Nucleated Red Blood Cells Are Predictive of In-Hospital Mortality for Pediatric Patients

Purpose We sought to establish whether nucleated red blood cells (NRBCs) are predictive of disposition, morbidity, and mortality for pediatric patients presenting to the emergency department (ED). Methods A single-center retrospective cohort study examining all ED encounters from patients aged younger than 19 years between January 2016 and March 2020, during which a complete blood count was obtained. Univariate analysis and multivariable logistic regression were used to test the presence of NRBCs as an independent predictor of patient-related outcomes. Results The prevalence of NRBCs was 8.9% (4195/46,991 patient encounters). Patient with NRBCs were younger (median age 4.58 vs 8.23 years; P < 0.001). Those with NRBCs had higher rates of in-hospital mortality (30/2465 [1.22%] vs 65/21,741 [0.30%]; P < 0.001), sepsis (19% vs 12%; P < 0.001), shock (7% vs 4%; P < 0.001), and cardiopulmonary resuscitation (CPR) (0.62% vs 0.09%; P < 0.001). They were more likely to be admitted (59% vs 51%; P < 0.001), have longer median hospital length of stay {1.3 (interquartile range [IQR], 0.22–4.14) vs 0.8 days (IQR, 0.23–2.64); P < 0.001}, and median intensive care unit (ICU) length of stay (3.9 [IQR, 1.87–8.72] vs 2.6 days [IQR, 1.27–5.83]; P < 0.001). Multivariable regression revealed presence of NRBCs as an independent predictor for in-hospital mortality (adjusted odds ratio [aOR], 2.21; 95% confidence interval [CI], 1.38–3.53; P < 0.001), ICU admission (aOR, 1.30; 95% CI, 1.11–1.51; P < 0.001), CPR (aOR, 3.83; 95% CI, 2.33–6.30; P < 0.001), and 30-day return to the ED (aOR, 1.15; 95% CI, 1.15–1.26; P < 0.001). Conclusions The presence of NRBCs is an independent predictor for mortality, including in-hospital mortality, ICU admission, CPR, and readmission within 30 days for children presenting to the ED.

N ucleated red blood cells (NRBCs) are early erythrocyte pre- cursors usually absent in the peripheral blood of healthy children and adults. 1 Disease processes involving extreme increases in erythropoiesis or failure of the normal bone marrow filtration mechanisms can lead to an influx of NRBCs into the circulation.][8][9] Limited data exist on the clinical utility of NRBCs as a biomarker for clinical outcomes in children. 4,9An association has been described between the length of cardiopulmonary bypass and presence of NRBCs, and presence of NRBCs has been described in a small case series of critically ill children after cardiopulmonary arrest. 9,10An observational prospective study of 670 patients in the pediatric intensive care unit (PICU) found in-hospital mortality was significantly higher among children with circulating NRBCs; however, the presence of NRBCs was not confirmed to be an independent predictor of composite outcome (death and/or ventilation and/or renal replacement therapy, and/or inotropic support) at multivariable analysis. 4Unfortunately, in this study, the number of children aged older than 28 days was limited.A recent neonatal intensive care unit (NICU) study showed infants with NRBCs had a higher risk of mortality and those with higher NRBC count had shorter survival. 11tudies on older children are scarce.Ascertaining whether NRBCs may predict outcomes in the pediatric emergency department (ED) could add value to existing knowledge and practice by improving systems for triaging, delegating resources, and deciding discharge with no additional cost by repurposing available laboratory information.In this study, we sought to establish whether NRBCs identified on routine complete blood count (CBC) screening in the pediatric ED could predict disposition, morbidity, and in-hospital mortality.

Study Design and Patient Selection
We performed a retrospective cohort study including all encounters of patients who presented to a single tertiary pediatric ED (Boston Children's Hospital [BCH], Boston, MA) between January 1, 2016 and March 31, 2020, from whom a CBC was obtained within 5 hours of arrival.The study end date was specifically selected to avoid any confounding from the emerging COVID-19 pandemic.The BCH ED has approximately 60,000 encounters per year from children and older patients with acute and chronic pediatric conditions.The study was approved by the local institutional review board with a waiver of informed consent.

Data Collection and Categorization
Demographic data, clinical characteristics, and outcomes of patients included were extracted from the electronic medical record (EMR) using a structured query language (SQL, DBeaver UI Tool, IBM Netezza Data Warehouse, IBM Inc, Armonk, NY) from time-stamped electronic health medical records.Collected variables were the following: age, race, sex, laboratory values (first absolute NRBC count, international normalized ratio, creatinine, aspartate transaminase, alanine transaminase within 5 hours of ED admission), emergency severity index score (ESI), admission category (hospital floor vs ICU), presence of sepsis or shock, use of oral or intravenous (IV) steroids within 72 hours of arrival, use of erythropoiesis-enhancing medication (darbepoetin alfa, erythropoietin, corticotropin IV, methylprednisolone IV, prednisone IV, hydrocortisone IV), use and type of vasoactive support (norepinephrine, dopamine, epinephrine, milrinone, and vasopressin), administration of blood products (platelets, packed red blood cells [PRBCs], cryoprecipitate, and fresh frozen plasma), need for dialysis, mechanical ventilation, noninvasive positive-pressure ventilation or supplemental oxygen, need for cardiopulmonary resuscitation (CPR), extracorporeal membrane oxygenation, hospital and ICU length of stay (LOS), and in-hospital mortality.6][17][18] Patients were classified to have renal dysfunction if they had a creatinine level greater than the upper limit for age based on BCH laboratory cut-offs and hepatic dysfunction if more than two or more of the following were present: aspartate transaminase greater than 100 units/L, conjugated bilirubin greater than 1.2 mg/dL, alanine transaminase greater than 500 units/L, and/or international normalized ratio greater than 1.5 in absence of warfarin anticoagulation.

Laboratory Methods
The NRBC counts were performed at the BCH laboratory, following a standardized protocol.The NRBCs were measured using a Sysmex XN-series multiparameter automated hematology analyzer (Sysmex Corporations, Kobe, Japan).Blood samples were mixed with the Lysercell WNR reagent and stains.Nucleated erythrocytes were counted by the WNR channel based on the intensity of forward-scattered light intensity and depth of fluorescent staining.The NRBCs were expressed as an absolute number of thousand cells per microliter (Kcells/μL) and not per 100 white blood cells because of the risk of confounding effects related to the leucocyte count.Nucleated red blood cell positivity (NRBC+) was defined as an absolute count greater than 0.

Outcomes
The primary outcome of this study was in-hospital mortality.Secondary outcomes included hospital admission, ICU admission, need for CPR, and return to the ED within 30 days.In addition, mortality was evaluated 1 year after presentation and through March 2021, that is, the time of data acquisition. 19

Statistical Analysis
Descriptive statistics include frequencies and percentages for categorical variables as well as median and interquartile range (IQR, 25th-75th percentile) for continuous variables.Demographic data, clinical characteristics, and outcomes were compared between patient encounters with NRBCs and those without NRBCs.The ED encounter was the unit of analysis; multiple encounters from the same patient were accounted for in all analyses using robust standard errors that cluster by patient.Categorical variables were compared using a Rao-Scott χ 2 test (clustering by patient) 20 and continuous variables were compared using the Mann-Whitney-Wilcoxon test (clustering by patient).
A Kaplan-Meier survival analysis was performed to visualize survival using single medical record numbers.The starting time for an NRBC+ patient was the encounter at which NRBC is greater than 0 even if a previous encounter existed without NRBC detection.To find the appropriate controls for these NRBC+ patients, we used the pool of patients without NRBC detection at any encounter during our study period.We then matched two NRBC− patients to an NRBC+ patient by encounter number (of the NRBC+ patient), age, ESI, and whether the patient was on glucocorticoids.Balance of the main confounders was then confirmed by comparing NRBC− and NRBC+ patients using a Rao-Scott χ 2 test (clustering by matched set) for categorical variables and Mann-Whitney-Wilcoxon test (clustering by matched set) for continuous variables.Here, the matched set consists of three patients (2 NRBC − patients matched to 1 NRBC+ patient).A Kaplan-Meier survival curve for NRBC− patients versus NRBC+ patients were compared using a log rank test, clustering by matched set. 21,22o test for the adjusted association between NRBCs and in-hospital mortality, we used all encounters in which patients were admitted to the hospital.With multiple hospital admissions from the same patients, we can consider this analysis a conditional analysis in which patients must survive the previous admission to be "at-risk" for in-hospital mortality at the next hospital admission.For each admission, a logistic regression was used to model the probability of in-hospital mortality as a function of NRBC status (present vs not present) at the current admission, as well admission number during our study period (coded as 1, 2, 3, 4, or ≥5 visits) and potential confounders: recent glucocorticoid therapy within 72 hours of ED presentation, 23 age (in years, as well as testing for a quadratic age effect) and ESI of 1 to 2 (reference ESI of 3-4). 12,14To assess if the effect of being NRBC+ increases the risk of in-hospital mortality as the number of admissions increases, we tested for the significance of an interaction between NRBC and admission number.Furthermore, to account for the multiple admissions from the same patient, we used a robust standard error. 24,25Lastly, we present results of a logistic regression model for in-patient mortality based on the first admission from each patient.Similar analyses were performed for testing the association between NRBCs and secondary outcomes (hospital admission, ICU admission, need for CPR, and return to ED in 30 days).A two-sided P value less than 0.05 was considered statistically significant.All analyses were performed using R version 4.0.3(R Foundation, Vienna, Austria) and SAS 9.4.

Cohort Description
A total of 46,991 ED encounters (51% White, 49% females, median age of 7.9 years [IQR, 2.7-14.1])during which a CBC was obtained formed the analytic cohort (Table 1).The cohort had 31,756 unique pediatric patients and 25,347 (79.8%) patients had a single encounter.Subjects who presented more than once to the ED were distributed as follows: 3656 (11.5%) twice, 1147 (3.6%) three times, 529 (1.7%) four times, with the remaining 1077 (3.4%) five or more times.Neonates (aged ≤28 days) comprised 1.9% (88/46,991) of the encounters.Almost half of patient encounters (21,525, 46%) had an ESI triage score of 1 or 2 of 5, indicative of high acuity.Fifty-two percent of the encounters resulted in admission, of whom 3399 (7.2%) were admitted to the ICU.The median LOS was 0.82 days (IQR, 0.23-2.77).with a median of 30 cells/μL (IQR, 20-90).The NRBC+ patients were younger than those without NRBCs (median age, 4.58 [IQR, 1.24-12.29]vs 8.23 [IQR, 2.96-14.24]years; P < 0.001).Patients with NRBC+ encounters more frequently had a lower ESI, re-ceived more than one inotrope, required CPR, received supplemental oxygen or positive-pressure support, required intubation, received blood products, had sepsis or shock, received glucocorticoids, needed dialysis, or had hepatic dysfunction (each P < 0.001, Categorical characteristics for the groups were compared using a Rao-Scott χ 2 test clustered by patient and presented as frequency (percentage); continuous characteristics are compared using a Mann-Whitney-Wilcoxon test (median) clustered by patient and are presented as a median with (25th-75th) percentiles.
Recent glucocorticoid use is defined as intravenous or oral steroids within 72 hours of presentation, kidney dysfunction is creatinine greater than upper limits for age, hepatic dysfunction is international normalized ratio greater than 1.5, and readmission to ED is within 30 days.
Pediatric Emergency Care • Volume 39, Number 12, December 2023 Pediatric In-Hospital Mortality Predictors Table 1).The NRBC+ encounters were more likely to lead to hospital admission, longer hospital LOS, more frequent admissions to the ICU, longer ICU stay LOS, and return to the ED within 30 days (each P < 0.001) (Table 1).

Association Between Absolute Number of NRBCs and Survival
Of the 31,757 individual pediatric patients who presented to the ED during the study period, 8.4% (2661) had detectable NRBCs during at least one encounter.Analysis was restricted to the first NRBC+ encounter for each patient and their two matched NRBC− encounters (matched by encounter number during the study period, age, ESI, and glucocorticoid use).In the matched analysis, there were 7948 patients (2661 NRBC+ patients and 5287 NRBC− patients) and 232 deaths (104 in the NRBC− group and 128 in the NRBC+ group).Supplemental Table 2 (http://links.lww.com/PEC/B116)gives the patient characteristics of the matched sample and shows that the NRBC+ and NRBC− patients are well balanced on the important possible confounders.A Kaplan-Meier estimate (Fig. 1) of survival comparing these groups gave a significant log rank test, accounting for clustering to matching ( P = 0.024), showing that NRBC are associated with in-hospital mortality.Particularly, there were no significant differences in mortality at 30 days (1.47% for NRBC+ vs 1.98% for NRBC−; P = 0.282); borderline significance at 90 days (3.53% for NRBC+ vs 2.37% for NRBC−; P = 0.053); and no significance at 1 year (9.7% for NRBC+ vs 7.3% for NRBC−; P = 0.405).The most significant difference (P = 0.001) between groups was at 2.5 years in which the mortality probability for NRBC+ patients was 20.6% (95% CI, 16.6-24.4)and 11.9% for NRBC− patients (95% CI, 9.3-14.4).

DISCUSSION
Among a large cohort of 31,757 children evaluated at a tertiary care ED, the presence of circulating NRBCs was found to be an independent predictor of in-hospital mortality, need for CPR, and need for ICU admission.Notably, however, the presence of NRBC was not associated with hospital admission for this high-risk group, but these patients were more likely to return to the ED within 30 days.This suggests that NRBCs may help identify a particular population at risk for critical outcomes that is not captured by our standard processes and thus may be discharged home.Therefore, NRBCs may have additive value in assisting the ED clinician considering disposition for borderline cases, as well as encouraging appropriate follow-up for patients with positive NRBCs on screening.
Previous studies have demonstrated the association of NRBCs with adverse outcomes in adults and more recently in neonates managed in critical care units; however, no such association has been demonstrated for the general pediatric population in the ED. 3,6,8In addition, the limited studies in the pediatric population outside the neonatal ICU have been unable to detect an  As opposed to other potential biomarkers, key advantages of using NRBCs are that it is typically included in a relatively inexpensive and routinely ordered screening test: a CBC with differential.Screening for NRBCs at the time of ED presentation, when a CBC is obtained, may serve as an additional clinical "clue", alerting the provider that the child may have a heightened risk of poor outcomes.Given the widespread availability of CBC, NRBC detection is a pragmatic variable to alert any busy health care provider to a subgroup of patients who may need admission, additional considerations, and closer follow-up.Although presence of NRBCs was not associated with in-hospital admission, it was significantly associated with higher criticality as need for ICU therapies and admission, as well as in-hospital mortality.Not surprisingly, NRBC+ patients were more likely to return to the ED within 30 days.This suggests that NRBCs may have an additive role in assisting the ED clinician in considering disposition for borderline cases, and arranging the appropriate level of follow-up to prevent adverse outcomes.
Previous studies in neonatal and adult patients have shown that even a modest rise in NRBCs was associated with increased risk of morbidity and mortality and that those with higher peak NRBC counts seemed to be at increased risk compared with those with lower nonzero counts. 11,26No such association between an absolute NRBC number and mortality was found in this study.Even patients with the lowest possible number of reported NRBCs (20 cells/uL) had an increased mortality when compared with NRBC− patients.Interestingly, the relative mortality risk for patients with NRBCs increased over time, being highest at 2.5 years, the longest time analyzed in the current study.Although this association may not be applicable to the ED setting, it demonstrates that detectable circulating NRBC counts may serve to identify a population of pediatric patients who might benefit from additional care.Although not addressed in this study, it is possible that for patients admitted or with prolonged follow-up, the trajectory of NRBC counts may prove to be most clinically useful.This is supported by some studies that have found that higher NRBC counts were associated with shorter survival in addition to increased overall mortality, 2,11,27 whereas the disappearance of circulating NRBC was associated with higher survival compared with those with persistent NRBCs. 2 At this point, the presence rather than a specific number of NRBCs should be considered sufficient to elicit increased vigilance.Future studies will be needed to investigate the trajectory of repeated NRBC counts and its relationship to risk of adverse outcomes.
Our study has several limitations.First, we did not account for all possible confounding variables thought to potentially influence NRBC counts such as genetic syndromes, congenital infections, systemic inflammation, hematologic malignancy, extramedullary hematopoiesis, or chronic hypoxemia. 1,28Although these may be contributing to observed differences in outcomes, we wanted our study to be generalizable to the population evaluated in the ED.Furthermore, reliance on ICD codes to determine these comorbidities, as well as sepsis and shock, will likely underestimate the true incidence. 29Second, we intentionally restricted our database inquiry to NRBCs within the time frame of the ED encounter to make the study applicable to the data available to the ED clinician.Complete blood counts are rarely repeated within the ED encounter, and therefore, clinical decisions are typically made from a singular data point.In addition, we chose not to control for PRBC transfusions before NRBC measurement, which can alter the NRBC measurement.We minimized the likelihood that a patient would have received a PRBC transfusion immediately before the NRBC measurement by using the first available blood sample after ED presentation.Finally, although we obtained detailed mortality data from the medical record, there is a possibility that an event occurred at an outside facility that was not reported to our institution, and therefore, deaths may be undercounted.

CONCLUSIONS
The presence of circulating NRBCs is an independent risk factor for need for higher level of care and adverse clinically relevant outcomes including in-hospital and long-term mortality in pediatric patients presenting to the pediatric ED.Our data support the clinical utility of NRBCs as an adjunctive biomarker to prognose children in a busy ED environment.Future studies should look at the trajectory of NRBC counts to assess the relationship to time to and recovery from adverse outcomes

TABLE 1 .
Clinical Characteristics and Outcomes of Pediatric Patients Presenting to the Pediatric ED and Comparison According to NRBC Finding (N = Encounters)

TABLE 2 .
Multivariable Logistic Regression for In-Patient Mortality (N = 24,270 Admissions; 94 Deaths.Model c-Statistic 0.79)Analysis for admitted patients at repeated encounter.This is a conditional analysis because, to have a future admission, a patient has to survive the previous admission.The a priori set of covariates were age, recent glucocorticoid use, ESI, and admission number.

TABLE 3 .
Adjusted and Unadjusted Logistic Regression for NRBC Association With Secondary Outcomes ,6,8 likely due to being underpowered.Our study design intentionally encompasses a large study population of more than 45,000 ED encounters from all comers who underwent laboratory studies including a CBC at the discretion of the treating clinician.It includes repeat encounters, making results generalizable to standard pediatric ED practice in which patients present repeatedly.