Estimation of Outcome and Quality of Life in ECMO Patients

The availability as well as the use of extracorporeal membrane oxygenation (ECMO) systems for severe acute respiratory (veno-venous, vv-ECMO) or cardiocirculatory (veno-arterial, va-ECMO) failure is steadily increasing. Despite increasing experience of specialized centres, mortality of ECMO patients remains high and only about 50% survive the initial hospital stay. For those patients who do leave the hospital alive, quality of life after this invasive therapy with long stays in the intensive care unit (ICU) is often limited and participation in social life can be difficult. Quality of life and the life-impact after ECMO therapy is, in contrast to hard endpoints such as mortality, insufficiently studied and currently only scarce data exist from large prospective cohort studies. Further, predictive scores and associated risk factors for patient-centred outcomes are not available.

The decision to initiate ECMO therapy is predominantly made in emergency situations, for which the focus is on acute survival. As such, the long-term implications in terms of quality of life and life-impact of ECMO treatment enjoy only limited consideration at an early time point. Further, the existing scores for prediction in ECMO were developed for mortality and reliable data on long-term life-impact are scarce. Therefore, while these decisions are influenced by empirical factors like patient-age, point-of-care laboratory parameters (e.g. lactate) or the neurological status of the patient, the clinical experience and educated guess in terms of prognosis and potential treatment futility by the attending team remains a crucial factor. This applies not only to ECMO initiation but also to decisions on continuation and termination. This approach to decision-making may be problematic since it has been shown for other settings that clinicians tend to overestimate the success of an intervention. Currently, there are no data evaluating in how far this also applies to ECMO therapy and if indeed, there is a mismatch between estimated and observed outcomes in ECMO patients. Considering the crucial role of subjective prognosis estimates, it becomes of major interest to quantify the potential mis-calibration between clinicians estimated and observed outcomes. Further, factors like Outcome uncertainties, potential doubts regarding treatment utility vs futility, and the immediate finality of these demanding decisions expose ICU health-care personnel to a relevant psychological burden. As shown by Johnson-Coyle and colleagues, both moral distress and burnout have a negative impact on job satisfaction. Moral distress occurs when one believes to know what is ethically right but something or someone limits their ability to do the right thing. Preliminary studies have shown a high incidence of moral distress in the care of patients with mechanical circulatory support systems, with particularly pronounced stress among nursing personnel. If these factors have significant impact on professional judgement is not clear in this context.

In a small single centre pilot study, we prospectively recruited 50 va-ECMO patients at the University Hospital Duesseldorf from March until November 2023 and investigated if ECMO care providers could predict in-hospital mortality in these patients. For these 50 patients we obtained 135 completed questionnaires within 24 hours and 111 answered questionnaires at day 4 to7 after initiation of ECMO therapy from consultants, residents and nursing personnel. Out of 50 patients 21 patients (42%) died during the initial hospital stay. Overall sensitivity and specificity of estimates were 57.9% and 85.9% respectively at 24 hours after start of ECMO therapy (precision: 75%, accuracy: 74.1%, F1 score: 65.3%). In a subgroup analysis, consultants showed highest agreement of estimated and actual in hospital mortality, whereas residents showed lowest agreement (see table below). At day 4 to 7 overall predictions had lower sensitivity, accuracy, precision and F1-score as compared to estimates on day 1 after ECMO initiation, however specificity slightly increased (sens: 35% spec: 91.5% acc: 71.1% F1-score: 46.7% precision: 70%). Highest values for sensitivity, accuracy and F1-score, were reached in subgroup of nursing personnel (see table below). Notably, years of experience in critical care were higher in consultants and nursing personnel as compared to residents (consultants: 10.8 ± 6.7 years versus residents: 1.4 ± 1.5 years versus nursing personnel: 15.8 ± 10.4 years).

Based on the results of our pilot data, we hypothesize that:

1. Subjective prognosis estimates by health care providers underestimate (low sensitivity) actual mortality. We expect that this will also account for reduced functionality and quality of life.
2. Estimations will vary significantly between different groups of health care professionals and levels of experience in critical care as well as between different time points of assessment. Specifically we expect, that prediction accuracy will improve with higher years of experience in critical care and that predictions for mortality will be more accurate immediately after ECMO initiation as compared to later study time points. This might be influenced by the fact that a substantial number of patients with poor prognosis will die within the first four days of ECMO therapy. Hence, estimation of outcome in the remaining patients is more difficult.
3. Agreement between estimated and observed outcomes will differ significantly between centres with different levels of experience (high-volume versus low-volume centres). Expecting higher accuracy of outcome estimates in high-volume centres with higher experience as compared to low-volume centres.
4. Additionally, we expect that high levels of moral distress influence professional judgement leading to negative estimated prognosis and low precisions of estimations.

Therefore, we aim to conduct the ESTRELLA study as large nationwide multicentre prospective cohort study to dissect usability of clinical estimates for outcome prediction in ECMO patients and to identify suitable factors for prediction of poor functional health in these patients.