Extracorporeal membrane oxygenation (ECMO) can provide temporary support for oxygenation, ventilation, and cardiac output in critically ill patients who are failing conventional treatment modalities. The benefits of ECMO have long been accepted in pediatric populations1 while evidence for the use of ECMO in adult populations was less promising.2,3 However, in recent years, technological improvements in the ECMO circuit design, promising survival data,4,5 and expanded indications for ECMO6–8 have led to renewed interest in ECMO for adults.
Many centers performing ECMO voluntarily report data to the Extracorporeal Life Support Organization (ELSO) registry. The ELSO releases regular summary reports of their registry data, which show that the worldwide use of ECMO in adult patients has increased substantially in recent years.9 A similar trend has been observed using data from the National Inpatient Sample (NIS) in the United States.10 This rapid increase in ECMO use has led some to worry that the technology is being used too liberally, arguing that other less invasive and less resource-intensive therapies would be more appropriate for many patients.11,12
Although some countries, such as Italy and the United Kingdom, have limited ECMO use to specific designated ECMO centers, no such system exists in the United States. As a result, there is marked heterogeneity in the delivery of intensive care services with advanced respiratory care capabilities.13–16 Like many complex, resource-intensive therapies, ECMO patients have better outcomes at centers that perform more cases annually17–19; however, neither the government nor insurance payers have set restrictions on ECMO use at smaller centers in the United States.
More data are needed about the patients receiving ECMO therapy, or the hospitals at which ECMO is being performed, in the United States. Our objective was to evaluate the rate of ECMO use within predefined categories of interest and then to evaluate for change over time within those groups using a standardized administrative data set. We used state-based ECMO use rates in the four census regions to examine regional variation in ECMO use. We sought to determine whether the worldwide trend toward increased ECMO use in adults was also present in the United States and to quantify any change in the rates of use in adult and pediatric populations. Because ECMO requires substantial resources and incurs high hospital costs, we also quantified the rate of ECMO use based on the primary insurance payer to assess whether there was potential bias against using the therapy in uninsured or governmentally insured patients. Similarly, we were interested in trends in ECMO use among for-profit, not-for-profit, and government-owned hospitals. Finally, because of the previously demonstrated relationship between annual volume and outcomes, we were interested to see how prevalent ECMO use was at small hospitals and whether this changed over time.
Materials and Methods
This study used only publicly available limited data sets and was granted an exemption from review by the Emory University Institutional Review Board.
Data were obtained from the Healthcare Cost and Utilization Project (HCUPnet, Healthcare Cost and Utilization Project. Agency for Healthcare Research and Quality, Rockville, MD. https://hcupnet.ahrq.gov/). Detailed data were obtained using the HCUPnet summaries of State Inpatient Databases (SID) from all 34 participating states for the years 2011–2014. This data set contains discharge abstracts from participating states regarding hospital inpatient stays. In this data set, counts <10 or with <2 hospitals reporting are censored to protect patient confidentiality.
Patients treated with ECMO were identified by querying international classification of diseases (ICD)-9 code 39.65 (ECMO). This code explicitly excludes cardiopulmonary bypass in the operating room (ICD-9 code 39.61) and thus would capture only those patients treated with ECMO outside of the operating room. This code does not differentiate between modes of cannulation (veno-arterial or veno-venous).
To control for the effects of population distribution, we calculated the annual per-capita rate of use within each state. State population data and census region were obtained from the US Census Bureau, Population division (https://www.census.gov/data/tables/2016/demo/popest/state-total.html). The total number of ECMO patients in each state was divided by the population of that state to determine a rate of ECMO use for that state (ECMO patients per 100,000 persons).
Aggregate data from the State Inpatient Dataset were extracted using the HCUPnet online query tool (hcupnet.ahrq.gov) for the 34 states contributing data for the years 2011–2014. Categories of interest for patient characteristics included age and primary payer; age data were grouped into adult (age ≥18 years) and pediatric (age <18 years). Categories of interest for hospital characteristics included census region, bed size, ownership, and teaching status. Definitions for these hospital characteristics were the standard American Hospital Association definitions used by HCUP (Appendix 1, Supplemental Digital Content, http://links.lww.com/ASAIO/A330).
The sample of 34 states was used to model ECMO rates over time, overall and within subcategories of age group, bed size, hospital ownership, teaching status, and payer type. Count data models were used to model the changes over time, main effects of subcategories, and subcategory by time interactions. Population counts for each state/year combination were used as the offset in the models to provide comparison of ECMO rates rather than counts. Significant overdispersion was present in all cases, thus standard negative binomial models were used instead of Poisson. Omnibus tests were computed to assess significance of the overall effects; where the effects were significant, post hoc linear contrasts of model estimates were used to target the nature of the effects.
Due to data reporting restrictions, values of counts were censored between 1 and 10 to avoid identification of specific subjects. This is a unique situation statistically as the zero counts are valid and numerous, and thus traditional censored models cannot be used. We, therefore, chose a sensitivity analysis approach to the overall analysis, wherein we imputed 1, 5, and 10 for the censored counts. The purpose of these analyses was to determine the strength of effects by evaluating the consistency of the testing results across these imputed counts. For ease of presentation, we present the data in graphs and tables for censored = 5. Any inconsistencies are thus reported in the results section when they occur.
During the study period, the overall rate of ECMO use was 1.34 (1.29–1.42) cases per 100,000 persons per year. Overall, there was a significant increase in ECMO use over time (Χ2 = 12.8, degrees of freedom = 3, p = 0.005). The ECMO use varied significantly by region, with the highest rates in the Midwest and the lowest in the West (Table 1; Figure 1)
The mean rate of ECMO use in adults and children during the 4-year period was not statistically different between groups. However, the data showed a significant increase in the rate of ECMO use among adult patients each year while the use among pediatric patients remained constant throughout the study period (Table 1; Figure 2).
The most common payer was private insurance, followed by Medicaid, Medicare, and other insurance (e.g., Tricare, Workman’s Compensation, etc.); a minority of patients were uninsured. Far more insured patients underwent ECMO during the time period than uninsured (Table 1). There was no difference in the rate of growth between insured and uninsured patients (Table 1; Figure 3).
Most cases of ECMO were performed at large hospitals, with a minority of cases at medium or small hospitals (Table 1). Although the overall number of cases increased, the relative proportion of cases at large, medium, and small hospitals remained unchanged (Table 1; Figure 4).
On average, the yearly rate of ECMO use was significantly higher in private, not-for-profit hospitals than in government-owned or private, for-profit (Table 1; Figure 5). Changes in ECMO rates over time did not vary significantly across hospital types. Likewise, the majority of ECMO cases were performed at teaching hospitals, and this pattern did not change significantly during the study period (Table 1)
Using a standardized administrative data set, we found a rapid growth in ECMO use among adults from 2011 to 2014 while use in pediatric populations remained relatively unchanged. This agrees with previous reports from the United States and elsewhere.9,10,20 The study period for this investigation starts shortly after the publication in prominent journals of several landmark papers in the ECMO literature, including the influential CESAR trial as well as the promising reports of ECMO use during the 2009 influenza H1N1 pandemic.4,5 This may have led to an increased awareness of and interest in ECMO for adults with respiratory failure and could explain the rapid growth of ECMO use in adult populations. By way of contrast, the long-standing acceptance of ECMO use in pediatric patients has likely resulted in a relatively stable rate of use in this population.
We found that the rate of ECMO use in the United States had increased from 1.06 to 1.77 cases per 100,000 persons by the year 2014. The only other wealthy nation for which similar data exist is Germany, where use of ECMO had increased from 1.1 cases per 100,000 in 2007 to 6.2 per 100,000 persons by the year 2014,20 more than three times the rate of increase we found in the United States. There was also significant regional variation in ECMO use within the United States, with the highest rate in the Midwest and the lowest in the West. Regional variation in access to centers with high-level respiratory care and mechanical circulatory support capabilities in the United States has been previously described.16 The current data cannot tell us whether these differences in ECMO rates reflect differences in access to ECMO-capable centers or differences in utilization rates between ECMO-capable centers.
As ECMO technology becomes more widely available and current reimbursement patterns incentivize ECMO use, overutilization has been cited as a concern.11 Given the clear relationship between hospital ECMO volumes and outcomes,17–19 overutilization by small hospitals performing only a few ECMO cases per year would be concerning. Based upon this analysis, the majority of ECMO cases were performed at large hospitals although the rate at medium and small hospitals was roughly one third of the total rate of ECMO use. Bed size is only a crude surrogate for overall ECMO volume, but further study into the impact of hospital size on patient outcomes may be warranted.
Conversely, because ECMO is a very resource-intensive therapy, we hypothesized that it may be relatively underutilized in uninsured populations. Uninsured patients are often ineligible for definitive therapies, such as heart transplants or ventricular assist devices, making ECMO relatively contraindicated in cases where no destination therapy is available.21,22 There may also be implicit bias among providers to avoid potentially expensive, unproven therapy that may leave hospitals and families with substantial financial burdens.
We found that ECMO rates were highest among insured patients, which is not surprising as the majority of Americans carry health insurance.23 Because of the small number of uninsured patients undergoing ECMO, these estimates varied considerably in sensitivity analysis. However, the rate of growth in ECMO use among uninsured patients appeared to be similar to the rate of growth among insured patients. Moreover, it should also be noted that overall rates of uninsurance in the United States declined during the study period,23 which began shortly after the passage of the Affordable Care Act in 2010. Taken together, this suggests that insurance status is not necessarily a determining factor in the expansion of ECMO use on a national level.
Strengths and Limitations
This study used deidentified limited data sets derived from the SID, available from the HCUP, a service of the Agency for Healthcare Research and Quality. This has several advantages over previous studies, which have used either the NIS or ELSO registry.9,10,24
Although the NIS comprises similar data to the data set used in this study, it contains only a 20% sample of admissions. Although the NIS uses a complex system of weighting to create generalizable estimates, for rare occurrences such as ECMO use, this may result in unreliable estimates. In contrast, the data set used here is derived directly from participating state inpatient databases, which contain exact count data for all discharges from the participating states.
The ELSO registry contains only cases that are self-reported by participating centers. Because many small centers may not perform ECMO often enough to be ELSO members, it is likely that there is at least some reporting bias in this sample. The data sets used in this study, unlike the ELSO registry, are derived from validated discharged abstracts and are not subject to the same reporting bias. Furthermore, the ELSO registry does not contain some of the detailed information related to hospital characteristics and patient insurance analyzed here.
This study and the use of the HCUP SID limited data sets have significant limitations. First, the number of participating states: only 34 states have agreements to make their data available. We do not have data from 16 states, which could substantially impact the findings. Next, the censoring of data within the data sets required the use of statistical techniques for handling missing data, introducing uncertain assumptions into the analysis. Although this prevents an exact calculation of ECMO usage rates, the sensitivity analyses showed that these findings were stable over a reasonable range of possible values. Finally, the data extracted are dependent on ICD-9 codes, which have been shown to have varying accuracy.25–27 For relatively rare modalities such as ECMO use, inappropriate coding may introduce systematic bias due to, for example, small hospitals undercoding (due to lack of familiarity) or large hospitals undercoding (due to the volume of these procedures or other factors).
ECMO use has grown rapidly in the United States between 2011 and 2014 to a rate of 1.77 cases per 100,000 persons, driven primarily by increased use in the adult population. Although the use of ECMO has grown in all regions of the United States, significant regional variation in the use of ECMO persists. Most ECMO patients are cared for at large hospitals, and at private, not-for-profit hospitals with teaching designation. Furthermore, most ECMO patients are insured although the growth in ECMO use did not differ between insured and uninsured patients. Further research is warranted to determine why these differences in ECMO use persist and what impact they have on patient outcomes.
1. Field DJ, Davis C, Elbourne D, Grant A, Johnson A, Macrae D. UK collaborative randomised trial of neonatal extracorporeal membrane oxygenation
. Lancet 1996.348:75–82.
2. Zapol WM, Snider MT, Hill JD, et al. Extracorporeal membrane oxygenation
in severe acute respiratory failure. A randomized prospective study. JAMA 1979.242: 2193–2196.
3. Morris AH, Wallace CJ, Menlove RL, et al. Randomized clinical trial of pressure-controlled inverse ratio ventilation and extracorporeal CO2 removal for adult respiratory distress syndrome. Am J Respir Crit Care Med 1994.149(2 pt 1): 295–305.
4. Noah MA, Peek GJ, Finney SJ, et al. Referral to an extracorporeal membrane oxygenation
center and mortality among patients with severe 2009 influenza A(H1N1). JAMA 2011.306: 1659–1668.
5. Peek GJ, Mugford M, Tiruvoipati R, et al.; CESAR Trial Collaboration: Efficacy and economic assessment of conventional ventilatory support versus extracorporeal membrane oxygenation
for severe adult respiratory failure (CESAR): A multicentre randomised controlled trial. Lancet 2009.374: 1351–1363.
6. Stub D, Bernard S, Pellegrino V, et al. Refractory cardiac arrest treated with mechanical CPR, hypothermia, ECMO and early reperfusion (the CHEER trial). Resuscitation 2015.86: 88–94.
7. Ouweneel DM, Schotborgh JV, Limpens J, et al. Extracorporeal life support during cardiac arrest and cardiogenic shock: A systematic review and meta-analysis. Intensive Care Med 2016.42: 1922–1934.
8. Sheu JJ, Tsai TH, Lee FY, et al. Early extracorporeal membrane oxygenator-assisted primary percutaneous coronary intervention improved 30-day clinical outcomes in patients with ST-segment elevation myocardial infarction complicated with profound cardiogenic shock. Crit Care Med 2010.38: 1810–1817.
9. Thiagarajan RR, Barbaro RP, Rycus PT, et al.; ELSO Member Centers: Extracorporeal Life Support Organization Registry International Report 2016. ASAIO J 2017.63: 60–67.
10. Sauer CM, Yuh DD, Bonde P. Extracorporeal membrane oxygenation
use has increased by 433% in adults in the United States
from 2006 to 2011. ASAIO J 2015.61: 31–36.
11. Quintel M, Gattinoni L, Weber-Carstens S. The German ECMO inflation: When things other than health and care begin to rule medicine. Intensive Care Med 2016.42: 1264–1266.
12. Li X, Scales DC, Kavanagh BP. Unproven and Expensive Before Proven and Cheap - Extracorporeal Membrane Oxygenation
vs. Prone Position in ARDS. Am J Respir Crit Care Med. 2018;197:991-993. doi:10.1164/rccm.201711-2216CP.
13. Rosenberg BL, Kellar JA, Labno A, et al. Quantifying geographic variation in health care outcomes in the United States
before and after risk-adjustment. PLoS One 2016.11: 1–21.
14. Angus DC, Barnato AE, Linde-Zwirble WT, et al.; Robert Wood Johnson Foundation ICU End-Of-Life Peer Group: Use of intensive care at the end of life in the United States
: An epidemiologic study. Crit Care Med 2004.32: 638–643.
15. Angus DC, Shorr AF, White A, Dremsizov TT, Schmitz RJ, Kelley MA; Committee on Manpower for Pulmonary and Critical Care Societies (COMPACCS): Critical care delivery in the United States
: Distribution of services and compliance with Leapfrog recommendations. Crit Care Med 2006.34: 1016–1024.
16. Wallace DJ, Angus DC, Seymour CW, et al. Geographic access to high capability severe acute respiratory failure centers in the United States
. PLoS One 2014.9: e94057.
17. Karamlou T, Vafaeezadeh M, Parrish AM, et al. Increased extracorporeal membrane oxygenation
center case volume is associated with improved extracorporeal membrane oxygenation
survival among pediatric patients. J Thorac Cardiovasc Surg 2013.145: 470–475.
18. Freeman CL, Bennett TD, Casper TC, et al. Pediatric and neonatal extracorporeal membrane oxygenation
: Does center volume impact mortality?*. Crit Care Med 2014.42: 512–519.
19. Barbaro RP, Odetola FO, Kidwell KM, et al. Association of hospital-level volume of extracorporeal membrane oxygenation
cases and mortality. Analysis of the extracorporeal life support organization registry. Am J Respir Crit Care Med 2015.191: 894–901.
20. Karagiannidis C, Brodie D, Strassmann S, et al. Extracorporeal membrane oxygenation
: Evolving epidemiology and mortality. Intensive Care Med 2016.42: 889–896.
21. Schmidt M, Bréchot N, Combes A. Ten situations in which ECMO is unlikely to be successful. Intensive Care Med 2016.42: 750–752.
22. Abrams DC, Prager K, Blinderman CD, Burkart KM, Brodie D. Ethical dilemmas encountered with the use of extracorporeal membrane oxygenation
in adults. Chest 2014.145: 876–882.
23. Clarke TC, Norris T, Schiller JS. Early release of selected estimates based on data from the 2016 national health interview survey. 2017. 1–120, Available at: https://www.cdc.gov/nchs/data/nhis/earlyrelease/earlyrelease201705.pdf
. Accessed May 5, 2018.
24. Natt BS, Desai H, Singh N, Poongkunran C, Parthasarathy S, Bime C. Extracorporeal membrane oxygenation
for ARDS: National trends in the United States
2008-2012. Respir Care 2016.61: 1293–1298.
25. Goldstein LB. Accuracy of ICD-9-CM coding for the identification of patients with acute ischemic stroke: Effect of modifier codes. Stroke 1998.29: 1602–1604.
26. Jolley RJ, Sawka KJ, Yergens DW, Quan H, Jetté N, Doig CJ. Validity of administrative data in recording sepsis: A systematic review. Crit Care 2015.19: 139.
27. Espino JU, Wagner MM. Accuracy of ICD-9-coded chief complaints and diagnoses for the detection of acute respiratory illness. Proc AMIA Symp. 2001.164–168.