Sepsis is a leading public health problem in children, affecting more than 75,000 children in the United States annually (1, 2). With earlier recognition, infectious source control, and resuscitation, more than 80% of children hospitalized for sepsis now survive (3). However, nearly a third of sepsis survivors who require intensive care suffer new morbidities and struggle to return to their pre-sepsis functional baseline (4–10). In the Life After Pediatric Sepsis (LAPSE) study, 35% of pediatric sepsis survivors did not regain their baseline health-related quality of life 1 year after their critical illness, with deficits in physical function being the most common (5).

Despite improved epidemiologic studies characterizing post-sepsis outcomes, there are limited data on the biological mechanisms underlying or risk factors associated with functional morbidities in children who survive sepsis. In adult sepsis survivors, the severity of the inflammatory response during critical illness has been proposed as a mechanism driving new functional morbidity after discharge (11, 12). However, the impact of systemic inflammation on post-sepsis pediatric outcomes beyond mortality has not been well described. To address this knowledge gap in pediatric sepsis, we sought to test the hypothesis that elevated levels of cytokines and chemokines during critical illness are associated with new functional morbidity at hospital discharge.

We conducted a secondary analysis of the Mitochondrial Dysfunction in Pediatric Sepsis (MitoPSe) study (13, 14). Briefly, MitoPSe investigated mitochondrial dysfunction as it pertains to the immune response in children with severe sepsis/septic shock in the pediatric intensive care unit (PICU) between May 2014 and June 2018 (13). MitoPSe included the measurement of inflammatory cytokines/chemokines at three timepoints during the acute phase of illness in 166 critically ill children aged <18 years with severe sepsis and/or septic shock, as defined by the International Pediatric Sepsis Consensus Conference (IPSCC) criteria (15). In this post hoc secondary analysis, we include the subset of MitoPSe participants who survived hospital discharge and provided consent to participate in future research (Supplementary Figure S1). This secondary analysis was determined to be exempt from human subject research by the Children's Hospital of Philadelphia Institutional Review Board.

A total of 119 sepsis survivors were included in this analysis. Patient characteristics for those with and without new functional morbidity at hospital discharge are shown in Table 1. Of the 119 survivors, 38 (32%) experienced new functional morbidity at hospital discharge. There were no differences in age, sex, race, severity of illness, lactate, primary pathogen, or site of infection between groups, but patients who developed new functional morbidity were more likely to have chronic comorbid conditions (29% vs. 11%, p = 0.02), exhibited a higher mean number of organ dysfunctions on day 1 [3 (SD ± 1.2) vs. 2.2 (SD ± 1.1); p < 0.01], exhibited a higher percentage of organ dysfunction on day 14 (58% vs. 20%, p < 0.01), and were more likely to have hospital-acquired infections at 28 days (24% vs. 9%, p = 0.02) than those without new functional morbidity. Patients with new functional morbidity also had significantly longer median PICU lengths of stay [13 (IQR 9, 30) days] than those without new morbidity [8 (3, 14) days; p < 0.001].

Most children had mild functional disability at baseline with a median baseline POPC of 2 (1, 3), but there was less variability in baseline POPC in children who developed new functional morbidity [2 (1, 2)] compared to those without [2 (1, 4)], p < 0.01; Table 2). In children with new morbidity, the magnitude of change in POPC was most often identified as an increase of 1 point (71%) from baseline to discharge. The median magnitude of change in POPC is detailed in Table 2. Patients with new functional morbidity were also more likely to experience an increase in FSS ≥ 3 from baseline to hospital discharge. Patients with new functional morbidity more often required new equipment (e.g., feeding tube), new medical services, and new outpatient services (Table 2; Supplementary Table S2) than patients without new functional morbidity. The proportion of readmissions at 6 months was not different between groups, but planned readmissions were more common among patients with new functional morbidity.

Ten biomarkers were associated with new functional morbidity after correction for multiple comparisons in unadjusted analyses: interleukin (IL)-6, IL-18, sIL-2Ra, MCP1, IL-8 (CXCL8), sIL-1RII, IL-10, MIP1a, RANTES (CCL5), and IL-2r (Table 3). At each time point, log-transformed levels of these biomarkers were increased in those with new morbidity, except for RANTES which exhibited lower levels, compared to those without new morbidity (Figure 1). However, the overall trajectory of these biomarkers was not different in unadjusted analyses between those with and those without new functional morbidity.

Based on differences in characteristics between groups, we included chronic comorbid conditions, hospital length of stay, number of organ dysfunctions on day 1, and either PELOD or PIM2 score as covariates in separate mixed-effects linear regression models to assess for differences in biomarker levels and trajectory between groups using time and group assignment as an interaction term. After adjusting for these covariates, none of the 10 inflammatory biomarker levels were associated with new functional morbidity. Although most biomarkers continued to demonstrate a significant decrease over time, there were no differences in biomarker trajectories over time between groups (Table 3).

In this post hoc secondary analysis of 119 critically ill children who survived sepsis, one-third experienced new functional morbidity at the time of hospital discharge. Our data show that children who survive sepsis have increased levels of cytokine and chemokine during their acute phase of illness and these levels are associated with the development of new functional morbidity. However, this association did not hold after controlling for clinical confounding variables. Specifically, patients who developed new functional morbidity presented with a higher burden of organ dysfunction were more likely to experience healthcare-acquired infections and had a longer ICU length of stay. Despite, this early trajectories of biomarker levels may be helpful as a proxy for later clinical variables which are associated with new morbidity.

Our morbidity findings are consistent with those in the LAPSE study, where one-third of children continued to experience functional morbidity up to 6- months after hospital discharge (6, 22). Identifying potential risk factors during the acute phase of sepsis may aid in improved prognostication of those children who may be at increased risk of having new functional morbidity after PICU discharge.

Patients who developed new functional morbidity generally had higher levels of circulating cytokines and chemokines than patients who did not develop new morbidity. However, inflammatory biomarkers decreased in all patients, and the downward trajectory did not differ by risk for new functional morbidity. Moreover, after controlling for differences in baseline characteristics and illness severity, inflammatory biomarker levels no longer differed between those with and without new functional morbidity at baseline. While the biomarkers themselves do not have independent predictive value when adjusted for confounding variables, they may act as surrogates for covariates like illness severity potentially contributing to new functional morbidity.

Within survivors of the MitoPSe cohort, the development of new functional morbidity at hospital discharge was associated with ten inflammatory biomarkers measured during a child's acute illness - IL-6, IL-18, sIL-2Ra, MCP1, IL-8 (CXCL8), sIL-1RII, IL-10, MIP1a, RANTES (CCL5), and IL-2r. Although children generally experienced improved levels of inflammatory biomarkers over time irrespective of the development of new functional morbidity, most biomarker levels were higher at each time point in those who developed new functional morbidity compared to those who did not.

Only one of the ten associated biomarkers, RANTES, consistently had an inverse relationship between groups regarding log-biomarker levels of inflammation. Lower levels of RANTES were observed in those with new functional mobility at each time point compared to those without. These findings are consistent with previous pediatric research and the severity of the illness in infectious diseases (23, 24). This is also found in adult and non-neonatal pediatric populations, which have concluded that low levels of RANTES (CCL5) have increased the predictive value of mortality (23, 25). Neonatal-specific research concluded that downregulation of RANTES over time may predict the development of sepsis-induced disseminated intravascular coagulation (24). In adults, a low level of RANTES is independently associated with cardiac mortality (26). Though to our knowledge while the association between RANTES (CCL5) and mortality is established, its association with morbidity is unknown. RANTES is important for immune response activation, including being a key chemoattractant of monocytes and lymphocytes during infection. Low levels may leave an individual vulnerable to an impaired immune response and re-infection. As such, low levels and downregulation of RANTES may be an important prognostic target for early morbidity recognition throughout the acute period of illness in critically ill children.

Contradictory to RANTES decreased levels, in adult ICU patients, increased inflammation and new morbidity yield similar results. While early high levels of biomarkers may have clinical significance in determining overall treatment plans, the association between early inflammatory biomarker levels and poor outcomes is inconsistent. A secondary analysis of the ALTOS trial included adult ICU survivors with acute respiratory distress syndrome and sepsis and identified subphenotypes (hypo- vs. hyper-inflamed). Peripheral inflammatory biomarkers were drawn, and assessed, early in the critical illness (at the time of trial randomization) (27). Those with hyperinflammatory subphenotypes experienced a higher overall 12-month mortality rate; yet, physical, cognitive, and mental outcomes at 6 and 12 months were similar between groups. A similar lack of association is highlighted in Brummel et al., exploring the association between elevated inflammatory biomarkers, cognitive impairment, and disability among adults (28). However, literature shows that follow-up of biomarkers beyond the immediate critical illness may aid in understanding morbidity development. In older adults, an elevated level of C-reactive protein, a non-specific inflammatory marker associated with IL-6, at and beyond ICU discharge is associated with functional impairment after an ICU admission (21), thereby indicating the potential need for clinicians to look beyond single timepoint biomarker level associations and fully assess the trajectory of biomarker levels, including those at ICU discharge. Such biomarkers and analyses may be surrogate indicators of post-ICU morbidity.

Overall, our findings support the use of early identification of those who may be at increased risk of new functional morbidity after discharge, irrespective of biomarker levels. Early identification and intervention may aid in recovery. In multivariate analyses, while biomarker levels were not significant between groups, several pre-PICU and acute phase factors were associated with increased risk of new morbidity—the presence of chronic comorbid conditions, increased PICU and hospital length of stay, and the number of organs with dysfunction on day 1 of sepsis illness. Within our cohort, there is also evidence of increased resource utilization including new equipment, medical services, outpatient services, and readmission within 6 months after PICU discharge in those with new functional morbidity. These risk factors and outcomes have been previously discussed in the PICU literature (29). As such, clinicians can provide continued support and resources for at-risk children and families throughout the critical illness period to augment a child's recovery.

Children who survive sepsis are at risk for new functional morbidity and increased resource utilization after discharge. Within our cohort, most inflammatory biomarkers remained elevated throughout the acute phase of illness in children who developed new morbidity, except RANTES. Despite this trend, the overall biomarker trajectories were non-significant, as inflammation resolution trajectories followed similar patterns between those who developed new morbidity and those who did not. The inflammatory milieu associated with new functional morbidity was confounded by more easily measurable clinical and illness severity variables, so we did not identify independent values for the use of cytokines/chemokines to predict or identify the risk of new functional morbidity in this study. Future studies with larger cohorts of PICU survivors, focused on the association of new functional morbidity and inflammatory biomarker profiles within the acute phase of illness and the use of more sensitive outcome measures may be useful to better assess for risk of post-PICU outcomes.