Inadequate Oxygen Delivery Dose and Major Adverse Events in Critically Ill Children With Sepsis

This retrospective, observational study was granted exempt status from the hospital and university institutional review boards; procedures were implemented so that the identities of human participants could not be easily ascertained.

The study sample consisted of critical care patients aged 0 to less than 18 years admitted to a 30-bed medical-surgical ICU or a 22-bed medical ICU at a large quaternary academic children’s hospital from January 1, 2017, to November 1, 2019, with an admission diagnosis of sepsis, severe sepsis, or septic shock. An additional inclusion criterion was the presence of an arterial catheter to obtain frequent blood pressure readings required for IDo₂ calculations. Exclusion criteria were preexisting cardiac dysfunction unrelated to sepsis and weight less than or equal to 2 kg.

The medical-surgical ICU and medical ICU patient databases were searched to identify patients meeting study criteria. Those with an admission diagnosis of sepsis, severe sepsis, or septic shock were evaluated by a member of the research team to confirm consistency with definitions established by the 2005 Pediatric Sepsis Consensus Conference.¹⁹  Specifically, sepsis requires the presence of at least 2 of the age-specific systemic inflammatory response syndrome criteria, including either abnormal temperature or abnormal leukocyte count, in the context of infection. Severe sepsis requires the presence of sepsis with either cardiovascular organ dysfunction or acute respiratory distress syndrome, or 2 or more noncardiovascular organ system dysfunctions. Septic shock requires the presence of sepsis with cardiovascular organ dysfunction including hypo-tension, need for vasoactive medications, or objective signs of impaired perfusion.

Medical records were reviewed to obtain data on demographics, admission diagnoses, MAEs, PICU LOS, and days receiving invasive ventilation. Maximum VIS and PELOD-2 scores were calculated on days 1 (0-24 hours) and 2 (24-48 hours) of admission. Additional clinical indicators including peak lactate level, minimum hemoglobin level, and total volume of red blood cell transfusions and fluid boluses were also collected for the first 48 hours of PICU admission. Data were entered into Research Electronic Data Capture (REDCap) version 9.5 (Vanderbilt University) and exported into SAS software version 9.4 (SAS Institute). The IDo₂ doses of deidentified study patients were retrospectively calculated by Etiometry for the 4 discrete and 4 cumulative time intervals. The IDo₂ doses were exported into Excel (Microsoft) and then combined with demographic and clinical characteristics in SAS for analysis.

Descriptive statistics were used to illustrate demographic and clinical characteristics. The Wilcoxon rank sum 2-sample test was used to compare the IDo₂ dose between groups with and without MAEs. The risk of MAEs for patients with an IDo₂ dose above versus below various percentile thresholds was quantified with odds ratios and 95% CIs. Spearman correlations were used to determine whether IDo₂ dose was related to illness severity metrics and outcomes, specifically maximum VIS, PELOD-2 scores, PICU LOS, and days receiving invasive ventilation. The threshold for statistical significance was set at P less than .05.

A total of 102 patients were identified for study inclusion. Demographic and clinical characteristics are presented in Table 1. Thirteen patients experienced MAEs, and 3 of the 13 experienced more than 1 MAE. All cases of cardiac arrest requiring chest compressions and cannulation for ECMO occurred within the first 48 hours of admission. Comparing the groups with (n = 13) and without (n = 89) MAEs, the MAE group had a significantly lower mean hemoglobin level, a higher volume of red blood cell transfusions, a higher day 1 VIS, higher day 1 and day 2 PELOD-2 scores, a longer mean PICU stay, and a greater mean number of days receiving invasive ventilation. Of note, PICU LOS and invasive ventilator days were exclusive of those who required invasive ventilatory support at baseline, as those patients were ineligible for discontinuation of invasive ventilation and were required to remain in the PICU in accordance with the policy of the study institution.

The IDo₂ doses for the sample ranged from 0 to 79.81. The IDo₂ doses were significantly higher in patients with MAEs than in those without MAEs for the time intervals of 0 to 12 hours (mean, 17.91 vs 3.36; P < .001), 12 to 24 hours (mean, 9.08 vs 1.84; P = .01), 0 to 24 hours (mean, 12.65 vs 2.56, P < .001), 0 to 36 hours (mean, 9.26 vs 2.33; P < .001), and 0 to 48 hours (mean, 7.54 vs 2.20; P < .001) (Figure 2, Table 2). Figure 3 shows odds ratios for MAEs based on an IDo₂ dose above versus below various percentile thresholds for the 4 discrete time intervals. Odds of a MAE were greatest when the IDo₂ dose at 0 to 12 hours was at or above the 80th percentile dose for the sample population (odds ratio, 23.6; 95% CI, 5.6-99.4).

Spearman correlations were used to examine the time interval of 0 to 12 hours in more detail. Significant correlations were observed between the IDo₂ dose at 0 to 12 hours and each of the following: day 2 maximum VIS (ρ = 0.27, P = .006), day 1 PELOD-2 score (ρ = 0.41, P < .001), day 2 PELOD-2 score (ρ = 0.44, P < .001), PICU LOS (ρ = 0.35, P < .001), and days receiving invasive ventilation (ρ = 0.42, P < .001) (Table 3). The relationship between IDo₂ dose at 0 to 12 hours and day 1 maximum VIS (ρ = 0.18, P = .07) approached but did not reach statistical significance. A significant negative correlation was observed between IDo₂ dose at 0 to 12 hours and age (ρ = −0.47, P < .001). Additional assessment of the age effect revealed that the majority (62%) of those who experienced MAEs were less than 2 years of age, and the majority (62%) of those with IDo₂ dose at 0 to 12 hours at or above the 80th percentile were also less than 2 years of age.

In this retrospective, observational study of critically ill children with sepsis, several statistically significant relationships were found between IDo₂ dose and MAEs, illness severity metrics, and outcomes. Specifically, patients with MAEs had higher IDo₂ doses over multiple time intervals, and an IDo₂ dose at 0 to 12 hours at or above the 80th percentile was associated with the greatest odds of experiencing a MAE. Additionally, there were significant correlations between IDo₂ dose at 0 to 12 hours and day 2 maximum VIS, organ dysfunction as measured by PELOD-2 score, PICU LOS, and days receiving invasive ventilation.

Although this is the first study to examine IDo₂ dose in critically ill children with sepsis, investigators have reported similar encouraging findings in other pediatric populations. Futterman et al²⁰  found that among neonates with congenital heart disease who had undergone cardiopulmonary bypass surgery, higher IDo₂ dose was associated with increased risk of cardiac arrest during a 120-minute monitoring window that terminated 10, 20, or 30 minutes before the cardiac arrest. Although the current study did not examine IDo₂ dose just proximal to the MAE, it did identify the time interval of 0 to 12 hours to be particularly valuable. Sankar et al¹⁰  similarly emphasized the importance of early Scvo₂ monitoring among children with septic shock, as they found that only those in whom Scvo₂ normalized within the first 6 hours of admission survived. In the future, IDo₂ monitoring at the bedside may be most critical early in the sepsis trajectory.

In contrast, Rogers et al²¹  examined the IDo₂ index as a predictor of adverse events associated with low cardiac output syndrome in children with congenital heart defects after cardiac bypass and found that IDo₂ dose for the 12 hours following surgery did not have a significant relationship with adverse event occurrence within 72 hours of surgery. Although the small sample of 28 paired cases and controls may have precluded more significant findings, the authors also suggested that high variability of IDo₂ index values in a 12-hour interval can restrict the discriminative capability of the dose, or average value. The current study also used retrospectively calculated IDo₂ doses as a research tool for summarizing the index values, although it is likely that real-time bedside visualization of IDo₂ index values will more accurately reflect the patient’s dynamic physiological state. In this way, the clinical team can be alerted to high-risk time periods and trends that prompt early interventions such as broadening antibiotics, administering fluid or inotropes, pursuing more aggressive source control, or electing early advanced mechanical support. Demonstrating this benefit, Salvin et al²²  found that implementation of the IDo₂ index in clinical practice resulted in a significant relative reduction in LOS among neonates after cardiac surgery. Future studies are imperative for measuring interventions and outcomes associated with bedside IDo₂ monitoring of critically ill children with sepsis.

The correlational findings in the current study were also informative. The significant positive correlations between IDo₂ dose at 0 to 12 hours and the outcomes of PICU LOS and days receiving invasive ventilation suggest that IDo₂ dose may facilitate early recognition of patients at risk of extended LOS and ventilator-associated morbidity. In previous research, PELOD-2 score on PICU day 1 has been highly predictive of PICU mortality among children with suspected infection.²³  In the present study, the statistically significant relationship between IDo₂ dose and PELOD-2 scores suggests that IDo₂ dose may similarly identify those children with sepsis who are at highest risk of mortality. The near-continuous generation of IDo₂ index values may be advantageous over PELOD-2 scores, which are calculated manually and only intermittently.

Curiously, the correlation observed between IDo₂ dose at 0 to 12 hours and day 1 maximum VIS approached but did not reach statistical significance. It is possible that treatment strategies for impaired Do₂, including crystalloid resuscitation and colloid administration, preceded and attenuated the need for vasoactive-inotropic support. Additionally, current consensus guidelines lack recommendations for specific mean arterial pressure targets for children with sepsis and septic shock.²⁴  Ensuing differences in clinical practice could account for variations in VIS and lack of a statistically significant correlation with IDo₂ dose.

The significant negative correlation between IDo₂ dose and age was of particular interest. A possible physiological explanation is that younger children have a tendency toward cold shock,²⁵  which may yield a lower Svo₂ than warm shock. They may also have subtle, nonspecific symptoms that make it challenging to diagnose infection,²  and delayed diagnosis may lead to worsening illness severity. Likewise, neonates carry unique physiological considerations including immunological immaturity, increased pulmonary vascular resistance, and immature mechanisms of thermogenesis.^26,27  Previous studies have found multiple organ dysfunction syndrome, often associated with sepsis, and subsequent PICU mortality to be highest in neonates and infants.^28,29  The Sepsis Prevalence, Outcomes, and Therapies (SPROUT) study, however, cited no significant difference in PICU or hospital mortality rates by age among children with severe sepsis.³⁰  Collectively, these findings demonstrate an ongoing need to investigate the relationship between IDo₂ dose and age among children with sepsis.

Finally, the MAE group had a significantly lower mean hemoglobin level (7.9 g/dL vs 8.8 g/dL, P = .03) and higher mean volume of red cell transfusions (22.3 mL/kg vs 3.7 mL/kg, P < .001) than the non-MAE group. The hemoglobin difference may not have been clinically meaningful, as the typical transfusion threshold at the study site is less than 7.0 g/dL. However, volume of red blood cell transfusions was explored as a potential contributor to the occurrence of MAEs, as red blood cell transfusion has been associated with morbidity and mortality in critically ill patients.³¹  Further review demonstrated that all but 1 patient requiring 30 mL/kg or greater of red blood cell transfusions had undergone cannulation for ECMO. The other patient suffered acute hemorrhage. Thus, transfusion volume was likely a consequence of the MAE and not a contributor.

This study had several limitations in addition to those inherent to a retrospective review of medical records. First, delayed placement or early removal of an arterial catheter, as well as early transfer from the PICU to an inpatient unit, occasionally resulted in less than 48 hours of IDo₂ data. Fortunately, data required to calculate the IDo₂ dose were available for most patients during at least 90% of the study period and no patients were excluded because of missing data. Also, each patient transferred to an inpatient unit within 48 hours of admission had an IDo₂ dose at 0 to 12 hours of less than 1 and experienced no MAEs, further supporting validation of the IDo₂ dose. Second, some study measures may have been influenced by concomitant diagnoses. For example, invasive ventilator days may have been attributable to an accompanying respiratory illness rather than the sepsis condition. Similarly, elevated PELOD-2 scores may have been related to comorbid conditions as opposed to sepsis pathophysiology. Finally, the 95% CI for the odds ratio statistic at the 80th percentile threshold was wide. A larger sample size may have produced a narrower CI, more precise results, and a stronger relationship between IDo₂ dose and VIS.

The results of this study suggest that routine IDo₂ monitoring of critically ill children with sepsis, particularly during the first 12 hours of PICU admission, may help identify those at highest risk so that interventions can be performed to optimize the clinical course. The ability of the IDo₂ index to continuously compile multiple data sources in an environment with a high cognitive workload, such as the PICU, may aid recognition of subtle patient changes that would otherwise go unnoticed. More research is needed to quantify the anticipated clinical benefits of real-time IDo₂ monitoring at the bedside. Future subgroup analysis of those with sepsis, severe sepsis, and septic shock may also offer guidance in the use of the IDo₂ index across the spectrum of sepsis diagnoses. Amid increasing availability of analytic tools to support clinicians in all realms of health care, it is essential to continue to explore how integration of these tools into practice can benefit patients and optimize outcomes.

This work was performed at Boston Children’s Hospital and Northeastern University, Boston, Massachusetts. We thank the following individuals for their help and support throughout the study: Adrianna Caraglia, Benjamin Cerrato, Patricia Hickey, Brian McAlvin, Shannon Meyer, Dimple Mirchandani, Mary O’Brien, and Julie Vincuilla.