Home‐operated ultrasound exam for detection of worsening heart failure (HOUSE‐HF)

Abstract

Aims

Acute decompensated heart failure (ADHF) is associated with a high degree of morbidity and mortality. Dynamic lung ultrasound artefact called B‐lines can be obtained at the bedside and directly correlate with pulmonary vascular congestion. Obtaining patient‐performed lung ultrasound images in the outpatient setting is novel. We assessed the feasibility of patients recently hospitalized for ADHF to self‐perform a limited lung ultrasound using a handheld ultrasound probe and upload the images to a secure cloud for physician interpretation.

Methods

This was a prospective observational convenience sample. Patients were enrolled from an urban academic tertiary care centre and were eligible if they had chronic left‐sided heart failure regardless of ejection fraction. While hospitalized, patients were educated for 20 min on a six‐lung‐zone image protocol, how to use the cloud archival system and given a handheld ultrasound transducer and smart tablet. A brief instructional video was also available to patients on the smart tablet throughout the study (https://www.dropbox.com/scl/fi/bii7ovdcv21ps7yxyqsy1/120‐21080‐00‐Rev‐01‐BNI‐041‐UPENN‐IN‐APP‐TRAINING‐VIDEO.mp4?rlkey=f5vu55xbnugdoz6jzyb8lv872&st=56es4qif&dl=0). Patients were asked to upload images three times weekly, for 3 weeks, for a total of nine studies. All images were reviewed and a B‐line score was calculated for each lung zone, and a total B‐line score for the entire exam. Additionally, patients completed a survey to assess the patient‐centred experience.

Results

A total of 15 patients were enrolled, all of whom completed seven or more studies (10 patients completed all 9). Median patient age was 63 years (range: 28 –86 years), the majority were male (73%), white (60%) and average body mass index was 33 kg/m². Of them,33.3% had an ejection fraction >50%, average hospital length of stay was 6.3 days. Of the 792 potential images, 788 were obtained (99.5%). Of these, a total of 637 scans were interpretable (80.8%). The right upper apical lung zone (zone 1R) was most often adequate for interpretation (96.2%), where left lower mid‐axillary (zone 3L) was least often interpretable (69.5%). The average number of B‐lines per six‐image scan was three (with a range of 0–13). Patient survey data identified zone 3L as the most challenging to obtain with overall high satisfaction with the study educational materials.

Conclusions

This pilot study demonstrates that patients with hospitalized ADHF can be taught to use a handheld portable ultrasound device and obtain and upload high quality lung ultrasound images. Compliance with the study protocol and ability to obtain some images were excellent. Further studies are needed to determine if patient‐performed lung ultrasound can help detect and manage acute worsening HF in this patient population.

Introduction

Heart Failure (HF) is a common condition affecting 2%–3% of the US population, is increasing in incidence and prevalence, and is associated with a high degree of morbidity and mortality. ¹ Despite significant advances in HF pharmacologic and device therapy, acute decompensated HF leading to hospitalization is associated with poor prognosis and high rates of subsequent readmission. ² The burden on the medical system is substantial and worsening over time. ³ Predicting and intervening on worsening HF before the patient is symptomatic and requires urgent intervention or hospitalization remains a major challenge. ⁴ Numerous studies have shown that intrathoracic and intracardiac pressures increase and herald decompensated HF prior to traditional clinical signs of congestion such as worsening dyspnoea, weight gain or oedema. ⁵ , ⁶ However, thoracic impedance monitors and implantable pulmonary artery sensors are invasive, which limit their availability and generalizability.

Point of care ultrasound (POCUS) can allow for rapid clinical assessment of pulmonary vascular congestion. ⁷ POCUS is low cost, repeatable and non‐invasive. Lung POCUS offers diagnostic information with respect to pulmonary congestion in patients with HF when a ring‐down artefact, or B‐line is present. ⁸ Previous studies have shown that B‐line prevalence is dynamic and correlates directly with pulmonary vascular congestion. ⁹ , ¹⁰ Although POCUS is traditionally performed by the clinician, performance of POCUS by non‐physicians augmented by expert tele‐guidance and artificial intelligence software has been demonstrated. ¹¹ , ¹²

The mission of patient‐centred care includes that it is collaborative, coordinated and accessible and that the right care is provided at the right time and place. ¹³ Historically, patients have been extensively empowered to measure and monitor medical data at home, from point‐of‐care COVID‐19 tests to blood pressure cuffs, scales and glucometers. Introducing ultrasound imaging into this patient‐centred toolbox is novel. Recently, there have been limited reports of patient self‐performed lung ultrasound for home monitoring of COVID‐19 infection. ¹⁴ The use of lung ultrasound by patients at home for monitoring of acute decompensated HF remains unexplored.

The aim of the current study is to determine the feasibility of ambulatory patients following a recent HF hospitalization to self‐perform lung POCUS and longitudinally upload the images to a secure cloud. This pilot is part of a larger proposed study to determine if these images can be then used to titrate medications before symptoms appear and require hospital‐based care.

Methods

This was a prospective observational convenience sample of patients with known HF admitted to an urban academic tertiary care centre under the management of an advanced HF cardiologist in a collaboration between the Division of Cardiology and the Department of Emergency Medicine. While initial enrolment criteria required patients to have an ejection fraction less than 40%, the enrolment criteria were broadened on 21 November 2022 to allow for all patients with clinical left‐sided HF regardless of ejection fraction. Full inclusion and exclusion criteria are shown in Table S1. Patients were enrolled from 10/12/2022 to 1/31/2023. The study was funded by an industry grant (Butterfly, Burlington, MA, USA) and involved investigational use of the Butterfly device under a University of Pennsylvania Institutional Review Board approved protocol (protocol 850789). Two research coordinators were involved in the study.

Patients were approached for enrolment when discharge was expected within 48 h. After the patient consented, a screening lung exam was performed, and images were uploaded to the cloud for review. If image quality was adequate (pleural line visualized, clear of A‐lines) and extra‐pulmonary findings that would preclude enrolment (pleural effusion and pleural consolidations) were excluded, the patient was cleared for enrolment. The patient was educated on the study imaging protocol and the cloud archival system and given imaging supplies including a study case with the handheld ultrasound transducer (ButterflyIQ+), iPad (Apple, Cupertino, CA, USA) and ultrasound gel. Multi‐modality educational documents were created to address different learning styles including a printed picture‐based scanning protocol (Figure S1), a 5 min video tutorial embedded on the iPad (https://www.dropbox.com/scl/fi/bii7ovdcv21ps7yxyqsy1/120‐21080‐00‐Rev‐01‐BNI‐041‐UPENN‐IN‐APP‐TRAINING‐VIDEO.mp4?rlkey=f5vu55xbnugdoz6jzyb8lv872&st=56es4qif&dl=0) and scanning prompts on the imaging platform that guided the patient through the six‐image exam.

A Health Insurance Portability and Accountability Act compliant image storage cloud was developed so that patients could upload personal images, but not see the studies of other participants. To ensure standardization, only the lung preset was available on the ultrasound device. Based on previous data on presets for lung imaging, the gain was set low, 15 cm depth, with 6 s clips that recorded prospectively. ¹⁵ After acquiring an image, the protocol would automatically advance to the next lung zone location. The lung zones included anterior upper (R1/L1), anterior lower (R2/L2) and mid axillary line (R3/L3) on both sides of the chest. Patients were instructed to upload images three times a week for 3 weeks immediately upon discharge from the hospital. All images were obtained by the patient without assistance from caregivers (Figure 1). Any interaction with the study participants was documented. A post‐participation survey was administered to the patients to assess confidence in scanning, identify challenges with the protocol and assess the quality of the training materials.

Figure 1.

Image of patient self‐scanning.

While traditional lung POCUS research involves a 28‐point lung exam, recent data have shown that six‐point lung point exam has similar diagnostic accuracy for pulmonary vascular congestion. ¹⁶ Thus, the latter protocol was used for patient simplicity. The lung zone nomenclature is described in Figure 2. Images of lung ultrasound are shown in Figure 3. A B‐line was defined as a pleural based ring‐down artefact that extended the 15 cm of the image and obliterated A‐lines in its path. ¹⁷ B‐lines were counted using the previously described multiple counting method. ¹⁸ B‐lines were counted within a single rib space, counted over the entire respiratory cycle, with confluent B‐lines counted as multiple based on the percentage of the rib space they occupy, at the moment when the most B‐lines are seen. Confluent B‐lines were scored as a percentage of the rib space occupied and then divided by 10 (e.g., 30% of the rib space = score of 3). B‐lines were manually counted by a trained provider (Emergency Medicine Ultrasound Director with >10 years experience) with a minimum score of 0 and maximum of 10 per rib space. For this study, the B‐line score was calculated for each image, as well as for the total six images in each study. When not all six lung zones provided interpretable images, the number of B‐lines were normalized to a six‐image study (by multiplying the number of measured B‐lines by the number of obtained images/6) to assist in comparisons between studies and between patients.

Figure 2.

Percentage of images obtained and interpretable for each lung zone and ease of patient capture. * = Easy or very easy to obtain image as rated by the patient.

Figure 3.

Lung ultrasound images as displayed on the smart tablet during image acquisition and as reported to the physician. (A) Real‐time lung ultrasound images during scanning; (B) viewable Images as reported to the physician.

The primary outcome variables included the number of patients completing all nine uploads and the percentage of interpretable images per study upload. Secondary outcomes included lung zone specific interpretability percentages, B‐line scores and patient perceptions of the imaging experience.

Results are presented as percentages or medians with IQR where applicable. All analyses were performed using SAS statistical software (version 9.4, SAS Institute, Cary NC). Figures were generated using GraphPad Prism (version 10.0.3, GraphPad Software, San Diego, CA, USA).

Results

A total of 210 admitted patients were screened for enrolment, 177 failed to screen in, 17 declined participation and 16 agreed to consent (Figure 4). No patients were excluded due to their lung findings on the initial screening ultrasound. Median patient age was 63 years (range: 28–86 years), 73% male, 60% White and 47% obese. Ten patients had ejection fraction <50%, and the average length of hospital stay was 6.3 days (Table 1).

Figure 4.

Consort diagram for patients screened and included in the study. (A) Initial protocol; (B) followed protocol revision to allow for patients with recent inotropes and any ejection fraction.

Table 1.

Patient characteristics.

Demographic		Frequency	(Percentage)
Sex	Male	11	(73.3)
	Female	4	(26.7)
Race	Black	5	(33.3)
	Pacific Islander	1	(6.7)
	White	9	(60.0)
Comorbidities	Hyperlipidaemia	9	(60.0)
	Chronic kidney disease	9	(60.0)
	Coronary artery disease	8	(53.3)
	Hypertension	8	(53.3)
	Diabetes	7	(46.7)
	Obesity (BMI > 30)	7	(46.7)
	Atrial fibrillation	7	(46.7)
	Obstructive sleep apnoea	7	(46.7)
	Ventricular tachycardia	6	(40.0)
	Chronic obstructive pulmonary disease	2	(13.3)
Ejection fraction	>50%	5	(33.3)
	≤50%	10	(66.7)
		Median	(IQR)
	Age (years)	63.0	(49.5, 70.0)
	Weight (kg)	85.4	(79.2, 105.1)
	BMI (kg/m2)	29.3	(26.5, 36.2)
	Length of stay (days)	5.0	(4.0, 7.5)
	NT‐ProBNP (n = 13)	1641.5	(913, 4326)

Abbreviations: BMI, body mass index; IQR, interquartile range; NT‐ProBNP, N terminal pro brain natriuretic peptide.

Of the 792 potential lung POCUS images, 788 were obtained (99.5%), and of these, 637 (80.8%) were interpretable. Of the 15 patients, 10 uploaded 9 studies or more, 3 uploaded 8 studies, and 2 uploaded 7 studies. Thus 67% completed the full nine‐scan protocol. Zone R1 was most often adequate for interpretation (96.2%), where zone L3 was least (69.5%) (Figure 2). The median number of interpretable images was high at five per scan. Figure 5 shows a graphical representation of the B‐line score standardized to a six images scan for each upload, and across all 15 patients. The average number of B‐lines per six‐image scan was three (with a range of 0–13). Figure S2 shows the temporal trend of B‐lines scores for each participant. Eight patients had changes to their diuretics during the study, of whom three had a decrease in diuretics and five had an increase. Only one patient was readmitted during the study, and this was for a non‐HF‐related issue.

Figure 5.

B‐line score standardized to a six‐image scan for each upload per patient.

Study coordinators provided telephone support for a total of 24 times across all 15 patients (average 1.6 per patient, range 0–4). 33.3% (8/24) were to help troubleshoot a technical issue, and 16% (4/24) of the interactions were prompted by the research coordinator in response to suboptimal image quality. Of the patients, 93% fully completed a post‐study survey. Patients identified zones R3 and L3 as the most difficult (3.1 out of 4) and zone R1 as the easiest to obtain (3.5 out of 4). Of the patients, 100% reported that the three training tools (verbal/hands‐on orientation, video tutorial and laminated training card) were simple or only slightly complicated to understand. The training materials were mostly rated as very useful (Figure 6). Of the 12 patients that received a phone call from the study coordinator, only 1 felt that the conversation was more than ‘slightly complicated’.

Figure 6.

Complexity and usefulness of training materials from patient survey.

Discussion

This is the first patient‐centred study to assess if patients admitted for acute decompensated HF can self‐perform lung POCUS after a brief in‐hospital training session and adhere to a scanning protocol over time. We found that with direct education and multi‐modal educational materials, patients could reliably be taught to perform a six‐point lung ultrasound exam with 100% of patients completing seven out of nine studies. In this sick patient population with no prior ultrasound experience, 80.8% of images obtained were interpretable, and 96.2% of studies had at least one interpretable image. Not surprisingly, patients reported that it was easier to scan the anterior chest than reaching across to the mid‐axillary chest, and this was concordant with image quality. Beyond image acquisition, this study demonstrates that patients can appropriately interact with and use a novel medical app. Patients reported high satisfaction with the device and the multi‐modal training materials.

Acute decompensated HF is associated with a high rate of subsequent readmission as well as a high short‐ and long‐term morbidity and mortality. ¹ Accordingly, reduction in HF hospitalization is a key metric in all HF pharmacologic and device clinical trials. ¹⁹ Recognition of worsening HF can often be challenging as clinical signs and symptoms of HF are often absent or lag despite presence of significantly elevated pressures on cardiac catheterization. ²⁰ For these reasons, telehealth initiatives that rely on patient‐reported metrics such as symptoms or physical findings such as changes in weight or lung auscultation abnormalities have often failed to reduce HF hospitalizations. ²¹ Invasive assessments of intravascular congestion often detect abnormalities prior to signs and symptoms of HF and have led to reduction in HF hospitalization and trends towards reduction in mortality. ²² , ²³ Unfortunately, these devices are invasive and not available to many patients. Integrating POCUS, with its low risk profile and cost and ability to dynamically quantify pulmonary vascular congestion, may be the right tool at the right time for this patient population.

The use of POCUS has increased significantly particularly in light of the COVID 19 pandemic and recent technologic advances in ultra‐portable handheld devices. Major society guidelines have advocated for the increased use of POCUS in several cardiovascular populations. ²⁴ , ²⁵ Current European HF guidelines suggest that the use of lung ultrasound to confirm the presence of acute HF or to assess congestion in a patient with chronic HF may be considered. ²⁶ Other consensus statements highlight the utility of lung ultrasound to assess congestion and also the need for further randomized, prospective studies to inform broader recommendation. ²⁷ , ²⁸ With the overall increase in POCUS utilization, patient‐performed POCUS is increasingly being studied. ²⁹ Recently, it has been shown that novice, non‐medically trained, healthy individuals can obtain high quality lung POCUS that is of similar quality to experts. ¹² In particular, lung POCUS has been shown to identify patients who are at risk for congestion despite lack of traditional physical exam findings such as crackles or oedema. ³⁰ A randomized study showed that use of a lung ultrasound‐guided treated strategy reduced episodes of decompensated HF and cardiovascular events in a high‐risk end‐stage renal disease population. ³¹ Lung ultrasound is increasingly being reported as a component of HF clinical investigations. ¹⁹ , ³² Abnormal lung ultrasound was shown to predict short‐ and long‐term mortality. ¹⁰ Furthermore, B‐line scores are dynamic and change rapidly with the initiation of HF therapy. ⁷ High baseline B‐line scores and high scores prior to discharge are individually predictive of poor outcomes. ⁷

Little is known about the ability to obtain B‐line scores from patient obtained images in an ambulatory setting. Chiem et al. assessed the ability of HF patients to self‐perform lung ultrasound after a brief video tutorial while in the emergency department. ³³ Similar to our study, they found that 85% of images were interpretable; however, this study did not assess the durability of the skill over time. We extend the findings of their study to the outpatient setting and also report B‐lines findings. Standardizing the B‐line score from each uploaded study to a six‐image scan allowed us to report temporal trends across patients and across scans. In our study, standardized B‐line scores varied both within an individual patient and also to some degree over the span of the study. The study was not designed to ascertain the nature of B‐line changes. This would require a different study design with longer follow‐up and assessment of patient symptoms.

While this pilot study was focused on HF patients specifically, there is evidence that B‐line monitoring for pulmonary vascular congestion may be valuable in other populations and misinterpreted in others. In a study by Noble et al., they found that B‐line resolution occurred dynamically in end‐stage renal patients receiving routine dialysis. ⁹ Patients with interstitial lung disease may have diffuse B‐lines secondary to fibrosis and unrelated to volume overload. ³⁴ In a study of patients with acute COVID‐19 infection, patients were mailed an ultrasound device and tele‐mentored to perform a lung POCUS exam. Without ever coming in direct contact with the physician team, 96% of images were interpretable, which could have significant impact on home‐bound patients and those in remote areas. ³⁵

A home‐based patient‐performed diagnostic test must be easy to perform, minimally invasive, and cost effective to achieve compliance and sustainability. However, to be actionable to a clinician, the results must be reliable, interpretable and intervenable. Our study demonstrates that even medically complex patients can generate interpretable images with minimal training with the majority of patients able to obtain five out of six images. Patient feedback in terms of imaging window difficulty was concordant with imaging findings that lateral lung zones had lower rates of interpretable images. As a patient‐centred study, extensive attention was given to training material development and included written, audio‐ and video‐based instructions. No patients found any of the training materials to be more than slightly complicated.

There are several limitations to this study. First, this study is a single‐centre convenience sample of patients cared for at a tertiary hospital by an advanced HF specialist. Compliance with scans and findings may not be applicable to a more generalized HF population. However, almost half of the patients were obese and had advanced HF and might be expected to have even more difficulty with protocol compliance than a healthier less medically complex population. Additionally, patients with significant lung disease were excluded due to challenges with interpreting the B‐line score. This also reduces the generalizability of the results as many patients with HF have significant underlying primary lung disease. While only 7.1% of patients admitted on the heart failure service were ultimately enrolled, the main reason for exclusion was the recent use of inotropes (prior to protocol revision) and being inotrope dependent. Thus, a much larger number of patients would have qualified for inclusion at other centres. Additionally, caregivers were not invited to assist with image generation, and while 80.8% of images were interpretable, this number may have improved with assistance from others. Finally, while this study demonstrates the feasibility of patient‐performed lung POCUS, more work is needed to determine clinically meaningful cutpoints of B‐lines that should prompt changes in management.

This is the first study assessing the feasibility of HF patients being able to self‐perform lung ultrasound and longitudinally comply with uploading images to a secure cloud for physician interpretation. We found that patients hospitalized with acute decompensated HF can quickly be taught to perform a six‐point lung ultrasound exam, maintain high rates of compliance with the study protocol and generate interpretable images over a 3 week period. Further studies are needed to determine if patient‐performed lung ultrasound exams can be used to guide HF medical therapy and whether there are downstream improvements in quality of life and reduced HF hospitalization.

Conflict of interest statement

None of the authors have financial, professional or advisory relationships with Butterfly. This study was initiated and conducted by the authors, and Butterfly had no involvement in the study design, data analysis or interpretation of the results.

Supporting information

Table S1: Inclusion and Exclusion Criteria.

Figure S1: Printed Patient Scanning Instructions.

Figure S2: B‐line score standardized to a 6 images scan for each patient over time.

Prenner, S. B. , Dougherty, K. , Shofer, F. S. , Goldberg, L. R. , and Panebianco, N. (2025) Home‐operated ultrasound exam for detection of worsening heart failure (HOUSE‐HF). ESC Heart Failure, 12: 3372–3380. 10.1002/ehf2.15343.