In this study, where every late-onset infection evaluation was included over an 11-year period, the nSOFA demonstrated good to very good discrimination for the outcome of mortality associated with a given evaluation in all NICU patients and very good to excellent discrimination in VLBWs. Early discrimination for evaluation-level mortality risk remained good even when analyses were restricted to evaluations in patients who ultimately died. The nSOFA identified life-threatening organ dysfunction at the evaluation level, which adds to the existing evidence [12, 14, 15] that supports the use of the nSOFA as a starting point for a consensus definition of neonatal sepsis.
Consensus sepsis definitions in children and adults require life-threatening organ dysfunction. To objectively quantify life-threatening organ dysfunction, the sequential organ failure assessment (SOFA) score is used for adults, while the Phoenix Sepsis Score is used for pediatric patients [4, 5]. The SOFA score includes assessments for organ dysfunction in six systems (respiratory, cardiovascular, hematologic, hepatic, renal, and neurologic) [16]. The Phoenix sepsis score includes assessments for organ dysfunction in four systems (respiratory, cardiovascular, hematologic, and neurologic) [17]. Like the Phoenix Score, the nSOFA includes assessments for organ dysfunction in three systems (respiratory, cardiovascular, and hematologic) and has been tested and validated for the NICU population, including term and preterm infants, in multiple cohorts. Multiple multicenter studies have demonstrated the utility of the nSOFA to accurately discriminate for mortality in NICU patients with confirmed early-onset [11] and late-onset bacteremia [12], necrotizing enterocolitis [13], and among the NICU population for all-cause mortality [10]. The current study showed the nSOFA had strong utility for mortality among all NICU patients (no exclusions) and included all evaluations for late-onset infection. Taken together, the nSOFA, like the SOFA in adults and the Phoenix score in children, has demonstrated utility in quantifying life-threatening organ dysfunction among those with suspected infection in the NICU population and could serve as a starting point for a consensus definition of neonatal sepsis.
These data suggest the nSOFA is likely to improve early patient classification and facilitate prognostic enrichment of NICU patients enrolled in future interventional clinical studies focused on the outcome of sepsis mortality. In this study, most neonates (65%) evaluated for LOIs had an nSOFA score of ≤1 at evaluation and at 6 h after, with a very low mortality rate (≤1%). Detailed patient classification is required for precision medicine approaches in any population [18]. A means to reduce patient exposure to potential deleterious side effects of an intervention when there is a very low risk of mortality would be an important contribution to neonatology clinical studies. Using the nSOFA could decrease group heterogeneity, facilitate equitable study arm enrollment, and reduce the number of neonates needed to achieve adequate power in clinical trials through better patient classification. Beyond the outcome of mortality, the nSOFA has demonstrated utility in identifying the risk of BPD, ROP, and NDI among neonates with sepsis [19, 20].
This study has limitations. This is a single-center study, and the practices unique to the center could influence the generalizability of the results. To mitigate the potential impact of bias, we included every LOI evaluation from over 11 years that occurred in every patient cared for in a level IV academic referral NICU, regardless of the duration of antibiotic therapy, presence of congenital anomalies, and the infant’s code status. We could not confirm that death was the direct result of LOI. However, we defined LOI mortality as death occurring while receiving antibiotics prescribed at the clinician’s discretion for the LOI evaluation. We could not determine the suspected infection source that prompted the LOI evaluation and thus could not perform primary site of suspected LOI-specific analyses.
