervals, focusing on the first week of life and beyond (up to [7, 8) days). Logistic regression analysis, incorporating categorized PCT levels, maternal perinatal factors, and neonatal disease characteristics, was utilized to develop a nomogram model for predicting sepsis.

RESULTS

Elevated PCT levels were significantly associated with an increased risk of neonatal sepsis (OR = 13.59, 95% CI: 9.01-20.50, p < 0.001) during the 2-3 day period. During this period, the area under the curve (AUC) for PCT was 0.853 (95% CI: 0.840-0.866), indicating strong predictive performance. The developed nomogram model achieved an AUC of 0.82 and an accuracy of 0.88 in the training set, effectively distinguishing between high- and low-risk sepsis cases.

CONCLUSION

This study confirms the critical role of PCT in the early diagnosis of neonatal sepsis and proposes a time-sensitive predictive model specifically tailored to the Hakka population, offering clinicians a personalized risk assessment tool. Nonetheless, further external validation is required to fully establish the clinical applicability of this model.