Delirium, despite its common occurrence in the ICU, remains ineffectively prevented and treated due to its multifactorial etiology and poorly understood pathophysiology. Our study confirms that the condition is nearly ubiquitous in the severe COVID-19 ARDS patients requiring VV-ECMO who have assessable neurological examinations. Furthermore, those who ultimately survived and those who did not, similarly, spent the majority of their VV-ECMO days heavily sedated or comatose, thereby limiting the role of delirium to represent the overall neurological status in our patients. With these provided restrictions, we utilized the proportion of days on VV-ECMO with delirium as a normalized measure to demonstrate that delirium was found more often when the patients were more lightly sedated, had more pain, and had fewer days in a coma or on neuromuscular blockade. A reasonable interpretation is that less sedation exacerbates pain and discomfort (i.e., patient-ventilator dyssynchrony), which can worsen delirium severity. Another explanation is that heavier levels of sedation masks clinical manifestations of delirium, as suggested by the similar total delirium days but the delayed median delirium days in the non-survivors over their more protracted ECMO therapy. COVID-19 patients in the ICU are known to have higher sedation requirements, likely related to younger age with fewer chronic comorbidities, high respiratory drive, and intense inflammatory responses linked to tolerance26. Paralytics were required in every patient in our cohort, as was proning for nearly all patients prior to cannulation to treat COVID-19 ARDS, and these extensive treatments were accompanied by heavy sedation. In this very sick population with both survivors and non-survivors’ mean RASS scores of below − 3 (moderate sedation), delirium characteristics were mostly similar in proportion to ECMO duration, and the detected duration of delirium did not predict mortality in our multivariable logistic regression analysis.
Our critically ill patients expectedly had a common occurrence of delirium, a well-known predictor of poor outcomes in the ICU2,4,8,10. Our finding in a relatively young cohort is consistent with a recent large-scale study on delirium in acute heart failure patients, in which delirium was similarly common in patients younger than 65 years old10. This highlights the importance of managing the underlying acute medical illness even in younger and healthier patients with a good cognitive reserve. Non-survivors and survivors had similar delirium characteristics and dosing of analgosedatives, antipsychotics, and paralytics but ended up being on VV-ECMO longer with more delirium-unassessable days. We found that non-survivors had trended towards lower RASS scores, closer to -4 (deep sedation), in the setting of the requirement of prolonged ECMO therapy. Their poorer neurological status can be related to the complicated medical course from severe COVID-19 (i.e. ARDS requiring anti-IL-6 therapy), acute cardiac dysfunction10, fatal ICH, and iatrogenic from heavy sedation, ventilator, and ECMO therapies. Secondary infection was similarly common in the two groups, but its severity can be variable. ICH, a known complication of VV-ECMO, occurred at a comparable frequency in both survivors and non-survivors, but five of the cases were fatal. Persistent or resolved delirium after ICH is strongly associated with unfavorable outcomes8, and, given the need for anticoagulation and the high incidence of ICH in patients on VV-ECMO, promoting regular neurological evaluations starting early in their hospital course, when feasible, should facilitate both diagnosis and treatment.
The authors believe that our analgosedation strategy was an effective way to manage severe COVID-19 ARDS, especially early in the hospital course. However, it inevitably clouded patient’s neurological status. While the mechanistic role of delirium here is not well elucidated, promoting lighter sedation may lead to improved neurological outcomes and reduced mortality. In two previously reported multicenter, observational studies, heavy sedation was found to predict mortality but was not associated with delirium, which occurred in 51% and 44% of patients11,13. Another study, for which the incidence of delirium was not reported, also identified a significant association between heavy sedation and mortality12. According to the Extracorporeal Life Support Organization, sedation to the point of light anesthesia should be the goal during cannulation and all sedation and opioids should be stopped for interval neurological exams27. The elevated medication clearance rate while on ECMO and the needs to achieve ventilator synchrony, optimize ECMO flows, lower metabolic demand, as well as early liberation28 make it difficult to lighten sedation.
There is no standard analgosedation practice for COVID-19 patients on VV-ECMO. In a recent publication on COVID-19 patients on VV-ECMO for predominantly moderate to severe ARDS, higher doses of ketamine, benzodiazepine, propofol, and dexmedetomidine, and less use of opioids were noted over the first seven days compared to ours over the full course of VV-ECMO therapy29. There is emerging evidence related to ECMO, such as the reduced dose of hydromorphone use due to its lower affinity to ECMO circuit compared to fentanyl27,30. More studies are needed to better understand the pharmacokinetic properties of commonly used agents over the course of ECMO therapy. We performed exploratory, linear regression analysis between RASS scores and the daily median doses of commonly used medications while on VV-ECMO, and medications appear to have variable effects on sedation (Table S1). Further studies are needed to validate analgosedation strategies (i.e., an awake VV-ECMO strategy31, use of adjunct agents29, incorporating non-pharmacological interventions for delirium prevention32, patient-specific factors to consider in ECMO27,33) as well as implementation of novel, practical management approaches for the unstable patients (i.e., multimodal neuro-monitoring24, bedside brain MRI34).
Our study results should be interpreted with caution. It is limited by its retrospective, observational design, which is subject to unmeasured confounding effects. We only included a specific population of COVID-19 patients requiring VV-ECMO support, so our findings are not generalizable to other ECMO populations (i.e., venoarterial ECMO). On the other hand, there was a consistency in the profile of analgosedatives and multimodality analgesia per the institution’s protocol. For instance, ketamine has not been shown to effectively prevent delirium3,5, but it was used in all our patients, given its safety profile and known benefits for hemodynamic support. The BPAS and RASS scores differed by less than 1 point between survivors and non-survivors, and the differences were likely of unclear clinical significance in daily clinical decision-making. Correlation does not imply causation, and the authors acknowledge that our interpretations based on clinical reasoning may not hold true. There were certain factors (i.e., mean RASS, fatal ICH) that showed near-significant trends between survivors and non-survivors, and although the study is not primarily designed to compare them, having a larger sample size could have resulted in statistical significance. Details on ventilator and ECMO management, withdrawal of care, outcomes post-discharge were not captured in this study. Utilization of alternative outcome scales, such as the CAM-ICU-716 and the Intensive Care Delirium Screening Checklist35, could provide additional details on delirium that may guide clinical management.