Citations

Abstract RATIONALE Propofol is the first-line sedative for critically-ill, mechanically-ventilated patients, but it can increase serum triglyceride levels. Machine-learning (ML) offers a promising approach for individualized risk stratification of propofol-associated hypertriglyceridemia, enabling proactive triglyceride monitoring and sedation optimization. This study aims to develop and validate ML models to predict hypertriglyceridemia risks in critically-ill patients receiving propofol. METHODS We analyzed retrospective data from adult ICU patients on mechanical ventilation receiving continuous propofol infusion for ≥24 hours across four Mayo Clinic sites from May 2018 to June 2023. The primary outcome was hypertriglyceridemia, defined as serum triglycerides >400 mg/dL within 10 days of starting propofol. Ten-fold cross-validation and leave-one-site-out nested cross-validation (LOSO-CV) were used for model tuning and external validation of our modeling process, respectively. Feature selection was conducted through a union of Boruta and least absolute shrinkage and selection operator (LASSO). While COVID status was a strong predictor of hypertriglyceridemia, we concurrently developed non-COVID models to ensure future relevance in non-pandemic settings. Six ML algorithms (logistic regression, support vector machines, random forests, LightGBM, XGBoost, and multilayer perceptrons) were tuned using Bayesian optimization to minimize cross-entropy loss. Classification thresholds were set to target sensitivity ≥80%. Model performance was evaluated using AUC-ROC, Brier score, sensitivity, and specificity. Models with the highest Youden's Index were designated the best-performing model. Model explainability was assessed using permutation importance and SHapley Additive exPlanations (SHAP). RESULTS We included 3922 eligible patients with triglyceride measurements. Of these, 3042 patients were COVID-negative and used for non-COVID modeling. The best-performing non-COVID models achieved an average AUC-ROC of 0.69 (95% confidence interval [CI] 0.68-0.70), Brier score of 0.12 (95% CI 0.09-0.14), sensitivity of 0.79 (95% CI 0.75-0.83), and specificity of 0.48 (95% CI 0.44-0.52) during LOSO-CV. The best-performing models that included COVID status achieved an average AUC-ROC of 0.71 (95% CI 0.70-0.72), Brier score of 0.13 (95% CI 0.10-0.16), sensitivity of 0.78 (95% CI 0.71-0.84), and specificity of 0.51 (95% CI 0.45 to 0.56) during LOSO-CV. The permutation importance results of the final non-COVID model are shown in the figure. CONCLUSION We have developed sensitive ML models for predicting hypertriglyceridemia risks in critically-ill patients receiving propofol. Although developed during the COVID-19 pandemic, the non-COVID models are more likely to have future relevance, and they have similar performance as models including COVID status. These models will be integrated into a web-based clinical calculator and undergo further validation in prospective studies.

Abstract Rationale Hyperchloremia, characterized by elevated chloride levels (>110 mEq/L) in the blood, frequently occurs in critically ill patients following aggressive fluid resuscitation with chloride-rich solutions. It is associated with metabolic acidosis, leading to negative clinical outcomes such as higher morbidity and mortality. However, the impact of hyperchloremia on the duration of mechanical ventilation remains unclear. This study evaluates the correlation between hyperchloremia and the duration of mechanical ventilation in critically ill patients. Methods A retrospective analysis was conducted on intensive care unit patients admitted between 2018 and 2023 who required mechanical ventilation. Patients were categorized into hyperchloremic (serum chloride >110 mEq/L) and non-hyperchloremic groups. The hyperchloremic group was further subdivided into those with metabolic acidosis (pH <7.35) and those without acidosis (pH >7.35). The primary outcome measured was the duration of mechanical ventilation across these groups. Results The study included 7,333 patients, 408 in the hyperchloremic group and 6,926 in the non-hyperchloremic group. Patients with hyperchloremia experienced significantly longer ventilation durations, with an average of 5.9 days compared to 2.6 days in the non-hyperchloremic group (p < 0.01). Among the hyperchloremic subgroups, those with metabolic acidosis had a mean ventilation duration of 4.9 days compared to 6.2 days in those without acidosis (p = 0.7), suggesting no apparent significant influence of acid-base status on ventilation duration. Discussion The study found that hyperchloremia correlates with longer ventilation duration (4.9-6.2 days vs. 2.6 days), independent of pH. While metabolic acidosis might impair respiratory function and increase ventilatory needs, evidence is mixed. Hyperchloremic metabolic acidosis can alter respiratory regulation, causing issues like tachypnea, ventilator dyssynchrony, and the need for sedation, complicating weaning from ventilation. However, hyperchloremia does not always impair respiratory muscle function. The study is limited by its retrospective nature, making causal conclusions difficult. However, the study suggests that better electrolyte management may reduce ventilation duration and improve patient outcomes. Conclusion This study demonstrates a significant association between hyperchloremia and prolonged mechanical ventilation, independent of pH status. The findings suggest that hyperchloremia may either be a marker of severe illness or aggressive resuscitation and directly contribute to prolonged ventilatory support. These findings offer hope for improved patient outcomes, as they suggest that chloride management strategies, including chloride-sparing resuscitation, could potentially reduce mechanical ventilation duration in critically ill patients.

Abstract RATIONALE: Many ICU patients experience physical, psychological, and cognitive impairments following resolution of critical illness, collectively referred to as post-intensive care syndrome (PICS). PICS has become widely recognized within the critical care field, but how to screen for PICS does not currently have consensus. Development of a validated, concise PICS screening tool would allow for rapid identification of this vulnerable and understudied population. Therefore, this study's goal was to develop and evaluate a screening tool to identify patients at risk for PICS. METHODS: We performed a prospective cohort study in a quaternary academic hospital. We included adults ≥ 18 years old with ≥48 hours spent in an ICU. Exclusion criteria were admission for suicide attempt, severe cognitive impairment/dementia, life expectancy <3 months, and non-English speaking. We assessed mental, cognitive, and physical function following ICU discharge and at 3 months using Hospital Anxiety and Depression Scale, Impact of Event Scale-Revised, Montreal Cognitive Assessment-blind, and Barthel Index. Our team, including critical care physicians, psychiatrists, psychologists, physical and occupational therapists, and nurses, developed a brief 15-item PICS screening tool (5 questions/domain). This tool was administered after patients were discharged from ICU stay to general care to assess its effectiveness in predicting impairments in all 3 domains at discharge and again at three months after hospital discharge, in comparison to previously mentioned validated questionnaires. Youden index statistical methods were used to determine potential “cutoff” values (scores) for the tool to maximize sensitivity and specificity. RESULTS: A total of 194 patients from 6 ICUs were recruited and completed the initial questionnaire, and 109 (56%) patients completed 3-month follow-up. Median age was 65 years (IQR 57-72) and 49% (95/194) were female. Screening tool is depicted in Figure 1. Overall, the screening tool weakly-to-moderately correlated with continuous outcomes of the validated measures upon ICU discharge and 3-month follow-up. When evaluating sensitivity and specificity for individual domains, the tool had an area under the receiver operating curve of >0.7 for the following domains: cognitive, anxiety, and physical impairments, indicating moderate-to-good sensitivity and specificity. CONCLUSIONS: This PICS screening tool was correlated with validated measures of PICS upon ICU discharge and at 3-month follow-up for all three domains. The tool's strength lies in its concise nature, making it easier for those caring for this patient population to screen for long-term effects of an ICU stay. Our proposed next step is validation of the PICS screening tool in an external cohort.

Abstract Objective: Patients with Opioid Use Disorder (OUD) are prone to Multidrug-Resistant Organism (MDRO) colonization and infections, thus at risk for worse outcomes during critical illness. Understanding the prevalence and predictors of MDRO infections is essential to optimize interventions and treatments. Design: Retrospective cohort study. Methods: The study evaluated the prevalence of MDRO isolation among adults with OUD admitted to an intensive care unit (ICU) between January 1, 2018, and July 31, 2023. It included adults admitted to an ICU with bacterial infections and positive cultures obtained within 48 hours of admission. Demographics, clinical traits, and MDRO isolation rates were analyzed using descriptive statistics, univariate methods, and Least Absolute Shrinkage and Selection Operator (LASSO) regression. Results: MDRO isolation occurred in 178 of 790 patients (22.5%), with methicillin-resistant Staphylococcus aureus as the most frequently isolated organism. LASSO regression identified housing insecurity (OR: 1.79, 95% CI 1.09–2.93, P = .022), no receipt of medications for OUD treatment (OR: 1.56, 95% CI 1.06–2.29, P = .023), positive hepatitis C virus (HCV) status (OR: 2.19, 95% CI 1.19–4.03, P = .012), and intravenous antibiotic use in the prior 90 days (OR: 1.04 per 24 h, 95% CI 1.01–1.07, P = .007) as significant predictors of MDRO isolation. Conclusions: The study highlights a high prevalence of MDRO isolation in critically ill OUD patients admitted for infection-related issues with positive cultures obtained within 48 hours of admission, influenced by factors like housing insecurity, no receipt of medications for OUD treatment, HCV status, and prior antibiotic use.

Background Recent technologies enable real‐time 3‐dimensional intracardiac echocardiography (ICE) with the advantages of 2‐dimensional ICE and imaging capabilities of transesophageal echocardiography. The purpose of this study is to evaluate the real‐world, multicenter safety and performance of a novel 3‐dimensional ICE system in a variety of cardiac interventions used in standard clinical practice. Methods This prospective, multicenter, observational, single‐arm study enrolled patients who were scheduled for an invasive cardiac procedure involving guidance with ICE and followed until discharge or ≤48 hours postprocedure. Intracardiac imaging was performed with the 3‐dimensional ICE catheter, VeriSight Pro (Philips Inc.). Fluoroscopy was used in all cases and transesophageal echocardiography was used in some cases. The primary safety end point was device‐related adverse events. The primary efficacy end point included technical success, imaging success, and clinical success. Results Between October 2021 and November 2022, 155 patients were enrolled and screened. Percutaneous interventions performed included atrial septal defect/patent foramen ovale closures, left atrial appendage occlusion, catheter ablations, and valve procedures. No adverse events were related to the device or procedure. The technical, imaging, and clinical success rates were 98.7% (95% CI, 95.4%–99.8%), 96.1% (95% CI, 91.8%–98.6%), 94.8% (95% CI, 91.8%–98.6%), respectively. Overall, the VeriSight Pro ICE image quality was assessed to be “acceptable” or greater in relation to reference image modality among 96.2% (149 of 155) of procedures. Conclusions VeriSight ICE imaging is safe and effective in the guidance of a wide variety of percutaneous cardiovascular procedures and has provided successful and high‐quality imaging of cardiac structures.

Objectives There is substantial practice variation in the management strategies for atrial fibrillation (AF) with rapid ventricular rate (RVR) and hemodynamic instability in critically ill patients. This study aimed to evaluate the use and effectiveness of direct current cardioversion (DCCV) for unstable AF RVR in the intensive care unit (ICU). Design Multicenter retrospective cohort study. Setting Non-cardiac/cardiovascular surgery ICUs of the Mayo Clinic Hospitals in Minnesota, Wisconsin, Florida, and Arizona. Patients Adult patients who developed paroxysmal AF RVR with hemodynamic instability (heart rate ≥ 130 beats per minute and a documented SBP ≤ 90 mmHg or the need for vasopressors) during ICU stay. Interventions Emergency DCCV. Measurements and Main Results Primary outcomes were in-hospital mortality, hospital-, ICU-, and organ support-free days. Secondary outcomes were return to sinus rhythm rate before and after discharge. Among 691 patients eligible for inclusion, 47 (6.8%) received emergent DCCV in the ICU. The frequency of DCCV was higher for patients located in surgical ICUs (14.5% vs 6.1%, P = .04). Patients were 1:2 propensity-score matched. The comparative analyses were conducted on 141 patients. Those who underwent DCCV were more likely to restore sinus rhythm (odds ratio [95% confidence interval], 5.2 [1.30 20.8]). However, DCCV was not associated with increased odds of having sinus rhythm at discharge, lower mortality, or higher number of ICU-, or hospital-free days. The number of organ support-free days was significantly lower in the DCCV group (estimate ± standard error, −5.0 ± 2.0, P = .012). Conclusions Emergency DCCV is rarely used to treat unstable AF in non-cardiac ICUs. Despite increased likelihood of immediate sinus rhythm restoration, DCCV was not associated with improvement in patient-centered outcomes. Prospective studies are warranted to evaluate harms and benefits of DCCV in non-cardiac critically ill patients with unstable AF where the driver of hemodynamic instability is often an underlying condition rather than arrhythmia per se.

OBJECTIVES: Acute hypoxemic respiratory failure in immunocompromised patients remains the leading cause of admission to the ICU, with high case fatality. The response to the initial oxygenation strategy may be predictive of outcome. This study aims to assess the response to the evolutionary profiles of oxygenation strategy and the association with survival. DESIGN: Post hoc analysis of EFRAIM study with a nonparametric longitudinal clustering technique (longitudinal K-mean). SETTING AND PATIENTS: Multinational, observational prospective cohort study performed in critically ill immunocompromised patients admitted for an acute respiratory failure. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: A total of 1547 patients who did not require invasive mechanical ventilation (iMV) at ICU admission were included. Change in ventilatory support was assessed and three clusters of change in oxygenation modality over time were identified. Cluster A: 12.3% iMV requirement and high survival rate, n = 717 patients (46.3%); cluster B: 32.9% need for iMV, 97% ICU mortality, n = 499 patients (32.3%); and cluster C: 37.5% need for iMV, 0.3% ICU mortality, n = 331 patients (21.4%). These clusters demonstrated a high discrimination. After adjustment for confounders, clusters B and C were independently associated with need for iMV (odds ratio [OR], 9.87; 95% CI, 7.26–13.50 and OR, 19.8; 95% CI, 13.7–29.1). CONCLUSIONS: This study identified three distinct highly performing clusters of response to initial oxygenation strategy, which reliably predicted the need for iMV requirement and hospital mortality.

OBJECTIVES: Estimated glomerular filtration rate is more accurate with combined creatinine and cystatin C equations (eGFRcr-cys) than creatinine alone. This study created and evaluated a cefepime dosing nomogram based on eGFRcr-cys for initial dosing in the critically ill. DESIGN: Pharmacokinetic modeling and simulation study. SETTING: Academic medical center. PATIENTS: Critically ill adults treated with cefepime. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Data from 120 patients with baseline cystatin C and follow-up cefepime levels were used to develop a nomogram based on eGFRcr-cys and weight for initial cefepime dosing. The predicted proportion of patients who achieved a free cefepime concentration above the minimum inhibitory concentration of the organism for 100% of the dosing interval in the first 24 hours (100% ƒT > MIC at 24 hr) was compared between administered doses and those predicted by the nomogram doses. Overall drug exposure was estimated with the free area under the concentration time curve from 0 to 24 hours (ƒAUC0–24) and compared between administered and nomogram doses. Achievement of 100% ƒT > MIC at 24 hours was predicted to be significantly better with the nomogram compared with the administered dose (76% vs. 38%; p < 0.001). The median ƒAUC0–24 as predicted by the nomogram (666 mg·hr/L) was slightly higher than the actual ƒAUC0–24 with administered doses (612 mg·hr/L; p = 0.01), but the nomogram led to fewer ƒAUC0–24 values which were either too high (> 900) or too low (< 300) (7% vs. 20%; p = 0.004). CONCLUSIONS: Use of a cystatin C-inclusive dosing nomogram for cefepime could improve target attainment without increasing the risk of potentially toxic levels in the critically ill.

Abstract Background The opioid crisis and the increasing prevalence of multi-drug resistant organism (MDRO) infections pose a threat to public health. Opioid Use Disorder (OUD) patient are prone to MDRO colonization and infections, thus potentially leading to adverse outcomes during periods of critical illness. Understanding the prevalence and predictors of MDRO infections within this vulnerable population is critical to improving patient outcomes by optimizing interventions and treatment plans. Methods We conducted a retrospective, multicenter, cohort study to evaluate the prevalence of MDRO isolation among adults with OUD admitted to an intensive care unit (ICU) with an infection across four Mayo Clinic regions. Patients were classified to the MDRO cohort if an MDRO was isolated from standard cultures or rapid diagnostic testing within 48 hours of admission. We utilized descriptive statistics, univariate and multivariate least absolute shrinkage and selection operator (LASSO) regression analysis including random effect to account for site variability, to analyze demographic data, clinical characteristics, and outcomes. Results 790 critically ill adult OUD patients with infection-related admission were included in the study cohort. The prevalence of MDRO isolation was 22.5% (n=178), with MRSA being the predominant isolated MDRO. In LASSO regression analyses, we identified homelessness (OR: 1.79, p = 0.022), no Medications for OUD (MOUD) treatment (OR: 1.56, p=0.023), and positive Hepatitis C Virus (HCV) status (OR: 2.19, p = 0.012) as significant predictors for isolation of an MDRO. Intravenous (IV) antibiotic use in the preceding 90 days was also positively associated (OR: 1.04 per 24 hours, p = 0.007), alongside mental health disorders (OR: 2.03, p &lt; 0.001) and diabetes (OR: 1.54, p = 0.023). Conclusion The high prevalence of MDRO isolation in critically ill OUD patients with infection is associated with complex social and clinical factors. Disclosures All Authors: No reported disclosures