Rapid-Response Extracorporeal Membrane Oxygenation to Support Cardiopulmonary Resuscitation in Children With Cardiac Disease
Background—Survival of children with in-hospital cardiac arrest that does not respond to conventional cardiopulmonary resuscitation (CPR) is poor. We report on survival and early neurological outcomes of children with heart disease supported with rapid-response extracorporeal membrane oxygenation (ECMO) to aid cardiopulmonary resuscitation (ECPR).
Methods and Results—Children with heart disease supported with ECPR were identified from our ECMO database. Demographic, CPR, and ECMO details associated with mortality were evaluated using multivariable logistic regression. Pediatric overall performance category and pediatric cerebral performance category scores were assigned to ECPR survivors to assess neurological outcomes. There were 180 ECPR runs in 172 patients. Eighty-eight patients (51%) survived to discharge. Survival in patients who underwent ECPR after cardiac surgery (54%) did not differ from nonsurgical patients (46%). Survival did not vary by cardiac diagnosis and CPR duration did not differ between survivors and nonsurvivors. Factors associated with mortality included noncardiac structural or chromosomal abnormalities (OR, 3.2; 95% CI, 1.3–7.9), use of blood-primed ECMO circuit (OR, 7.1; 95% CI, 1.4–36), and arterial pH <7.00 after ECMO deployment (OR, 6.0; 95% CI, 2.1–17.4). Development of end-organ injury on ECMO and longer ECMO duration were associated with increased mortality. Of pediatric overall performance category/pediatric cerebral performance category scores assigned to survivors, 75% had scores ≤2, indicating no to mild neurological injury.
Conclusions—ECPR may promote survival in children with cardiac disease experiencing cardiac arrest unresponsive to conventional CPR with favorable early neurological outcomes. CPR duration was not associated with mortality, whereas patients with metabolic acidosis and noncardiac structural or chromosomal anomalies had higher mortality.
Survival with conventional cardiopulmonary resuscitation (CPR) after in-hospital cardiac arrest (CA) is poor in the pediatric population.1,2 The use of extracorporeal membrane oxygenation (ECMO) to support failed conventional CPR (ECPR) has been shown to rescue some patients who would face imminent mortality without this intervention.3–7 The American Heart Association 2005 Pediatric Advanced Life Support guidelines recommend the consideration of ECPR in patients when the conditions leading to the arrest are thought to be potentially reversible or amenable to heart transplantation. The use of ECPR has been increasing in adult and pediatric populations.8
Recent series report survival to hospital discharge for patients supported with ECPR to be variable, with a range of 33% to 55%.4–7 This variability in survival outcomes may be related to differences in ECPR deployment systems, variability in indications for ECPR, and postdeployment care among reporting institutions. A clear population of patients who are best suited for ECPR has not been identified. In addition, neurological outcomes in ECPR survivors are rarely reported. Given the interest and recommendation for the use of ECPR and the considerable institutional resources needed to run and sustain an ECPR program, it is important to critically review outcomes and try to identify factors associated with mortality to help guide patient selection for ECPR.
In this large, single institution study, we report on the survival to hospital discharge after ECPR used to support children with cardiac disease at Children’s Hospital Boston. We also describe the structure of our ECPR program, report on predictors of mortality in our ECPR patients, and describe early neurological outcomes for ECPR survivors.
Subjects and Methods
We conducted a retrospective review of the Children’s Hospital Boston ECMO database during 1995 to 2008 and selected all patients with cardiac disease (congenital and acquired) supported with ECMO for CA after failed conventional CPR therapies. Review of patient medical records was approved by the Children’s Hospital Committee on Clinical Investigation and the need for informed consent was waived. For purposes of this study, we defined ECPR as ECMO instituted during active CPR with chest compressions. Patients (n=8) who had return of spontaneous circulation during resuscitation but were deemed to be unstable and requiring ECMO support immediately (<30 minutes) after return of spontaneous circulation were also included in this cohort.
Rapid Response ECPR Program at Children’s Hospital Boston
Children with cardiac disease who experience an in-hospital CA that does not respond to conventional CPR techniques are eligible for ECMO support. The decision to deploy ECPR is made by the patient’s primary bedside physician based on the patient’s diagnosis, etiology of the event, and likely response to resuscitation, and it is usually initiated if there is failure of return of spontaneous circulation after 2 rounds of resuscitation drugs. Contraindications for ECPR include existing irreversible neurological injury and multi-organ failure before CA, irreversible or nonoperable primary disease, extreme prematurity (<32 weeks gestation), and previous parental refusal for the use of mechanical support. The ECPR team consists of an in-house cardiac intensive care physician (pool of 13), a cardiac surgery resident or fellow (pool of 5), ECMO specialists, and the bedside nursing staff. Two crystalloid-primed (Plasma Lyte A; Baxter) ECMO circuits (1 for use in children <15 kg and 1 for those ≥15 kg) and ECMO equipment carts with necessary surgical instruments and ECMO cannulae are available at all times in the cardiac intensive care unit. The cardiac intensive care unit attending is the event manager responsible for overseeing all aspects of CPR and ensuring resource availability, whereas the surgical team cannulates the patient for ECMO. Cannulation sites may be peripheral (neck or femoral vessels) or intrathoracic, via the right atrium and aorta, depending on patient site, anatomy, and postoperative status. All ECPR patients are supported with veno-arterial ECMO at our institution and a mandatory “time-out” is completed before starting ECMO flow to insure that cannulae are connected properly. After initiation, ECMO flow is increased to achieve a goal of 100 to 150 mL/kg per minute. Sweep gases for the membrane oxygenator are adjusted to contain fraction of inspired oxygen concentration of 0.21. Controlled hypothermia is maintained at 34°C to 35°C for 48 hours to lower metabolic demands and to provide central nervous system (CNS) protection. Cannulae position are confirmed using chest x-rays and echocardiography. Anticoagulation is provided using heparin infusion to maintain an activated clotting time of 180 to 240 seconds. Decisions regarding the care of the patient administered ECMO, investigating the cause of CA, and weaning from ECMO are patient-specific. In general, patients who are unable to be weaned from ECMO because of persistent myocardial dysfunction and who have no major end-organ injury are typically evaluated by our transplant team to determine candidacy for heart transplantation. All ECMO patients receive daily consultation with a pediatric neurologist, and evaluations for neurological injury may include a combination of serial head ultrasounds, electroencephalography, and portable head computerized tomography scans. All ECPR deployment events are debriefed within 24 hours to identify issues with personnel, equipment, and the conduct of CPR to identify factors that are modifiable, to improve the quality and safety of ECPR deployment, and to decrease deployment times. ECPR survivors are referred routinely for neuro-developmental follow-up. The ECPR team at Children’s Hospital Boston is available to deploy ECMO 24 hours per day and is staffed with 9 full-time-equivalent ECMO specialists.
Data Collection and Categorization
Data collected for the purposes of this study include patient demographic information, CPR details, ECMO details, course, and complications. The primary outcome variable used for purposes of this study is survival to discharge from our institution to either home or another facility.
Patients were divided into 1 of 4 diagnostic categories, including single ventricle lesions (including any patient with only 1 pumping chamber), 2 ventricle lesions, primary myocardial disease (including cardiomyopathy and patients who have undergone transplantation), and primary pulmonary hypertension. Patients were designated as postoperative if they had undergone a cardiac surgery during the admission before CA. Noncardiac structural malformations were defined as major anomalies that were present in other organ systems that may have significantly influenced the course of the patient (eg, tracheo-esophageal fistula). The presence of a known genetic syndrome was noted. Post-ECMO blood gases represent the first arterial sample that was sent after ECMO was deployed.
Information on complications or end-organ injury that became evident during the ECMO run was also collected. Bleeding was defined as cannulation site or surgical bleeding requiring surgical exploration. Mechanical circuit complications included air emboli, thrombus formation in the circuit, breaks or leaks developing in any part or component of the ECMO circuit, and circuit or oxygenator change. Renal injury was categorized based on peak serum creatinine values >1.5 mg/dL (for all age groups) after deployment of ECPR or use of renal replacement therapy, including peritoneal dialysis, continuous veno-venous hemodialysis, or hemodialysis. CNS injury was defined by the presence of clinical injury (eg, seizures, hypoxic-ischemic encephalopathy, movement disorder) or an abnormal neurological imaging study result. Patients with hepatic injury were defined as those with a peak aspartate aminotransferase or alanine aminotransferase ≥500 IU/dL that developed in the period after deployment of ECPR. The presence of a positive blood culture was used to define infection during ECMO use. Respiratory complications included ventilator associated pneumonia, acute respiratory distress syndrome, pneumothorax, and pulmonary hemorrhage.
To assess global neurological status, we used pediatric overall performance category (POPC) and pediatric cerebral performance category (PCPC) scores.9 These scores were assigned to survivors older than age 2 months at discharge and at follow-up if there was adequate neurological information available in the medical record.
For the patients who underwent multiple ECPR runs during the same admission (n=7), only the first ECPR run was included in this analysis. Demographics, CPR variables, pre-ECMO data, ECMO support details, and ECMO complications were compared between survivors and nonsurvivors. Pearson χ2 tests were used to compare categorical data and Fisher exact tests were used when the expected count in any category was <5. Continuous data were compared using the Mann-Whitney test. For patients with multiple ECPR runs, cumulative duration of ECMO based on all ECMO runs was used. Data are shown as median values with the interquartile range or as numbers with proportion (%).
Three separate multivariable logistic regression models were developed in an effort to identify factors independently associated with mortality after ECPR. The first model explored the influence of pre-ECMO and ECMO variables, whereas the second model evaluated the association of ECMO duration and complications on mortality. Because our patient cohort spanned over the course of 13 years, and because some variables (eg, blood lactate levels) were only collected for patients in the recent cohort, we developed a third model exploring the association of pre-ECMO and ECMO factors at the time of deployment using data from patients treated with ECPR during 2000 to 2008. Variables with a univariate P<0.1 were selected for consideration in the model. A forward selection procedure was used for variable entry into the multivariable model. Variables were retained in the multivariable model if adjusted P<0.05. Continuous variables (eg, post-ECMO arterial blood pH) retained in the multivariable model were refitted as categorical variables based on quartile or quintiles values of their distribution to test the presence of a linear association with the outcome variable. If a linear increase in the odds ratio across categories could not be demonstrated, then linearity assumptions were considered violated and the variable was retained as a categorical variable.
From 1995 to 2008, 172 patients with cardiac disease underwent 180 ECPR runs (41% of all ECMO runs) at Children’s Hospital Boston (Figure 1). The median age (interquartile range) of the study population was 5.7 months (0.4, 43.6) and the median weight was 6.0 kg (3.2, 14.0). Eight patients with congenital heart disease were older than 18 years (5%). Seventy patients (41%) had single-ventricle congenital heart disease, 65 (38%) had 2-ventricle congenital heart disease, 31 patients (18%) had primary myocardial disease, and 6 (3%) had primary pulmonary hypertension. A total of 103 (60%) patients underwent ECPR after cardiac surgery. Median duration of CPR before ECMO flow for the entire cohort was 33 minutes (23, 44) and the trend decreased significantly over time (P<0.001; Figure 2). Twenty-seven of 172 patients (16%) were deemed eligible and, subsequently, were listed for transplantation, with 6 of these patients transitioned to a ventricular assist device and 13 ultimately undergoing transplantation (12 patients underwent orthotopic heart transplantation and 1 received a lung transplant for primary pulmonary hypertension). The patients listed for transplantation who did not receive an organ died (n=14). Overall, 88 patients (51%) survived to hospital discharge.
Table 1 illustrates the demographic details of survivors compared to nonsurvivors after ECPR use. There were more females and fewer children with noncardiac structural malformations among survivors compared to nonsurvivors. Notably, survivors and nonsurvivors did not differ by age, cardiac diagnosis, and year of ECPR. Details of the resuscitation and pre-ECMO deployment are outlined in Table 2. There was a significant difference in survivors and nonsurvivors when the location of CA was compared between the 2 groups. In this cohort, 129 (75%) of the CA occurred in the cardiac intensive care unit with 49% (n=63) of the patients surviving to discharge, patients who required ECPR in our catheterization laboratory (n=27) had a survival to discharge rate of 73%, and patients whose CA occurred in other locations (inpatient ward, emergency department, and operating room) had a survival to discharge rate of 35%. Duration of CPR, initial rhythm, and medications used during CPR were not significantly different between survivors and nonsurvivors.
ECMO and Factors After CA
Table 3 outlines the differences between survivors and nonsurvivors of factors after CA and ECMO management. There was no survival difference between patients who were cannulated in the neck as compared to those who underwent chest cannulation. There were more survivors among patients who underwent decompression of the left atrium compared to those who did not. A smaller number of patients went on to ECMO with a blood prime (n=11) compared to crystalloid (n=161), and 82% (n=9) of these patients died compared to only 47% (n=75) of patients with a crystalloid-primed ECMO circuit (P=0.03). Arterial blood pH and standardized bicarbonate levels obtained after establishing ECMO flows were higher in survivors than in nonsurvivors. In the cohort of patients supported with ECPR after January 1, 2000, peak lactate level after ECMO deployment was lower in survivors compared to nonsurvivors.
Eighty-three (48%) patients underwent an interventional procedure either in the operating room or in the catheterization laboratory after deployment of ECPR. The most common intervention performed after ECPR was manipulation of the systemic to pulmonary shunt (n=20) for either presumed thrombosis or pulmonary overcirculation. The performance of an additional procedure after ECPR deployment did not vary between survivors and nonsurvivors. The duration of ECMO was shorter in survivors compared to nonsurvivors. The survival rate of 29% for the 7 patients who needed >1 ECPR run was lower but not significantly different than the 52% survival rate for patients supported with 1 ECPR run (Fisher exact P=0.29).
The incidence of complications occurring during the ECMO run in survivors and nonsurvivors are shown in Table 4. The incidence of ECMO circuit complications, respiratory complications, sepsis, CNS complications, and end-organ dysfunction were significantly higher in nonsurvivors compared to survivors. Eighty-nine patients in the cohort (52%) had clinical or radiological evidence of CNS injury.
Multivariable Logistic Regression Models Predicting Mortality for ECPR Patients
Multiple models were constructed to evaluate predictors of in-hospital mortality in ECPR users as shown in Table 5. The first model shows the association of mortality with demographic variables and ECMO factors. Noncardiac structural anomalies, use of blood-primed ECMO circuit for ECPR deployment, and post-ECPR deployment arterial blood pH <7.00 compared to a pH ≥7.380 were significantly associated with higher mortality. The second model evaluated the association of ECMO duration and complications and mortality. Longer ECMO duration and injury to the CNS, liver, or kidney were significantly associated with increased mortality. The need for longer ECMO support was associated with increased mortality odds compared to ECMO duration of ≤49 hours. The third multivariable model using data from the most recent cohort of patients showed that higher peak lactate levels (>13 mmol/L) after ECPR deployment was significantly associated with higher mortality. Arterial blood pH was not retained in the model.
Neurological Outcomes Among Survivors
POPC/PCPC scores were assigned to 87 of 88 survivors at hospital discharge or at follow-up after discharge. Sixty-five survivors (75%) had POPC scores ≤2, and 69 (79%) had PCPC scores ≤2, indicating no to mild neurological injury. Neurological scores for single-ventricle patients were similar to the remaining survivors with 2-ventricle circulation. Seventy-one percent (26/35) of the single-ventricle patients had POPC scores ≤2 compared to 77% (40/52) of the remaining survivors with 2-ventricle circulations. Among ECPR survivors who had a cavopulmonary connection (n=7), 4 (57%) had moderate to severe CNS impairment (POPC/PCPC score ≥3). The patients assigned scores in follow-up (n=63) compared to those assigned scores at discharge (n=24) were similar in baseline characteristics and in their median POPC (P=0.67) and PCPC (P=0.77). Based on the patients seen in follow-up at a median duration of 25 months (11, 62), 80% of our ECPR survivors to hospital discharge are alive at 5 years.
The use of rapid-response ECMO to support failed conventional CPR in our institution rescued 51% of the patients who would have most likely died, and 75% of these survivors had no or mild early neurological impairment. Despite a heterogeneous compilation of cardiac disease ranging from single-ventricle physiology to primary myocardial disease, there were no significant differences in mortality based on the underlying cardiac diagnosis. Children with significant chromosomal or noncardiac structural anomalies and patients who have severe metabolic acidosis based on arterial blood gas pH or blood lactate levels immediately after ECMO deployment had increased mortality. Another factor that was predictive of increased mortality in our cohort was the use of a blood-primed ECMO circuit. We believe that use of a blood-primed circuit may indicate severity of patient illness before ECPR deployment and that an ECMO circuit was being prepared for deployment to support significant refractory cardiorespiratory dysfunction before the CA. Conversely, it may also be a surrogate for a complicated and prolonged cannulation procedure. As reported in previous studies, need for longer duration of ECMO support and evidence of end-organ injury with ECMO were associated with increased mortality.4
The overall survival rate of 51% for ECPR compares favorably to the overall survival for pediatric cardiac patients administered ECMO as reported by the Extracorporeal Life Support Organization (42%)10 and by other smaller single-center studies.5,6 Optimizing CPR, early deployment of ECPR, and refining a system for rapid deployment that includes immediate staff and resource availability are important factors contributing to our improved survival and outcomes. This is evidenced by the continued reduction in the duration of CPR from the onset of CA to the establishment of ECMO flows as our experience has grown over the years (Figure 2). However, despite decreases in CPR duration, our survival rates have not improved with time. Our analysis, as well as those of others, has shown that CPR duration may not influence eventual survival.5 We showed that arterial blood pH after ECPR deployment and peak lactate levels within 72 hours after ECMO are strongly associated with increased mortality. This association may reflect those patients who had severely compromised circulation before CA and thus may have benefited from earlier introduction of ECMO support before CA, those who did not receive adequate CPR, and those who were not adequately supported with ECMO soon after their CA. Although speculative, it appears that patient selection for ECPR and the quality of CPR are more likely to influence mortality than CPR duration. In our study, the efficacy of CPR could not be evaluated because of the heterogeneity of locations, personnel, and monitoring equipment present, but it should be an important focus of future research to improve outcomes for ECPR users.
The Extracorporeal Life Support Organization study of pediatric cardiac patients requiring ECPR found that single-ventricle physiology, history of stage 1-type palliation, and arterial pH <7.01 were associated with increased mortality, whereas right carotid artery cannulation was associated with improved survival.7 We found similar results in regard to metabolic acidosis after the initiation of ECMO, but cannulation site did not predict mortality. Chest cannulation allows for easier access to the vasculature, but open cardiac massage is frequently interrupted, reducing the effectiveness of CPR. We speculate that it is this conflict between administering ECMO quickly and delivering uninterrupted CPR that creates similar mortality rates between patients who undergo chest and neck cannulation. There have been a number of studies with contrary survival outcomes regarding single-ventricle patients supported with ECPR. Chan et al7 found lower survival rates for single-ventricle patients, whereas a smaller case series found single-ventricle patients had a 47% survival rate compared to 27% of 2-ventricle lesions.6 Our data show similar survival rates between single-ventricle and 2-ventricle lesions of 51% and 48%, respectively. The subset of single-ventricle patients with cavopulmonary anastomoses on ECMO has not been well-studied in the literature. Booth et al11 evaluated 20 patients from Children’s Hospital Boston with either bidirectional Glenn or Fontan circulation who were supported by ECMO and found that the survival rate was 30%. Seven of our 16 patients (44%) with cavopulmonary anastomoses survived to hospital discharge, with a tendency to have more severe neurological injury than shunted single-ventricle patients in this cohort. Passive systemic venous return as a source of pulmonary blood flow is vulnerable to conventional CPR because the increased intrathoracic pressure with chest compressions can actually restrict pulmonary blood flow and increase systemic venous pressure. Thus, we suggest the presence of single-ventricle lesions alone may not be used as a contraindication for ECPR deployment. These patients are at higher risk for CA because of their physiology and are less likely to respond to conventional CPR therapies.
Location of the CA was not predictive of mortality in our series despite our belief that the cardiac catheterization laboratory is often an ideal location for ECPR because of a higher level of invasive patient monitoring that can help detect deterioration quickly and the presence of invasive vascular access that can be used to rapidly place percutaneous ECMO cannulae. The finding that a similar number of survivors and nonsurvivors underwent a procedure after CA demonstrates that CA in these patients may be related to unrecoverable myocardial dysfunction despite the perception that a “correctable” anatomic lesion may be present. Early listing for cardiac transplantation must be considered in some of these patients as an optimal rescue strategy.
Neurological injury has been well-documented as a complication of ECMO and has been reported to be higher in patients undergoing ECPR.12 Studies from the Extracorporeal Life Support Organization registry have used the incidence of brain death, brain infarction, and hemorrhage to define the frequency of CNS injury, but they lack functional assessments and follow-up evaluation of neurological impairment.12–13 Utilizing the Extracorporeal Life Support Organization registry, Barrett et al13 reported a 22% incidence of CNS injury in ECPR patients. Patients with less severe metabolic acidosis before ECMO and an uncomplicated ECMO course were more likely to avoid neurological injury.13 Our series had a higher incidence of CNS injury (52%). This may have resulted from the use of broader clinical and radiological criteria to define neurological injury, because we wanted to capture all neurological insults to determine if they predicted functional impairment during follow-up. Our use of POPC/PCPC metric of functional status is limited because it lacks a detailed objective assessment of the patient’s true neurological status. However, it has been shown to accurately predict performance in more rigorous psychometric testing.9
This study has several limitations. The retrospective observational design precluded collection of important predictors that may have influenced our outcome. In addition, the generalizability of these findings is limited because of the heterogeneity of our study population and the report of a single-center experience with ECPR. Neurological assessment would have been more robust if our survivors were evaluated by standardized neuro-developmental testing rather than POPC/PCPC categorization. Furthermore, 28% of our survivors were lost to follow-up, and assignment of POPC/PCPC scores at discharge limits our ability to truly assess functional impairment and overall capability. However, the similar baseline characteristics and median POPC/PCPC scores between the patients assigned scores in follow-up and at discharge suggest that there is likely little or no bias in this distribution. The POPC/PCPC measures also tend to not capture the behavioral and attention issues that these patients may be at severe risk for. Our neurological outcomes are only descriptive, and there needs to be a primary outcome variable in future studies to determine predictors of favorable outcomes.
The 51% overall survival and the preserved long-term survival without severe early neurological impairment demonstrate that ECPR is a useful modality to rescue children with cardiac disease who experience an in-hospital CA that does not respond to conventional CPR therapies. The presence of a structured ECPR program with continuous quality improvement can help decrease deployment times. Patients with noncardiac structural and chromosomal anomalies may not be suitable for ECPR because their risk of mortality is higher. The presence of significant metabolic acidosis increases risk of mortality and future studies on ECPR should focus on methods to improve quality of CPR and early management after ECPR to help provide adequate support to these patients. Although early crude indicators suggest preserved neurological function in ECPR survivors, CNS injury remains high and future studies should also focus on reducing the burden of neurological injury in these patients. An active transplant and mechanical support program is necessary because 11% of our survivors underwent successful heart and lung transplantation. A pediatric cardiovascular program that performs surgical and interventional catheterization procedures should consider developing ECPR capabilities to best serve their patients.
Ravi R. Thiagarajan received an honorarium from Seattle Children’s Hospital for a lecture on ECPR and served as a consultant to the ECMO program at Children’s National Medical Center.
Presented at the 2009 American Heart Association meeting in Orlando, Fla, November 14–18, 2009.
del Nido PJ, Dalton HJ, Thompson AE, Siewers RD. Extracorporeal membrane oxygenator rescue in children during cardiac arrest after cardiac surgery. Circulation. 1992; 86: 11300–11304.
Thiagarajan RR, Laussen PC, Rycus PT, Bartlett RH, Bratton SL. Extracorporeal membrane oxygenation to aid cardiopulmonary resuscitation in infants and children. Circulation. 2007; 116: 1693–1700.
Alsoufi B, Al-Radi OO, Nazer RI, Gruenwald C, Foreman C, Williams WG, Coles JG, Caldarone CA, Bohn DG, Van Arsdell GS. Survival outcomes after rescue extracorporeal cardiopulmonary resuscitation in pediatric patients with refractory cardiac arrest. J Thorac Cardiovasc Surg. 2007; 134: 952–959.
Chen YS, Lin JW, Yu HY, Ko WJ, Jerng JS, Chang WT, Chen WJ, Huang SC, Chi NH, Wang CH, Chen LC, Tsai PR, Wang SS, Hwang JJ, Lin FY. Cardiopulmonary resuscitation with assisted extracorporeal life-support versus conventional cardiopulmonary resuscitation in adults with in-hospital cardiac arrest: an observational study and propensity analysis. Lancet. 2008; 372: 554–561.
Fiser DH, Long N, Roberson PK, Hefley G, Zolten K, Brodie-Fowler M. Relationship of pediatric overall performance category and pediatric cerebral performance category scores at pediatric intensive care unit discharge with outcome measures collected at hospital discharge and 1- and 6- month follow-up assessments. Crit Care Med. 2000; 28: 2616–2620.
Alsoufi B, Al-Radi OO, Gruenwald C, Lean L, Williams WG, McCrindle BW, Caldarone CA, Van Arsdell GS. Extra-corporeal life support following cardiac surgery in children; analysis of risk factors and survival in a single institution. Eur J Cardiothorac Surg. 2009; 35: 1004–1011.