Costs and complications of extracorporeal membrane oxygenation therapy in Australia and New Zealand Authors: Kate R Wilcox, Lane Meredith, Aidan Burrell, Bentley Fulcher, Carol L Hodgson, Alisa M Higgins Published: 2026 ABSTRACT AbstractBackgroundExtracorporeal membrane oxygenation (ECMO) is a life-saving form of cardiopulmonary support for patients with severe refractory cardiopulmonary disease. It is a resource-intensive intervention requiring specialised equipment and personnel, with significant associated costs. This study aimed to determine the costs of care for patients who received ECMO in Australia and New Zealand, and to determine the impact of ECMO-related complications on costs.MethodsA prospective observational clinical registry analysis was performed using data from 1,345 patients commenced on ECMO in Australia and New Zealand between 2019 and 2022. Per patient resource use was costed based on data from hospital admission to hospital discharge. Key outcomes were mean total patient cost and costs associated with complications. Costs were reported in 2022 Australian Dollars.ResultsMean costs and daily ECMO costs varied significantly between ECMO modes. While total costs for venovenous (VV) ECMO was significantly more expensive (mean $287,264, standard deviation [SD] $251,193) than venoarterial (VA) ECMO (mean $215,128, SD $162,418) and extracorporeal cardiopulmonary resuscitation (ECPR) (mean $121,274, SD $131,893), daily costs while receiving ECMO were least costly for VV ECMO. Complications occurred in 57% of patients. Cost incrementally increased with number of complications. Mean costs were significantly higher in patients with infection (mean $292,197, SD $217,806) compared to those without (mean $161,674, SD $165,175). Key cost drivers were intensive care and hospital length of stay, and ECMO-related procedure costs.ConclusionsECMO costs were high, with venovenous ECMO being most expensive due to an increased length of stay. Total patient cost increased with additional ECMO complications. Accurate ECMO costs and complications data can enable informed health budgeting and identification of areas for clinical improvement. (Trial registration: NCT03793257).Supplementary InformationThe online version contains supplementary material available at 10.1186/s13054-026-06086-7.Keywords: Critical care, Intensive care units, Costs and cost analysis, Artificial respiration Background Extra-corporeal membrane oxygenation (ECMO) is a form of cardiopulmonary support for patients with severe cardiac and/or respiratory failure that is refractory to conventional treatment. Although ECMO can be effective at improving short-term survival, it is resource intensive and associated with significant costs [1, 2]. Given the diversity of ECMO indications and patient population it is unsurprising that cost varies between ECMO patients and by mode [2]. A comprehensive investigation into the underlying drivers of these cost differences has not been undertaken to date. ECMO patients with complications have higher costs than those without complications [3]. Complications due to ECMO are common, though the prevalence and types of complications can also vary by ECMO mode [4]. Major ECMO complications include bleeding, equipment failure, and infection [4]. The aim of this costing analysis was to quantify the total patient cost of ECMO and determine the main cost drivers across three ECMO modes, including venovenous (VV), venoarterial (VA), and extracorporeal cardiopulmonary resuscitation (ECPR), with particular focus on the impact of complications on cost. Some of the results of these studies have been previously reported in the form of abstracts [5, 6]. Methods Study design and data collection A prospective observational clinical registry analysis was performed using data from a comprehensive binational registry of ECMO patients across 29 hospitals in Australia and New Zealand (the EXCEL registry; Trial registration: NCT03793257). Eligible patients were those consecutively enrolled in the EXCEL registry, were commenced on ECMO between 19 February 2019, the date of registry commencement, and 31 December 2022 and had been discharged from hospital prior to 31 December 2022. Patients without baseline demographic details or ECMO start dates at the time of data review were excluded from analysis. Patients receiving extracorporeal life support without oxygenation (e.g. patients with right ventricular assist devices only) were excluded. The EXCEL registry received ethical approval, including a waiver of consent for hospital data and opt-out consent for 180- and 365-day follow-up, from Monash University Human Research Ethics Committee 43,134, in accordance with the National Health and Medical Research Council’s National Statement on Ethical Conduct in Human Research (2007). The data collection processes associated with the EXCEL registry have been outlined previously [7]. Briefly, data were collected by trained research coordinators and data monitoring is also conducted to ensure data quality. Complications are reported within the registry from ECMO commencement until seven days following ECMO discontinuation or death according to the ECMO core data set [8]. Costing Healthcare resource utilisation was measured from hospital admission until discharge, using a healthcare provider perspective (Table 1). Surgical procedures other than those directly related to ECMO cannulation and decannulation and costs associated with hospital readmissions were not included in this analysis as they were not collected. Post-discharge data including location at 180-days and mortality at 180- and 365-days were identified based on long-term follow up. Further costing methodology is outlined in Additional File 2. Table 1Resources Costed by CategoryCosting CategoryCosting InclusionsICU CostDaily ICU cost weighted by number of organ supports (renal replacement therapy, vasopressors, mechanical ventilation) by length of stay, including readmission to ICUDaily additional staffing requirements whilst patients were on ECMO by length of stayDaily ECMO machine costs by length of stayWard CostDaily ward costs by length of stay, including ward days prior to transfer to primary ECMO care facilityRehabilitation CostTotal rehabilitation cost with adjustment for key complications (stroke, amputation, anterior spinal artery infarct)Emergency Department CostEmergency department admission cost by ECMO indicationProcedure CostEquipment and staffing for key ECMO-related procedures including central or peripheral cannulation and decannulation, cannula change and circuit changeRetrieval CostRetrieval by air or road accounting for whether patient was on or off ECMOIntra-Hospital Transfers CostStaffing cost for transfers to radiology and theatre whilst on ECMOBlood Product CostTotal blood product use costed per patient including red blood cells, platelets, cryoprecipitate, Prothrombinex, albumin, fibrinogen, fresh frozen plasmaECMO – extracorporeal membrane oxygenation; ICU – intensive care unit. Resources Costed by Category Daily ICU cost weighted by number of organ supports (renal replacement therapy, vasopressors, mechanical ventilation) by length of stay, including readmission to ICU Daily additional staffing requirements whilst patients were on ECMO by length of stay Daily ECMO machine costs by length of stay ECMO – extracorporeal membrane oxygenation; ICU – intensive care unit. Per patient costs were calculated by multiplying resource use by associated unit costs for each resource. A complete list of unit costs is presented in Additional File 3. All costs are reported in 2022 Australian dollars (AUD). Where required costs were adjusted using the Consumer Price Index [9]. Costs were not discounted due to the short (< 1 year) time horizon Outcomes The primary outcome of the analysis was total mean patient cost of ECMO by ECMO mode. The secondary outcomes included total mean cost of ECMO by number and type of complications. Statistical analysis Statistical analysis was performed using Stata MP 17 (College Station, Texas, USA). We performed a descriptive analysis and generated several mixed effects generalised linear models to examine the effects of complications on cost, adjusting for site as a random effect and adjusting for ECMO mode, sex, age, comorbidities and length of stay as fixed effects. We compared those with the complication versus those without. Both of these groups may have had other complications, so some patients may have contributed to multiple comparisons. Patients with missing data for a given variable were excluded from the corresponding model. Subgroup analyses were performed with patients who survived to hospital discharge and for costs incurred from ECMO commencement. We considered a p value of < 0.05 as statistically significant. Results The costing analysis included 1,345 patients who were commenced on ECMO in Australia and New Zealand between 2019 and 2022 (see Additional File 4 for PRISMA flow chart of exclusions, missing data and long-term follow up). There were 615/1,345 (46%) who underwent VA ECMO, 464/1,345 (34%) who underwent VV ECMO, and 266/1,345 (20%) who underwent ECPR. The sample population was predominantly male (890/1,345, 66%), with a median age of 52 years (interquartile range [IQR] 40 – 61 years) (Table 2). In-hospital mortality was significantly higher for patients in the ECPR cohort (181/266, 68%) than those who underwent VA (263/615, 43%) or VV (139/464, 30%) ECMO (p < 0.001). Mortality at 365 days did not significantly rise above in-hospital mortality, though there was notable missing data for long-term follow up (Additional File 4). Table 2Patient DemographicsTotal(n = 1,345)VA ECMO(n = 615)VV ECMO(n = 464)ECPR(n = 266)Age, years52 (40 – 61)55 (43 – 64)48 (36 – 57)52 (43 – 60)Sex, male66% (890/1,345)64% (393/615)66% (304/464)73% (193/266)Total days in hospital26 (11 – 48)27 (13 – 47)35 (20 – 59)8 (2 – 23)Total days in hospital, survivors38 (23 – 61)39 (25 – 60)42 (23 – 67)30 (20 – 50)Total days in ICU16 (7 – 31)15 (8 – 27)25 (13 – 45)6 (1 – 15)Total days in ICU, survivors21 (12 – 38)19 (12 – 31)28 (14 – 48)16 (10 – 23)In-hospital mortality43% (583/1,345)43% (263/615)30% (139/464)68% (181/266)Mortality at day 18047% (608/1,298)46% (273/593)34% (150/442)70% (185/263)Mortality at day 36552% (603/1,163)51% (273/532)38% (149/388)74% (181/246)Total days on ECMO6 (3 – 11)5 (3 – 9)9 (5 – 20)3 (1 – 6)Total days on ECMO, survivors6 (4 – 11)5 (4 – 9)9 (5 – 17)5 (3 – 7)Data are presented as median (IQR) for continuous measures and % (n/N) for categorical measures. Survivors only data is presented for patients who survived to hospital discharge (n = 762). Long-term follow-up data is presented with denominators for complete follow-up data. ECMO – extracorporeal membrane oxygenation; ECPR – extracorporeal membrane oxygenation; ICU – intensive care unit; VA – venoarterial; VV – venovenous Patient Demographics Data are presented as median (IQR) for continuous measures and % (n/N) for categorical measures. Survivors only data is presented for patients who survived to hospital discharge (n = 762). Long-term follow-up data is presented with denominators for complete follow-up data. ECMO – extracorporeal membrane oxygenation; ECPR – extracorporeal membrane oxygenation; ICU – intensive care unit; VA – venoarterial; VV – venovenous Intensive care unit (ICU) and hospital length of stay were significantly shorter in the ECPR group (median ICU length of stay 6 days, IQR 1 – 15 days; median hospital length of stay 8 days, IQR 2 – 23 days) compared to the VA group (median 15 days, IQR 8 – 27 days; median 27 days, IQR 13 – 47 days) and the VV group (median 25 days, IQR 13 – 45 days; median 35 days, IQR 20 – 59 days) (p < 0.001). This was consistent in the subgroup of survivors to hospital discharge with ECPR patients having significantly shorter median ICU and hospital lengths of stay (median 16 days, IQR 10 – 23 days; median 30 days, IQR 20 – 50 days) when compared to both VA patients (median 19 days, IQR 12 – 31 days; median 39 days, IQR 25—60) and VV patients (median 28 days, IQR 14 – 48 days; median 42 days, IQR 23 – 67 days). Notably in the subgroup of survivors to hospital discharge, ECPR patients spent a similar time on ECMO as VA patients (both median 5 days), with both groups having significantly shorter ECMO runs than VV patients (median 9 days) (p < 0.001) (Additional File 5). The mean total patient cost varied considerably across ECMO modes. VV ECMO was significantly more expensive (mean $287,264, SD $251,193) than VA ECMO (mean $215,128, SD $162,418) and ECPR (mean $121,274, SD $131,893) (p < 0.001) (Table 3 and Additional File 5). This remained true in the subgroup of survivors to hospital discharge. Subgroup analysis of cost from ECMO commencement found significant differences between mean cost across modes of ECMO with VV ECMO patients costing significantly more (mean $259,141, SD $237,882) than VA ECMO (mean $197,241, SD $156,180) and ECPR (mean $115,749, SD $129,218) patients (Additional File 5). However, the cost per day on ECMO was significantly lower in VV ECMO (mean $11,676 SD $14,067) compared to VA ECMO (mean $24,741 SD $81,301) and ECPR (mean $49,655 SD $177,021), indicating that the higher mean total cost in VV ECMO is driven by increased LOS (Additional File 5). ICU costs (including organ supports, daily ECMO machine costs and additional staffing requirements) made up the highest proportion of costs with 80% of VV ECMO costs, 69% of VA ECMO costs, and 70% of ECPR costs (Fig. 1), with additional staff and machine costs comprising 7% and 3% of overall ICU costs, respectfully (Additional File 5). Further analysis of the components of cost, both overall and by ECMO mode is shown in Additional File 5. Table 3Total Patient Cost By Survivors and By Number of ComplicationsTotal (n = 1,345)VA ECMO(n = 615)VV ECMO(n = 464)ECPR(n = 266)P valueTotal Patient Cost$221,452 ($201,787)$215,128 ($162,418)$287,264 ($251,193)$121,274 ($131,893) < 0.001Total Patient Cost For Survivors To Hospital Discharge$243,060 ($196,022)$228,601 ($159,312)$269,458 ($235,475)$202,005 ($151,342)0.003Total Patient Cost By Number of Complications0 (n = 576)$153,667 ($160,641)$159,641 ($122,196)$215,764 ($229,294)$72,433 ($69,606) < 0.0011 (n = 236)$231,266 ($199,025)$235,251 ($201,673)$260,622 ($201,009)$170,573 ($178,135)0.0642 (n = 154)$253,279 ($222,505)$223,945 ($116,100)$313,938 ($295,800)$147,033 ($105,300)0.0033 (n = 106)$288,369 ($200,047)$276,610 ($161,605)$346,254 ($227,892)$180,930 ($166,534)0.008 ≥ 4 (n = 273)$312,052 ($220,801)$292,423 ($174,075)$365,511 ($260,421)$208,410 ($169,296) < 0.001Data are presented as mean (SD)ECMO – extracorporeal membrane oxygenation; ECPR – extracorporeal membrane oxygenation; VA – venoarterial; VV – venovenous Total Patient Cost By Survivors and By Number of Complications Data are presented as mean (SD) ECMO – extracorporeal membrane oxygenation; ECPR – extracorporeal membrane oxygenation; VA – venoarterial; VV – venovenous Breakdown of ECMO Costs by Mode. ECMO – extracorporeal membrane oxygenation; ECPR – extracorporeal membrane oxygenation; ICU – intensive care unit; VA – venoarterial; VV – venovenous See Table 1 for inclusions within each subcategory of costing. See Additional File 5 in the online data supplement for costs for each category ECMO complications were common, with 57% (769/1,345) of patients having at least one complication and 40% (533/1,345) having more than one complication (Table 3). The main complications were infection, acute kidney injury and cardiovascular complications (including arrhythmia, cardiac tamponade, acute myocardial ischaemia and cardiopulmonary resuscitation). The mean total cost of patients without complications was $153,667 (SD $160,641), whilst patients with at least one complication had a mean total cost of $272,225 (SD $214,264). Patients with complications cost 78% more than patients without complications when adjusting for age, sex, ECMO mode and Charlson score (Fig. 2 and Additional File 5). Total Patient Cost By ECMO Mode and Complications – Cost Ratios From Mixed Effects Generalised Linear Model. *Adjusted for age, sex, Charlson comorbidity index.**Adjusted for age, sex, Charlson comorbidity index, ECMO mode. See Additional File 5) in the online data supplement for numerical cost ratios and confidence intervals Compared to patients without any complications, the additional cost for patients with one, two, three and four or more complications was 48%, 67%, 93% and 109%, respectively. When adjusting for hospital length of stay, costs still consistently increased by number of complications though to a lesser degree (one complication, 24%; two complications, 29%; three complications, 36%; four or more complications, 41%). Total patient cost significantly increased in patients with infection (mean $292,197, SD $217,806) compared to those without infection (mean $161,674, SD $165,175) (Additional File 5). Patients with neurological complications cost significantly less than those without neurological complications (mean $154,207, SD $134,178 versus mean $223,887, SD $203,434, respectively); however, among survivors to hospital discharge, costs were higher in those with a neurological complication though this was not statistically significant. Notably, costs did not significantly differ between patients with bleeding or cardiovascular complications or those who developed acute kidney injury. There were 28 patients in which ECMO was used as a bridge to transplant (Additional File 5). They had an average ECMO duration of 9.4 days (SD 7.3 days), which was not significantly different to patients who were not bridged to transplant. Despite this, patients bridged to transplant spent significantly longer in ICU and in hospital than other ECMO patients. In-hospital mortality for these patients was 29% (compared to 43% in the total sample population). The cost of those bridged to transplant was significantly higher ($396,262, SD $420,812) than those not bridged to transplant ($221,020, SD $201,308) (p < 0.001) (Additional File 5). Half (669/1,345) of the patients in this dataset underwent retrieval to be transferred to their primary ECMO centre. Patients who were transferred cost significantly more (mean $241,656, SD $197,475) than patients who were not transferred (mean $201,158, SD $204,656) (p < 0.001). Discussion In this retrospective analysis of 1,345 ECMO patients across Australia and New Zealand, we found that the mean total patient cost of an admission with ECMO was $221,452 (SD $201,787), with significant variation by ECMO mode. VV ECMO had the highest mean patient cost followed by VA ECMO then ECPR. Cost increased incrementally by number of complications. The main total patient cost drivers were length of stay (particularly in ICU), number of complications, and ECMO-related procedure costs. ICU costs represented the largest proportion of total patient costs (71%), largely driven by organ support costs (64%) rather than additional staffing (5%) or ECMO machine costs (2%). Ward stays, procedures and blood products represented 14%, 8% and 3%, respectively, of the total patient costs. The cost per day on ECMO was lowest for VV ECMO, followed by VA ECMO, with ECPR having the highest per ECMO day cost. Current literature on the cost of ECMO varies widely with total hospital costs ranging from $39,652 to $1,372,176 AUD 2022 per ECMO patient (inflated and converted from $25,533 USD 2019 and $1,009,617 USD 2024) [2, 10]. In the Australian context, we note a single costing analysis of 62 patients who underwent ECPR found a mean total patient cost of $87,744 AUD 2022 (inflated from $75,165 AUD 2016), with notable limitations due to reliance on Australian Refined Diagnosis Related Groups to assign costs [11]. This finding is substantially lower than our mean total ECPR patient cost of $121,274. This is likely due to our inclusion of rehabilitation, emergency department and intra-hospital trip costs. Additionally, our analysis is a granular assessment of costs based on individual patients’ daily resource use. This approach provides insight into the drivers for cost variations, specifically the role of complications and ECMO mode on costs. Patients on VV ECMO had significantly higher total costs than their VA ECMO and ECPR counterparts even among patients who survived to hospital discharge. Length of stay was one of the key cost drivers identified in this study. While the cost per day on ECMO was lowest in patients on VV ECMO, they had longer stays in hospital and ICU, highlighting the protracted recovery time required for those with significant respiratory illness. ECPR was the least expensive mode, predominantly driven by the shorter length of stay and higher mortality rates, though even among the subset of patients who survived to hospital discharge, ECPR remained the least expensive ECMO mode. Most ECMO studies—and ICU studies more broadly—measure mortality reduction as a primary outcome [12, 13]. Yet, most patients (66%) who survive ECMO have a moderate-to-severe disability at 180 days [7]. This study’s findings suggest that moving beyond mortality outcomes alone in ECMO research may benefit the healthcare system in terms of cost savings and patients in terms of complications and other factors that influence recovery and quality of life. Although ECMO probably confers a mortality benefit when compared to conventional cardiopulmonary resuscitation [14], the significance of complications on both patient outcomes and cost must not be understated. We have highlighted that cost increases with each additional complication. Given the costly nature of ECMO as an intervention, a proportional decrease in cost with prevention of a single complication could have meaningful cost savings. In our analysis, infection was consistently identified as a key complication associated with increased costs. Reduction in hospital-associated infections through continued emphasis on prevention at both a patient and systemic level would not only have clinical implications but also cost savings. We found that patients transferred to an ECMO centre, both prior to ECMO commencement and following commencement of ECMO at the primary centre, cost significantly more than their counterparts. With around half of the patients in our study being retrieved from an alternative centre, further investigation into models of care to reduce transfers and costs is clearly warranted. Data from 29 hospitals across Australia and New Zealand were included in this analysis, with a mix of high volume and low volume centres. This diverse population allowed ECMO costs and complications to be studied across a wide variety of clinical settings with different clinicians and likely variable ECMO protocols. We reviewed data from over 1,300 patients across four years. This sample size enabled comprehensive analysis of cost differences between ECMO modes. The specificity of costing by individual resource use is a step forward from the majority of current ECMO costing literature. This included costing duration of ECMO and other organ supports, time in ICU and on the ward, admission to the emergency department, inter- and intra-hospital transfers, blood product use, and specific ECMO-related procedures such as cannulation and decannulation. This analysis of ECMO-specific data from a binational registry of patients receiving ECMO lacked a non-ECMO comparator group, limiting the ability to determine costs and complications due to ECMO specifically versus those that are inherent to the underlying illness these patients present with. The limited time horizon of this costing analysis means that long-term data regarding complications and long-term morbidity and mortality was not captured. Further investigation into the long-term outcomes of ECMO, including costs incurred post-discharge and ongoing complications, would give a more comprehensive picture of the efficacy and outcomes of ECMO. An APACHE score is a useful measure for assessing illness severity in the general ICU population; however, SAVE and RESP scores are more useful in VA and VV ECMO patients. Neither SAVE nor RESP scores can be applied to the entire heterogenous ECMO cohort, and there was notable missing APACHE score data across the patient cohort. This limited our ability to use these scores for illness severity adjustment. Several key assumptions may have limited the accuracy of our costing. Specialist clinician estimates were used to determine average retrieval duration for retrievals in the Australian state of Victoria, meaning over- or under-estimation of retrieval costing is possible. Given the lack of data surrounding definitive rehabilitation discharge date, rehabilitation length of stay was based on a composite of data from the National Hospital Cost Data Collection for similar patient populations and location data at 180 and 360 days. Notably, cost of surgical procedures unrelated to ECMO cannulation or decannulation (including those relating to management of ECMO complications such as amputation) were not included in the scope of this analysis, meaning likely underestimation of ECMO-related complication costs. Costs were sourced specifically from Australian data with application and assumptions that may not fit New Zealand costs and practices. Retrieval services within New Zealand are notably different to those in Australia and, as such, these costs may not accurately reflect the New Zealand cohort. Given that even within ECMO modes there is significant population heterogeneity, investigating costs by diagnosis may be more useful. For example, bridge to transplant patients had significantly higher costs than those not bridged to transplant, with markedly longer ICU and hospital length of stay. Notably these patients did not spend longer periods on ECMO than their non-bridge to transplant counterparts. Assessing the cost drivers within patient groups such as this would provide a more accurate estimation of costs that could be used for health budgeting and resource allocation. Conclusion The use of ECMO continues to grow as a treatment option for patients with refractory cardiovascular or respiratory compromise. Total patient costs were heavily influenced by mode of ECMO, with VV ECMO patients having significantly higher total costs than patients who received VA ECMO or ECPR, despite lower costs per day on ECMO. The major cost drivers for patients receiving ECMO therapy was their ICU length of stay and use and duration of organ supports. ECMO complications were common and led to more time on ECMO, and more time in ICU, hospital and rehabilitation. Total mean patient cost increased with each additional complication. Supplementary Information Additional file 1. Additional file 2. Additional file 3. Additional file 4. Additional file 5. Acknowledgements The authors would like to acknowledge the contributions of the EXCEL Management Committee, consumers, outcome assessors, and collaborators for their assistance in bringing this publication to fruition. We would also like to acknowledge New South Wales Ambulance and Ambulance Victoria for their contributions to costing data. Thank you also to clinicians and costing staff from: St Vincent’s Hospital, Sydney; The Alfred, Melbourne; Barwon Health, Geelong; Eastern Health, Box Hill; Royal Melbourne Hospital, Melbourne; Prince Charles Hospital, Brisbane; Royal Prince Alfred Hospital, Sydney; Austin Hospital, Heidelberg; Gold Coast University Hospital, Southport for their contributions. A comprehensive list of those we wish to acknowledge can be found in the online Supplement. Abbreviations Australian dollars Extracorporeal membrane oxygenation Extracorporeal cardiopulmonary resuscitation Intensive care unit Interquartile range Venoarterial Venovenous Author contributions All authors contributed to the final manuscript with edits and review. KW drafted the initial manuscript and performed the preliminary data analysis. AMH conceptualised and supervised the project. LM and AMH were involved in initial draft reviews, assessment of results, and provided advice regarding data analysis and presentation. AB provided expert clinical insights and advice. BF was the primary liaison regarding the EXCEL registry data. CMH provided oversight as Chair of the EXCEL registry management committee. Funding CLH, AMH and AB are supported by an NHMRC Investigator Grant (#2033103, #2008447 and #2010110). The EXCEL database is funded by the Medical Research Future Fund, National Health and Medical Research Council, International ECMO Network, Heart Foundation, The Dicker Family, Alfred Health, Barwon Health, Critical Care Research Group, St Vincent’s NSW, Royal Prince Alfred Hospital, NHMRC CRE-ICU (2006514) Data availability The datasets analysed during the current study are available from the EXCEL registry. Access requests can be made on their website. Declarations Ethics approval and consent to participate The EXCEL registry received ethical approval, including a waiver of consent for hospital data and opt-out consent for 180- and 365-day follow-up, from Monash University Human Research Ethics Committee 43134, in accordance with the National Health and Medical Research Council’s National Statement on Ethical Conduct in Human Research (2007). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes The EXCEL management committee membership and the list of all collaborators is shown in Additional file 1. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Additional file 1. Additional file 2. Additional file 3. Additional file 4. Additional file 5. Data Availability Statement The datasets analysed during the current study are available from the EXCEL registry. Access requests can be made on their website.