Journal Logo

Online Clinical Investigations

Lactate, Base Excess, and the Pediatric Index of Mortality: Exploratory Study of an International, Multicenter Dataset

Morris, Kevin P. MD1,2; Kapetanstrataki, Melpo MSc3; Wilkins, Barry MD4; Slater, Anthony J. MD5; Ward, Victoria MD4; Parslow, Roger C. PhD3

Author Information
Pediatric Critical Care Medicine: June 2022 - Volume 23 - Issue 6 - p e268-e276
doi: 10.1097/PCC.0000000000002904


Mortality risk prediction is central to case-mix adjustment and assessment of ICU performance. Two scoring systems are used widely in pediatric practice—the Pediatric Index of Mortality (PIM), first published in 1997 (1), and the Pediatric Risk of Mortality Score (PRISM), first published in 1988 (2). In the latest versions, PIM3 adjusts for admitting diagnosis, type of admission, and various physiologic variables over the first hour of ICU admission (3), whereas PRISM III includes the most abnormal values of a number of prespecified physiologic and laboratory variables over the first 4 hours of admission (4). Each score incorporates a marker of acid-base status: pH for PRISM III and “absolute” base excess (BE) for PIM3, reflecting the fact that in the development of PIM, both a positive BE and a negative BE were associated with greater mortality risk. The PIM3 score is the scoring system used to compare PICU performance across the United Kingdom and Republic of Ireland, and Australia and New Zealand (ANZ).

Monitoring of blood lactate was not common practice in ICUs at the time that PRISM and PIM scores were derived, and so its performance was not evaluated in their development, nor evaluated in subsequent revisions. However, since the scores were developed, a large body of literature has emerged supporting a strong relationship between blood lactate and mortality in neonatal, pediatric, and adult critically ill patients (5–9). A recent study in adults with sepsis found lactate to be a more powerful predictor than even refractory hypotension (10). A number of these studies have highlighted the superior predictive power of blood lactate over other acid-base markers, specifically pH and BE (6,7).

Many different factors can affect blood lactate concentration, and our understanding has moved on beyond simple interpretation of lactate as a marker of anaerobic metabolism in situations of hypoperfusion; both the rate of production and of clearance, particularly in the liver and kidney, will influence the level (11). Clearance is often compromised in the critically ill (12). In addition, situations which result in an increase in glucose metabolism (glycolysis) will result in increased lactate generation, in the absence of anaerobic metabolism or hypoperfusion. Drugs, including epinephrine and salbutamol, have this effect (13). Hyperventilation-induced alkalosis can also result in an increase in blood lactate concentration (14). Not all of these clinical situations would be expected to be associated with increased mortality. In keeping with this complexity, the relationship between blood lactate concentration and markers of acidosis (pH, bicarbonate, BE) is not a simple one, and there may not be a tight relationship between lactate concentration and pH or BE (15).

Single-center studies have suggested that inclusion of blood lactate results in improved performance of PIM2 (6) and PRISM III (16). Point-of-care measurement of blood lactate is now available using blood gas analysers, and lactate has been incorporated into key pediatric and adult ICU guidelines (17–19), and in many ICUs is now monitored as standard ICU practice.

In this study, we combine two large international PICU databases from the United Kingdom and Republic of Ireland (ROI) and ANZ to investigate relationships between blood lactate, BE, and ICU mortality. In addition, we explore the effect of incorporating blood lactate into PIM and consider how best to deal with missing values.


Study Population and Study Design

This was a retrospective cohort study based on data prospectively collected on PICU admission. BE and lactate measurements were all taken in the time period between first contact with an ICU doctor and 1 hour after ICU admission. All admissions between January 1, 2012, and December 31, 2015, were included. Data were obtained from the Pediatric Intensive Care Audit Network (PICANet) (20) and the Australian and New Zealand Pediatric Intensive Care (ANZPIC) Registry (21). PICANet is an audit database, established in 2002, which collects detailed information on all children admitted to 33 designated PICUs in the United Kingdom and the ROI. Collection of personally identifiable data was approved by the Patient Information Advisory Group (now the Health Research Authority Confidentiality Advisory Group), and ethical approval for the use of data for research purposes was granted by the Trent Medical Research Ethics Committee (REC 2018/EM/0267). The ANZPIC registry was established in 1997 and contains information on all admissions to nine PICUs as well as 20 general ICUs in ANZ. Ethical approval for maintaining the registry and using the data for research and quality improvement was granted by the Children’s Health Queensland Hospital and Health Service Human Research Ethics Committee (HREC/2016/QRCH/338). This study, using retrospective data, was undertaken in accordance with the above Ethics/IRB approvals. No personally identifiable data were used.

A series of analyses were undertaken which are summarized in Figure 1 and described in detail below.

Figure 1.:
Flow diagram describing analyses and numbers of admissions.

Analysis 1—Descriptive Statistics

All admissions were included in this analysis which examined the characteristics of patients with and without a lactate value obtained at ICU admission. Normality of continuous variables was assessed with the use of histograms and the Shapiro-Wilk statistic. Association between the availability of a lactate value (yes/no) and continuous variables was investigated with the use of the nonparametric Mann-Whitney U test and between categorical variables with the chi-square test.

Analysis 2—Relationship Between Lactate, BE, and Mortality

This analysis was confined to admissions with both lactate and BE values obtained at ICU admission. Arterial, capillary, and venous BE and lactate values were included. Correlation between lactate and BE was assessed with the Spearman correlation coefficient. Generalized additive modeling (GAM) functions were used to visualize the trend of mortality against lactate and BE values. A generalized additive model is a generalized linear model in which smoothing functions of the linear predictors are used instead. The smoothing function used was based on penalized regression splines.

Odds ratios (ORs) and 95% CI were calculated to determine the risk of mortality per unit change in lactate and BE. In addition, standardized ORs (sORs) were calculated to assess the effect on mortality of an increase of 1 sd in lactate and BE.

Analysis 3—Investigating Different Approaches for Missing Lactate Values

In order to assess how best to handle missing lactate values in subsequent modeling, we modified PIM2 by substituting lactate for absolute BE, whenever lactate was available, and assessed the impact on model performance, measured by area under the receiver operating characteristic curve (AUROC) and Akaike information criterion (AIC). Three models were explored: 1) setting missing lactate to 0, 2) setting missing lactate to 1, and 3) setting missing lactate to 1, replacing values between 0 and less than 1 with the value 1 and using the term (lactate–1) in the model. Missing BE was handled according to established PIM guidance, substituting a value of zero.

Analysis 4—Evaluating PIM Performance Using Absolute BE, Lactate, or Both Variables

PIM2 was introduced in 2003; therefore, PIM2 data were available for the whole period of our study. Since we had a large amount of data on PIM2, we recalibrated it to more accurately represent the data and predict mortality. PIM3 was introduced in 2013 but data collection did not start immediately, so complete PIM3 data were available only for admissions in 2015. PIM3 was not recalibrated, since we only had 1 year’s worth of data and so it was appropriate to use the original coefficients.

A modified PIM model was created without BE, and the impact on model performance of stepwise addition of absolute BE alone, lactate alone, and a combination of absolute BE and lactate was assessed. The analysis was performed for both PIM2 (whole cohort) and PIM3 (2015 only) models. As per PIM guidance, a missing BE was assigned a value of zero. Handling of missing lactate was determined by the findings of Analysis 3.

The fit of the models was assessed with the use of the AUROC, AIC, and the calculation of residuals and influence measures (Pearson, standardized Pearson and Deviance Residuals, change in chi-square, change in deviance, and Pregibon Delta-Beta). Calibration plots were created to visualize how well the predicted mortality probabilities compared with the observed mortality. The Hosmer-Lemeshow goodness of fit statistic was used, grouping the patients in 20 groups according to their estimated probability of mortality and then plotting the observed against the predicted mortality of these groups. Analysis with the use of the GAM functions was carried out in RStudio 0.99.486 (R Foundation for Statistical Computing, Vienna, Austria). All other analyses were conducted in Stata 14.1 (StataCorp, College Station, TX).

Additional Analysis

Historically, PIM guidance has been to exclude BE values derived from a venous sample. An analysis was undertaken to explore and compare the performance of PIM with and without venous BE values using the ANZPIC registry cases only, since submission of venous BE values was supported in the ANZPIC registry but not in PICANet.


In years 2012–2015, there were a total of 123,252 admissions recorded in both datasets; 81,576 (66.2%) in the UK/ROI and 41,676 in ANZ (33.8%). Of these, 75,070 (61%) had a BE recorded, and 63,316 (51%) had a lactate recorded. Of note the percentage with lactate recorded increased from 47% in 2012 to 61% in 2015, whereas the proportion with a BE recorded remained constant. Lactate measurements of less than or equal to 0 or greater than equal to 30 mmol/L (n = 7) were recoded as missing.

Within the cohort of 63,316 admissions with an admission lactate, there were 2,440 with no corresponding BE value, leaving a cohort of 60,876 admissions with both lactate and BE available (66.9% from UK/ROI, 33.1% from ANZ) (Fig. 1).

Analysis 1 (n = 123,252)

Median lactate value was 1.5 mmol/L (interquartile range, 1–2.4 mmol/L) (UK/ROI: 1.5 [1–2.5]; ANZ: 1.4 [1–2.3]). Table 1 shows descriptive characteristics for children with and without a lactate measurement. Median length of stay for children with a lactate measurement was 2.26 days, compared with 1.74 days for children without a lactate measurement (p < 0.001). Children with a valid lactate measurement had a higher illness severity based on PIM2 score, were more likely to be invasively ventilated, admitted after cardiac surgery (p < 0.001), and had a higher mortality rate (4% vs 2.5%; OR, 1.57 [1.47–1.67]; p < 0.001).

TABLE 1. - Patient Characteristics by Lactate Measurement Status
Variables Lactate Measurement, N = 63,316 No Lactate Measurement, N = 59,936 p
Age (yr), median (IQR) 1.59 (0.28–6.94) 1.43 (0.29–5.84) < 0.001
Planned admission after surgery, n (%) 23,679 (37.4) 16,161 (27.0) < 0.001
Invasive mechanical ventilation, n (%) 44,292 (70.0) 25,748 (43.0) < 0.001
Length of stay (d), median (IQR) 2.26 (1.01–5.33) 1.74 (0.84–4.07) < 0.001
Pediatric Index of Mortality 2 (%), median (IQR) 1.25 (0.61–3.97) 0.99 (0.27–2.23) < 0.001
Outcome, n (%)
 Alive 60,771 (96.0) 58,454 (97.5) < 0.001
 Died 2,545 (4.0) 1,482 (2.5)
IQR = interquartile range.

Analysis 2 (n = 60,876)

The relationship between lactate, actual, and absolute BE is shown in Figure 2, for the subset of cases with a lactate and BE measurement. Overall the relationship is relatively weak but stronger between lactate and negative BE (rho = –23; p < 0.001) than positive BE (rho = 0.008; p = 0.03).

Figure 2.:
Relationship between lactate and actual base excess (A) and absolute base excess (B).

Figures 3 and 4 show the relationship between lactate, BE, and mortality, for the subset of cases with a lactate and BE measurement, with the use of GAM models. Negative BE is associated with higher mortality (OR, 1.17; 95% CI, 1.16–1.18) than positive BE (OR, 1.01; 95% CI, 0.99–1.02) (Fig. 4). The relationship between lactate and mortality is stronger (OR, 1.32; 95% CI, 1.31–1.34), per unit increase, than between absolute BE and mortality (OR, 1.13; 95% CI, 1.12–1.14).

Figure 3.:
Relationship between lactate and mortality.
Figure 4.:
Relationship between mortality and actual base excess (A) and absolute base excess (B).

This finding is confirmed by comparison of the sORs for mortality for an increase of 1 sd in lactate and BE (lactate sOR, 1.86 [1.82–1.91]; BE sOR, 1.74 [1.70–1.79]; p < 0.001).

Analysis 3 (n = 123,252)

Table 2 explores the effect of substituting absolute BE with lactate in a PIM2 model using three different treatments for handling missing lactate. The highest AUROC and lowest AIC were found if missing lactate was assigned a value of 1. This handling of missing lactate values was therefore applied in Analysis 4.

TABLE 2. - Effect of Substituting Lactate for Absolute Base Excess in Pediatric Index of Mortality 2 Using Three Different Treatments for Missing Lactate
Models Area Under the Receiver Operating Characteristic Curve 95% CI AIC
Model 1: original PIM2 model 0.8682 0.8628–0.8737 25424.649
Model 2: substitute lactate for BE in PIM2 0.8699 0.8645–0.8753 25401.247
Set missing lactate to 0
Model 3: substitute lactate for BE in PIM2 0.8712 0.8659–0.8765 25348.149
Set missing to 1
Model 4: Substitute lactate for BE in PIM2 0.8710 0.8657–0.8764 25355.850
Set missing lactate to 1, replace values between 0 and < 1 with 1, put the term (lactate–1) in the model
BE = base excess, PIM = Pediatric Index of Mortality.

Analysis 4 (n = 123,252 [PIM2]; n = 30,297 [PIM3])

Addition of lactate to modified PIM2 (without BE) resulted in a small improvement of performance over addition of absolute BE, whereas adding both lactate and BE achieved the highest AUROC of 0.8729 (Table 3). Similar effects were seen when the analysis was repeated on a modified PIM3 score for the 2015 cohort of patients (Table 3).

TABLE 3. - Modeling the Effect of Adding in Absolute Base Excess, Lactate, or Both, to a Modified Pediatric Index of Mortality 2 and Pediatric Index of Mortality 3 (without Base Excess)
Model Likelihood Ratio Test p Area Under the Receiver Operating Characteristic Curve 95% CI
Modified PIM2 (without BE) n = 123,252 Add absolute BE 432.41 < 0.0001 0.8682 0.8628–0.8737
Add lactate 508.91 < 0.0001 0.8712 0.8659–0.8765
Add both absolute BE and lactate 680.89 < 0.0001 0.8729 0.8676–0.8783
Modified PIM3 (without BE) n = 30,297 Add absolute BE 68.28 < 0.0001 0.8670 0.8614–0.8726
Add lactate 138.37 < 0.0001 0.8694 0.8639–0.8750
Add both absolute BE and lactate 155.34 < 0.0001 0.8713 0.8658–0.8768
BE = base excess, PIM = Pediatric Index of Mortality.
Likelihood ratio test against modified PIM without BE, area under the receiver operating characteristic curve. Missing BE set to zero, missing lactate set to 1.

Exploration of Adding in Venous BE Values (n = 41,676)

We repeated the modeling undertaken in Analysis 4 on a subset of patients from the ANZPIC registry (n = 41,676), with an additional step of allowing venous as well as arterial and capillary BE samples. Missing BE was set at zero, as per established PIM guidance, and missing lactate set to 1, as per the findings in analysis 3. We excluded the subset of patients from PICANet as guidance to date has been to not enter a value for BE for venous samples. As a result, only 3.0% (2,441/81,576) of PICANet BE values were documented as being venous, compared with 15.0% (6,253/41,676) of ANZPIC registry BE values, corresponding to 5.6% (2,302/40,821) and 28.9% (5,798/20,055) of admissions with both lactate and BE values, respectively.

The analysis demonstrated that performance of PIM2 was slightly improved if venous BE values were allowed in addition to arterial and capillary values (AUROC 0.9098 vs 0.9086) (Supplementary Table,


In this retrospective study of a large international dataset, PICU admission lactate level was found to be more strongly associated with mortality than BE. Exploratory analysis of this dataset suggests that incorporation of lactate into the PIM model results in a small improvement in model performance.

A major strength of the study is the large sample of patients from the UK and ROI and ANZ, with excellent systems of audit in place that capture information on every child admitted to a PICU in these countries (PICANet and ANZPIC). This allowed a comprehensive evaluation of the research question.

Despite the work of Weil et al (22) as early as the 1970s demonstrating the importance of blood lactate as a marker of the severity of shock and critical illness (20,21), widespread measurement of lactate was infrequently undertaken in ICU until comparatively recently. Evaluation of lactate was not incorporated into the work to develop PIM, undertaken in the early 1990s, to develop PRISM, undertaken in the mid-1980s, or included in the development of the most widely used severity of illness measure in adult intensive care (Acute Physiology and Chronic Health Evaluation) (1,2,22).

Findings of single-center studies have suggested that lactate adds prognostic value to existing scoring systems and should be considered for inclusion into PIM and PRISM (6,16,23). These studies add to a large body of evidence supporting a consistent association between blood lactate and mortality in critically ill patients across neonatal, pediatric and adult critical care populations (5–10). For example, in the Australasian Resuscitation In Sepsis Evaluation (ARISE) study of adults with sepsis those randomized to goal-directed therapy because of isolated hyperlactatemia had 1.7 times the risk of 90-day mortality compared with patients with isolated hypotension (10). Blood lactate has recently been incorporated into new prognostic models that have been developed for severe pediatric sepsis (18) and mortality prediction in sick African children (19).

Although the normal lactate concentration in unstressed individuals is 1.0 ± 0.5 mmol/L (20), patients with critical illness are generally considered to have normal lactate levels at concentrations of less than 2 mmol/L. Nichol et al (25) evaluated the relationship between lactate and hospital mortality in a cohort of 7,155 consecutive admissions to ICU and confirmed a relationship between both admission lactate (OR, 2.1 [1.3–3.5]) and time-weighted lactate concentration (OR, 3.7 [1.9–7.0]) and hospital mortality (24). This significant association was first detectable at lactate concentrations greater than 0.75 mmol/L, leading the authors to suggest that the current reference range for lactate in the critically ill may need to be reassessed.

A significant limitation of this study is the number of cases without a lactate and/or BE measurement. During the first year of the study (2012), a higher proportion of cases had a BE available compared with lactate (62.5% vs 46.6%), whereas by 2015, the proportions were similar (61.3% vs 60.7%) in keeping with widespread incorporation of lactate sensors into ICU point of care blood gas analysers. Overall however ~50% of cases did not have a measurement or BE or lactate, most likely because these patients did not have a blood sample taken for blood gas analysis in the first hour of ICU admission. However, the handling of missing data is something that is considered in the development of any prediction model, and the original PIM model, and subsequent revisions (PIM2, PIM3), was validated in the setting of comparable levels of missing values of Po2 and BE on blood gas analysis. Another relative limitation of our study is that we had complete PIM3 data only for 2015, although this cohort still amounted to over 30,000 admissions, so had to undertake most of the analysis using PIM2. In addition, we were only able to explore the relationship between BE, lactate, and mortality, as these are recorded routinely, and did not have data to include or explore other variables such as pH, Hco3, anion gap, albumin, or strong-ion gap.

A challenge with any severity of illness score is how to deal with missing values, with the usual assumption being that missing values are replaced with a “normal” value for that variable. In the case of PIM2, a value of zero would be assigned for a missing absolute BE value. We know that children who have a blood gas taken within the first hour of ICU admission are a more at-risk population with higher probability of death than children in whom a blood gas is not undertaken. By considering venous BE values as missing values in PIM, and thereby assigning a zero value for BE, we have until now potentially underestimated the risk of death for these cases. The analysis undertaken in a subset of over 40,000 ANZ admissions suggests that allowing venous BE values results in a slightly improved PIM2 performance. In this study, we also explored the impact of adopting a value of either 0 or 1 for missing lactate and found slightly better performance if missing values were set to 1.


In this large international, retrospective study, PICU admission blood lactate was more strongly associated with ICU mortality than absolute BE. Adding lactate into the PIM model results in a small improvement in performance. Any potential improvement in PIM performance must be balanced against the added burden of data capture when considering potential incorporation into the PIM model.


We would like to thank all participating ICUs for their support and submission of data to Pediatric Intensive Care Audit Network (PICANet) and the Australia and New Zealand Pediatric Intensive Care Registry. The PICANet Audit is commissioned by the Healthcare Quality Improvement Partnership as part of the National Clinical Audit Program. The PICANet Audit is funded by NHS England, NHS Wales, NHS Lothian/National Service Division NHS Scotland, the Royal Belfast Hospital for Sick Children, The National Office of Clinical Audit, Republic of Ireland, and HCA Healthcare.


1. Shann F, Pearson G, Slater A, et al.: Paediatric index of mortality (PIM): A mortality prediction model for children in intensive care. Intensive Care Med 1997; 23:201–207
2. Pollack MM, Ruttimann UE, Getson PR: Pediatric risk of mortality (PRISM) score. Crit Care Med 1988; 16:1110–1116
3. Straney L, Clements A, Parslow RC, et al.; ANZICS Paediatric Study Group and the Paediatric Intensive Care Audit Network: Paediatric index of mortality 3: An updated model for predicting mortality in pediatric intensive care*. Pediatr Crit Care Med 2013; 14:673–681
4. Pollack MM. Pediatric risk of mortality score: Update 2015. Pediatr Crit Care Med. 2016: 17: 2–9
5. Phillips LA, Dewhurst CJ, Yoxall CW: The prognostic value of initial blood lactate concentration measurements in very low birthweight infants and their use in development of a new disease severity scoring system. Arch Dis Child Fetal Neonatal Ed 2011; 96:F275–F280
6. Morris KP, McShane P, Stickley J, et al.: The relationship between blood lactate concentration, the paediatric index of mortality 2 (PIM2) and mortality in paediatric intensive care. Intensive Care Med 2012; 38:2042–2046
7. Deshpande SA, Platt MP: Association between blood lactate and acid-base status and mortality in ventilated babies. Arch Dis Child Fetal Neonatal Ed 1997; 76:F15–F20
8. Duke TD, Butt W, South M: Predictors of mortality and multiple organ failure in children with sepsis. Intensive Care Med 1997; 23:684–692
9. Jansen TC, van Bommel J, Woodward R, et al.: Association between blood lactate levels, sequential organ failure assessment subscores, and 28-day mortality during early and late intensive care unit stay: A retrospective observational study. Crit Care Med 2009; 37:2369–2374
10. Gotmaker R, Peake SL, Forbes A, et al.; ARISE Investigators*: Mortality is greater in septic patients with hyperlactatemia than with refractory hypotension. Shock 2017; 48:294–300
11. Hernandez G, Bellomo R, Bakker J. The ten pitfalls of lactate clearance in sepsis. Intensive Care Med 2019; 45:82–85]
12. Tapia P, Soto D, Bruhn A, et al.: Impairment of exogenous lactate clearance in experimental hyperdynamic septic shock is not related to total liver hypoperfusion. Crit Care 2015; 19:188
13. Greene NM: Effect of epinephrine on lactate, pyruvate, and excess lactate production in normal human subjects. J Lab Clin Med 1961; 58:682–686
14. Garcia AC, Lai YI, Attebery BA, et al.: Lactate and pyruvate accumulation during hypocapnia. Respir Physiol 1971; 12:371–380
15. Hatherill M, McIntyre AG, Wattie M, et al.: Early hyperlactataemia in critically ill children. Intensive Care Med 2000; 26:314–318
16. Bai Z, Zhu X, Li M, et al.: Effectiveness of predicting in-hospital mortality in critically ill children by assessing blood lactate levels at admission. BMC Pediatr 2014; 14:83
17. Shankar-Hari M, Phillips GS, Levy ML, et al.; Sepsis Definitions Task Force: Developing a new definition and assessing new clinical criteria for septic shock: For the third international consensus definitions for sepsis and septic shock (Sepsis-3). JAMA 2016; 315:775–787
18. Schlapbach LJ, MacLaren G, Festa M, et al.; Australian & New Zealand Intensive Care Society (ANZICS) Centre for Outcomes & Resource Evaluation (CORE) and Australian & New Zealand Intensive Care Society (ANZICS) Paediatric Study Group: Prediction of pediatric sepsis mortality within 1 h of intensive care admission. Intensive Care Med 2017; 43:1085–1096
19. George EC, Walker AS, Kiguli S, et al.: Predicting mortality in sick African children: The FEAST paediatric emergency triage (PET) score. BMC Med 2015; 13:174
20. Paediatric Intensive Care Audit Network (PICANet). Available at: Accessed February 3, 2020
21. Australia and New Zealand Paediatric Intensive Care (ANZPIC) Registry. Available at: Accessed February 3, 2020
22. Weil MH, Afifi AA: Experimental and clinical studies on lactate and pyruvate as indicators of the severity of acute circulatory failure (shock). Circulation 1970; 41:989–1001
23. Cady LD Jr, Weil MH, Afifi AA, et al.: Quantitation of severity of critical illness with special reference to blood lactate. Crit Care Med 1973; 1:75–80
24. Knaus WA, Zimmerman JE, Wagner DP, et al.: APACHE-acute physiology and chronic health evaluation: A physiologically based classification system. Crit Care Med 1981; 9:591–597
25. Nichol AD, Egi M, Pettila V, et al. Relative hyperlactataemia and hospital mortality in critically ill patients: A retrospective multi-centre study. Critical Care 2010: 14: R25–R34

base excess; lactate; mortality prediction; Pediatric Index of Mortality; severity of illness

Supplemental Digital Content

Copyright © 2022 by the Society of Critical Care Medicine and the World Federation of Pediatric Intensive and Critical Care Societies