The NIS hospital discharge database for the period 2001-2011 was obtained from the HCUP of the Agency for Healthcare Research and Quality. This database represents 20% of all inpatient discharges from nonfederal hospitals in the United States. Patients included in this study were selected using the International Classification of Diseases 9th Revision (ICD-9) procedure code for extracorporeal membranous oxygenation, 3965.

The demographic information analyzed included age and gender. Age was categorized into neonates (<1 month), infants (1 month to 1 year), children (1-17 years), adults (18-64 years), and seniors (≥65 years). These age groups were similar to those in the Extracorporeal Life Support Organization (ELSO) registry, with the exception that adults were defined as aged ≥18 years in the present study and a separate group of patients aged ≥65 years was also included.8

Comorbidities and complications included in this study were congenital heart disease [clinical classification software code (CCS): 213], sepsis (CCS: 2), cardiac arrest (CCS: 107), shock (CCS: 249), respiratory distress syndrome (CCS: 221), and respiratory failure, insufficiency, arrest (CCS: 131). The prevalence of these comorbidities among ECMO patients in each age group was studied. The Charlson Comorbidity Index was calculated for each patient.9

The primary endpoints of this study were the following neurologic complications: ICH (ICD-9: 430, 431, 4320, 4321, 4329), acute ischemic stroke (ICD-9: 433.X1, 434.X1, 436, 437.0, 437.1, 437.4, 437.5, 437.7, 437.8, and 437.9), and seizure (CCS: 83). The discharge status (discharge to home or discharge to long-term facility), in-hospital mortality, length of stay, and 2011-adjusted hospitalization costs of patients with neurologic complications with patients who did not suffer neurologic complications were compared. Rates of neurologic complications were also compared across age groups and comorbidities/indications. A separate analysis of the rate of neurologic complications was also performed by age group, and trends in the utilization of ECMO during the period 2001-2011 were examined.

Chi-squared tests were used to compare categorical variables, and the Student's t-test was used to compare continuous variables, assuming a cutoff for statistical significance of p<0.05. Probability values for outcomes were obtained using the no-neurologic-complications group as a reference group. Discharge weights were applied to make the data nationally representative. Two multivariate analyses were performed; one was designed to determine the comorbidities and demographic characteristics associated with neurologic complications (stroke, ICH, or seizure) and in-hospital mortality; all comorbidities, indications, and demographic factors were forced into this model. Except where stated otherwise, the data are presented as mean±SD values. All statistical analyses were performed using the SAS-based statistical package JMP (www.jmp.com).

In total, 23,951 patients were included in this study. The number of patients receiving ECMO increased from 1,613 patients in 2001 to 3597 patients in 2011 (Fig. 1). The number of pediatric patients receiving ECMO increased from 1,027 to 1,434 during the same time period, while the number of adult ECMO patients increased from 309 to 2004 (Fig. 1). Age data were available for 21,304 patients; these patients were 18.5±56.3 years old, and 7,741 patients (36.3%) were younger than 1 month and 8,398 patients (39.4%) were ≥18 years old. The most common indication/comorbidity was congenital heart disease (n=11,712, 48.9%) followed by adult respiratory failure (n=11,407, 47.6%) and shock (n=7,369, 30.8%). These data are summarized in Table 1. The most common indications/comorbidities by age group were congenital heart disease for patients aged <1 month (86.8%) and aged between 1 month and 1 year (71.0%). Respiratory failure was the most common indication in patients aged 1-17 years (70.5%), 18-64 years (77.0%), and ≥65 years (66.6%). These data are summarized in Supplementary Table 1 (in the online-only Data Supplement).

A total of 2,604 patients (10.9%) suffered neurologic complications of seizure (n=986, 4.1%), stroke (n=980, 4.1%), or ICH (n=862, 3.6%). When compared to patients without neurologic complications, acute ischemic stroke patients had a significantly higher rate of discharge to a long-term facility (12.2% vs. 6.8%, p<0.0001), a significantly longer mean length of stay (41.6 days vs. 31.9 days, p<0.0001), and significantly higher costs (US$ 639,528 vs. US$ 501,184, p<0.0001). When compared to patients without neurologic complications, ICH patients had a significantly higher rate of discharge to a long-term facility (9.5% vs. 6.8%, p=0.007), a significantly higher mortality rate (59.7% vs. 50.0%, p<0.0001), a significantly longer mean length of stay (41.8 days vs. 31.9 days, p=0.01), and significantly higher hospitalization costs (US$ 641,636 vs. US$ 501,184, p=0.002). The outcomes did not differ significantly between seizure patients and patients without neurologic complications. These data are summarized in Table 2.

The results of univariate analyses of variables associated with neurologic complications are presented in Table 3. Patients aged <1 month had the lowest rate of acute ischemic stroke (2.5%), while patients aged between 1 month and 1 year had the highest rate (6.4%, p<0.0001). Patients with shock had the highest rate of acute ischemic stroke (6.0% vs. 3.2%, p<0.0001), while those with respiratory distress syndrome had the lowest rate (2.5% vs. 4.2%, p<0.0001). Patients aged <1 month also had the lowest rate of ICH (1.8%), while those aged between 1 month and 1 year had the highest rate (9.9%, p<0.0001). Respiratory distress syndrome was associated with the lowest rate of ICH (1.2% vs. 3.7%, p<0.0001). Patients aged <1 month had the lowest rate of seizure (1.7%), while those aged between 1 month and 1 year had the highest (13.9%, p<0.0001). Cardiac arrest was the comorbidity associated with the highest rate of seizure (8.3% vs. 3.1%, p<0.0001), and respiratory distress syndrome was associated with the lowest rate (2.0% vs. 4.2%, p<0.0001). The results of analyses of variables associated with neurologic complications by age are summarized in Supplemental Table 2, 3, 4, 5, 6 (in the online-only Data Supplement).

Variables predictive of neurologic complications (stroke, ICH, or seizure) on multivariate analysis included cardiac arrest [odds ratio (OR)=1.30, 95% confidence interval (CI)=1.17-1.45, p<0.0001], sepsis (OR=1.12, 95%CI=1.01-1.24, p=0.03), and shock (OR=1.24, 95%CI=1.12-1.37, p<0.0001). Compared to neonates, all age groups had significantly lower odds of neurologic complications on multivariate analysis, with the risk being highest for infants aged between 1 month and 1 year. These data are summarized in Table 4.

Variables predictive of mortality on multivariate analysis included cardiac arrest (OR=1.49, 95%CI=1.38-1.60, p<0.0001), sepsis (OR=1.12, 95%CI=1.05-1.20, p=0.0009), shock (OR=1.20, 95%CI=1.13-1.29, p<0.0001), and respiratory failure (OR=1.16, 95%CI=1.08-1.25, p<0.0001). ICH was the only neurologic complication associated with higher odds of mortality (OR=2.31, 95%CI=1.89-2.84, p<0.0001). Adult patients (aged 18-64 years) had significantly higher odds of mortality compared with the pediatric cohort (OR=1.88, 95%CI=1.69-2.09, p<0.0001), as did seniors (aged ≥65 years; OR=2.34, 95%CI=2.04-2.68, p<0.0001). These data are summarized in Table 4.

This study was subject to certain limitations. First, it was not possible to distinguish between patients receiving venoarterial and venovenous ECMO. It is possible that complication rates differ according to the type of ECMO received by patients. However, many studies comparing venoarterial to venovenous ECMO have found no difference in complication rates.192021 Second, it was also not possible to determine whether the duration of ECMO support was associated with neurologic complications, since this information was not available in the NIS database. Neurologic complications were instead identified using discharge data in the form of ICD-9 codes; therefore, the data are potentially subject to coding errors. Third, long-term outcomes among patients included in this database could not be studied, and it was not possible to determine whether any particular neurologic complication seen in this study resulted from the original insult that triggered the use of ECMO or if it was a complication of the ECMO procedure itself. Whether ECMO is a risk factor for neurologic complications per se, or if patients with these specific neurologic conditions are just more likely to require ECMO has yet to be ascertained. Fourth, it is possible that many of these patients were not examined with neuroimaging; thus, the incidence of stroke or ICH could be underestimated. One neuroimaging and neuropathologic study of 74 children receiving ECMO found that more than 20% of patients had ischemic infarctions,22 and another study using T2-weighted MRI among ECMO patients found that nearly all imaging patients had microhemorrhages on MRI.11 Finally, no adjustment for multiple comparisons was made in the present analyses, and so it is possible that some of the findings were due to chance. For example, multivariate analysis of mortality predictors revealed that stroke and seizure patients had lower odds of death than those who did not suffer these complications; this is almost certainly a chance finding.

The utilization of ECMO is increasing exponentially in the United States, especially among adult patients. Neurologic complications such as ICH, acute ischemic stroke, and seizures affect approximately 11% of patients. ICH is an independent predictor of mortality, while stroke is associated with a higher rate of discharge to a long-term facility. Given the increasing utilization of ECMO and the high costs and poor outcomes associated with neurologic complications, more research is needed to determine the best way to prevent neurologic complications in this patient population.