Design and rationale of the B-lines lung ultrasound guided emergency department management of acute heart failure (BLUSHED-AHF) pilot trial

Transfer Learning-Based B-Line Assessment of Lung Ultrasound for Acute Heart Failure

OBJECTIVE

B-lines assessed by lung ultrasound (LUS) outperform physical exam, chest radiograph, and biomarkers for the associated diagnosis of acute heart failure (AHF) in the emergent setting. The use of LUS is however limited to trained professionals and suffers from interpretation variability. The objective was to utilize transfer learning to create an AI-enabled software that can aid novice users to automate LUS B-line interpretation.

METHODS

Data from an observational AHF LUS study provided standardized cine clips for AI model development and evaluation. A total of 49,952 LUS frames from 30 patients were hand scored and trained on a convolutional neural network (CNN) to interpret B-lines at the frame level. A random independent evaluation set of 476 LUS clips from 60 unique patients assessed model performance. The AI models scored the clips on both a binary and ordinal 0-4 multiclass assessment.

RESULTS

A multiclassification AI algorithm had the best performance at the binary level when applied to the independent evaluation set, AUC of 0.967 (95% CI 0.965-0.970) for detecting pathologic conditions. When compared to expert blinded reviewer, the 0-4 multiclassification AI algorithm scale had a reported linear weighted kappa of 0.839 (95% CI 0.804-0.871).

CONCLUSIONS

The multiclassification AI algorithm is a robust and well performing model at both binary and ordinal multiclass B-line evaluation. This algorithm has the potential to be integrated into clinical workflows to assist users with quantitative and objective B-line assessment for evaluation of AHF.

Multi-modality assessment of congestion in acute heart failure: Associations with left ventricular ejection fraction and prognosis

BACKGROUND

Integrating clinical examination with ultrasound measures of congestion could improve risk stratification in patients hospitalized with acute heart failure (AHF).

AIM

To investigate the prevalence of clinical, echocardiographic and lung ultrasound (LUS) signs of congestion according to left ventricular ejection fraction (LVEF) and their association with prognosis in patients with AHF.

METHODS

We pooled the data of four cohorts of patients (N = 601, 74.9±10.8 years, 59 % men) with AHF and analysed six features of congestion at enrolment: clinical (peripheral oedema and respiratory rales), biochemical (BNP/NT-proBNP≥median), echocardiographic (inferior vena cava (IVC)≥21 mm, pulmonary artery systolic pressure (PASP)≥40 mmHg, E/e'≥15) and B-lines ≥25 (8-zones) in those with reduced (<40 %, HFrEF), mildly reduced (40-49 %, HFmrEF and preserved (≥50 %HFpEF) LVEF.

RESULTS

Compared to patients with HFmrEF (n = 110) and HFpEF (n = 201), those with HFrEF (N = 290) had higher natriuretic peptides, but prevalence of clinical (39 %), echocardiographic (IVC≥21 mm: 56 %, E/e'≥15: 57 %, PASP≥40 mmHg: 76 %) and LUS (48 %) signs of congestion was similar. In multivariable analysis, clinical (HR: 3.24(2.15-4.86), p < 0.001), echocardiographic [(IVC≥21 mm (HR:1.91, 1.21-3.03, p=0.006); E/e'≥15 (HR:1.54, 1.04-2.28, p = 0.031)] and LUS (HR:2.08, 1.34-3.24, p = 0.001) signs of congestion were significantly associated with all-cause mortality and/or HF re-hospitalization. Adding echocardiographic and LUS features of congestion to a model than included age, sex, systolic blood pressure, clinical congestion and natriuretic peptides, improved prediction at 90 and 180 days.

CONCLUSIONS

Clinical and ultrasound signs of congestion are highly prevalent in patients with AHF, regardless of LVEF and their combined assessment improves risk stratification.

Usefulness of Combined Ultrasound Assessment of E/e' Ratio, Pulmonary Pressure, and Cava Vein Status in Patients With Acute Heart Failure

Congestion is poorly investigated by ultrasound scans during acute heart failure (AHF) and systematic studies evaluating ultrasound indexes of cardiac pulmonary and systemic congestion during early hospital admission are lacking. We aimed to investigate the prevalence of ultrasound cardiac pulmonary and systemic congestion in a consecutive cohort of hospitalized patients with AHF, analyzing the relevance of each ultrasound congestion component (cardiac, pulmonary, and systemic) in predicting the risk of death and rehospitalization. This is a prospective research study of a single center that evaluates patients with an AHF diagnosis who are divided according to the left ventricular ejection fraction in patients with heart failure with preserved ejection fraction or reduced ejection fraction. We performed a complete bedside echocardiography and lung ultrasound analyses within the first 24 hours of hospital admission. The ultrasound congestion score was preliminarily established by measuring the following parameters: cardiac congestion, which was defined as the contemporary presence of E/e' >15 and pulmonary systolic pressure >35 mm Hg and the pulmonary congestion, defined as the total B-line number >25 at the lung ultrasound performed in 8 chest sites; moreover, the systemic congestion was defined if the inferior vena cava (IVC) was >21 mm and if it was associated with a reduced inspiratory collapse >50%. We thoroughly assessed 230 patients and evaluated their results. Of these patients, 135 had heart failure with reduced ejection fraction and there were 95 patients with heart failure with preserved ejection fraction; 122 patients experienced adverse events during the 180-day follow-up. The receiver operating characteristic curve analysis showed that the tricuspid annular peak systolic excursion (TAPSE) (area under the curve [AUC] 0.34 [0.26 to 0.41], p <0.001), E/e' (AUC 0.62 [0.54 to 0.69], p = 0.003), and IVC (AUC 0.70 [0.63 to 0.77], p <0.001) were all significantly related to poor prognosis detection. The univariate Cox regression analysis revealed that cardiac congestion in terms of E/e' and pulmonary systolic pressure (hazard ratio [HR] 1.49 [1.02 to 2.17], p = 0.037), TAPSE (HR 0.90 [0.85 to 0.94], p <0.001), and systemic congestion (HR 2.64 [1.53 to 4.56], p <0.001) were all significantly related to the 180-day outcome. After adjustment for potential confounders, only TAPSE (HR 0.92 [0.88 to 0.98], p = 0.005) and IVC (HR 1.92 [1.07 to 3.46], p = 0.029) confirmed their prognostic role. The multivariable analysis of multiple congestion levels in terms of systemic plus cardiac (HR 1.54 [1.05 to 2.25], p = 0.03), systemic plus pulmonary (HR 2.26 [1.47 to 3.47], p <0.001), and all 3 congestion features (HR 1.53 [1.06 to 2.23], p = 0.02) revealed an incremental prognostic role for each additional determinant. In conclusion, among the ultrasound indexes of congestion, IVC and TAPSE are related to adverse prognosis, and the addition of pulmonary and cardiac congestion indexes increases the risk prediction accuracy. Our data confirmed that right ventricular dysfunction and systemic congestion are the most powerful predictive factors in AHF.

The prognostic value of improving congestion on lung ultrasound during treatment for acute heart failure differs based on patient characteristics at admission

BACKGROUND

Lung ultrasound congestion scoring (LUS-CS) is a congestion severity biomarker. The BLUSHED-AHF trial demonstrated feasibility for LUS-CS-guided therapy in acute heart failure (AHF). We investigated two questions: 1) does change (∆) in LUS-CS from emergency department (ED) to hospital-discharge predict patient outcomes, and 2) is the relationship between in-hospital decongestion and adverse events moderated by baseline risk-factors at admission?

METHODS

We performed a secondary analysis of 933 observations/128 patients from 5 hospitals in the BLUSHED-AHF trial receiving daily LUS. ∆LUS-CS from ED arrival to inpatient discharge (scale -160 to +160, where negative = improving congestion) was compared to a primary outcome of 30-day death/AHF-rehospitalization. Cox regression was used to adjust for mortality risk at admission [Get-With-The-Guidelines HF risk score (GWTG-RS)] and the discharge LUS-CS. An interaction between ∆LUS-CS and GWTG-RS was included, under the hypothesis that the association between decongestion intensity (by ∆LUS-CS) and adverse outcomes would be stronger in admitted patients with low-mortality risk but high baseline congestion.

RESULTS

Median age was 65 years, GWTG-RS 36, left ventricular ejection fraction 36 %, and ∆LUS-CS -20. In the multivariable analysis ∆LUS-CS was associated with event-free survival (HR = 0.61; 95 % CI: 0.38-0.97), while discharge LUS-CS (HR = 1.00; 95%CI: 0.54-1.84) did not add incremental prognostic value to ∆LUS-CS alone. As GWTG-RS rose, benefits of LUS-CS reduction attenuated (interaction p < 0.05). ∆LUS-CS and event-free survival were most strongly correlated in patients without tachycardia, tachypnea, hypotension, hyponatremia, uremia, advanced age, or history of myocardial infarction at ED/baseline, and those with low daily loop diuretic requirements.

CONCLUSIONS

Reduction in ∆LUS-CS during AHF treatment was most associated with improved readmission-free survival in heavily congested patients with otherwise reassuring features at admission. ∆LUS-CS may be most useful as a measure to ensure adequate decongestion prior to discharge, to prevent early readmission, rather than modify survival.

Impact of the MEESSI-AHF tool to guide disposition decision-making in patients with acute heart failure in the emergency department: a before-and-after study

OBJECTIVES

To determine the impact of risk stratification using the MEESSI-AHF (Multiple Estimation of risk based on the Emergency department Spanish Score In patients with acute heart failure) scale to guide disposition decision-making on the outcomes of ED patients with acute heart failure (AHF), and assess the adherence of emergency physicians to risk stratification recommendations.

METHODS

This was a prospective quasi-experimental study (before/after design) conducted in eight Spanish EDs which consecutively enrolled adult patients with AHF. In the pre-implementation stage, the admit/discharge decision was performed entirely based on emergency physician judgement. During the post-implementation phase, emergency physicians were advised to 'discharge' patients classified by the MEESSI-AHF scale as low risk and 'admit' patients classified as increased risk. Nonetheless, the final decision was left to treating emergency physicians. The primary outcome was 30-day all-cause mortality. Secondary outcomes were days alive and out of hospital, in-hospital mortality and 30-day post-discharge combined adverse event (ED revisit, hospitalisation or death).

RESULTS

The pre-implementation and post-implementation cohorts included 1589 and 1575 patients, respectively (median age 85 years, 56% females) with similar characteristics, and 30-day all-cause mortality was 9.4% and 9.7%, respectively (post-implementation HR=1.03, 95% CI=0.82 to 1.29). There were no differences in secondary outcomes or in the percentage of patients entirely managed in the ED without hospitalisation (direct discharge from the ED, 23.5% vs 24.4%, OR=1.05, 95% CI=0.89 to 1.24). Adjusted models did not change these results. Emergency physicians followed the MEESSI-AHF-based recommendation on patient disposition in 70.9% of cases (recommendation over-ruling: 29.1%). Physicians were more likely to over-rule the recommendation when 'discharge' was recommended (56.4%; main reason: need for hospitalisation for a second diagnosis) than when 'admit' was recommended (12.8%; main reason: no appreciation of severity of AHF decompensation by emergency physician), with an OR for over-ruling the 'discharge' compared with the 'admit' recommendation of 8.78 (95% CI=6.84 to 11.3).

CONCLUSIONS

Implementing the MEESSI-AHF risk stratification tool in the ED to guide disposition decision-making did not improve patient outcomes.

Comparison of pulmonary congestion severity using artificial intelligence-assisted scoring versus clinical experts: A secondary analysis of BLUSHED-AHF

AIM

Acute decompensated heart failure (ADHF) is the leading cause of cardiovascular hospitalizations in the United States. Detecting B-lines through lung ultrasound (LUS) can enhance clinicians' prognostic and diagnostic capabilities. Artificial intelligence/machine learning (AI/ML)-based automated guidance systems may allow novice users to apply LUS to clinical care. We investigated whether an AI/ML automated LUS congestion score correlates with expert's interpretations of B-line quantification from an external patient dataset.

METHODS AND RESULTS

This was a secondary analysis from the BLUSHED-AHF study which investigated the effect of LUS-guided therapy on patients with ADHF. In BLUSHED-AHF, LUS was performed and B-lines were quantified by ultrasound operators. Two experts then separately quantified the number of B-lines per ultrasound video clip recorded. Here, an AI/ML-based lung congestion score (LCS) was calculated for all LUS clips from BLUSHED-AHF. Spearman correlation was computed between LCS and counts from each of the original three raters. A total of 3858 LUS clips were analysed on 130 patients. The LCS demonstrated good agreement with the two experts' B-line quantification score (r = 0.894, 0.882). Both experts' B-line quantification scores had significantly better agreement with the LCS than they did with the ultrasound operator's score (p < 0.005, p < 0.001).

CONCLUSION

Artificial intelligence/machine learning-based LCS correlated with expert-level B-line quantification. Future studies are needed to determine whether automated tools may assist novice users in LUS interpretation.

New International Guidelines and Consensus on the Use of Lung Ultrasound

Following the innovations and new discoveries of the last 10 years in the field of lung ultrasound (LUS), a multidisciplinary panel of international LUS experts from six countries and from different fields (clinical and technical) reviewed and updated the original international consensus for point-of-care LUS, dated 2012. As a result, a total of 20 statements have been produced. Each statement is complemented by guidelines and future developments proposals. The statements are furthermore classified based on their nature as technical (5), clinical (11), educational (3), and safety (1) statements.

Right ventricular dysfunction in acute heart failure from emergency department to discharge: Predictors and clinical implications

BACKGROUND

Among acute heart failure (AHF) inpatients, right ventricular dysfunction (RVD) predicts clinical outcomes independent of left ventricular (LV) dysfunction. Prior studies have not accounted for congestion severity, show conflicting findings on echocardiography (echo) timing, and excluded emergency department (ED) patients. We describe for the first time the epidemiology, predictors, and outcomes of RVD in AHF starting with earliest ED treatment.

METHODS

Point-of-care echo and 10-point lung ultrasound (LUS) were obtained in 84 prospectively enrolled AHF patients at two EDs, ≤1 h after first intravenous diuresis, vasodilator, and/or positive pressure ventilation (PPV). Echo and LUS were repeated at 24, 72, and 168 h, unless discharged sooner (n = 197 exams). RVD was defined as <17 mm tricuspid annulus plane systolic excursion (TAPSE), our primary measure. To identify correlates of RVD, a multivariable linear mixed model (LMM) of TAPSE through time was fit. Possible predictors were specified a priori and/or with p ≤ 0.1 difference between patients with/without RVD. Data were standardized and centered to facilitate comparison of relative strength of association between predictors of TAPSE. Survival curves for a 30-day death or AHF readmission primary outcome were assessed for RVD, LUS severity, and LVEF. A multivariable generalized linear mixed model (GLMM) for the outcome was used to adjust RVD for LVEF and LUS.

RESULTS

46% (n = 39) of patients at ED arrival showed RVD by TAPSE (median 18 mm, interquartile range 13-23). 18 variables with p ≤ 0.1 unadjusted difference with/without RVD, and 12 a priori predictors of RVD were included in the multivariable LMM model of TAPSE through time (R² = 0.76). Missed antihypertensive medication (within 7 days), ED PPV, chronic obstructive pulmonary disease history, LVEF, LUS congestion severity, and right ventricular systolic pressure (RVSP) were the strongest multivariable predictors of RVD, respectively, and the only to reach statistical significance (p < 0.05). 30-day death or AHF readmission was associated with RVD at ED arrival (hazard ratio {HR} 3.31 {95%CI: 1.28-8.53}, p = 0.009), ED to discharge decrease in LUS (HR 0.11 {0.01-0.85}, p < 0.0001 for top quartile Δ), but not LVEF (quartile 2 vs. 1 HR 0.78 {0.22-2.68}, 3 vs. 1 HR 0.55 {0.16-1.92}, 4 vs. 1 HR 0.32 {0.09-1.22}, p = 0.30). The area under the receiver operating curve on GLMM for the primary outcome by TAPSE (p = 0.0012), ΔLUS (p = 0.0005), and LVEF (p = 0.8347) was 0.807.

CONCLUSION

In this observational study, RVD was common in AHF, and predicted by congestion on LUS, LVEF, RVSP, and comorbidities from ED arrival through discharge. 30-day death or AHF-rehospitalization was associated with RVD at ED arrival and ΔLUS severity, but not LVEF.

Lung Ultrasound-Guided Emergency Department Management of Acute Heart Failure (BLUSHED-AHF): A Randomized Controlled Pilot Trial

OBJECTIVES

The goal of this study was to determine whether a 6-hour lung ultrasound (LUS)-guided strategy-of-care improves pulmonary congestion over usual management in the emergency department (ED) setting. A secondary goal was to explore whether early targeted intervention leads to improved outcomes.

BACKGROUND

Targeting pulmonary congestion in acute heart failure remains a key goal of care. LUS B-lines are a semi-quantitative assessment of pulmonary congestion. Whether B-lines decrease in patients with acute heart failure by targeting therapy is not well known.

METHODS

A multicenter, single-blind, ED-based, pilot trial randomized 130 patients to receive a 6-hour LUS-guided treatment strategy versus structured usual care. Patients were followed up throughout hospitalization and 90 days' postdischarge. B-lines ≤15 at 6 h was the primary outcome, and days alive and out of hospital (DAOOH) at 30 days was the main exploratory outcome.

RESULTS

No significant difference in the proportion of patients with B-lines ≤15 at 6 hours (25.0% LUS vs 27.5% usual care; P = 0.83) or the number of B-lines at 6 hours (35.4 ± 26.8 LUS vs 34.3 ± 26.2 usual care; P = 0.82) was observed between groups. There were also no differences in DAOOH (21.3 ± 6.6 LUS vs 21.3 ± 7.1 usual care; P = 0.99). However, a significantly greater reduction in the number of B-lines was observed in LUS-guided patients compared with those receiving usual structured care during the first 48 hours (P = 0.04).

CONCLUSIONS

In this pilot trial, ED use of LUS to target pulmonary congestion conferred no benefit compared with usual care in reducing the number of B-lines at 6 hours or in 30 days DAOOH. However, LUS-guided patients had faster resolution of congestion during the initial 48 hours. (B-lines Lung Ultrasound-Guided ED Management of Acute Heart Failure Pilot Trial; NCT03136198).

B-line quantification: comparing learners novice to lung ultrasound assisted by machine artificial intelligence technology to expert review

BACKGROUND

The goal of this study was to assess the ability of machine artificial intelligence (AI) to quantitatively assess lung ultrasound (LUS) B-line presence using images obtained by learners novice to LUS in patients with acute heart failure (AHF), compared to expert interpretation.

METHODS

This was a prospective, multicenter observational study conducted at two urban academic institutions. Learners novice to LUS completed a 30-min training session on lung image acquisition which included lecture and hands-on patient scanning. Learners independently acquired images on patients with suspected AHF. Automatic B-line quantification was obtained offline after completion of the study. Machine AI counted the maximum number of B-lines visualized during a clip. The criterion standard for B-line counts was semi-quantitative analysis by a blinded point-of-care LUS expert reviewer. Image quality was blindly determined by an expert reviewer. A second expert reviewer blindly determined B-line counts and image quality. Intraclass correlation was used to determine agreement between machine AI and expert, and expert to expert.

RESULTS

Fifty-one novice learners completed 87 scans on 29 patients. We analyzed data from 611 lung zones. The overall intraclass correlation for agreement between novice learner images post-processed with AI technology and expert review was 0.56 (confidence interval [CI] 0.51-0.62), and 0.82 (CI 0.73-0.91) between experts. Median image quality was 4 (on a 5-point scale), and correlation between experts for quality assessment was 0.65 (CI 0.48-0.82).

CONCLUSION

After a short training session, novice learners were able to obtain high-quality images. When the AI deep learning algorithm was applied to those images, it quantified B-lines with moderate-to-fair correlation as compared to semi-quantitative analysis by expert review. This data shows promise, but further development is needed before widespread clinical use.

Effects of Lung Ultrasonography-Guided Management on Cumulative Fluid Balance and Other Clinical Outcomes: A Systematic Review

Lung ultrasonography is accurate in detecting pulmonary edema and overcomes most limitations of traditional diagnostic modalities. Whether use of lung ultrasonography-guided management has an effect on cumulative fluid balances and other clinical outcomes remains unclear. In this systematic review, we included 12 studies using ultrasonography guided-management with a total of 2290 patients. Four in-patient studies found a reduced cumulative fluid balance (ranging from -0.3 L to -2.4 L), whereas three out-patient studies found reduction in dialysis dry weight (ranging from -2.6 kg to -0.2 kg) compared with conventionally managed patients. None of the studies found adverse effects related to hypoperfusion. The use of lung ultrasonography-guided management was not associated with other clinical outcomes. This systematic review shows that lung ultrasonography-guided management, exclusively or in concert with other diagnostic modalities, is associated with a reduced cumulative fluid balance. Studies thus far have not shown a consistent effect on clinical outcomes.

Assessing congestion in acute heart failure using cardiac and lung ultrasound - a review

Introduction: Acute heart failure (AHF) is one of the leading causes of hospital admissions and is characterized by systemic and pulmonary congestion, which often precedes the overt clinical signs and symptoms. Echocardiography in the management of chronic HF is well described; however, there are less evidence regarding echocardiography and lung ultrasound (LUS) in the acute setting.Areas covered: We have summarized current evidence regarding the use of echocardiography and LUS for assessing congestion in patients with AHF. We discuss the value and reliability of handheld/pocketsize ultrasound devices in AHF.Expert opinion: Echocardiography is an essential tool for the diagnostic work up in patients with AHF. No individual parameter reliably detects congestion, thus the physician must integrate several measurements from the right and left heart. Novel methods and advances in cardiac imaging and clinical chemistry make it possible to detect congestion at an early stage. LUS is particularly helpful in assessing congestion, and it has demonstrated diagnostic, therapeutic, and prognostic value in AHF. LUS is relatively easy to learn and allows for quick assessment of the presence of pulmonary congestion and pleural effusion. We recommend integration of LUS for routine management of patients with AHF.

Are pocket sized ultrasound devices sufficient in the evaluation of lung ultrasound patterns and aeration scoring in pulmonary ICU patients?

Lung ultrasound (LUS) is a practical diagnostic tool for several lung pathologies. Pocket sized USG devices (PSUDs) are more affordable, accessible, practical, and learning to use them is easier than standard ultrasound devices (SUDs). Their capability in image quality have been found as comparable with standard USG machines. Studies have been showing that these devices can be useful as much as SUDs in the evaluation of heart, abdomen, vascular structures, diaphragm and optic nerve. The aim of this study is to compare PSUD with a standard ultrasound devices (SUD) in the evaluation of LUS patterns such as alveolar, interstitial syndromes and lung aeration score (LAS). Study performed in an University Hospital Pulmonary ICU. All patients older than 18 years old were included in this study. The sector probe of SUD (Vivid-Q) and PSUD (Vscan) were used for investigation of A lines, interstitial (B lines), alveolar syndromes (consolidation, hepatisation, air bronchograms) and pleural effusion. 33 patients were included in the study. When PSUD was compared with SUD in terms of total B2 count, and LAS in the right, left and both lung, there was an agreement without proportional bias according to Bland Altman test. There was also good inter class correlation coefficient value as greater than 0.8 and 0.7 between two physicians in terms of counting of total B1, B2, total B lines and calculating of total LAS for SUD and PSUD respectively. PSUDs is a reliable and valid method for evaluation of LUS patterns like SUDs.

Prognostic value of pulmonary oedema assessed by lung ultrasound in patient with acute heart failure

It is very important to assess pulmonary oedema in patients with acute heart failure. The aim of the study was to investigate the accuracy of lung ultrasound in evaluating pulmonary oedema and to explore lung ultrasound in predicting the prognosis. One hundred twenty-four acute heart failure patients were divided into 3 groups, according to the total number of lung ultrasound B-lines groups: B-lines < 15 was the mild pulmonary oedema group (33 cases), 15 ≤ B-lines < 30 was the moderate pulmonary oedema group (33 cases), and B-lines ≥ 30 was the severe pulmonary oedema group (58 cases). The PiCCO monitoring system was used in 11 patients and measured 26 times in different clinical situations. EVLWI have a higher positive correlation with B-lines (r = 0.95), compared with NT-proBNP and E/e' (r = 0.72, r = 0.62). During 1 year of follow-up, a multivariate cox regression analysis showed that age, E/e' and B-lines ≥ 30 at admission (C-index of 75%) were risk factors for prognosis. 12-month event-free survival showed a significantly worse outcome was observed in patients with ≥ 30 B-lines at admission. B-lines have a good correlation with EVLWI; age, E/e' and B-lines ≥ 30 at admission were risk factors for prognosis.

Point-of-Care Lung Ultrasound for Detecting Severe Presentations of Coronavirus Disease 2019 in the Emergency Department: A Retrospective Analysis

OBJECTIVES

Analyze the diagnostic test characteristics of point-of-care lung ultrasound for patients suspected to have novel coronavirus disease 2019.

DESIGN

Retrospective cohort.

SETTING

Two emergency departments in Detroit, Michigan, United States, during a local coronavirus disease 2019 outbreak (March 2020 to April 2020).

PATIENTS

Emergency department patients receiving lung ultrasound for clinical suspicion of coronavirus disease 2019 during the study period.

INTERVENTIONS

None, observational analysis only.

MEASUREMENTS AND MAIN RESULTS

By a reference standard of serial reverse transcriptase-polymerase chain reactions, 42 patients were coronavirus disease 2019 positive, 16 negative, and eight untested (test results lost, died prior to testing, and/or did not meet hospital guidelines for rationing of reverse transcriptase-polymerase chain reaction tests). Thirty-three percent, 44%, 38%, and 17% had mortality, ICU admission, intubation, and venous or arterial thromboembolism, respectively. Receiver operating characteristics, area under the curve, sensitivity, and specificity with 95% CIs were calculated for five lung ultrasound patterns coded by a blinded reviewer and chest radiograph. Chest radiograph had area under the curve = 0.66 (95% CI, 0.54-0.79), 74% sensitivity (95% CI, 48-93%), and 53% specificity (95% CI, 32-75%). Two lung ultrasound patterns had a statistically significant area under the curve: symmetric bilateral pulmonary edema (area under the curve, 0.57; 95% CI, 0.50-0.64), and a nondependent bilateral pulmonary edema pattern (edema in superior lung ≥ inferior lung and no pleural effusion; area under the curve, 0.73; 95% CI, 0.68-0.90). Chest radiograph plus the nondependent bilateral pulmonary edema pattern showed a statistically improved area under the curve (0.80; 95% CI, 0.68-0.90) compared to either alone, but at the ideal cutoff had sensitivity and specificity equivalent to nondependent bilateral pulmonary edema only (69% and 77%, respectively). The strongest combination of clinical, chest radiograph, and lung ultrasound factors for diagnosis was nondependent bilateral pulmonary edema pattern with temperature and oxygen saturation (area under the curve, 0.86; 95% CI, 0.76-0.94; sensitivity = 77% [58-93%]; specificity = 76% [53-94%] at the ideal cutoff), which was superior to chest radiograph alone.

CONCLUSIONS

Lung ultrasound diagnosed severe presentations of coronavirus disease 2019 with similar sensitivity to chest radiograph, CT, and reverse transcriptase-polymerase chain reaction (on first testing) and improved specificity compared to chest radiograph. Diagnostically useful lung ultrasound patterns differed from those hypothesized by previous, nonanalytical, reports (case series and expert opinion), and should be evaluated in a rigorous prospective study.