Outcomes of Simplified Lung Ultrasound Exam in COVID-19: Implications for Self-Imaging.

OBJECTIVES: Lung ultrasound B-lines represent interstitial thickening or edema and relate to mortality in COVID-19. As B-lines can be detected with minimal training using point-of-care ultrasound (POCUS), we examined the frequency, clinical associations, and outcomes of B-lines when found using a simplified POCUS method in acutely ill patients with COVID-19.
METHODS: In this retrospective cohort study, hospital data from COVID-19 patients who had undergone lung imaging during standard echocardiography or POCUS were reviewed for an ultrasound lung comet (ULC) sign, defined as the presence of ≥3 B-lines from images of only the antero-apex of either lung (ULC+). Clinical risk factors, oximetry and radiographic results, and disease severity were analyzed for associations with ULC+. Clinical risk factors and ULC+ were analyzed for associations with hospital mortality or the need for intensive care in multivariable models.
RESULTS: Of N = 160 patients, age (mean ± standard deviation) was 64.8 ± 15.5 years, and 46 (29%) died. ULC+ was present in 100/160 (62%) of patients overall, in 81/103 (79%) of severe-or-greater disease versus 19/57 (33%) of moderate-or-less disease (P < .0001) and was associated with mortality (odds ratio [OR] = 2.4 [95% confidence interval [CI]: 1.1-5.4], P = .02) and the need for intensive care (OR = 5.23 [95% CI: 2.42-12.40], P < .0001). In the multivariable models, symptom duration and severe-or-greater disease were associated with ULC+, and ULC+, diabetes, and symptom duration were associated with the need for intensive care.
CONCLUSIONS: B-lines in the upper chest were common and related to disease severity, intensive care, and hospital mortality in COVID-19. Validation of a simplified lung POCUS exam could provide the evidence basis for a self-imaging application during the pandemic.

confidence interval

infection with SARS coronavirus 2

computed tomography

U.S. Food and Drug Administration

point‐of‐care ultrasound

saturation of oxygen

ultrasound lung comet sign

COVID‐19 has resulted in greater than 3 million deaths worldwide and more than 550,000 deaths in the United States within its first year alone. Most of the mortality has been due to respiratory failure that usually begins 10 days after exposure, with the progression of the viral infection from the nasopharynx to the lung. Methods to identify those patients in whom disease will progress have medical and social implications for quarantine, triage, and early treatments. Notably, once disease progresses to the lungs, point‐of‐care ultrasound (POCUS) can readily detect B‐lines, , , , , an ultrasound artifact presumably generated by the visceral pleura or neighboring interstitium thickened from infection and edema. Outcome studies in COVID‐19 have demonstrated that a comprehensive lung B‐line score relates to a worse prognosis. , , , , , ,

Earlier during the pandemic, healthcare professionals who were infected with COVID‐19 were able to image their own lungs and demonstrate the development of B‐lines using small POCUS devices while isolating at home. , , Although few data exist to guide how to image one's self, a small study has shown that the upper chest is a region that is easily accessed and amenable to the simplified imaging that could facilitate self‐imaging by patients, family members, or caregivers. However, before a patient self‐imaging strategy can be studied, B‐lines detected at the upper chest will first need to show a significant prevalence and a relationship to COVID‐19 disease severity and outcomes. Therefore, we sought to determine the frequency, clinical associations, and mortality of a simple B‐line finding from only the upper chest sites across a clinical spectrum of acutely infected patients who tested positive for COVID‐19.

Materials and Methods

This retrospective cohort study took place at a 300‐bed community teaching medical center utilizing data from April 2020 to March 2021 and was approved by the Scripps Institutional Review Board (Scripps Health, San Diego, #20‐7691) with waiving of patient consent. At this hospital, two lung ultrasound antero‐apical views in the second or third intercostal space in the mid‐clavicular line, previously shown to have prognostic value, had been incorporated before the pandemic into the institution's routine echo imaging protocol using standard echocardiographs (S5‐1MHz, Phillips iE33; Phillips Healthcare, Andover, MA) and were also a part of routine POCUS bedside examination using one of many pocket‐sized U.S. Food and Drug Administration‐approved POCUS devices, including Lumify (S4‐1MHz transducer, Phillips Healthcare), Vave (4 MHz transducer, Vave Health, Santa Clara, CA), and Vscan (1.7–3.8 MHz transducer, GE Healthcare, Wauwatosa, WI) in use at the institution. All patients, who had tested positive for COVID‐19 by polymerase chain reaction testing via nucleic acid amplification (Hologic, San Diego, CA) and had received echo or lung POCUS, had their antero‐apical images noted for an ultrasound lung comet (ULC) sign, defined as the presence or coalescence of three or more vertical ring‐down linear B‐line artifacts in a single image in either lung , (Figure 1). Patients were considered ULC+ if either lung showed a comet sign with images confirmed by an expert physician reviewer with over a decade of experience in interpretation and acquisition of lung images who was blinded to patient outcome.

Figure 1

Composite of lung imaging: abnormal study with ≥3 B‐lines (arrows), ULCs+; inset, example of probe placement on upper anterior chest, intercostal space #2.

Electronic medical records (Epic Systems, Verona, WI) were examined for patient clinical data while blinded to the ULC result, including age, gender, body mass index, presence of diabetes or hypertension, results of radiographic imaging (chest X‐ray or chest computed tomography [CT]), and the estimated time since onset of symptoms to ultrasound imaging (symptom duration). The results of chest X‐ray or chest CT from within 48 hours of ultrasound imaging were considered abnormal if the formal report mentioned diffuse, interstitial, lobar, or ground‐glass infiltrates and stratified by whether the infiltrates were considered extensive (eg, >50% of the lung). The Centers for Disease Control Clinical Severity Scale was used to categorize COVID‐19 disease severity at the time of ultrasound evaluation as follows :

Asymptomatic or pre‐symptomatic infection: Individuals who tested positive for the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS‐CoV‐2) using a virologic test, but who had no symptoms that were consistent with COVID‐19;

Mild illness: Individuals who had any of the various signs and symptoms of COVID‐19 (eg, fever, cough, sore throat, malaise, headache, muscle pain, nausea, vomiting, diarrhea, loss of taste, and smell) but who did not have shortness of breath, dyspnea, or abnormal chest imaging;

Moderate illness: Individuals who showed evidence of lower respiratory disease during clinical assessment or imaging and who had a saturation of oxygen (SpO2) ≥94% on room air at sea level;

Severe illness: Individuals who had SpO2 < 94% on room air at sea level, (or, for patients with chronic hypoxemia, a decrease from baseline of >3%), a ratio of arterial partial pressure of oxygen to fraction of inspired oxygen (PaO2/FiO2) < 300 mmHg, respiratory rate > 30 breaths/min, or lung infiltrates >50%, or;

Critical illness: Individuals who had respiratory failure, septic shock, and/or multiple organ dysfunction.

Two clinical outcomes, mortality and the need for intensive care, were considered. Mortality was defined by patient death or disposition to hospice during hospitalization. The need for intensive care was present if the patient was transferred to the intensive care unit or had critical disease severity.

Statistical Analysis

Continuous raw data are expressed as a mean ± standard deviation. Based upon the need for significant hospital resources (ie, supplemental oxygen and intravenous therapies) in the severe‐or‐greater categories, data from the severe and critical categories were grouped and compared to a moderate‐or‐less category group, defined as the combination of the asymptomatic, mild, and moderate categories, for outcome analysis. Clinical risk factors (age ≥ 65 years, gender, diabetes, hypertension, obesity defined as a body mass index >30, and time from symptom onset to the imaging study of >10 days), clinical test results (abnormal chest X‐ray/CT, SpO2 < 94%, or requiring supplemental oxygen), and disease severity category were analyzed for univariable associations with ULC+, reported as odds ratios (ORs) with 95% confidence intervals [95% CIs]. Clinical risk factors and ULC+ were analyzed for associations with the two outcomes in reduced multivariable models using backward elimination. P‐values were considered significant when ≤.05. All data were analyzed using R version 4.0.3 (2020).

Results

Of N = 160 patients, age was 64.8 ± 15.5 years, 64% were male, and 94% were hospital admissions while 6% were discharged from the emergency department. ULC+ was present in 100/160 (62%) of patients overall and in 81/103 (79%) of severe‐or‐greater disease versus 19/57 (33%) of moderate‐or‐less disease (P < .0001) (Figure 2). In the 100 patients who were ULC+, the duration of symptoms before ULC detection was <24 hours in 8%, <5 days in 37%, and ≤10 days in 61%. Patients with shorter symptom durations had less significant disease severity (Figure 3). Of 19 patients who were in the moderate‐or‐less group and ULC+, only 1/19 (5%) had hypoxemia due to reasons not directly attributed to COVID‐19 disease and most, 16/19 (84%), had demonstrated ULC+ within 10 days of symptom onset. Age ≥ 65 years, a history of hypertension, symptom duration >10 days, SpO2 < 94%, abnormal chest X‐ray/CT, and severe‐or‐greater disease severity (Table 1) were associated with ULC+. In the multivariable analysis, ULC+ was best related to severe‐or‐greater disease (OR = 6.43 [95% CI: 3.11–13.71], P < .0001) and duration of symptoms >10 days (OR = 2.66 [95% CI: 1.13–6.68], P = .03).

Figure 2

Bar graph of ULC+ frequency and inpatient mortality (%) versus disease severity category. The ULC frequency and mortality were significantly lower between moderate‐or‐less versus severe‐or‐greater groups (see text). ASx, asymptomatic.

Figure 3

Bar graph of percent disease severity category versus symptom duration. Disease severity was less in patients with shorter durations (see text).

Table 1

Clinical Risk Factors and Tests and Their Association with ULC+

N = 160	n (% of Total)	COMETS (+) n = 100	COMETS (−) n = 60	OR [95% CI]	P Value (Univariable)
Age ≥ 65 years	90 (56.3%)	63 (63%)	27 (45%)	2.08 (1.09–4.02)	.0273
Gender (% male)	102 (63.8%)	68 (68%)	34 (56.7%)	1.63 (0.84–3.16)	.1500
Diabetes	90 (56.3%)	62 (62%)	28 (46.7%)	1.86 (0.98–3.59)	.0596
Hypertension	112 (70%)	76 (76%)	36 (60%)	2.11 (1.06–4.24)	.0340
Obesity (BMI > 30)	67 (41.9%)	46 (46%)	21 (35%)	1.58 (0.82–3.09)	.1730
Time from onset of symptoms > 10 days				3.62 (1.66–8.61)	.0020
0	22 (13.8%)	8 (8%)	14 (23.3%)
1–4	50 (31.3%)	29 (29%)	21 (35%)
5–10	40 (25%)	24 (24%)	16 (26.7%)
>10	48 (30%)	39 (39%)	9 (15%)
SpO2 < 94%	102 (63.8%)	80 (80%)	22 (36.7%)	6.91 (3.42–14.45)	<0.0001
Abnormal chest X‐ray or CT	127 (79.4%)	90 (90%)	37 (61.7%)	5.68 (2.44–14.14)	.0001
Severity classification				7.36 (3.63–15.52)	<.0001*
Asymptomatic	24 (15%)	7 (7%)	17 (28.3%)
Mild	10 (6.3%)	2 (2%)	8 (13.3%)
Moderate	23 (14.4%)	10 (10%)	13 (21.7%)
Severe	57 (35.6%)	43 (43%)	14 (23.3%)
Critical	46 (28.8%)	38 (38%)	8 (13.3%)

BMI, body‐mass index; CI, confidence interval, CT, computed tomography; OR, odds ratio.

relates to the p‐value of the Severity Classification where OR 7.36 (3.63‐15.52) <.0001 . P‐value for Severity classification derived from univariable logistic regression model with aggregated categories: Severe/critical relative to asymptomatic/mild/moderate.

In outcome analysis, patient mortality was 29% (46/160) overall and 42% (43/103) in the severe‐or‐greater group versus 5% (3/57) in the moderate‐or‐less group (P < .0001). ULC+ was associated with mortality, present in 76% (35/46) of those who died vs. 57% (65/114) in survivors (OR = 2.4 [95% CI: 1.1–5.4], P = .02). In the multivariable model that included ULC+ and clinical risk factors, mortality was best related to age ≥65 years (OR = 3.06 [95% CI: 1.43–6.97], P = .0053), male gender (OR = 2.71 [95% CI: 1.22–6.44], P = .02), and symptom duration >10 days (OR = 1.86 [95% CI: 1.33–6.23], P = .007). ULC+ was associated with the need for intensive care (OR = 5.23 [95% CI: 2.42–12.40], P < .0001), which occurred in 57/160 (36%) patients, 28 of whom had transferred from an initial ward admission. In the multivariable model, the need for intensive care was best related to ULC+ (OR = 3.88 [95% CI: 1.69–9.61], P = .002), diabetes (OR = 4.04 [1.86–9.32], P = .0006), and symptom duration >10 days (OR = 3.35 [95% CI: 1.52–7.63], P < .003).

Discussion

This study, by analyzing a clinical cross‐section of patients with COVID‐19, demonstrated that a simple sign of 3 B‐lines found while imaging the upper chest was common and associated with disease severity, intensive care, and hospital mortality. Notably, in the subgroup with non‐severe illness, nearly a third of patients demonstrated the ULC+ sign at this site, without hypoxemia and usually within 10 days of symptom onset. The findings of this study provide the evidence basis for simplified imaging at easily accessible upper chest sites and have implications for increasing the application of POCUS in assessing prognosis, disease progression, triage, and early therapies in COVID‐19 disease.

In alveolar‐interstitial syndromes, ultrasound imaging of the lung has been shown to produce a specific reverberation or “ring‐down” artifact, called the B‐line, presumably generated from surface interfaces on the lung that have been thickened by inflammation, edema, or fibrosis. , , , Since the early description of lung ultrasound imaging over 20 years ago, the antero‐apical lung region has been repeatedly recognized as a distinct imaging zone for B‐lines. , , In a retrospective analysis of outcomes in N = 486 hospitalized patients referred for echo, using the same antero‐apical site as the current study, ULC+ had a frequency of 29% and was associated with an 11% in‐hospital mortality before the pandemic, as compared to the current study of 63% frequency and 29% mortality in similarly‐referred COVID‐19 patients. The increased frequency and mortality of B‐lines in COVID‐19 is supported by reports of postmortem pathology, which commonly demonstrates edematous lungs and findings of diffuse alveolar damage with hyaline membranes and type 2 pneumocyte hyperplasia, consistent with widespread virus‐induced acute lung injury.

The current findings are consistent with recent lung ultrasound outcome studies in COVID‐19 , , , , , , that have demonstrated a relationship of patient mortality with complex lung ultrasound scores involving B‐line and pleural abnormalities over the entirety of both lungs. However, the current investigation is unique in its validation of a simple POCUS sign in only the antero‐apical lung zones and by using outcomes across the full spectrum of COVID‐19 severity. Recently, based on localization of findings by chest X‐ray, CT imaging, , and ultrasound, the posterior basal segments have been suggested for imaging COVID‐19, but the involvement of these regions could represent ongoing development of alveolar edema, atelectasis, and mucociliary impairment that are subject to gravity‐dependent forces in the lung bases. As opposed to the simplified protocol, imaging these hard‐to‐reach posterior areas can be a difficult task, as would be to score the 8–16 lung zones that were used in previous ultrasound investigations. In comparison with a study using 8 imaging zones that examined disease severity in 106 consecutive COVID‐19 patients, the prevalence of B‐lines was 31% in mild disease and 70% in severe‐or‐greater disease, which is similar to the 33 and 79% found in the current study using only the 2 apical zones. Our data show a surprisingly high prevalence of a simple upper lobe finding often in the absence of hypoxemia, which may represent an early phase of a progressive infection. Self‐inoculation or tracheobronchial spread to the upper lungs from a virus‐laden nasopharynx infection could be plausible explanations, given that the upper lung regions have been modeled to have a higher deposition and dispersion of submicrometer airborne particles, are suspected to have delayed lymphatic drainage, and are subject to inflammatory oxidative stress from oxygen‐free radicals. Lung ultrasound findings in this infection have been associated with interleukin 6 levels and may herald a stage of cytokine activation. In the current study, the reported duration of symptoms of ≤10 days in the majority of ULC+ patients is consistent with the time‐course of initial lung involvement observed in COVID‐19. ,

Importantly, our data validate an imaging site for patients to self‐image remotely in COVID‐19. The rationale for studying a simplified lung ultrasound protocol during the pandemic is to ultimately increase the availability of outpatient ultrasound examination by making it easier for nonexpert users to apply the technique in the community. Searching for comets in the upper chest is one of the simplest ultrasound maneuvers to learn and perform. The imaging process described does not require patients to undress and takes less than a minute. In a recent feasibility study using a single page of patient instruction, the same lung imaging technique employed in the current study was successfully taught to patients with cardiac disease, unexplained dyspnea, and COVID‐19 pneumonia. Over 85% of patients under 78 years of age obtained adequate images on themselves and overall accuracy was 88% compared to expert imaging. In conjunction with the current findings, our data would project the superiority of ultrasound to self‐monitor initial COVID‐19 infection over other methods, as SpO2 assessment would identify a later, more severe stage of disease and chest X‐ray, which is less sensitive than ultrasound, would require transport to a medical facility. In a quick‐look POCUS application using smartphone devices, the appearance of upper lobe B‐lines could result in immediate quarantine and serum testing of the individual with minimal or atypical symptoms, identify outpatients who may benefit from immunologic therapies, and assist in triage of acutely ill patients arriving at the hospital. More data are needed to validate a self‐imaging telemedicine pathway, confirm the timing and location of initial lung involvement, and investigate whether B‐lines represent higher airborne infectivity. In the future, proof‐of‐concept studies with simplified imaging protocols, patient‐friendly instruction, and artificial intelligence could usher in a novel public telehealth methodology for various diseases and extend ultrasound imaging to nonphysician personnel in homes and hard‐to‐reach, resource‐poor communities.

This retrospective analysis of “front‐line” care during the COVID‐19 pandemic is subject to multiple limitations and biases. Nonconsecutive patients with COVID‐19 who were referred for echo or underwent POCUS were categorized by disease severity, but still likely represented a clinically heterogeneous group, as data may reflect more severe and complicated disease in those referred for echo or lesser disease in others who had been routinely imaged due to the POCUS capabilities of their examining physician. Our unadjusted overall mortality rate of 29% for patients who had been imaged with ultrasound suggests a modest effect of selection bias, as this rate is high but comparable to that of unselected patients in similar‐sized U.S. hospitals during the initial wave of the pandemic and similar to other COVID‐19 ultrasound outcome studies of older patients. , , Concomitant cardiogenic pulmonary edema was not excluded and virus‐induced heart failure exacerbation may have been a significant and prevalent comorbidity, which was nonetheless considered attributable to the COVID‐19 infection. As a single‐center study, our findings represent the prevalence and virulence of cases encountered by a hospital in the downtown San Diego region which has a high proportion of admissions from the surrounding Hispanic community and an underserved homeless population and occurred before widespread vaccination, the appearance of the delta variant, and the development of any standardized outpatient treatments.

In conclusion, a ULC sign consisting of 3 B‐lines, observed upon scanning the upper chest, was common, related to disease severity, intensive care treatment, and patient mortality in COVID‐19. Future studies may clarify the utility of simplified lung ultrasound imaging in staging and monitoring COVID‐19 infection, potentially using data obtained early by affected individuals from within the community.