PURPOSE: The late effects of treatment with extracorporeal membrane oxygenation (ECMO) in nonneonatal pediatric patients remain unclear. The aims of our study were to better characterize the long-term survival and hospital readmission rates for pediatric patients after ECMO treatment.
PATIENTS AND METHODS: From 1999 to 2006, data on children aged 1 month to 18 years who underwent ECMO were extracted from the California Patient Discharge Database. Data from patients with diagnoses of congenital cardiac disease were excluded. We analyzed patient data on initial hospital course, subsequent readmissions, development of long-term morbidities, and long-term survival.
RESULTS: The study cohort consisted of 188 children from 13 California hospitals. The median age was 3 years, and 46% of the patients survived to hospital discharge. ECMO indications included acquired heart disease in 81 patients, pneumonia in 56, other respiratory failure in 22, sepsis in 8, trauma in 8, and other indications in 12. Of the 87 survivors, 56 were tracked for a median period of 3.7 years. The readmission rate was 62%, and the mean time to first readmission was 1.2 years. Readmissions for respiratory infections were observed in 34% of the patients and for reactive airway disease in 7%. Neurologically debilitating conditions (epilepsy [7%] and developmental delay [9%]) occurred in 16%. Late deaths occurred in 5% of the children. Readmitted survivors had a cumulative length of readmission hospitalization of 8 days and hospital charge of $43 000. Cox proportional hazard regression demonstrated a positive relationship between treatment-center case volume and long-term survival outcomes (hazard ratio: 0.98 per case; P < .01).
CONCLUSIONS: Pediatric ECMO survivors suffered from significant long-term morbidities after initial hospital discharge. More than 60% of these children required subsequent readmissions, and late deaths were observed in 5%. Furthermore, hospitals with high case volumes were associated with improved long-term survival.
WHAT'S KNOWN ON THIS SUBJECT:
Pediatric ECMO is a rarely performed, potentially life-saving procedure used for severe respiratory failure. Experience with this procedure has been limited and associated with high rates of morbidity and mortality.
WHAT THIS STUDY ADDS:
No known assessments have investigated outcomes related to long-term survival, morbidity, and hospitalization after ECMO in nonneonatal pediatric patients. This 8-year population-based longitudinal study was the largest of its kind to investigate the ECMO procedure in this patient population.
Extracorporeal membrane oxygenation (ECMO) has become an accepted rescue therapy for children with acute cardiopulmonary failure. Since ECMO was first introduced in the 1970s, nearly 29 000 patients worldwide have benefited from this technology.1,2 Recently, the indications for ECMO have gradually expanded so that this treatment is used in patients who would have been excluded in the past, such as those suffering from traumatic injuries as well as children with multiple organ-system failures.1 Although overall survival rates are favorable after ECMO, the long-term outcomes for these children are less clear.3
In neonates, long-term studies have shown that ECMO survivors may suffer from significant respiratory and neurologic morbidities after hospital discharge.4,–,14 ECMO treatment has also been associated with behavioral problems and poor school performance in survivors without significant neurologic complications.7 For these neonatal ECMO survivors the rate of disability stabilized at ∼15% at age 3 years and seemed to be associated with the child's pre-ECMO mortality risk and degree of prematurity.15
In contrast to neonatal ECMO, only a handful of studies have investigated the long-term outcomes of ECMO in the older pediatric population. The majority of these studies focused on children who had congenital heart disease and children who had undergone cardiac surgery.16,–,20 In comparison, only a few studies have evaluated the long-term outcomes of ECMO for other indications in this older age group.16,21 Furthermore, the effect of hospital case volume on the outcomes of pediatric ECMO has not been well documented despite the regionalization of pediatric ECMO to tertiary specialty hospitals.
In this study we examined the long-term outcomes of nonneonatal pediatric ECMO from a population-based perspective. The main outcome measures were hospital readmissions and development of long-term morbidities after ECMO. Long-term survival was also analyzed with respect to case volume at treatment centers that provide ECMO.
We extracted patient cohort data from the California Patient Discharge Database (CaPDD) after we obtained approvals from the institutional review boards of both the University of California Los Angeles and the California Office of Statewide Health Planning and Development. The CaPDD contains annual data on inpatients discharged from all nonfederal hospitals licensed in California. The recorded data contain a discharge abstract for each inpatient hospitalization and include patient demographic, admission, and discharge details; and up to 24 diagnoses and 20 procedure codes.22 All diagnoses and procedures were coded according to the International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9-CM) standards.23 The Office of Statewide Health Planning and Development internally validates individual records in the CaPDD through 9 levels of checkpoints with an error-tolerance level of <2%.24 Most important is that individual records were flagged and records that contained illogical or erroneous relationships between ICD-9-CM diagnosis and procedure codes were reviewed. This database has been used previously to examine health outcomes in both adult and pediatric populations.25,–,27
From 13 of the 516 licensed California hospitals, we extracted data from the CaPDD on children between 1 month and 18 years old who had undergone ECMO (ICD-9-CM code 39.65) during the period from 1999 to 2006. Children with a diagnosis of congenital heart disease were excluded because of the complex nature of their disease and treatment. For each child the first hospitalization for which the record contained the ECMO procedure code was identified as the index hospitalization. Subsequent hospital readmissions and procedures were tracked for each child by means of a unique patient identifier.
Definition of Variables
The indications for ECMO were categorized according to the admission and discharge diagnoses from the index hospitalizations, with particular focus on the principal-diagnosis code. Baseline neurologic status and the presence of reactive airway disease were established from all records before and during the index ECMO hospitalization. Also identified were cases that involved later development of neurologically debilitating conditions and/or reactive airway disease. Additional procedures performed during the index hospitalization as well as during readmissions were identified by using ICD-9-CM codes. In-hospital deaths that occurred during readmissions were characterized as “late deaths,” and individual hospital case volumes for nonneonatal pediatric ECMO during the study period were determined for individual hospitals by using deidentified hospital codes in the database.
Bivariate statistical comparisons were made by using the χ2 test for categorical variables and the Wilcoxon-Mann-Whitney test for continuous variables. Because of the skewed distribution of the continuous variables, the nonparametric Wilcoxon-Mann-Whitney test was used. Kaplan-Meier survival curves were generated for long-term survival after pediatric ECMO, and Cox proportional-hazard multivariate regression was used to determine the hazard ratio of survival, which was controlled for ECMO indications, hospital case-volume status, and cases in which the child was transferred from an outside hospital. By using the cluster option in Stata (Stata Corp, College Station, TX), we performed nonhierarchical cluster analysis to take into account patient clustering within individual hospitals. Observations were right censored at the end of the study period (December 31, 2006) or when the patient reached the age of 18 years. Statistical analyses were performed by using SAS 9 (SAS Institute, Inc, Cary, NC) and Stata/IC 10.1 (Stata Corp).
The CaPDD included data on 1313 children who had undergone ECMO during the 8-year study period (1999–2006). After we excluded 896 neonatal patients who received ECMO and 229 patients with congenital heart disease who received ECMO, data for 188 nonneonatal pediatric patients who received ECMO were included in the study cohort. The median age at the time of ECMO was 3 years, 48% of the patients were girls, 55% of the children had public insurance, and 44% were transferred from an outside hospital. There were no differences in patient characteristics, ECMO indications, and initial hospital outcomes between the patients for whom long-term outcome data were available compared with those for whom such data were not available. The indications for ECMO consisted of acquired heart disease in 43% of patients, pneumonia in 30%, other respiratory failure in 12%, sepsis in 4%, trauma in 4%, and other indications such as asphyxia, drowning, and ingestion of a foreign body in 7%. A total of 13 different California hospitals provided ECMO treatment to the study patients, with a mean hospital case volume of 14 nonneonatal pediatric ECMO (range: 1–43, SD: 12) during the 8-year period.
Initial Hospital Outcomes of Pediatric Patients Who Received ECMO
The median length of stay for the pediatric ECMO patients during the index hospitalization was 23 days, with a median hospital charge of $423 000. Renal failure along with the need for dialysis and hemofiltration was present in up to 19% of the study cohort. The survival-to-initial discharge rate was 46%. This initial survival rate was similar to that reported in the Extracorporeal Life Support Organization (ELSO) registry data for nonneonatal pediatric ECMO (46% vs 49%; P = .29). Although the percentage of patients who survived to discharge from the index hospitalization ranged from 25% to 52% for different ECMO indications, these differences were not statistically significant (P = .65). Figure 1 shows a comparison of the survival rates of nonneonatal pediatric ECMO between our patient cohort and data from the ELSO registry.
Long-term Outcomes of Pediatric ECMO Survivors
There were 56 ECMO survivors whose long-term outcomes were assessed for a median period of 3.7 years (range: 1–8 years). More than one half of these children (62%) required subsequent readmissions after initial discharge. The median time to readmission was 1.2 years. Children who were readmitted required a median of 2 readmission episodes, with a median cumulative hospital stay of 8 days and a cumulative hospital charge of $43 000.
In 16% of the children who survived ECMO, neurologically debilitating conditions occurred, consisting of developmental delay in 9% and epilepsy in 7%. Reactive airway disease occurred in 7% of the ECMO survivors. Subsequent respiratory infections developed in 34% of the children after hospital discharge and necessitated readmissions for treatment.
Late deaths occurred in 5% of the study patients and were observed only in children with acquired heart disease and who also required cardiac transplantation in addition to ECMO during the initial hospital admission. These late deaths occurred on days 40, 565, and 1147 after initial ECMO. The overall long-term survival rate for nonneonatal pediatric ECMO in our cohort was 45%. Figure 2 illustrates the unadjusted long-term survival rates according to hospital case volumes. Cox proportional-hazard multivariate regression showed a positive case-volume–outcome relationship after we controlled for differences in ECMO indications, patient age, and whether patients were transferred from an outside facility (hazard ratio: 0.98 per case; P < .01; Table 1). When we compared differences in patient survival rates between hospitals with a mean ECMO case volume of 14 cases and hospitals with a case volume 1 SD above the mean (26 cases), we observed a 1.34-times higher likelihood of patient survival with increased hospital case volume. Figure 3 demonstrates the Kaplan-Meier survival curve of this difference in long-term survival associated with the difference between the ECMO case volumes of the 2 hospitals. Other factors in the model did not seem to affect long-term survival.
From this large population-based cohort of nonneonatal pediatric patients without congenital cardiac disease, we found that the long-term survival rate was 45%, and late deaths after initial discharge occurred only in children with acquired heart disease. These ECMO survivors suffered from significant long-term morbidities as indicated by hospital readmissions. More than 60% of the nonneonatal pediatric ECMO survivors required hospital readmissions. Survival analysis also demonstrated a definitive positive relationship between hospital ECMO case volume and outcomes for long-term survival after pediatric ECMO, for which higher hospital case volumes were associated with improved long-term survival after we controlled for ECMO indications and other patient factors.
We identified 3 important factors when we assessed the outcomes of these pediatric patients who received ECMO. First, long-term survival after nonneonatal pediatric ECMO has not been studied beyond small single-institution case series. Recent data from the ELSO registry demonstrated a decreasing overall short-term survival rate for patients treated with ECMO compared with previous analysis.1 This decrease was attributed in part to recent advances in pediatric and neonatal critical care, such as inhaled nitric oxide and high-frequency oscillatory ventilation, which have led to alternative therapies for effective treatment of respiratory failures. Children in whom treatment with these less-invasive therapies was successful would not need ECMO support, and patients treated with ECMO after multiple alternative therapies failed may have been “sicker” than patients treated with ECMO before these alternatives were available. The majority of this recent decrease in survival rate has been attributed to the neonatal population, although short-term survival rates for nonneonatal pediatric patients who received ECMO have actually improved slightly compared with previously reported rates.1 Children in our cohort had similar short-term survival rates compared with nonneonatal pediatric patients in the ELSO registry who received ECMO.1 Because the CaPDD allows longitudinal tracking of these patients, we were able to establish a population-based long-term survival rate for nonneonatal pediatric ECMO that had not been described previously. The long-term survival rate for these children in our study was 45%, and late deaths were rare. Late deaths after initial discharge occurred in only 5% of the surviving children, all of whom had undergone heart transplantation for acquired heart disease. These late deaths likely reflected the morbidities of their underlying conditions and not necessary complications from ECMO therapy. On the other hand, late mortality did not occur in children with other ECMO indications. In contrast to neonates and children with heart disease who suffered late mortality as a consequence of their underlying illness,2,12,13 older children who required ECMO for other indications such as traumatic injury or severe invasive pneumonia were generally healthy before their acute-illness episode. For these children the risk of death after recovery from their acute illness requiring ECMO would be low, a theory that was supported by our findings.
A second factor we identified was that outcome assessment in nonneonatal pediatric ECMO survivors must also consider long-term morbidity. Despite sustained survival, these children suffered from significant morbidities after their ECMO treatment. In 188 near-term infants who had ECMO for neonatal respiratory failure, Schumacher et al10 found that 31% required hospital readmission within 12 months. In that study population, respiratory infections and wheezing-associated respiratory illness were the most common indications for hospital readmissions. In the same study, patient follow-up revealed the occurrence of a significant rate of neurologic complications and growth failures. Results of the UK collaborative ECMO trial revealed moderate-to-severe disability in 13% of the children who were assigned to the ECMO group when they were 4 years old.13 Similar long-term morbidities were observed in children who received ECMO support after cardiac surgery.17,20 Ibrahim et al17 found that although 80% of the patients who received cardiac surgery and survived ECMO reported their health to be good or excellent, follow-up examinations revealed moderate-to-severe neurologic impairment in >60% of them. Our data showed that similar rates of long-term morbidity occurred in nonneonatal pediatric patients who received ECMO compared with neonates and children with congenital cardiac disease. Children in our cohort also suffered from significant neurologic and respiratory conditions after discharge. Neurologically debilitating conditions occurred in 16% of these children, and 41% had long-term respiratory sequelae. More than 60% of the nonneonatal pediatric patients who survived ECMO required hospital readmissions, and the cumulative hospital stays and monetary costs after initial discharge were substantial.
The third important factor we identified was a positive relationship between ECMO case volume of the treatment center and outcomes in nonneonatal pediatric patients treated with ECMO. Although increased surgical-case volume has been shown to influence the outcomes of some adult and pediatric surgical procedures,28,–,30 this relationship had not been previously established for pediatric ECMO. To our knowledge, ours it the first study to have compared outcomes after pediatric ECMO across different hospitals. Despite the existing regionalization of pediatric ECMO in California, where only 13 hospitals performed pediatric ECMO during the study period, we still observed a definitive relationship between hospital case volume and the outcomes in these patients. Our multivariate-analysis results indicated that the likelihood of long-term survival was increased by 1.34 times with a 1-SD increase of hospital case volume above the mean hospital case volume. In our study, similarly to other studies, hospital case volume was defined as a continuous variable because this definition enable us to avoid the pitfalls of deciding a priori what constitutes “small” or “large” volumes.31 Nonetheless, even if hospital case volume was dichotomized at the mean hospital volume of 14 cases, we still found that hospitals with a high ECMO case volume were associated with a 1.6-times higher likelihood of long-term survival rates in nonneonatal pediatric patients after ECMO compared with hospitals with low case volumes. This difference in long-term survival rate also seemed to occur early in the course of treatment. ECMO is an extremely resource-intensive technology, because a multidisciplinary team of physicians, nurses, perfusionists, and ECMO specialists are required to provide optimum support for a child being treated with ECMO. Hospitals that perform a higher volume of ECMO treatments are more likely to have a more cohesive and efficient ECMO team as well as established clinical care protocols. We believe that higher hospital case volume may have been a proxy for improved coordination and more effective care, which led to increased long-term survival. Although both of these aspects of care may lead to improve outcomes, their exact effect on the care of pediatric patients who receive ECMO remains unknown. Future studies that focus on the impact of hospital characteristics on long-term survival can help to provide answers to these questions.
Other factors may also have contributed to the observed positive relationship between case volume and outcome in pediatric patients who received ECMO. We observed a difference in the underlying distribution of transferred versus nontransferred patients that was related to differences in hospital case volumes. Differences in patients transferred from an outside hospital compared with those who were not may have reflected differences in disease severity and affected survival rates at different hospitals. In addition, different indications for performing ECMO treatment may also affect long-term survival. In our model, however, long-term survival rates showed a persistent positive relationship between case volume and outcome even after we adjusted for patient age, whether the patient was transferred from an outside hospital, and patient indications for ECMO.
Similarly to other studies that used an administrative database, our study had a few limitations. Administrative data such as those obtained from the CaPDD may lack important clinical information not accounted for by ICD-9-CM coding, and observational data may not capture potential sources of bias arising from patient-selection factors. The indications for ECMO were extrapolated from admission and discharge diagnoses. Furthermore, details of each survivor's neurologic status were limited to those coded at the time of hospital readmission. Despite these limitations, however, this report of our study results makes several unique contributions to the literature, including defining the long-term outcomes for nonneonatal pediatric ECMO in a patient cohort from a broad population base. By using multivariate survival analysis adjusted for other patient factors, we also identified for the first time a definitive relationship between case volume and outcome for pediatric ECMO. Finally, our study results demonstrated robust findings from a large (12% of the US population), diverse population that included all nonneonatal pediatric patients undergoing ECMO. Thus, we avoided biases that might have arisen if data collection had been limited to a single institution.
Although ECMO has been shown to be efficacious and cost-effective compared with conventional therapy,13,32 the ECMO-treated children in our study suffered from significant long-term morbidities. Our findings in children without congenital heart disease were comparable to those reported in the neonates and children with heart disease. Our study patients suffered from long-term neurologic and respiratory conditions that frequently required hospital readmissions.
A relationship between hospital case volume and long-term survival in patients who undergo ECMO was also demonstrated in our study. This relationship highlights the importance of continued ECMO performance monitoring at ECMO centers so that optimum care is provided to those children who require ECMO support. Future studies are needed to better determine the impact of these late effects on the health of nonneonatal pediatric ECMO survivors, as well as to define hospital characteristics that are associated with improved long-term survival.
- Accepted February 8, 2010.
- Address correspondence to Stephen B. Shew, MD, Division of Pediatric Surgery, Mattel Children's Hospital, David Geffen School of Medicine at UCLA, 10833 Le Conte Ave, CHS Building, MC 957098, Los Angeles, CA 90095-7098. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
- ECMO =
- extracorporeal membrane oxygenation •
- CaPDD =
- California Patient Discharge Database •
- ICD-9-CM =
- International Classification of Disease, Ninth Revision, Clinical Modification •
- ELSO =
- Extracorporeal Life Support Organization
- Schumacher RE,
- Palmer TW,
- Roloff DW,
- et al
- Hamutcu R,
- Nield TA,
- Garg M,
- et al
- 12.↵UK Collaborative ECMO Group. The collaborative UK ECMO (Extracorporeal Membrane Oxygenation) trial: follow-up to 1 year of age. Pediatrics. 1998;101(4). Available at: www.pediatrics.org/cgi/content/full/101/4/e1
- Van Meurs K,
- Lally KP,
- Peek G,
- Zwischenberger JB
- Glass P,
- Brown J
- Van Meurs K,
- Lally KP,
- Peek G,
- Zwischenberger JB
- Dalton HJ,
- Day SE
- Hamrick SE,
- Gremmels DB,
- Keet CA,
- et al
- 22.↵State of California Office of Statewide Health Planning and Development. Patient Discharge Data File Documentation January-December 2006. Available at: www.oshpd.ca.gov/HID/Products/PatDischargeData/PublicDataSet/index.html. Accessed November 5, 2008
- 23.↵US Department of Health and Human Services. International Classification of Disease, Ninth Revision, Clinical Modifications (ICD-9-CM). Hyattsville, MD. National Center for Health Statistics. Available at: www.cdc.gov/nchs/icd/icd9cm.htm. Accessed March 22, 2010
- 24.↵Medical Information and Reporting for California. The Basics of MIRCal Edit Programs for Inpatient, Emergency Department, and Ambulatory Surgery DataSacramento, CA: Office of Statewide Health Planning and Development, 2006. Available at: www.oshpd.cahwnet.gov/HID/MIRCal/Text_pdfs/Bulletins/QuickNotesV16.pdf. Accessed March 22, 2010
- Schmitt SK,
- Sneed L,
- Phibbs CS
- Burgos AE,
- Schimitt SK,
- Stevenson DK,
- et al
- Bazzani LG,
- Marcin JP
- Petrou S,
- Edwards J
- Copyright © 2010 by the American Academy of Pediatrics