Update in Critical Care 2020

In this review, we highlight the major scientific output in 2020 related to critical care in American Thoracic Society journals. Scientific focus surrounded acute respiratory distress syndrome (ARDS), sepsis, acute kidney injury (AKI), ICU organization, post-ICU recovery, critical care research, and education. This review focuses predominantly on original research articles and related commentary, but there were also several excellent review articles (1–3) and letters to the editor (4–12) related to critical care. Furthermore, there is a dedicated update on coronavirus disease (COVID-19), which also contains pertinent information related to critical care.

ARDS

ARDS Pathogenesis

The identification of subphenotypes of ARDS has led to a more nuanced approach to clinical trials, but biomarkers to identify subphenotypes are often not available for clinical use. In a secondary analysis of the NHLBI ARDS Network clinical trials, a machine learning technique called “gradient-boosted machines” using routinely available clinical laboratory parameters reliably reidentified hyperinflammatory and hypoinflammatory subphenotypes initially identified using the latent classifier approach (13, 14). Genetic predisposition to ARDS may be an important factor for future clinical trials. In a Mendelian randomization study on two well-phenotyped cohorts of critically ill patients with sepsis and ARDS, plasma sRAGE was genetically determined and identified as a potential causal factor in sepsis-associated ARDS (15). Logical next steps would include studying patients at risk for ARDS (16).

The renin–angiotensin system (RAS) has been implicated in the pathogenesis of ARDS. A normal angiotensin I–to–angiotensin II ratio was inverted in individuals with early ARDS but not in surgical control subjects, and it was associated with poor outcomes in patients with ARDS (17). In addition, biomarkers of the alternative RAS pathway stimulated by neprilysin, Ang 1–7, and Ang 1–5 were higher in patients with ARDS than in control subjects, and RAS dysregulation was noted to persist in later ARDS.

Lung repair after ARDS continued to be a focus in preclinical studies. Regulatory T cells (Tregs) are elevated in human blood and BAL fluid from patients with ARDS; however, the mechanism by which Tregs alter alveolar type II IFN responses remains unclear. Using gene ontology analysis in an animal model of ARDS, gene sets composing type I IFN, intracellular transport, and cellular localization were upregulated in the Treg-depleted samples (18). Tregs may be a potential therapeutic target in the resolution phase of ARDS (19). Permissive hypercapnia may have unintended consequences for cellular metabolism and may slow down wound healing in epithelial monolayers through the inhibition of Rac1/Cofilin via AMPK (20). Upregulation of NF-κB suppressed CXCL12 and activated Rac1 via the binding to CXCR4 and prevented the adverse effects of hypercapnia in an orthotopic tracheal transplantation model of wound healing (20).

Lung-Protective Ventilation and Ventilator-induced Lung Injury

Positive end-expiratory pressure (PEEP) management in ARDS remains controversial, with a heterogenous response being based on lung recruitability. Through a physiologic study in 45 adults with ARDS, a single-breath method to assess recruitability by calculating the recruitment-to-inflation ratio provided a simple way to identify patients who may benefit from a higher-PEEP strategy (21, 22). The addition of expiratory resistance in both spontaneously breathing and mechanically ventilated pigs with lung injury resulted in less heterogeneous expiratory time constants, preventing end-expiratory lung collapse. The method may provide an alternative or adjuvant approach to PEEP management to prevent atelectrauma and regional overdistention in heterogenous lung disease (23, 24).

A pilot study demonstrated that a strategy to limit driving pressure in adult patients with ARDS, compared with conventional low Vt ventilation, was feasible (25). Although providing a foundation for future trials, deep sedation or paralysis was required and may limit its widespread use (26). Mechanistically, stretch-related ventilator-induced lung injury may be decreased by the activation of Piezo1, a nonselective cation channel that allows the influx of calcium and sodium into endothelial cells, stabilizes endothelial adherens junctions and lung vascular barriers (27), and activates calpain (28).

Patient Self-inflicted Lung Injury and Ventilator-induced Diaphragm Dysfunction

Although ventilator-induced lung injury has been well recognized over the last several decades, it is becoming increasingly clear that lung injury can occur when patients have a strong respiratory drive, a term collectively known as patient self-inflicted lung injury (P-SILI). P-SILI is exacerbated when patients achieve high Vts with correspondingly high transpulmonary pressure, particularly when lung compliance is impaired. Strategies have been developed to identify patients with significant respiratory effort who are at high risk of P-SILI, both on invasive and noninvasive mechanical ventilation. It is also important to identify mechanically ventilated patients with very low respiratory effort, as this is a risk factor for ventilator-induced diaphragm dysfunction.

Through secondary analysis of 132 mechanically ventilated patients with moderate-to-severe ARDS, neuromuscular blockade was only beneficial in the subgroup with higher baseline diaphragm thickness and was harmful in the subgroup with lower baseline diaphragm thickness, suggesting that patients with higher baseline thickness may have the strength to generate more injurious transpulmonary pressures and are at higher risk of P-SILI (29).

Airway occlusion maneuvers during expiration or inspiration can be used to estimate patient effort or driving pressure. As compared with esophageal pressure, excessively high or low respiratory effort can be reliably detected using the ventilator-calculated change in pressure during the first 100 ms of a breath during airway occlusion (30, 31). During assisted ventilation on pressure support, driving pressure measured during inspiratory or expiratory airway occlusion maneuvers are highly correlated with one another. Specifically, the negative deflection in airway pressure during an expiratory hold multiplied by 0.75 plus the set pressure support level represents the dynamic driving pressure. This dynamic driving pressure was found to be nearly identical to the static driving pressure measured as the difference between the plateau pressure during an inspiratory hold and the PEEP level, after controlling for the airway resistance.

Tissue doppler imaging of the diaphragm using ultrasound is a novel technique to monitor respiratory effort and identify patients likely to experience weaning failure. The technique was highly reproducible and identified differences between patients with weaning failure and patients with weaning success or healthy volunteers, and parameters were correlated with gold standard measures of respiratory effort, such as the transdiaphragmatic pressure time product (33, 34). Electrical activity of the diaphragm, often used to measure patient effort and gauge patient–ventilator synchrony, may have limitations in detecting ineffective respiratory effort, as compared with esophageal manometry or surface electromyography (35).

In patients with hypoxemic respiratory failure treated with noninvasive ventilation (NIV), strong spontaneous breathing effort (large swings in esophageal pressure) were associated with worsening of lung injury on chest radiographs and an increased risk of intubation. Minimal reduction in esophageal pressure swings (<10 cm H₂O) after 2 hours of NIV was more strongly associated with intubation than large expiratory Vts; the Heart Rate, Acidosis, Consciousness, Oxygenation, and Respiratory Rate score; or other physiologic parameters (36, 37). Higher pressure support was strongly associated with decreasing spontaneous breathing effort (38).

In a physiologic crossover study on 15 patients with acute hypoxemic respiratory failure, use of high-PEEP helmet NIV was associated with improved oxygenation and work of breathing compared with use of a high-flow nasal canula (HFNC). Patients with high inspiratory effort during HFNC use were more likely to have improvement with helmet NIV use, but patients with low inspiratory effort during HFNC use had increased transpulmonary pressure swings during helmet NIV use, which may have resulted in larger and harmful Vts (39, 40). Although Vt is generally not measurable during helmet NIV use, it may be accurately measured by using a turbine-driven ventilator connected to the inspiratory port of the helmet by using a single-limb circuit with an intentional calibrated leak from the expiratory port of the helmet (41).

Adjuvant ARDS Therapies and Extracorporeal Membrane Oxygenation

Mesenchymal stem cells have been a recent focus in ARDS. The 1-year and 5-year follow-up results of two patients who had received a single intravenous infusion of allogenic bone marrow–derived mesenchymal stromal cells were reported and showed full resolution of lung parenchymal injuries on computed tomography scans (42). In an animal model of ARDS treated with extracorporeal membrane oxygenation (ECMO), endobronchial infusion of induced pluripotent stem cell–derived human mesenchymal stem cells was associated with decreased lung inflammation, histologic lung injury, and shock severity. However, no differences in oxygenation or respiratory mechanics were observed, and stem cells were associated with pulmonary arterial thromboses and impairment of the membrane oxygenator (43, 44).

Adjuvant therapies for ARDS are used with significant variability among institutions. An observational study conducted in 100 international pediatric ICUs identified that continuous neuromuscular blockade was used in 30% of patients with pediatric ARDS and in 50% of patients with severe pediatric ARDS, whereas the remainder of other adjuvant therapies were used less than 15% of the time. There were significant regional and institutional differences in the application of adjuvant therapies and their order of implementation (45, 46). Some of these institutional differences may relate to comfort with the therapy and training, particularly for prone positioning. According to a survey conducted in 54 acute care hospitals in Massachusetts, only 44% of hospitals were “prone ready,” and the absence of a protocol for prone positioning was associated with the use of other therapies (47).

Data from the Extracorporeal Life Support Organization registry suggested that a relative drop in Pa_CO2 > 50% in the first 24 hours of ECMO was associated with a 70% increased risk of neurologic complications. These findings suggest that Pa_CO2 or the related cerebral vasoconstriction should be monitored after ECMO initiation to guide sweep gas flow (48, 49). Interestingly, Bourcier and colleagues (50) reported favorable outcomes in four out of five young patients with refractory obstructive shock secondary to mediastinal compressive tumors with ECMO and chemotherapy.

Translational and Preclinical Studies on the Lung

Electronic cigarette– or vaping-associated lung injury is a recently recognized cause of ARDS in young individuals. Vitamin E acetate caused dose-dependent increases in pulmonary edema as well as organizing and lipoid pneumonia in a mouse model of electronic cigarette– or vaping-associated lung injury through the upregulation of the gene expression of inflammatory pathways and through a direct toxic effect on alveolar type II cells (51). α-Tocopherol, a metabolite of Vitamin E acetate is a free radical scavenger that has been proposed to treat ARDS. Pretreatment of endothelial and epithelial cells with α-tocopherol reduced paracellular permeability induced by Pseudomonas aeruginosa exoenzyme (type III secretion system) and prevented mortality in a mouse model of P. aeruginosa pneumonia (52).

Severe meconium aspiration syndrome (MAS) is a neonatal form of ARDS. A translational study using nonbronchoscopic BAL fluid from neonates with MAS compared with that of control subjects found that MAS was associated with significant changes in the surfactant phospholipid profile, with an increase in lysophosphatidylcholine and surfactant proteins B and C being shown. In addition, MAS was also associated with alteration of surfactants’ nanostructure, possibly due to concurrent lung inflammation (53).

Viral and bacterial products activate the innate immune system through TLR (Toll-like receptor) signaling. TLR8 protein levels were lower in blood samples of patients with ARDS compared with blood samples of healthy control subjects. The authors discovered that E3 ligase RNF216 (ring finger protein 216) ubiquitinates TLR8 in response to microRNAs. Gene expression profiles obtained from patients with ARDS had reduced RNF216. Furthermore, RNA extracted from plasma of human ARDS samples activated TLR8 in a macrophage cell line in vitro, suggesting that TLR8–RNF216 signaling plays an important role in ARDS (54) and that the hyperinflammatory phenotype is perhaps due to reduced RNF216 (55). The time course of immune system activation is critical in the development of ARDS. Harnessing epithelial antiviral properties has the potential to decrease the pulmonary viral burden and hence the occurrence of ARDS. Inhaled Pam2-ODN, a TLR 2/6 and 9 agonist, reduced the lung Sendai virus (the murine form of the human parainfluenza virus) burden and decreased mortality in mice. In addition, Pam2-ODN blocked the increase in CD8⁺ T cells induced by viral challenge, leading to reduced mortality (56). However, the mechanism for viral inactivation due to Pam2-ODN and its effects on established viral pneumonia are unknown (57).

Respiratory viral infections are common in children and can be associated with poor outcomes. In neonates with respiratory syncytial virus, higher disease severity was related to CD8⁺ T cells expressing IL-4 (Tc2) and reduced proportions of CD8⁺ T cells expressing IFNg (Tc1) in nasal aspirates. Less severe disease was associated with greater frequencies of Tc1, CD8⁺ T cells expressing IL-17 (Tc17), and CD4⁺ T cells expressing IL-17 (Th17), highlighting the importance of T-cell subsets in the pathogenesis and severity of respiratory syncytial viral infection (58). The nasal washing cytokine composition of pediatric allogeneic hematopoietic cell-transplant patients in the absence of viral respiratory tract infection is different from that of healthy control subjects and produced higher concentrations of IL-4, IL-12p40, and IL-1ra during a viral infection (59).

The lung microbiome was a focus of several investigations. Using droplet digital PCR and bacterial 16S quantification and sequencing in mini-BAL fluid before the receipt of antibiotics, differences in the lung microbiome were predictive of ventilator-free days and were associated with the diagnosis of ARDS (60). Enterobacteriaceae predominance was a risk factor for poor outcomes (61). Similarly, three clusters of endotracheal aspirate flora and two clusters of oral flora were identified in ICU patients with respiratory failure before administration of antibiotics. Individual clusters were associated with bacterial diversity, the prevalence of pathogenic bacteria, inflammation, and outcomes (62). Together, these findings suggest that the alteration of the airway microbiome of ICU patients may be a modifiable factor or therapeutic target (63).

The incidence of bronchopulmonary dysplasia is increasing. Maternal vitamin D deficiency impaired lung development in rats, causing outcomes including a simplified distal lung structure and higher resistance and airway hyperactivity in experimental animals compared with control animals (64). These changes mimicked those seen after postnasal hyperoxia exposure and may be explained by an abnormal reduction of HIF-1α/VEGF signaling (64, 65). Basal autophagy is impaired during neonatal hyperoxia-induced lung injury and in a baboon model of bronchopulmonary dysplasia. Autophagy-deficient mice had reduced alveolarization, increased expression of proinflammatory genes, and enhanced NLRP3 inflammasome activation in alveolar macrophages, increasing their susceptibility for neonatal hyperoxia-induced lung injury (66).

Other Studies of Mechanical Ventilation

Ciliary dyskinesia was common in mechanically ventilated patients suspected of having pneumonia and was associated with the severity of respiratory failure but not with ventilator-associated events or outcomes (67). The prevalence of probable Aspergillus infection using the algorithm proposed by the AspICU investigators (68) was as high as 12% in patients without neutropenia who had suspected ventilator-associated pneumonia and was associated with longer ICU stays but was not associated with higher mortality (69, 70). In contrast, invasive tracheobronchial aspergillosis, as suggested by the presence of endotracheobronchial lesions in patients with Aspergillus infection, was diagnosed in 29% of patients with severe influenza and pulmonary aspergillosis, and their in-ICU mortality rate was 90% despite antifungal treatment (71, 72).

From 2010 to 2017, NIV was used as the first-line treatment of asthma exacerbation in 25% of cases and increased over time. Although it failed in 22% of cases, use of NIV was independently associated with lower risks of intubation and mortality (73, 74). On the basis of administrative diagnosis data of ICU admissions in Australia and New Zealand, it was shown that the percentage of ICU admissions with acute exacerbations of chronic obstructive pulmonary disease increased between 2005 and 2017 but that mortality decreased (75, 76).

Among patients with respiratory failure requiring mechanical ventilation in Taiwan, more than 90% of successful weaning outcomes occurred in the first 30 days of mechanical ventilation, and the probability of death surpassed that of weaning after the 27th ventilator day (77). It remains to be seen whether the online prognosis tool derived from this study informs shared decision-making and whether these results are generalizable (78). The characteristics of high-performing centers specialized in the management of patients with prolonged mechanical ventilation included hospital leadership engagement in quality improvement, adequate staffing, detailed yet flexible protocols to increase the autonomy of bedside providers, and interdisciplinary team meetings. Key factors associated with effective care included the coordination among caregivers to balance aggressiveness and responsiveness and the involvement of patients and their families in decision-making (79, 80).

Delirium is common among ICU patients, particularly among those on mechanical ventilation. In a secondary analysis of two randomized controlled trials on delirium management, five trajectories of delirium, integrating symptom severity and duration, were identified as being associated with patient characteristics and 30-day mortality (81).

In ventilated patients, the optimal nutritional target remains as a subject of debate because delivery of 100% or 70% of the recommended calorie intake in the acute phase of critical illness did not result in significant differences in the long-term mortality, disability, quality of life, or return-to-work rate of survivors after 6 months (82, 83).

Sepsis

Clinical Studies in Sepsis

Hydrocortisone, high-dose ascorbic acid, and thiamine therapy has been considered in septic shock (84). From 2015 to 2018, its use in adults admitted to the ICU for septic shock increased from use in 0.03% of patients to use in over 2.5% of patients. There was marked heterogeneity in its use among centers, with use in up to 20% of patients in some hospitals, especially in patients with more severe organ dysfunction. In a retrospective cohort analysis in adults, its use was associated with worse outcomes despite extensive multivariable modeling and propensity matching (85). In contrast, in a single-institution pediatric study, hydrocortisone, high-dose ascrobic acid, and thiamine therapy were associated with improved clinical outcomes when compared to control therapy after propensity score-matched and inverse probability-weighted analyses (86). Differences in patient selection, criteria for allocation of treatment, and unmeasured confounding likely contribute to the seemingly conflicting results of these two studies. Interestingly, baseline levels of cortisol, aldosterone, or ascorbic acid were not related to the treatment effect of hydrocortisone on mortality, shock resolution, or other clinical outcomes in a secondary analysis of the Adjunctive Corticosteroid Treatment in Critically Ill Patients With Septic Shock (ADRENAL) trial, a large, multinational, randomized, placebo-controlled trial of stress steroids in septic shock (87, 88).

In a secondary analysis of the randomized controlled trial of exogenous angiotensin II in vasodilatory shock (ATHOS-3) (89), the response to treatment was heterogeneous. Higher renin levels were associated with kidney injury and with a higher angiotensin I/II ratio (90). Compared with placebo treatment, treatment with exogenous angiotensin II improved outcomes in the subgroup of patients with higher renin levels (91).

Through a Bayesian reanalysis of the randomized clinical trial comparing peripheral perfusion to a lactate-clearance resuscitation strategy (ANDROMEDA-SHOCK), there was a >90% likelihood that the peripheral perfusion strategy was better, using a wide range of assumptions about the prior probability. Bayesian networks were also used to analyze patterns of change in Sequential Organ Failure Assessment (SOFA) scores, with death as a competing risk (92, 93).

Premorbid (baseline) blood pressure influenced the duration of vasopressor medication use in the ICU (94). Compared with normal premorbid blood pressure, a lower premorbid blood pressure was associated with a longer duration of vasopressor use and higher mortality, whereas higher premorbid blood pressure was associated with a shorter duration of vasopressor use, suggesting that differential blood pressure targets stratified by premorbid blood pressure should be evaluated (95). In addition, the depth and duration of the blood pressure deficit during critical illness compared with a premorbid measurement was associated with risk of a major adverse kidney event (96, 97).

Methods to improve delivery of care to patients with sepsis was a focus. Compliance with a 3-hour sepsis treatment bundle varied between daytime and nighttime and between emergency and nonemergency areas, suggesting the need to incorporate the setting of diagnosis into quality-improvement initiatives (98). In a pilot trial, coated devices (endotracheal tubes, central venous catheters, and urinary catheters) were found to be possibly associated with more days alive and free of antibiotics (99). The provision of palliative care consultation for older patients who died with a diagnosis of septic shock was associated with shorter lengths of stay and lower hospital costs compared with no provision of palliative care (100). Sepsis-associated mortality rates improved in patients with cancer but not in patients without cancer (101), highlighting the need to understand the interplay between cancer and sepsis and the validity of administrative data surrounding cancer and sepsis (101, 102). In selected patients labeled as being allergic to penicillin who had a history of low-risk clinical manifestations, direct oral amoxicillin challenge without prior desensitization or skin testing was shown to be safe and allowed for the removal of inaccurate penicillin allergy labels (103).

Diagnostic tools in sepsis may need modification to achieve optimal performance in resource-limited settings. In a prospective study on patients with suspected infection in Brazil, a quick SOFA (qSOFA) score ≥1 predicted mortality with a sensitivity of 84.9% (vs. 53.9% for a qSOFA score ≥2). In resource-limited settings, using a SOFA score ≥1 or using a qSOFA score ≥1 plus the lactate status to identify any organ dysfunction could be alternatives to the usual method of identification using a qSOFA score ≥2 (104, 105). In patients with sepsis hospitalized in Malawi, sonographic lung B-lines indicating supleural interstitial edema were strongly associated with eventual hypoxemia (area under the curve, 0.86 [95% confidence interval, 0.69–0.97]), with a dose–response increase in the number of B-lines and a dose–response decrease in oxygen saturation occurring with each liter of intravenous crystalloid administered (106).

Translational Studies in Sepsis

Compared with placebo and vasopressin administration, norepinephrine administration increased IL-10 levels in human leukocytes, increased bacterial growth after cecal ligation in a mouse model, and altered the cytokine response to an LPS challenge in healthy patients (107). However, the immune balance of patients admitted with sepsis remains incompletely understood, and large randomized trials have not demonstrated increased harm with norepinephrine administration (108).

In a proof-of-concept study in pediatric patients with sepsis, a microfluidic multianalyte system had acceptable agreement with biomarkers measured with standard ELISA techniques, which were associated with mortality. This may facilitate nearly real-time assessment of biomarker profiles for precision medicine approaches and trials (109).

Analysis of the differential RNA expression in the cortical gray matter of older patients with dementia who died of infectious versus noninfectious causes revealed clusters of gene transcripts that were associated with infection, including classical innate immune-related genes, damage-associated molecular patterns, complement factors, and cytokines (110). These results provide new targets for sepsis-associated brain injury research.

Higher levels of alternative complement pathway activity in the serum of critically ill patients were associated with higher levels of other complement-associated proteins, a hypoinflammatory phenotype, and better outcomes (111). Differential complement activation may represent differing host complement reserves or differential sampling (112).

AKI

Early identification of patients likely to develop AKI may have therapeutic implications. The development of severe kidney injury after ICU admission could be accurately identified through the measurement of admission serum creatinine and the Pediatric Sepsis Biomarker Risk Model (PERSEVERE) biomarker panel within the first 24 hours of sepsis diagnosis. This may help identify patients who may benefit from renal protective interventions (113, 114).

Patients with high serum proenkephalin (a renal biomarker) had a higher mortality rate than those without, but this mortality rate remained lower than that of those with overt AKI, suggesting that proenkephalin may help identify subclinical kidney injury (115, 116). Through secondary analysis of a randomized clinical trial comparing resuscitation strategies for patients at risk of septic shock, the product of tissue inhibitor of metalloproteinases-2 and insulin-like growth factor–binding protein 7 in urine was associated with the up-front severity of illness and was predictive of Kidney Disease Improving Global Outcomes stage 3 kidney injury, renal replacement therapy, or death within 7 days (117, 118). The genetic predisposition for developing AKI, associated with SNPs at the RF2 and TBX1 loci, was not confirmed in a large Finnish ICU population (119), tempering the enthusiasm for genome-wide association studies that do not have rigorous prior evidence for implicated pathways.

Using principles of the Stewart strong ion difference, an extracorporeal mechanism to remove blood chloride successfully decreased blood chloride, improved pH and urinary chloride and ammonium excretion, and was well tolerated in an animal model of respiratory and metabolic acidosis without shock (120). This may represent an alternative treatment for acidosis, although advantages over traditional dialysis are not clear (121).

ICU Organization

Nearly one in six ICU admissions are potentially preventable, with wide variation across U.S. states (122), suggesting that up-front investments in ambulatory care may enable the avoidance of ICU admissions and that preventable ICU admissions could serve as a metric of population health (123). Hospital strain, measured by a novel composite strain index, was strongly associated with ICU admission for patients with sepsis and/or acute respiratory failure, setting up a within-hospital variable to assess the marginal benefit of ICU admission (124). In Australia and New Zealand, ICU discharge delay for more than 6 hours occurred in 25% of patients and was associated with lower ICU readmission and hospital mortality, especially in the subgroup of patients with the highest predicted risk of mortality at ICU admission (125, 126). Identifying factors associated with ICU readmission is important because clinicians had only fair accuracy in predicting it (127). This may reflect the complexity of ICU admission and/or readmission as an organizational construct (128).

The structure and makeup of the ICU may have important implications for patient outcomes. In a before and after cohort study, the rollout of an emergency department–embedded critical care unit was not associated with improved outcomes (129), which puts into question whether the potential benefits are offset by the incurred harm of such units (130, 131). However, nurse staffing models and network relationships of nursing teams (132), as well as interactions between staff members (measured through the analysis of wearable badges) (133), seemed to play an important role in the outcomes of ICU patients. Furthermore, in rural hospitals, initiatives, attitudes, and ownership of change are key determinants for implementing evidence-based practice in the ICU, leading to better program engagement (134). In addition, methods to extend the influence of critical care physicians, including telemedicine approaches that enable physicians to participate as “copilots” guiding resuscitation during in-hospital cardiac arrest, may be beneficial on the wards of community hospitals (135).

Unfortunately, increasing demands on critical care practitioners may have consequences, as nearly 80% of division directors in pulmonary, sleep, and critical care believed that burnout was a problem in their division. Many drivers of burnout were inherent to personality characteristics of individuals or characteristics of the job, but many potentially modifiable factors, such as misalignment between the clinicians and administration, the electronic health records, and the workplace climate, were identified (136).

Disparities have a large impact on patient outcomes. Among critically ill patients, 25% of African American patients and 48% of Hispanic patients were concentrated in only 14 out of 208 U.S. hospitals. Hospital mortality and ICU and hospital stays decreased over time but remained higher in minority-serving hospitals than in non–minority-serving hospitals and did not change in the subgroup of African American patients (137, 138). Another study identified that female survivors of shock or respiratory failure had poorer physical and psychological function than male survivors and were less likely to live at home 3 months after discharge, although most of these differences attenuated after 12 months (139).

Novel therapeutics have improved outcomes for many patients with cancer, although they may have adverse events prompting ICU admission. Patients who received chimeric antigen receptor T-cell therapy for lymphoma had higher rates of ICU admission for neurotoxicity and longer hospital stays than control patients with lymphoma, although ICU lengths of stay were shorter and the 60-day mortality was lower for these patients than for control patients with lymphoma admitted to the ICU (140).

Post-ICU Trajectories and Patient/Family Satisfaction

Critical illness survivors’ priorities evolved during recovery from basic survival concerns (e.g., mobility, participating in self-care) immediately after leaving the ICU to more aspirational concerns (e.g., seeking new experiences) 2 months after discharge (141). Identifying factors associated with post-ICU morbidity has therefore been a focus. In patients older than 65 years of age with severe traumatic brain injury, age and an Injury Severity Score ≥25 were independently associated with functional outcomes, with only 6% of patients recovering functional independence by 6 months (142, 143). Furthermore, at the transition out of the hospital (144), adherence to expert recommendations in the postsepsis setting (medication optimization, screening for common impairments, monitoring for common preventable causes of deterioration, and treatment alignment with patient preferences) (145) occurred in only 11% of patients recovering from sepsis but was associated with an 88% reduced odds of death or hospital readmission (146). Although there is often a concern that opioid use in the ICU may lead to dependence after ICU discharge, persistent opioid use 1 year after hospital discharge was reported in only 2.6% of ICU survivors treated with mechanical ventilation (147, 148).

Family satisfaction is a crucial component of delivering high-quality critical care, as illustrated by the need for a family-centered discharge approach focused on the coordination of care and teaching of emergency preparedness as keys to the safe discharge of tracheostomized children or young adults (149). These themes suggest opportunities to use simulation for education on airway emergencies and to use telemedicine for minimizing the risks of travel (150). During the second and third weeks of patients’ ventilation, both surrogate decision-makers and physicians were less likely to prefer that mechanical ventilation be continued for patients with poor prognoses, but surrogates required until Week 3 of mechanical ventilation before incorporating the prognosis into their decision-making (151). Interestingly, the concepts of futile or inappropriate treatment were difficult to understand for community members, but most were able to recognize inappropriate treatment when presented with a case (152). Inadequate physician–patient communication may underpin the difficulty with the concept, together with fears that the judgment of medical futility could undermine patient/family autonomy. Nevertheless, for patients who died in the hospital, complete ICU or mixed ICU/general-floor care at the end of life was consistently associated with higher family satisfaction, highlighting the importance of ICU care at the end of life (153, 154).

Critical Care Research

Disparities are apparent in the gender breakdown of critical care authorship over the last 10 years. Only 30% of critical care publications have a female first author, with less than 20% having a female senior author, although the presence of a female senior author was associated with a higher likelihood of having a female co-author. This may be contributing to there being fewer women in leadership positions in critical care (155, 156).

Improving methods to perform critical care research was a focus of several investigations. Many collaborative critical care research networks are focused on conducting high-quality randomized controlled trials but are highly dependent on external sources of funding to support network infrastructure (157). The creation of common protocols and approaches is crucial for the success of critical care trials. Recommendations on the maximal blood volume allowable for research and on the safe lower limit of Hb at the time of blood draws for critically ill patients have been proposed (158). Trial recruitment is often challenging in the ICU (159), and a collection of “nudges” presented to surrogate decision-makers as a preconsent survey was not an effective means of increasing enrollment in a sham trial (160). This prompts a need to better understand how surrogates perceive the risk of trial participation.

Administrative data are frequently used in critical care research investigations, and analyses using such data have more accurately identified the use of mechanical ventilation, but not the use of vasoactive medications or acute renal replacement therapy, than clinical chart review (161). Integration of electronic health record data may be required to answer some questions (162).

Education in Critical Care

This past year, the greatest accomplishment in critical care education was helping nonintensivists care for patients in the ICU. Educational content was rapidly created and widely disseminated through publications as well as on social media platforms such as Twitter, which had been widely used by intensivists before the pandemic and became an even more powerful tool for education during the pandemic (163). Much of the educational content aimed at helping care for patients with COVID-19 was published in ATS Scholar, a new open-access journal disseminating medical education focused on pulmonary, critical care, and sleep medicine. A series of videos (164–167) on mechanical ventilation for those with limited or nonexistent training proved to be an invaluable resource for clinicians redeployed to work in ICUs during the pandemic. In light of severely stretched staffing, the concept of “respiratory therapy extenders” was created, and a framework for how to train medical students as such extenders was reported (168). Online resources with quick tips and algorithms for nonintensivists to reference when caring for critically patients (169, 170) as well as methods to keep intensivists up to date on the rapidly changing evidence related to the care of patients with COVID-19, such as the “Fast Literature Assessment and Review Initiative,” were shared widely (171).