Evaluation of the Neonatal Sequential Organ Failure Assessment and Mortality Risk in Preterm Infants With Late-Onset Infection

Key Points Question How useful is the neonatal Sequential Organ Failure Assessment for identification of preterm infants at high risk for late-onset, infection-associated mortality? Findings In this multicenter cohort study of 653 preterm infants with late-onset infection, the neonatal Sequential Organ Failure Assessment score was associated with infection-attributable mortality. Analyses stratified by sex or Gram stain of pathogen class or restricted to less than 25 weeks’ completed gestation did not reduce the association of the neonatal Sequential Organ Failure Assessment score with infection-related mortality. Meaning In a large, multicenter cohort, the single-center–validated neonatal Sequential Organ Failure Assessment score was associated with mortality risk with late-onset infection in preterm infants, implying generalizability.


Introduction
Although overall survival rates among preterm infants have improved, sepsis remains a significant factor associated with long-term morbidity and mortality in this population. [1][2][3][4] Late-onset sepsis (sepsis that occurs after 72 hours of life during the birth hospitalization) affects 20% to 30% of the most preterm infants and has a 15% mortality rate. [5][6][7][8][9][10] Despite multiple efforts, the methods of diagnosis, clinical management, and outcomes for preterm infants with late-onset sepsis have been largely unchanged over several decades. These findings are in part due to a lack of an accepted consensus definition for sepsis and a paucity of validated metrics reliably associated with sepsisrelated mortality in this population. 11 Consensus definitions are needed (1) for the development of standard research protocols, (2) to determine sample size, (3) to prevent misclassification of patients, and (4) to facilitate replication and improvement in future studies. Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection. 12 Life-threatening organ dysfunction is substantiated by a Sequential Organ Failure Assessment (SOFA) score that categorically defines the extent of specific organ dysfunction that increases the risk of death or intensive care unit (ICU) admission. 12,13 In contrast to adult and pediatric ICU environments, the distinction of infection from sepsis and the incorporation of organ dysfunction are not widely recognized in the neonatal ICU (NICU). Definitions of sepsis used in neonatology are mostly predicated on the isolation of bacterial pathogens from blood and/or the associated length of antimicrobial treatment. 5,7 A failure to incorporate the presence or severity of organ dysfunction in preterm infants with late-onset sepsis prevents grouping of patients according to severity and mortality risk, which is necessary for advancements in research and improved outcomes.
The neonatal SOFA (nSOFA) was developed to address the need for a consensus definition for sepsis applicable to neonates. 14 Similar to the adult SOFA and pediatric SOFA, the nSOFA serves as an operational definition of organ dysfunction that can identify those with a high risk of mortality among preterm infants with infection. Because there is no consensus definition for neonatal sepsis, 11 the development of the nSOFA was informed by the progression of organ failure among preterm infants with lethal late-onset bacteremia and then validated in a single-center, retrospective cohort of preterm (<33 weeks) infants with late-onset bacteremia, fungemia, or intestinal perforation (included as an irrefutable source of infection). 14,15 This validation study reported an association of higher nSOFA scores with infection-related mortality. With the goal of establishing the nSOFA score as a criterion for a consensus definition of sepsis in this population, in this study we determined the generalizability of the nSOFA for neonatal infection-related mortality risk in a multicenter, retrospective cohort of preterm (<33 weeks), very low-birth-weight (VLBW) infants with late-onset bacteremia, fungemia, or intestinal perforation. board because the research was retrospective and data were deidentified; thus, the study involved no more than minimal risk to participants.

Late-Onset Infection Episode
Because there is no consensus definition for neonatal sepsis, 11 we considered a new bacteremic or fungemic event in an infant aged greater than 72 hours who received 5 or more days of antibiotic treatment or died before completing at least 5 days of treatment with intention to treat as a lateonset infection (LOI) episode. The onset of the LOI episode was defined as the time of clinician declaration of suspected infection, herein defined as the time of blood culture acquisition. Late-onset episodes of surgical peritonitis (perforated bowel associated with necrotizing enterocolitis or spontaneous intestinal perforation) with negative results of blood cultures were included as an irrefutable source of infection. Infants with positive blood cultures with organisms that are frequently considered contaminants, including Bacillus species and Corynebacterium species, were excluded.

Inclusion and Exclusion Criteria
All sites used a standardized set of criteria to independently identify eligible events for inclusion.
Inclusion criteria were new LOI (>72 hours of life) in VLBW, preterm (<33 weeks completed gestation) infants. Only the first episode of LOI was studied. An absence of antibiotic exposure for 48 hours before the LOI episode evaluation was required to facilitate characterization of an uninfected baseline in each patient. Each site abstracted and deidentified consecutive clinical and laboratory data from their electronic health records for all patients who met the inclusion criteria until a minimum of 5 patients who died with the LOI episode were identified. The frequency of LOIs as well as the availability of data at each center resulted in variation in the study period between centers, which overall ranged from 2010 to 2020. The primary outcome was episode mortality, defined as death that occurred during ongoing antibiotic therapy for an LOI episode. Birth weight, gestational age at birth, sex, age at event, pathogen identified, time from culture to death, and the suspected primary infection site were collected. Clinical data required to calculate the nSOFA score were collected and included the receipt of intubation and mechanical ventilation, the peripheral oxygen saturation, the fraction of inspired oxygen to achieve the peripheral oxygen saturation, the most current platelet count, and any requirement for glucocorticoid, inotropic, or vasoactive drugs.

Application of the nSOFA Score
The nSOFA components and scoring paradigms were modeled after the adult SOFA. 12,13 The nSOFA uses only objective and available clinical standard-of-care data to provide an operational definition of organ dysfunction that facilitates mortality risk stratification among VLBW infants with unequivocal LOI (bacteremia or intestinal perforation). 14 The nSOFA uses categorical scores (total score range from 0 [best] to 15 [worst]) to objectively describe dynamic changes in (1) receipt of mechanical ventilation and oxygen to maintain a physiologic peripheral saturation (score range, 0-8); (2) inotropic or vasoactive drug support, including the use of corticosteroids for presumed adrenal insufficiency or catecholamine-resistant shock (score range, 0-4); and (3) the presence and severity of thrombocytopenia based on the most recent platelet measure (score range, 0-3) ( Table 1; eAppendix in the Supplement).
The nSOFA score was validated in an independent, single-center cohort of infants with LOI. 14 In this multicenter, retrospective study to determine the generalizability of the nSOFA, each center identified qualifying infants and calculated the nSOFA score at 9 discrete time points, including and surrounding the LOI episode, using a web-based calculator and data from the electronic health records. The time the blood sample was drawn from the patient represented time 0 (T0). The time points that preceded (T minus 48 hours [T-48], T-24, T-12, and T-6) and followed (T plus 6 hours [T6], T12, T24, and T48) the infection evaluation were based on the progression of organ dysfunction in previous studies. 14   increased with nSOFA score category at all time points analyzed (eTable 3 in the Supplement).

Discussion
In adults with suspected infection, the definition of sepsis is based on the presence of lifethreatening organ dysfunction, reflected in the SOFA score, which is associated with ICU admission or mortality. 12 To reach a consensus definition of sepsis in the neonatal population, it is essential to establish an operational definition of life-threatening organ dysfunction applicable to the preterm population (nSOFA). Because preterm neonates are not often cared for outside of the NICU, the most severe outcome (mortality) among those with an infection must be used. The nSOFA, an adaptation of the SOFA specific to neonates, demonstrated generalizability in aggregate and in each center as a risk assessment tool that identified LOI-associated mortality risk among VLBW infants in this multicenter cohort study. Taken together, the nSOFA score may serve as a useful criterion on which to build a consensus definition of sepsis in this population.
A lack of developmental age-specific criteria for life-threatening organ dysfunction applicable to the preterm population prevents meaningful classification and study of infection as well as other clinical entities in this population. Infants at the lowest extremes of gestational age are at disproportionately higher risk for late-onset infection owing to physiologic immaturity as well as critical illness that requires prolonged NICU hospitalizations. 16,17 However, the nSOFA demonstrated consistent associations with LOI-associated mortality even in the lowest gestational age strata.
Stratified analyses by sex showed negligible differences in the nSOFA performance. Similarly, the analyses stratified by Gram-stain class of the isolated pathogen showed similar extents of organ dysfunction among LOI survivors and nonsurvivors.
We identified several relevant parallels of our study with studies on SOFA performance. An increase in the SOFA score over time (Ն2 points) was predictive of mortality risk. 18 A systematic review and meta-regression of SOFA as a predictor of mortality in adults found the change in SOFA score from baseline to be the metric most reliably associated with mortality 19 ; this finding was reflected in our study of preterm infants, in which a 2-point increase in the nSOFA score was associated with a 2-fold increase in the odds of mortality. Similarly, organ dysfunction scores are widely used as metrics associated with mortality as well as clinical course in pediatric and adult ICU populations and have been validated across multiple disease states. [20][21][22][23][24][25][26] No risk of mortality, clinical course description, or severity of illness scoring paradigms are routinely used in neonatal critical care.
Although life-threatening organ dysfunction is a requirement to distinguish sepsis from infection, organ dysfunction accompanies multiple disease states that commonly occur in the NICU. As an objective measure of clinical course, the nSOFA has potential utility as both a surrogate variable for illness severity and a metric for risk assessment of short-and long-term clinical end points.
We have developed an international consortium of academic centers that have integrated the nSOFA into their electronic health records for all NICU patients. This consortium may enable nSOFAfocused prospective studies, as well as quality improvement initiatives, aimed at short-and longterm outcomes in the NICU. We do not advocate for use of the nSOFA in clinical practice at this time.
Rather, these data are evidence that the nSOFA can be used to classify mortality risk among preterm infants with suspected infection for future studies. A precision medicine approach that uses patientspecific stratification of infection mortality risk is needed to establish useful diagnostic testing that will both facilitate effective antibiotic stewardship and identify those that are most likely to benefit from potential adjunctive treatments.
deaths considered infection-related occurred in close proximity to the reference culture and while receiving antibiotics for the infection episode evaluated. Taken together, these attributes suggest infection was a significant contributor to mortality. We could not include a focus on meningitis because of a limited number of documented cases (n = 3), consistent with previous reports in this population indicating that meningitis is rare and frequently associated with bacteremia. 27 In contrast to adults and children, who may experience high rates of late mortality (after antimicrobial treatment is completed) following sepsis, the frequency of late mortality during the birth hospitalization among LOI episode survivors in this multicenter study was low, it occurred several weeks after the reference culture, and it was most common among infants born extremely preterm with severe bronchopulmonary dysplasia. It could not be determined whether the cause of late mortality was primarily the result of complications of extreme prematurity, the impact of the previous infection, or a combination of factors. These questions may be answered via study of larger cohorts. This was a study to assess the generalizability of the nSOFA score in its current iteration. We concede that the score can always be improved as more data become available. We acknowledge the potential limitation that changes in demographics and standards of care may have occurred over the study period.

Conclusions
Among preterm VLBW infants with confirmed LOI episode, the nSOFA showed generalizable utility as an operational definition of organ dysfunction associated with mortality risk unaffected by sex, pathogen, specific center, or extreme prematurity. The nSOFA may provide the requisite foundation on which to build a consensus definition for sepsis in preterm neonates.