Mortality of Patients With Sepsis Administered Piperacillin-Tazobactam vs Cefepime.

Importance: Experimental and observational studies have suggested that empirical treatment for bacterial sepsis with antianaerobic antibiotics (eg, piperacillin-tazobactam) is associated with adverse outcomes compared with anaerobe-sparing antibiotics (eg, cefepime). However, a recent pragmatic clinical trial of piperacillin-tazobactam and cefepime showed no difference in short-term outcomes at 14 days. Further studies are needed to help clarify the empirical use of these agents. Objective: To examine the use of piperacillin-tazobactam compared with cefepime in 90-day mortality in patients treated empirically for sepsis, using instrumental variable analysis of a 15-month piperacillin-tazobactam shortage. Design, Setting, and Participants: In a retrospective cohort study, hospital admissions at the University of Michigan from July 1, 2014, to December 31, 2018, including a piperacillin-tazobactam shortage period from June 12, 2015, to September 18, 2016, were examined. Adult patients with suspected sepsis treated with vancomycin and either piperacillin-tazobactam or cefepime for conditions with presumed equipoise between piperacillin-tazobactam and cefepime were included in the study. Data analysis was conducted from December 17, 2022, to April 11, 2023. Main Outcomes and Measures: The primary outcome was 90-day mortality. Secondary outcomes included organ failure-free, ventilator-free, and vasopressor-free days. The 15-month piperacillin-tazobactam shortage period was used as an instrumental variable for unmeasured confounding in antibiotic selection. Results: Among 7569 patients (4174 men [55%]; median age, 63 [IQR 52-73] years) with sepsis meeting study eligibility, 4523 were treated with vancomycin and piperacillin-tazobactam and 3046 were treated with vancomycin and cefepime. Of patients who received piperacillin-tazobactam, only 152 (3%) received it during the shortage. Treatment groups did not differ significantly in age, Charlson Comorbidity Index score, Sequential Organ Failure Assessment score, or time to antibiotic administration. In an instrumental variable analysis, piperacillin-tazobactam was associated with an absolute mortality increase of 5.0% at 90 days (95% CI, 1.9%-8.1%) and 2.1 (95% CI, 1.4-2.7) fewer organ failure-free days, 1.1 (95% CI, 0.57-1.62) fewer ventilator-free days, and 1.5 (95% CI, 1.01-2.01) fewer vasopressor-free days. Conclusions and Relevance: Among patients with suspected sepsis and no clear indication for antianaerobic coverage, administration of piperacillin-tazobactam was associated with higher mortality and increased duration of organ dysfunction compared with cefepime. These findings suggest that the widespread use of empirical antianaerobic antibiotics in sepsis may be harmful.

Understanding How Clinicians Personalize Fluid and Vasopressor Decisions in Early Sepsis Management.

Importance: Recent sepsis trials suggest that fluid-liberal vs fluid-restrictive resuscitation has similar outcomes. These trials used generalized approaches to resuscitation, and little is known about how clinicians personalize fluid and vasopressor administration in practice. Objective: To understand how clinicians personalize decisions about resuscitation in practice. Design, Setting, and Participants: This survey study of US clinicians in the Society of Critical Care Medicine membership roster was conducted from November 2022 to January 2023. Surveys contained 10 vignettes of patients with sepsis where pertinent clinical factors (eg, fluid received and volume status) were randomized. Respondents selected the next steps in management. Data analysis was conducted from February to September 2023. Exposure: Online Qualtrics clinical vignette survey. Main Outcomes and Measures: Using multivariable logistic regression, the associations of clinical factors with decisions about fluid administration, vasopressor initiation, and vasopressor route were tested. Results are presented as adjusted proportions with 95% CIs. Results: Among 11 203 invited clinicians, 550 (4.9%; 261 men [47.5%] and 192 women [34.9%]; 173 with >15 years of practice [31.5%]) completed at least 1 vignette and were included. A majority were physicians (337 respondents [61.3%]) and critical care trained (369 respondents [67.1%]). Fluid volume already received by a patient was associated with resuscitation decisions. After 1 L of fluid, an adjusted 82.5% (95% CI, 80.2%-84.8%) of respondents prescribed additional fluid and an adjusted 55.0% (95% CI, 51.9%-58.1%) initiated vasopressors. After 5 L of fluid, an adjusted 17.5% (95% CI, 15.1%-19.9%) of respondents prescribed more fluid while an adjusted 92.7% (95% CI, 91.1%-94.3%) initiated vasopressors. More respondents prescribed fluid when the patient examination found dry vs wet (ie, overloaded) volume status (adjusted proportion, 66.9% [95% CI, 62.5%-71.2%] vs adjusted proportion, 26.5% [95% CI, 22.3%-30.6%]). Medical history, respiratory status, lactate trend, and acute kidney injury had small associations with fluid and vasopressor decisions. In 1023 of 1127 vignettes (90.8%) where the patient did not have central access, respondents were willing to start vasopressors through a peripheral intravenous catheter. In cases where patients were already receiving peripheral norepinephrine, respondents were more likely to place a central line at higher norepinephrine doses of 0.5 µg/kg/min (adjusted proportion, 78.0%; 95% CI, 74.7%-81.2%) vs 0.08 µg/kg/min (adjusted proportion, 25.2%; 95% CI, 21.8%-28.5%) and after 24 hours (adjusted proportion, 59.5%; 95% CI, 56.6%-62.5%) vs 8 hours (adjusted proportion, 47.1%; 95% CI, 44.0%-50.1%). Conclusions and Relevance: These findings suggest that fluid volume received is the predominant factor associated with ongoing fluid and vasopressor decisions, outweighing many other clinical factors. Peripheral vasopressor use is common. Future studies aimed at personalizing resuscitation must account for fluid volumes and should incorporate specific tools to help clinicians personalize resuscitation.

Acute Respiratory Distress Syndrome: Definition, Diagnosis, and Routine Management.

Acute respiratory distress syndrome (ARDS) is an acute inflammatory lung injury characterized by severe hypoxemic respiratory failure, bilateral opacities on chest imaging, and low lung compliance. ARDS is a heterogeneous syndrome that is the common end point of a wide variety of predisposing conditions, with complex pathophysiology and underlying mechanisms. Routine management of ARDS is centered on lung-protective ventilation strategies such as low tidal volume ventilation and targeting low airway pressures to avoid exacerbation of lung injury, as well as a conservative fluid management strategy.

Shift-Level Team Familiarity Is Associated with Improved Outcomes in Mechanically Ventilated Adults.

Due to a typesetter error, this is a previous version of this article; It will be replaced shortly by the final accepted version. Rationale: Organizing ICU interprofessional teams-nurses, physicians, and respiratory therapists-is high priority because of workforce crises, but how often clinicians work together (i.e., interprofessional familiarity) remains unexplored. Objectives: Determine if mechanically ventilated patients cared for by teams with greater familiarity have lower mortality, shorter duration of mechanical ventilation, and greater spontaneous breathing trial (SBT) implementation. Methods: Using electronic health records from five ICUs (2018-2019), we identified the interprofessional team that cared for each mechanically ventilated patient each shift, calculated familiarity, and modeled familiarity exposures separately on ICU mortality, duration of mechanical ventilation, and SBT implementation using encounter-level generalized linear regression models with a log-link, unit-level fixed effects adjusting for cofounders, including severity of illness. Measurements and Main Results: Familiarity was defined as how often clinicians worked together for all patients in an ICU (i.e., coreness) and for each patient (i.e., mean team value). Among 4,292 patients (4,485 encounters, 72,210 shifts), unadjusted mortality was 12.9%, average duration of mechanical ventilation was 2.32 days, and SBT implementation was 89%. An increase in coreness and mean team value, by the SD of each, was associated with lower probability of dying (coreness: adjusted marginal effect, -0.038; 95% confidence interval [-0.07 to -0.004]; mean team value: adjusted marginal effect, -0.0034 [-0.054 to -0.014]); greater probability of receiving SBT when eligible (coreness: 0.45 [0.007 to 0.083]; mean team value: 0.012 [-0.017 to 0.042]), and shorter duration of mechanical ventilation (coreness: -0.23 [-0.321 to -0.139]). Conclusions: Interprofessional familiarity was associated with improved outcomes; assignment models that prioritize familiarity might be a novel solution.

Rural-Urban Differences in Mortality among Mechanically Ventilated Patients in Intensive and Intermediate Care.

Rationale: Intermediate care (also termed "step-down" or "moderate care") has been proposed as a lower cost alternative to care for patients who may not clearly benefit from intensive care unit admission. Intermediate care units may be appealing to hospitals in financial crisis, including those in rural areas. Outcomes of patients receiving intermediate care are not widely described. Objectives: To examine relationships among rurality, location of care, and mortality for mechanically ventilated patients. Methods: Medicare beneficiaries aged 65 years and older who received invasive mechanical ventilation between 2010 and 2019 were included. Multivariable logistic regression was used to estimate the association between admission to a rural or an urban hospital and 30-day mortality, with separate analyses for patients in general, intermediate, and intensive care. Models were adjusted for age, sex, area deprivation index, primary diagnosis, severity of illness, year, comorbidities, and hospital volume. Results: There were 2,752,492 hospitalizations for patients receiving mechanical ventilation from 2010 to 2019, and 193,745 patients (7.0%) were in rural hospitals. The proportion of patients in rural intermediate care increased from 4.1% in 2010 to 6.3% in 2019. Patient admissions to urban hospitals remained relatively stable. Patients in rural and urban intensive care units had similar adjusted 30-day mortality, at 46.7% (adjusted absolute risk difference -0.1% [95% confidence interval, -0.7% to 0.6%]; P = 0.88). However, adjusted 30-day mortality for patients in rural intermediate care was significantly higher (36.9%) than for patients in urban intermediate care (31.3%) (adjusted absolute risk difference 5.6% [95% confidence interval, 3.7% to 7.6%]; P < 0.001). Conclusions: Hospitalization in rural intermediate care was associated with increased mortality. There is a need to better understand how intermediate care is used across hospitals and to carefully evaluate the types of patients admitted to intermediate care units.

Early Deep Sedation Practices Worsened During the Pandemic Among Adult Patients Without COVID-19: A Retrospective Cohort Study.

BACKGROUND: There is substantial evidence that patients with COVID-19 were treated with sustained deep sedation during the pandemic. However, it is unknown whether such guideline-discordant care had spillover effects to patients without COVID-19. RESEARCH QUESTION: Did patterns of early deep sedation change during the pandemic for patients on mechanical ventilation without COVID-19? STUDY DESIGN AND METHODS: We used electronic health record data from 4,237 patients who were intubated without COVID-19. We compared sedation practices in the first 48 h after intubation across prepandemic (February 1, 2018, to January 31, 2020), pandemic (April 1, 2020, to March 31, 2021), and late pandemic (April 1, 2021, to March 31, 2022) periods. RESULTS: In the prepandemic period, patients spent an average of 13.0 h deeply sedated in the first 48 h after intubation. This increased 1.9 h (95% CI, 1.0-2.8) during the pandemic period and 2.9 h (95% CI, 2.0-3.8) in the late pandemic period. The proportion of patients that spent over one-half of the first 48 h deeply sedated was 18.9% in the prepandemic period, 22.3% during the pandemic period, and 25.9% during the late pandemic period. Ventilator-free days decreased during the pandemic, with a subdistribution hazard ratio of being alive without mechanical ventilation at 28 days of 0.87 (95% CI, 0.79-0.95) compared with the prepandemic period. Tracheostomy placement increased during the pandemic period compared with the prepandemic period (OR, 1.41; 95% CI, 1.08-1.82). In the medical ICU, early deep sedation increased 2.5 h (95% CI, 0.6-4.4) during the pandemic period and 4.9 h (95% CI, 3.0-6.9) during the late pandemic period, compared with the prepandemic period. INTERPRETATION: We found that among patients on mechanical ventilation without COVID-19, sedation use increased during the pandemic. In the subsequent year, these practices did not return to prepandemic standards.

The Lung Microbiome Predicts Mortality and Response to Azithromycin in Lung Transplant Recipients with Chronic Rejection.

Rationale: Chronic lung allograft dysfunction (CLAD) is the leading cause of death after lung transplant, and azithromycin has variable efficacy in CLAD. The lung microbiome is a risk factor for developing CLAD, but the relationship between lung dysbiosis, pulmonary inflammation, and allograft dysfunction remains poorly understood. Whether lung microbiota predict outcomes or modify treatment response after CLAD is unknown. Objectives: To determine whether lung microbiota predict post-CLAD outcomes and clinical response to azithromycin. Methods: Retrospective cohort study using acellular BAL fluid prospectively collected from recipients of lung transplant within 90 days of CLAD onset. Lung microbiota were characterized using 16S rRNA gene sequencing and droplet digital PCR. In two additional cohorts, causal relationships of dysbiosis and inflammation were evaluated by comparing lung microbiota with CLAD-associated cytokines and measuring ex vivo P. aeruginosa growth in sterilized BAL fluid. Measurements and Main Results: Patients with higher bacterial burden had shorter post-CLAD survival, independent of CLAD phenotype, azithromycin treatment, and relevant covariates. Azithromycin treatment improved survival in patients with high bacterial burden but had negligible impact on patients with low or moderate burden. Lung bacterial burden was positively associated with CLAD-associated cytokines, and ex vivo growth of P. aeruginosa was augmented in BAL fluid from transplant recipients with CLAD. Conclusions: In recipients of lung transplants with chronic rejection, increased lung bacterial burden is an independent risk factor for mortality and predicts clinical response to azithromycin. Lung bacterial dysbiosis is associated with alveolar inflammation and may be promoted by underlying lung allograft dysfunction.

Higher versus lower oxygenation targets in adult ICU patients: A rapid practice guideline.

The aim of this Intensive Care Medicine Rapid Practice Guideline (ICM-RPG) was to provide evidence-based clinical guidance about the use of higher versus lower oxygenation targets for adult patients in the intensive care unit (ICU). The guideline panel comprised 27 international panelists, including content experts, ICU clinicians, methodologists, and patient representatives. We adhered to the methodology for trustworthy clinical practice guidelines, including the use of the Grading of Recommendations Assessment, Development, and Evaluation approach to assess the certainty of evidence, and used the Evidence-to-Decision framework to generate recommendations. A recently published updated systematic review and meta-analysis constituted the evidence base. Through teleconferences and web-based discussions, the panel provided input on the balance and magnitude of the desirable and undesirable effects, the certainty of evidence, patients' values and preferences, costs and resources, equity, feasibility, acceptability, and research priorities. The updated systematic review and meta-analysis included data from 17 randomized clinical trials with 10,248 participants. There was little to no difference between the use of higher versus lower oxygenation targets for all outcomes with available data, including all-cause mortality, serious adverse events, stroke, functional outcomes, cognition, and health-related quality of life (very low certainty of evidence). The panel felt that values and preferences, costs and resources, and equity favored the use of lower oxygenation targets. The ICM-RPG panel issued one conditional recommendation against the use of higher oxygenation targets: "We suggest against the routine use of higher oxygenation targets in adult ICU patients (conditional recommendation, very low certainty of evidence). Remark: an oxygenation target of SpO2 88%-92% or PaO2 8 kPa/60 mmHg is relevant and safe for most adult ICU patients."

Measuring the Impact of AI in the Diagnosis of Hospitalized Patients: A Randomized Clinical Vignette Survey Study.

Importance: Artificial intelligence (AI) could support clinicians when diagnosing hospitalized patients; however, systematic bias in AI models could worsen clinician diagnostic accuracy. Recent regulatory guidance has called for AI models to include explanations to mitigate errors made by models, but the effectiveness of this strategy has not been established. Objectives: To evaluate the impact of systematically biased AI on clinician diagnostic accuracy and to determine if image-based AI model explanations can mitigate model errors. Design, Setting, and Participants: Randomized clinical vignette survey study administered between April 2022 and January 2023 across 13 US states involving hospitalist physicians, nurse practitioners, and physician assistants. Interventions: Clinicians were shown 9 clinical vignettes of patients hospitalized with acute respiratory failure, including their presenting symptoms, physical examination, laboratory results, and chest radiographs. Clinicians were then asked to determine the likelihood of pneumonia, heart failure, or chronic obstructive pulmonary disease as the underlying cause(s) of each patient's acute respiratory failure. To establish baseline diagnostic accuracy, clinicians were shown 2 vignettes without AI model input. Clinicians were then randomized to see 6 vignettes with AI model input with or without AI model explanations. Among these 6 vignettes, 3 vignettes included standard-model predictions, and 3 vignettes included systematically biased model predictions. Main Outcomes and Measures: Clinician diagnostic accuracy for pneumonia, heart failure, and chronic obstructive pulmonary disease. Results: Median participant age was 34 years (IQR, 31-39) and 241 (57.7%) were female. Four hundred fifty-seven clinicians were randomized and completed at least 1 vignette, with 231 randomized to AI model predictions without explanations, and 226 randomized to AI model predictions with explanations. Clinicians' baseline diagnostic accuracy was 73.0% (95% CI, 68.3% to 77.8%) for the 3 diagnoses. When shown a standard AI model without explanations, clinician accuracy increased over baseline by 2.9 percentage points (95% CI, 0.5 to 5.2) and by 4.4 percentage points (95% CI, 2.0 to 6.9) when clinicians were also shown AI model explanations. Systematically biased AI model predictions decreased clinician accuracy by 11.3 percentage points (95% CI, 7.2 to 15.5) compared with baseline and providing biased AI model predictions with explanations decreased clinician accuracy by 9.1 percentage points (95% CI, 4.9 to 13.2) compared with baseline, representing a nonsignificant improvement of 2.3 percentage points (95% CI, -2.7 to 7.2) compared with the systematically biased AI model. Conclusions and Relevance: Although standard AI models improve diagnostic accuracy, systematically biased AI models reduced diagnostic accuracy, and commonly used image-based AI model explanations did not mitigate this harmful effect. Trial Registration: ClinicalTrials.gov Identifier: NCT06098950.

Learning Concept Credible Models for Mitigating Shortcuts.

During training, models can exploit spurious correlations as shortcuts, resulting in poor generalization performance when shortcuts do not persist. In this work, assuming access to a representation based on domain knowledge (i.e., known concepts) that is invariant to shortcuts, we aim to learn robust and accurate models from biased training data. In contrast to previous work, we do not rely solely on known concepts, but allow the model to also learn unknown concepts. We propose two approaches for mitigating shortcuts that incorporate domain knowledge, while accounting for potentially important yet unknown concepts. The first approach is two-staged. After fitting a model using known concepts, it accounts for the residual using unknown concepts. While flexible, we show that this approach is vulnerable when shortcuts are correlated with the unknown concepts. This limitation is addressed by our second approach that extends a recently proposed regularization penalty. Applied to two real-world datasets, we demonstrate that both approaches can successfully mitigate shortcut learning.