Single-center nomogram internally predicts sepsis risk in neonatal pneumonia

A six-variable clinical nomogram developed from retrospective records predicted sepsis risk among neonates hospitalized with pneumonia, with an area under the receiver operating characteristic curve of 0.902 in the training cohort and 0.861 in the internal test cohort. The model incorporates sex, gestational age, procalcitonin, hematocrit, white blood cell count and C-reactive protein. The findings support further evaluation of the nomogram for risk stratification at admission, but its use beyond the studied hospital remains unestablished because the research included external validation as a requirement for assessing generalizability and clinical utility.

The single-center retrospective study reviewed clinical records of 817 neonates admitted to the Department of Neonatology at Jiangsu Province (Suqian) Hospital between January 2021 and March 2026. Investigators collected demographic characteristics, age at admission, gestational age, birth weight, Apgar scores and routine laboratory findings, then evaluated the occurrence of sepsis as the outcome. Sepsis developed in 109 of the 817 neonates, corresponding to an incidence of 13.3%. The records were divided in an approximate 7:3 ratio into a training cohort of 571 neonates and a test cohort of 246. Sepsis occurred in 13.3% of the training cohort and 13.4% of the test cohort, with no significant difference between them (P = 0.968).

Candidate predictors were first screened with least absolute shrinkage and selection operator regression, which identified eight preliminary variables: sex, gestational age, procalcitonin, albumin, urea, hematocrit, white blood cell count and C-reactive protein. The researchers then constructed the nomogram with multivariable logistic regression. Six variables were retained in the final model: sex, gestational age, procalcitonin, hematocrit, white blood cell count and C-reactive protein. Within this retrospective cohort, male sex, lower gestational age, higher procalcitonin, lower hematocrit, increased white blood cell count and higher C-reactive protein were each associated with a higher probability of sepsis, with all P values below 0.05. These associations describe the fitted prediction model; the study did not test whether changing any individual factor would prevent sepsis.

The internal test results were lower than the training results but remained in the range reported by the researchers. The training cohort achieved an AUC of 0.902 (95% confidence interval, 0.862–0.943), while the test cohort achieved an AUC of 0.861 (95% confidence interval, 0.768–0.954). Calibration plots showed acceptable agreement between predicted and observed risks. Decision curve analysis demonstrated positive net benefit across a defined range of threshold probabilities, indicating that the model retained clinical net benefit within the evaluated range. Because both cohorts came from the same center, however, this was an internal split rather than validation in a separate hospital or population, and the test cohort’s wider confidence interval indicates greater uncertainty in the smaller validation sample.

The nomogram is intended to convert six admission-time clinical and laboratory variables into an individualized estimate of sepsis risk for neonates with pneumonia. Its favorable discrimination, calibration and decision-curve results apply to the single-center retrospective population studied and therefore do not establish the model’s performance in other hospitals or clinical settings. The authors characterize the findings as hypothesis-generating and call for external validation to assess generalizability and clinical utility. Accordingly, the study provides an internally validated prediction model rather than evidence that nomogram-guided decisions improve clinical outcomes, and no clinical recommendation can be drawn beyond the need for further validation.

Subject of Research: Internal validation of a nomogram for predicting sepsis in neonates with pneumonia

Article Title: Development and internal validation of a nomogram for predicting sepsis risk in neonates with pneumonia

Article References: Zhang, Q., Zhou, J., Dong, N., Wang, F., & Zhou, L. (2026). Development and internal validation of a nomogram for predicting sepsis risk in neonates with pneumonia. BMC Pediatrics. https://doi.org/10.1186/s12887-026-07595-8

Image Credits: AI Generated

DOI: 10.1186/s12887-026-07595-8

Keywords: neonatal sepsis, neonatal pneumonia, clinical prediction model, nomogram, internal validation, procalcitonin, C-reactive protein, hematocrit, retrospective cohort, risk stratification, Development, internal