Predictors of early mobilization in patients requiring VV ECMO for greater than 7 days: An international cohort study

Abstract

Background:

Despite the demonstrated benefits of rehabilitation, active physical therapy and early mobilization are not universally performed during critical illness, especially among patients receiving extracorporeal membrane oxygenation (ECMO), with variation among sites.

Objective:

What factors are predictive of physical mobility during venovenous (VV) ECMO support?

Methods:

We performed an observational analysis of an international cohort using data from the Extracorporeal Life Support Organization (ELSO) Registry. We analyzed adults (≥18 years) supported with VV ECMO who survived for at least 7 days. Our primary outcome was early mobilization (ICU Mobility Scale score >0) at day 7 of ECMO support. Hierarchical multivariable logistic regression models were utilized to identify factors independently associated with early mobilization at day 7 of ECMO. Results are reported as adjusted odds ratios (aOR) with 95% confidence intervals (95%CI).

Results:

Among 8,160 unique VV ECMO patients, factors independently associated with early mobilization included cannulation for transplantation (aOR 2.86 [95% CI 2.08–3.92]; p<0.001), avoidance of mechanical ventilation (aOR 0.51 [95% CI 0.41–0.64]; p<0.0001), higher center level patient volume (6–20 patients annually: aOR 1.49 [95% CI 1 to 2.23] and >20 patients annually: aOR 2 [95% CI: 1.37 to 2.93]; p<0.0001 for group), and cannulation with a dual-lumen cannula (aOR 1.25 [95% CI 1.08–1.42]; p=0.0018). Early mobilization was associated with a lower probability of death (29 vs 48%; p<0.0001).

Conclusions:

Higher levels of early mobilization on ECMO were associated modifiable and non-modifiable patient characteristics, including cannulation with a dual-lumen cannula, and with high center level patient volume.

Introduction

Rehabilitation during critical illness, including early mobilization during mechanical ventilation, is both well described and can mitigate deconditioning during prolonged critical illness ^1–4. Despite the demonstrated benefits of rehabilitation in preventing muscle atrophy, a highly prevalent complication of prolonged critical illness, active physical therapy and early mobilization are not universally performed during critical illness ⁵. This is even more true among the most critically ill patients, such as those receiving extracorporeal membrane oxygenation (ECMO) ⁶.

While early mobilization during ECMO have been described ^7,8, for most patients supported with ECMO, rehabilitation remains an in bed activity ⁶. Patients on ECMO receive full cardiopulmonary support (for venoarterial) or pulmonary support (for venovenous [VV]) and are often considered too critically ill to be awake, let alone to mobilize or to ambulate ^6–8. While notable studies have demonstrated that early mobilization during ECMO is possible at high volume centers ^9,10, data are limited and a recent scoping review of physical rehabilitation during ECMO confirmed that most studies were limited in size and to a single center ¹¹. Marhong et al reported that around 35% of VV ECMO patients at high volume ECMO centers are mobilized on ECMO ⁶.

The recognition that certain centers can achieve a high prevalence of early mobilization during ECMO suggests important differences between centers in either patient illness or in the care provided. In an effort to understand predictors of early mobilization during ECMO support, we examined the largest international ECMO database, the Extracorporeal Life Support Organization (ELSO) Registry, to identify patient-level and institutional characteristics that predicted early mobilization at day 7 of VV ECMO support.

METHODS

Study Design

We conducted a registry-embedded, international cohort study using data from the Extracorporeal Life Support Organization (ELSO) Registry. Our primary objective was to identify factors predictive of early mobilization at day 7 of VV ECMO support. Our secondary objectives included describing the level of early mobilization on ECMO at day 7 and the maximum level of mobilization achieved on ECMO. This analysis was approved by the ELSO Registry Scientific Oversight Committee. Institutional Review Board review was not required given this was an analysis of de-identified registry data, which did not meet criteria for human subject’s research and was determined to be exempt (IRB #00133307).

Data Source

Data came from the ELSO Registry, which has been previously described, but briefly, is an international voluntary registry of patients on ECMO ¹². The ELSO Registry includes over 175,000 patients from over 463 centers who are located over 6 continents ^13,14. Data quality and minimizing bias was assured through the use of a standardized data dictionary, training and certification for data abstractors prior to data submission, validation guardrails on allowable data ranges during entry, and limited external auditing ¹⁵. Patients were followed until their date of death or hospital discharge.

Participants

Patients were eligible for our analysis if they met all of the following characteristics: (1) ≥18 years of age; (2) were entered into the ELSO Registry and had received ECMO from January 15^th, 2018, through April 20^th, 2021; (3) received ECMO for 7 days or more; (4) received venovenous (VV) ECMO; (5) had level of mobilization at day 7 of ECMO recorded in the Registry. Data were extracted from the Registry for analysis on April 20^th, 2021, such that all patients entered prior to this date were assessed for inclusion. Limiting patient data to after January 15^th, 2018, was due to the addition of the physical mobilization variable to the ELSO Registry at this time. For individual patients with more than one ECMO run per hospital admission, only the first ECMO run was analyzed.

Outcomes

The primary outcome was early mobilization at day 7 of ECMO, measured with the ICU Mobility Scale (IMS)¹⁶, which is recorded in the ELSO Registry. The ICU Mobility Scale is a previously validated scale (0–10) to quantify level of physical activity among critically ill patients; the ICU mobility scale is responsive to changes in level of mobility during critical illness ¹⁷. For the purposes of this analysis, early mobilization was defined as an IMS >0. Secondary outcomes included the highest level of mobilization achieved at Day 7 of ECMO and maximum level of mobilization achieved during ECMO.

Demographic and Clinical Variables

We reported demographic and clinical characteristics of patients, including age, weight, sex, race, comorbidities, ventilation status and settings at ECMO initiation, medications and therapies prior to ECMO, center level adult VV patient volume, cardiac arrest prior to ECMO, hospital survival, discharge location, reason for ECMO discontinuation, and duration from ECMO start until death.

Candidate variables for predicting mobility at day 7 of ECMO were selected a priori among available variables, and were informed by recent analyses of early mobility during ECMO support ¹⁰; these candidate predictive variables included: age (in years); sex (male vs female); weight (in kilograms); primary diagnosis (by ICD10 code); admission year; intended bridge-to-transplant for the use of ECMO; ECMO cannula sites; cardiac arrest before ECMO; center level patient volume (<6 runs/year, 6–20 runs/year, >20 runs/year), Charlson comorbidity index (CCI); duration (in hours) from intubation to initiation of ECMO; the following variables, which were obtained during the 6–24 hours prior to ECMO: use of prone positioning, partial pressure of arterial oxygen (PaO₂) (in millimeters of mercury [mmHg]); partial pressure of arterial carbon dioxide (PaCO₂) (mmHg); pH; use of neuromuscular blockers, inhaled epoprostenol, systemic steroids; intravenous use of any of the following medications: epinephrine, norepinephrine, vasopressin, bicarbonate, epoprostenol.

Sample Size

With a minimum of 1800 patients mobilized early and a ratio of 1:3 non mobilized patients, this study had >95% power (2 sided p-value of 0.05) to detect a difference in proportion of 5% and a difference in continuous normally distributed variables of 10% of one standard deviation.

Statistical Analysis

All data were initially assessed for normal distribution. Group comparisons were performed using chi-square tests for equal proportion, student t-tests for normally distributed data and Wilcoxon rank sum tests otherwise with results reported as percent (n), mean (standard deviation), median (interquartile range) respectively. Hierarchical multivariable logistic regression models were used to identify factors independently associated with patient mobility on day 7 with patients nested within sites and sites included as a random effect and results reported as odds ratios (95%CI). A list of candidate predictors of mobility was determined a priori and included only variables with a known or potential association with mobility. After fitting a full model using all predetermined variables, a reduced model was then developed using a least absolute shrinkage and selection operator (lasso) approach and confirmed using backwards elimination before undergoing a final assessment for clinical and biological plausibility. Linearity assumptions were assessed using Locally Weighted Scatterplot Smoothing (Loess) with potential multi-collinearity assessed using variance-inflation factors. Presenting diagnosis was collapsed into 19 categories based on available frequencies for ICD10 codes. To increase the robustness of the analysis, a two-sided p-value of 0.01 was used to indicate statistical significance. Missing data were not imputed. Adjusted odds ratios (ORs), 95% confidence intervals (CIs) and p-values were reported from all models. All analysis was performed using SAs version 9.4 (SAS Institute Inc., Cary, NC, USA)

RESULTS

Study Population and mobility scores on ECMO

From January 15^th, 2018, through April 20^th, 2021, 12,692 VV ECMO runs were identified in the ELSO Registry, in which the mobilization data field was filled out. After filtering, 8,160 unique patients who were hospitalized for at least 7 days after starting ECMO remained for analysis (Figure 1). Among the 8,160 patients in the final analysis, 2,403 patients (30%) achieved some level of physical activity during ECMO support, with 845 (10%) achieving standing while receiving ECMO (Figure 2). At 7 days of ECMO support, 1,830 patients (22%) achieved some level of physical mobilization, with 557 (6.8%) achieving an IMS of 4 (standing) or greater by this time.

Figure 1. Study enrollment flowchart.

Flowchart depicting the inclusion and exclusion of patients from the Extracorporeal Life Support Organization (ELSO) Registry into the study.

Abbreviations: ECMO-extracorporeal membrane oxygenation; ELSO-Extracorporeal Life Support Organization

Patient selection method for adult mobility during veno-veno extracorporeal membrane oxygenation (ECMO) from the Extracorporeal Life Support Organization (ELSO) Registry

Figure 2. ICU Mobility Level Achieved (Frequency).

Logarithmic scale depicting the frequency of mobility level achieved using the ICU Mobility Scale, at 7 days (red) and max level achieved (blue).

Abbreviations: ICU-intensive care unit.

At baseline, there were significant differences in patient characteristics between those who mobilized at Day 7 vs those who did not; patients who mobilized weighed less (91 kg vs 95 kg; p<0.0001) and were less likely to be male (63% vs 68%; p<0.0001) (Table 1). Patients who mobilized early were more likely to be Caucasian (53% vs 48%) and less likely to be Black (10% vs 12%; p<0.0001 for group). Overall, patients who mobilized early had chronic comorbidities, including a history of a transplanted organ (11% vs 3%; p<0.0001), a greater count of comorbidities (4 [IQR 2, 7] vs 2 [IQR 2, 7]; p=0.007) and a higher CCI score (0 [0, 1] vs 0 [0, 0]; p=0.014).

Table 1.

Patient characteristics

Variable, % (n)	No early mobilization N=6330		Early mobilization N=1830	P-value
Baseline Variables
Patient agea	48.1 (13.7)		48.1 (14)	0.85
Weighta	94.8 (29.5)		91.2 (28.4)	<0.0001
Male gender	67.5% (4274)		62.5% (1143)	<0.0001
Race b
Asian	11.7% (738)		10.2% (186)	0.08
Black	12% (759)		9.6% (175)	0.004
Hispanic	15.2% (959)		14.4% (263)	0.41
Middle Eastern	2.7% (168)		5.2% (96)	<0.0001
Native American	1.3% (85)		0.546% (10)	0.005
Native Hawaiian / Pacific Islander	0.237% (15)		0.273% (5)	0.78
Caucasian	48.1% (3044)		52.7% (964)	0.001
Multiple races	3.5% (223)		3.6% (65)	0.95
Other	1.9% (120)		1.1% (21)	0.03
Unknown	3.5% (219)		2.5% (45)	0.03
Number of comorbidities	3 (2–7)		4 (2–7)	0.007
Charlson Comorbidity Index Scoreb	0 (0–0)		0 (0–1)	0.01
Comorbiditiesb
Myocardial infarction	2.2% (139)		1.4% (25)	0.02
Congestive heart failure	5.3% (332)		6.8% (123)	0.02
Peripheral vascular disease	0.771% (48)		1% (19)	0.25
Cerebrovascular disease	2.7% (170)		2.1% (38)	0.14
Dementia	0.016% (1)		0.055% (1)	0.4
Chronic pulmonary disease	8.3% (517)		9.8% (178)	0.04
Renal disease	2.2% (134)		2.4% (44)	0.48
Cancer	2.7% (171)		2.5% (46)	0.64
COVID-19	27.7% (1752)		16.4% (301)	<0.0001
Transplant	2.7% (161)		10.9% (188)	<0.0001
Ventilator Type Prior to ECMO
Conventional	90% (5695)		83% (1518)	<0.0001
HFO	0.363% (23)		0.656% (12)	0.09
No ventilator	1.7% (106)		5.1% (93)	<0.0001
Other	1.3% (84)		4% (74)	<0.0001
Other HFV	0.095% (6)		0.164% (3)	0.43
Ventilation Settings Prior to ECMO
FiO2 (%)a	93.2 (13.8)		92.3 (15)	0.02
PEEP (cmH20)a	13.1 (4.8)		12.5 (4.81)	<0.0001
PO2 (mmHg)b	66 (54.8–81)		69 (57–85)	<0.0001
Duration of mechanical ventilation prior to ECMO (hours)b	62 (17–141)		43 (10–134)	<0.0001
Medications Prior to ECMO
Narcotics	66.9% (4236)		67.2% (1230)	0.81
Neuromuscular blocker	66% (4176)		59.8% (1095)	<0.0001
Systemic steroids	27.6% (1744)		25.6% (469)	0.10
Inhaled epoprostenol	15.1% (955)		11% (201)	<0.0001
Systemic epoprostenolc	3.7% (232)		2.1% (39)	0.001
Nitric oxide	13.3% (844)		14.7% (269)	0.13
Vasopressin	15.2% (961)		15.2% (279)	0.95
Phenylephrine	5.5% (348)		5.2% (95)	0.61
Intravenous bicarbonate	10.6% (671)		8.2% (150)	0.003
Pre-ECMO Therapies
Prone positioning	37.1% (2350)		31.6% (579)	<0.0001
Renal replacement therapy	10% (634)	9.4% (172)		0.44
Therapeutic hypothermia	0.32% (20)	0.16% (3)		0.28
Annual Patient Volume
<6 patients	75.9% (4802)	80.9% (1481)		<0.0001
6–20 patients	20% (1265)	16.3% (298)		<0.0001
>20 patients	4.2% (263)	2.8% (51)		0.007
Cannulation Typesd
Return Cannula	49.1% (2894)	49.1% (839)		0.97
Drainage Cannula	74.2% (4371)	65.8% (1125)		<0.0001
Dual-Lumen Cannula	26.6% (1565)	32.6% (558)		<0.0001
Duration of ECMO (hours)b	383 (257–628)	380 (250–674)		0.63
Duration of Hospitalization prior to mechanical ventilation (hours)a	25.7 (228)	40.9 (345)		0.04
Level of Mobilizationb	0 (0–0)	2 (1–4)		<0.0001
Max Level of Mobilizationb	0 (0–0)	3 (1–5)		<0.0001
Cardiac arrest prior to ECMO	5.4% (341)	4.9% (90)		0.43
ECMO Discontinuation Reason
Death or poor prognosis	40.9% (2559)	23.6% (427)		<0.0001
Expected recovery	56.5% (3537)	70.3% (1274)		<0.0001
Patient died before discontinuation	47.5% (3005)	29.4% (583)		<0.0001
Duration from ECMO until Death (days)e	18.3 (12–29.3)	22.5 (13.8–36)		<0.0001
Death	47.5% (3005)	29.4% (538)		<0.0001
Discharge Location
Home	14.7% (930)	26.5% (485)		<0.0001
Other/Unknown	8.2% (516)	4.8% (88)		<0.0001
Transferred to another hospital	13.9% (877)	13.3% (244)		0.57
Transferred to LTAC	4.4% (281)	3.3% (60)		0.029
Transferred to LTAC or Rehab	13.6% (862)	19.2% (352)		<0.0001
Transferred to rehab	4.3% (275)	6.1% (112)		0.002
Transferred to hospice	0.695% (44)	0.328% (6)		0.08

^amean (standard deviation)

^bmedian (interquartile range)

^cIncludes other synthetic prostacyclin analogues

^d8 patients were listed as also having a distal limb cannula

^eAmong patients who received ECMO until Day 7.

Abbreviations: %-Percent; cmH₂0-centimeters of water; ECMO-extracorporeal membrane oxygenation; mmHg-millimeters of mercury; LTAC-Long term acute care facility

Patients who mobilized early were more likely not have not received invasive mechanical ventilation at the time of ECMO initiation (5% vs 2%; p<0.0001) and were less likely to have utilized a conventional ventilator (83% vs 90%; p<0.0001). Physiologically, patients who mobilized early had lower ventilatory settings, higher oxygen levels (Table 1) and had undergone shorter durations of mechanical ventilation prior to ECMO initiation (43 hours [10, 134] vs 62 hours [17, 141]; p<0.0001). Patients who received early mobilization were less likely to die during the hospitalization (29 vs 48%; p<0.0001).

Factors associated with early mobilization at day 7 of ECMO

Univariate associations between a priori identified patient-level and institutional factors and early mobilization are listed in Table 2. After multivariable adjustment and model simplification, a number of patient-level and institutional factors remained significant predictors of early mobilization at day 7 of ECMO support (Table 2). Full multivariable model results are also presented in the e-Appendix.

Table 2.

Adjusted probability of physical mobility during the first 7 days of ECMO

Variable	Unadjusted OR	95% CI	P-value	Adjusted OR	95% CI	P-value
Patient Characteristics
Planned bridge to transplant	4.49	(3.61–5.58)	<0.0001	2.86	(2.08–3.92)	<0.0001
Use of Conventional Ventilatora	0.54	(0.47–0.63)	<0.0001	0.51	(0.41–0.64)	<0.0001
Annual Patient Volume
1–5 / year	reference		<0.0001	reference		<0.0001
6–20 / year	1.21	(0.88–1.68)		1.49	(1–2.23)
> 20 / year	1.59	(1.17–2.16)		2	(1.37–2.93)
pH prior to ECMO	2.03	(1.31–3.13)	0.0014	1.94	(1.2–3.13)	0.0069
Medications given prior to ECMO
Systemic epoprostenolb	0.57	(0.41–0.81)	0.0014	0.53	(0.35–0.81)	0.0029
Inhaled epoprostenol	0.69	(0.59–0.82)	<0.001	0.73	(0.6–0.88)	0.0008
Epinephrine	0.84	(0.71–1)	0.0488	0.76	(0.64–0.96)	0.0172
Dual-Lumen Cannulation	1.34	(1.19–1.5)	<0.001	1.254	(1.08–1.42)	0.0018
Diagnosis
COVID-19	reference		<0.001	reference		<0.0001
Sepsis	2.33	(1.82–2.98)		2.29	(1.72–3.06)
Complications of Diabetesc	5.59	(3.15–9.91)		2.02	(0.95–4.29)
Complications of CVAd	2.77	(1.87–4.12)		2.67	(1.66–4.28)
Complications of AMI	2.69	(1.71–4.21)		2.07	(1.21–3.54)
Acute URI	3.05	(2.15–4.34)		2.46	(1.64–3.7)
Influenza, NOS	2.39	(1.84–3.11)		2.76	(2.06–3.69)
Novel Influenza / H5N1 / Swine Flu	1.85	(1.24–2.76)		2.18	(1.42–3.34)
Viral pneumonia, NOS	1.47	(1.13–1.92)		1.42	(1.03–1.97)
Bacterial pneumonia, NOS (J15)	2.77	(1.87–4.12)		3.11	(2.01–4.83)
Pneumonia, unknown organism (J18.9)	1.8	(1.14–2.84)		1.85	(1.11–3.08)
Asthma (J45)	0.83	(0.37–1.85)		1.01	(0.45–2.29)
Acute respiratory distress syndrome (J80)	1.68	(1.42–1.99)		1.89	(1.56–2.3)
Other interstitial pulmonary diseases (J84)	5.66	(3.99–8.02)		3.07	(1.9–4.94)
Acute respiratory failure (J96)	1.77	(1.5–2.1)		1.67	(1.37–2.04)
Complications of transplanted organs and tissue	3.46	(1.93–6.18)		3.03	(1.55–5.93)
Lung transplant	2.2	(1.18–4.12)		0.84	(0.26–2.68)
Cardiogenic shocke	1.9	(1.05–3.44)		0.92	(0.4–2.16)
Poisoning	1.86	(1.24–2.81)		1.86	(1.16–2.96)
Other diagnosis	1.52	(1.18–1.94)		1.64	(1.22–2.2)

^aReference level: no invasive mechanical ventilation prior to ECMO.

^bIncludes other synthetic prostacyclin analogues

^cIncludes other endocrine complications

^dIncludes other vascular diseases

^eIncludes other heart disease

Abbreviations: AMI-acute myocardial infarction, COVID-19-coronavirus disease 2019, CVA-cerebrovascular accident, ECMO-extracorporeal membrane oxygenation, NOS-not otherwise specified, PaCO2-partial pressure of arterial carbon dioxide, PaO2-partial pressure of arterial oxygen, URI-upper respiratory infection

Patient

A diagnosis of COVID-19 was strongly associated with lower probability of early mobilization at day 7 compared to all other diagnoses, except for pneumonia due to COVID-19 (ICD-10 J12.81) (adjusted OR [aOR] 0.95 compared to COVID-19 alone [ICD-10 U07.1] [95% CI 0.57 to 1.57]; p<0.0001).

Clinical patient factors negatively associated with early mobilization at day 7 included the use of any of the following medications for at least 6 hours in the 24 hours prior to starting VV ECMO: epinephrine infusion (aOR 0.76 [95% CI 0.62–0.93]; p=0.0083), epoprostenol inhaled (aOR (aOR 0.73 [95% CI 0.6–0.88]; p=0.0011), and epoprostenol infusion (aOR 0.53 [95% CI 0.35–0.8]; p=0.0025).

Higher pH prior to ECMO was associated with increased probability of early mobilization (aOR 2 [95% CI 1.23–3.25]; p=0.005), as was cannulation in anticipation of lung transplantation (aOR 2.64 [95% CI 1.91–3.65]; p<0.001).

Clinical Management/Cannulation

Use of a conventional ventilation (vs oscillatory or other) was associated with lower probability of early mobilization (aOR 0.51 [95% CI 0.41–0.64]; p<0.0001), whereas cannulation using a dual-lumen cannula (aOR 1.25 [95% CI 1.09–1.43]; p=0.0014) was associated with higher probability of early mobilization.

Institutional

At an institutional level, higher annual adult VV patient count was associated with increased probability of early mobilization (6–20 patients annually: aOR 1.48 [95% CI 0.99 to 2.22] and >20 patients annually: aOR 2 [95% CI: 1.36 to 2.93] vs 1–5 patients annually [reference]; p<0.0001 for group).

DISCUSSION

In this international analysis of adult patients managed with VV ECMO, we identified variables at a patient, cannulation and institutional level that were independently associated with early mobilization. Among institutional factors associated with early mobilization, the finding of higher center-level annual adult VV ECMO patient volume suggests that higher patient volumes, or regionalization of VV ECMO patients, may improve levels of mobilization by increasing the skillset mix of staff, and providing opportunities for professional teams to establish protocols and practices that enhance mobilization.

While the identified patient factors associated with early mobilization are non-modifiable, including intended pre-transplant status, and select diagnoses, our findings are an important reminder that select indications for VV ECMO in adults may be better suited for early mobilization during ECMO, and thus could prompt increased therapy. Further, we found that modifiable factors, including cannulation approach with a dual-lumen cannula, was associated with increased levels of mobilization, although this association may reflect clinician preference in anticipation of patient mobilization during ECMO.

Notably, we found a diagnosis of COVID-19 was associated with lower probability of mobility compared to other diagnoses. This may not be surprising considering the transmissibility of the COVID-19 virus, which requires airborne isolation to protect workers. Prior to the availability of a vaccine, the risk of serious illness and death from contracting COVID-19 during the care of patients was much higher, likely influencing the decision to minimize exposure to additional staff and limiting the capacity to mobilize patients. We also hypothesize that these patients may be sicker overall, with higher mortality and longer ECMO runs, and therefore it may be more difficult overall to perform early mobilization in this population, especially early on, as has been previously observed ¹⁸.

Our results build upon previous studies of physical therapy during ECMO support that demonstrate that intended bridge to transplant status is an identified factor associated with early mobilization at high volume centers ¹⁰. The survival benefit from higher center level patient volume for ECMO patients has been demonstrated among subgroups of ECMO patients, including neonates and adults ¹⁹, and adult extracorporeal cardiopulmonary resuscitation ²⁰. Our findings support the concept that for complex therapies, higher patient volume may lead to improved survival through refined care processes.

In contrast to previous survey results ⁶, our study found that only a minority of patients achieved any early mobilization during ECMO support, with an even smaller minority achieving an IMS of 4 (standing) or greater. This discrepancy could reflect the large number of sites included in the study, which increases the generalizability of the results.

As we found that the dual-lumen cannula was associated with increased mobilization, it is reasonable to consider that this finding is not applicable to patients on VA ECMO, who cannot solely utilize a dual-lumen venous cannula. Much of the previously published literature has examined patients on ECMO with both VA and VV. All of our other predictors could be present within patients on VA, though as we did not analyze patients on VA ECMO, this would need to be tested.

Despite the strengths of our study, we acknowledge several limitations. Firstly, as mobility status was only collected on day 7, to ensure an equal exposure period for all patients, we included only patients that were on ECMO for 7 days. Given the median VV ECMO duration for ARDS within the ELSO Registry is 13.67 days ¹⁴, the majority of patients excluded by this step had died. Secondly, the ELSO Registry utilizes voluntary reporting, and thus may have reporting bias. Finally, it is increasingly understood that the race variable is socially constructed and not fully representative of the differences that it purports to distinguish, especially within registry studies^21–23.

Across an international registry of adult patients treated with ECMO, we demonstrated that higher levels of physical mobility on VV ECMO are associated with both modifiable and non-modifiable patient characteristics, including cannulation with a dual-lumen cannula, and with high center level patient volume. Acknowledging the increasing use of adult ECMO, our data represent a foundational study identifying predictors of physical mobilization for patients on VV ECMO, including potentially early referral to a high volume ECMO center.

Highlights.

• Among adults on VV ECMO, 22% achieved some level of physical mobilization.

• Early mobilization was associated with modifiable and non-modifiable factors.

• Modifiable factors included dual lumen cannulation, and higher center case volume.

TAKE HOME POINT.

Study Question:

Are there patient, cannulation, and center level factors that are predictive of physical mobility during extracorporeal membrane oxygenation (ECMO) support?

Results:

Among adult patients supported on VV ECMO, higher levels of early mobilization at day 7 of ECMO were associated with both modifiable and non-modifiable patient characteristics. Modifiable patient characteristics included cannulation with a dual-lumen cannula, and with higher center level patient volume.

Interpretation:

Higher levels of early mobilization on VV ECMO were associated with both modifiable and non-modifiable patient characteristics, including cannulation with a dual-lumen cannula, and with high center level patient volume.

Abbreviation List

aOR

adjusted odds ratio

CCI

Charlson comorbidity index

CI

confidence interval

ECMO

extracorporeal membrane oxygenation

ELSO

Extracorporeal Life Support Organization

IQR

interquartile range

VV ECMO

veno-veno extracorporeal membrane oxygenation