Septic shock mortality
Description
Objective We lack data to timely identify the severity of late-onset sepsis (LOS) in its early stages, contributing to treatment delay and poor prognosis. Our objective was to develop a quantitative model for estimating the probability of death and/or brain injury in LOS in preterm infants, based on objective and measurable data available at onset (i.e. time of blood collection for culture). Methods The base population consisted of neonates at ≤ 36 weeks’ gestation admitted to the Neonatal Intensive Care Unit of Modena (2010–2022) with a positive (blood or cerebrospinal fluid) culture for pathogens, after 72 hours from birth (LOS). We defined poor outcomes as sepsis-related death or brain injury first detected at neuroimaging after the sepsis. Results Among 3217 preterm live births, 94 cases were included (median gestational age 26.5 weeks’), of whom 26 (27.7%) had poor outcomes (17 fatal cases, 18.1%; 9 brain injuries, 9.6%). At onset, infants with poor outcomes showed lower postnatal age (11.5 vs 12.5 days, p<0.001), lower mean blood pressure (30.5 vs 43 mmHg, p < 0.001) and higher lactate levels (4.4 vs 1.5 mmol/l, p < 0.001). Our multivariable model showed good discrimination and calibration (c statistic=0.8618 and Hosmer-Lemeshow p=0.8532). We defined a risk stratification scheme to divide the population into 3 groups: low-risk (sensitivity 97%, specificity 52%), middle-, and high-risk (sensitivity 77%, specificity 80%). Conclusions This predictive model performs well as practical and easy-to-use tool to aid clinicians in tailoring neonatal management based on individual risks.