If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Corresponding author. Department of Circulation and Medical Imaging, Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology, Trondheim, Norway.
Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, NorwayClinic of Cardiology, St. Olavs University Hospital, Trondheim, Norway
Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, NorwayDepartment of Internal Medicine, Levanger Hospital, Nord-Trøndelag Hospital Trust, Levanger, Norway
Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, NorwayClinic of Cardiology, St. Olavs University Hospital, Trondheim, NorwayDepartment of Internal Medicine, Levanger Hospital, Nord-Trøndelag Hospital Trust, Levanger, Norway
Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, Trondheim, NorwayClinic of Cardiology, St. Olavs University Hospital, Trondheim, Norway
Early and correct heart failure (HF) diagnosis is essential to improvement of patient care. We aimed to evaluate the clinical influence of handheld ultrasound device (HUD) examinations by general practitioners (GPs) in patients with suspected HF with or without the use of automatic measurement of left ventricular (LV) ejection fraction (autoEF), mitral annular plane systolic excursion (autoMAPSE) and telemedical support. Five GPs with limited ultrasound experience examined 166 patients with suspected HF (median interquartile range = 70 (63–78) y; mean ± SD EF = 53 ± 10%). They first performed a clinical examination. Second, they added an examination with HUD, automatic quantification tools and, finally, telemedical support by an external cardiologist. At all stages, the GPs considered whether the patients had HF. The final diagnosis was made by one of five cardiologists using medical history and clinical evaluation including a standard echocardiography. Compared with the cardiologists’ decision, the GPs correctly classified 54% by clinical evaluation. The proportion increased to 71% after adding HUDs, and to 74 % after telemedical evaluation. Net reclassification improvement was highest for HUD with telemedicine. There was no significant benefit of the automatic tools (p ≥ 0.58). Addition of HUD and telemedicine improved the GPs’ diagnostic precision in suspected HF. Automatic LV quantification added no benefit. Refined algorithms and more training may be needed before inexperienced users benefit from automatic quantification of cardiac function by HUDs.
Symptoms of heart failure (HF) are non-specific and present a challenge to the diagnostic workflow. One of six patients older than 65 years presenting to their general practitioner (GP) with dyspnea on exertion will have undiagnosed HF [
2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
]. Early and correct diagnostics are essential to improve patient care and to reduce the burden on the health care system. Echocardiography is the cornerstone of HF diagnostics and the method of choice when evaluating function of the left and right ventricles [
2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure 2012: the Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of the European Society of Cardiology. Developed in collaboration with the Heart Failure Association (HFA) of the ESC.
]. The European Association of Cardiovascular Imaging (EACVI) supports the use of HUDs to screen for cardiac pathology under the condition that proper training has been performed [
We believe that inexperienced HUD users would benefit from automatic quantification of left ventricular (LV) function and telemedical support by an expert when assessing patients with possible HF. Evaluation of LV ejection fraction (EF) and mitral annular plane systolic excursion (MAPSE) are two methods for quantification of LV function [
Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
]. Automatic measurement of EF is commercially available and implemented on HUDs. An algorithm for automatic quantification of MAPSE has been implemented on HUDs for research purposes [
Clinical validation of an artificial intelligence-assisted algorithm for automated quantification of left ventricular ejection fraction in real time by a novel handheld ultrasound device.
The primary aim was to evaluate the clinical influence of HUD examinations by GPs in patients with suspected HF without or with the use of supportive tools such as automatic quantification of EF/MAPSE and telemedical support compared with experienced cardiologists’ decisions based on comprehensive assessment of medical history, clinical evaluation and echocardiography as reference.
Methods
The study was conducted at the outpatient cardiology clinic at Levanger Hospital, Levanger, Norway, from January 2018 until June 2020. Inclusion was paused during part of the COVID-19 pandemic (March–May 2020).
Study population
Patients with suspected HF referred to the outpatient clinic at Levanger Hospital were invited to participate in the study. Eligible patients were contacted by study researchers and gave their oral and written consent. The inclusion criteria were suspicion of HF, N-terminal pro-brain natriuretic peptide (NT-pro-BNP) ≥125 ng/L, age >18 y and ability to give consent. The exclusion criteria were known HF or known results from cardiac imaging examination within the last 10 y. The study was performed in conformity with the policy statement for the use of human subjects of the Declaration of Helsinki. It was approved by the regional committee for medical and health research ethics (REK 2017/2054) and registered in the ClinicalTrial.gov database (identifier: NCT03547076).
Study design
On arrival at the outpatient clinic, blood samples (NT-pro-BNP, creatinine, sodium, potassium and hemoglobin), blood pressure and an electrocardiogram (ECG) were taken. Five GPs with limited experience in echocardiography were randomly selected by the study administration to participate in the study. They had access to the initial patient referrals and the ECGs, but not in-hospital medical records. The GPs examined patients in chronological order. A standard clinical evaluation including patient history and a physical examination was performed, followed by a focused cardiac ultrasound examination by HUD, addition of automatic quantification tools and, finally, telemedical supportive image analyses by out-of-hospital cardiologists. At each stage, the GPs considered whether the patients had HF and whether they would refer them for a cardiac examination. Because of laboratory delay, NT-pro-BNP was not always available for the GPs during their examination. After the GP evaluation, all patients were examined by one of five in-hospital cardiologists who performed a complete reference echocardiography (Fig. 1). Five to eight patients were included per 30 inclusion days.
Figure 1Flowchart of study participants. Real time response from cardiologist used in GPs’ decision making. GPs decided whether the patients had heart failure at each stage. ECG, electrocardiogram; autoEF, automatic analyses of ejection fraction; autoMAPSE, automatic analyses of mitral annular plane systolic excursion; GP, general practitioner; HUD, handheld ultrasound device.
The GPs received six days (6 h/d) of practical training and two theoretical lectures. The training focused on visualizing parasternal long- and short-axis, apical four-chamber and subcostal views by HUD, as well as the inferior vena cava (IVC) and pleural cavities with respect to pleural effusion. On average, seven HUD examinations were performed per day. In addition, The GPs performed on average 13 unsupervised focused ultrasound examinations by HUD in their daily practice.
Handheld ultrasound
The focused ultrasound was performed using Vscan Extend (GE Ultrasound, Horten, Norway) with the capability of storing cine loops of one cardiac cycle without the need for ECG [
]. The commercially available LVivo application (DiA Imaging Analysis Ltd, Be'er Sheva, Israel) for automatic EF quantification (autoEF) was implemented on the HUDs. It detects the LV endomyocardial wall in the apical four-chamber view (Fig. 2) [
Accuracy of left ventricular ejection fraction determined by automated analysis of handheld echocardiograms: a comparison of experienced and novice examiners.
]. A customized research version of the automatic MAPSE (autoMAPSE) application was implemented on the HUD. MAPSE was calculated by tracking the basal LV points’ movement using a customized method implemented on the HUDs—real-time contour tracking library (RCTL) (Fig. 2). The method was originally developed by our group a decade ago [
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
]. The RCTL provides segmentation of the left ventricle using a model composed of 12 control points which are updated by detecting the LV border in 75 equally spaced edge profiles. When tracking is enabled, the RCTL returns the septal and lateral points of the mitral annulus. The operators are unable to see, control or adjust the segmentation process. However, the tracking of the basal LV points and a line between these two points during the whole cardiac cycle is presented to the operator of the HUD (Fig. 2). There was no automatic feedback of the quality of the recordings or the robustness of the measurements for the autoEF or the autoMAPSE software. The tracking of the regions of interest throughout a cardiac cycle was available to the users. Examinations were performed with patients in the left lateral supine position and included the main cardiac views (parasternal long and short axis, apical four chamber and subcostal) and identification of IVC and pleural cavities. LV function was categorized by the GPs as normal, moderately reduced or severely reduced, while the cardiologists categorized LV function by EF (≤40%, 41%–49% or ≥50%) [
Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
]. The IVC was described as dilated or not. The presence of pericardial and pleural effusion was evaluated. GPs were instructed to measure autoMAPSE and autoEF three times each per patient.
Figure 2Tools for automatic quantification of left ventricular function. Automatic quantification of left ventricular function by mitral annular plane systolic excursion (left panel) and ejection fraction (right). Values of the systolic excursion of the mitral annulus are annotated at top of the left panel, and values for the automatic quantification of ejection fraction and volumes are at top of the right panel. EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume.
Pseudonymized images stored on the HUDs, were transferred in near real time to a cloud-based server (Trice Imaging, Inc., Del Mar, CA, USA), an integrated software that allows for secure and anonymous sharing of medical images [
in: Tayal VS Blaivas M Foster TR Ultrasound program management: a comprehensive resource for administrating point-of-care, emergency, and clinical ultrasound. Springer,
Cham2018: 281-299
]. One of two external cardiologists (both localized at St. Olav's Hospital, Trondheim University Hospital, Trondheim, Norway) downloaded the recordings to EchoPAC SWO (Version 203, GE Ultrasound, Horten, Norway) for interpretation. The cardiologists had access to the initial referrals, but not the results from the clinical evaluation, ECG, blood samples or hospital records. They provided feedback to the GPs electronically.
Reference examinations by cardiologists
A comprehensive echocardiographic evaluation was performed within 1 h after the GP`s examination. The echocardiograms were recorded using high-end equipment (Vivid E9 or E95, GE Ultrasound) and included the main cardiac views [
Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
]. EF was measured using Simpson's biplane in apical four- and two-chamber views. Systolic LV function was defined as normal if the EF was >50%, mild to moderately reduced and significantly reduced if the EF was 40%–49% and <40%, respectively. MAPSE was measured in the septal and lateral mitral annular points by M-mode or reconstructed M-mode. All measurements represent the average of three consecutive cardiac cycles [
Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
]. The echocardiographic analyses were performed using EchoPAC SWO (Version 202, GE Vingmed Ultrasound).
Statistics
Continuous data are expressed as the mean ± standard deviation (SD) or median (interquartile range [IQR]) as appropriate. Categorical data are expressed as frequencies and percentages.
A difference of <15% of correctly identified patients with and without HF between the GPs and reference cardiologists was considered of little importance. Use of Sample Power (SPSS, Inc., Chicago, IL, USA) resulted in a sample size of 104 and power of 0.80 (p < 0.05). The number of patients with significant pathology was expected to be small; furthermore, we expected failure of both GPs’ recordings and some of the automatic measurements. Thus, a sample size of 150 was chosen. Throughout the study period, the autoEF software was upgraded and the sample size was adjusted to 170 to account for the new version.
McNemar's test was used to compare paired nominal data. An indication of the prognostic gain was calculated using net reclassification improvement (NRI), which is aimed at quantifying whether a new method or marker improves the prediction of a disease [
]. For this test, the clinical examination was used as a reference. NRI combines NRI event and NRI non-event. NRI event represents the net reclassification proportion of HF diagnosis among those with HF after each stage (clinical and HUD, clinical, HUD and addition of automatic tools, clinical, HUD, automatic tools and addition of telemedicine) compared with clinical examination alone. Similarly, an NRI non-event represents the net reclassification proportion in non-events (non-HF). The overall NRI is the sum of the net proportions of correctly reclassified exams. The positive (PPV) and negative (NPV) predictive values were calculated as the probability of correct classification.
A p value <0.05 was considered to indicate statistical significance for all analyses. SPSS (version 26, IBM Corp, Armonk, NY, USA) and Excel were used for the analyses.
Results
Study population
Of 185 patients invited, 170 agreed to participate in the study. After exclusion, 166 (78 women) completed the exams (Fig. 1). Table 1 outlines the population characteristics. Median (IQR) age was 70 (63–78) y, mean BMI 29 ± 5 kg/m2 and median NT-pro-BNP 298 (65–870) ng/L. Patients with HF had a median NT-pro-BNP of 1302 (866–2626) ng/L compared with 148 (53–525) ng/L in patients without HF. Sinus rhythm was present in 126 (76%) patients. Systolic function was preserved in 131 (79%) patients; 21 (13%) had a mild reduction in LV function and 14 (8%) had moderately or severely reduced function. Mean EF was 53 ± 10%. Of the 166 patients, 118 (71%) presented with dyspnea, 24 (14 %) had reduced physical capacity, 40 (24 %) had peripheral edema, 22 (13 %) had fatigue and 16 (10 %) experienced palpitations.
Table 1Baseline characteristics of the study population
Reference cardiologists diagnosed 28 patients with HF, excluded HF in 130 patients and were uncertain about the diagnosis in the remaining 8. Of the 28 patients, 13 (46%) had HF with preserved EF (HFpEF). After the clinical examination, the GPs correctly classified 92 (55%) patients (15 with HF, 77 without). The corresponding numbers increased to 118 (71%) after HUD examinations (19 with HF, 99 without) and 123 (74%) after telemedical evaluation (20 with HF, 103 without) (Fig. 3). The difference between the clinical examination and HUD or telemedicine was highly significant (p <0.001), but that between HUD and telemedicine was not (p = 0.44). There was no improvement in diagnostic precision after adding automatic quantification (57% correctly classified after autoMAPSE and 55% after autoEF). AutoMAPSE and autoEF were not performed in 16 and 34 cases, respectively. In these cases, the GPs failed to run the applications because of suboptimal ultrasound images. On the basis of symptoms and physical examination, the GPs suspected 7 of the 13 patients with HFpEF of having HF. The NRI for HUD and telemedicine was 0.10 and 0.19, respectively. These data were based on 3.6% and 12.0% of correctly reclassified HF patients after HUD and telemedicine, respectively (NRI for events) and 6.5% (both HUD and telemedicine) correct reclassification of non-events (Fig. 4). The GPs were uncertain of the diagnosis in 43 patients after the initial assessment. There was a statistically significant decrease (p < 0.05) in uncertain cases after adding HUD and telemedicine (20 and 24, respectively), with no significant difference between the two (p = 0.44). There was non-significant reduction after autoMAPSE and autoEF (36 and 40 uncertain cases, respectively). The NPV was high at all stages (>0.91), while the PPV was lowest for autoMAPSE (0.40) and highest for telemedicine (0.71) (Table 2). When we evaluated the importance of the autoEF upgrade, we found no improvement in NPV and PPV (0.96 vs. 0.86 and 0.58 vs. 0.38, respectively). However, similar differences were found for autoMAPSE (data not shown). Coefficients of variation (COVs) for the GP recordings according to the reference were recently reported [
Real-time automatic quantification of left ventricular function by hand-held ultrasound devices in patients with suspected heart failure: a feasibility study of a diagnostic test with data from general practitioners, nurses and cardiologists.
]. The COVs for autoEF and autoMAPSE were 15.4% and 24.3%, respectively.
Figure 3Diagnostic precision of heart failure diagnosis by general practitioners. Total number of patients with correct and incorrect classification after each stage of the examination. The columns include patients with and without heart failure. Black reflects incorrect classification; white reflects correctly classified individuals. Agreement between the stages was calculated with McNemar's test. The group with uncertain diagnoses is not included and reflects the proportion between 166 and the presented sum of correct and incorrect classifications. autoEF, automatic analyses of ejection fraction; autoMAPSE, automatic analyses of mitral annular plane systolic excursion; HUD, handheld ultrasound device.
Figure 4Reclassification of heart failure diagnosis. Number for correct/incorrect classification of patients with heart failure diagnosis dichotomized to “yes” or “no.” Net reclassification improvement is shown at each stage (clinical + HUD, clinical + HUD + autoMAPSE/autoEF, clinical + HUD + autoMAPSE/autoEF + telemedicine) compared with clinical examination alone. autoEF, automatic analyses of ejection fraction; autoMAPSE, automatic analyses of mitral annular plane systolic excursion; HUD, handheld ultrasound device; NRI, net reclassification improvement.
Even though all patients underwent a reference echocardiogram, the GPs still had to state whether they would refer the patients for a cardiac examination. Because of logistics, they were not presented with this possibility on the first day of inclusion. They intended to refer 113 (68%) patients after the clinical examination. They suspected HF in 35 of these patients, while in the remaining 78 (69%) there were other reasons for the referral. There was a significant decrease in the total number of referrals after addition of HUD, automatic quantification and telemedicine (all p values <0.02) (Fig. 5).
Figure 5General practitioners’ referrals of patients for cardiac examinations. The orange columns represent the total number of patients the GPs decided to refer for a cardiac examination no matter the diagnosis. The black columns represent how many of the total number of referred patients had suspected heart failure. autoEF, automatic analyses of ejection fraction; autoMAPSE, automatic analyses of mitral annular plane systolic excursion; HF, heart failure; HUD, handheld ultrasound device; NRI, net reclassification improvement.
There was a non-significant decrease in the proportion of referred patients with suspected HF after adding HUDs, autoMAPSE and autoEF (31, 35 and 26 patients, respectively). After telemedical support, 23 of 98 (60%) referred patients were suspected of having HF (difference vs. clinical examination p = 0.02). Of the 23 patients, 21 (91%) were diagnosed with HF by the reference cardiologist. Of the 8 patients with a missed HF diagnosis, all but one was referred for cardiac evaluation (Table 3).
Table 3Characteristics of patients with heart failure incorrectly classified at one or more of the diagnostic stages
In patients with suspected HF, the proportion of patients correctly diagnosed by GPs improved by ≥25% after HUD and telemedical support. The number of uncertain cases was reduced by approximately 50%. Adding automatic quantification tools for MAPSE or EF did not improve diagnostic precision. After telemedical support, the GPs would refer 35% fewer patients with suspected HF, and among those selected for referral, >90% were finally diagnosed with HF.
Population
In the present population, 28 (17%) were diagnosed with HF. The results agree with previously published studies in which one in six patients older than 65 years presenting to their primary care physician with dyspnea had undiagnosed HF [
]. In our study, 71% of the patients experienced dyspnea, either at rest or on exertion. The proportion of HF highlights that HF symptoms are unspecific and overlap with other disorders [
2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
]. Improved selection of patients would allow for better use of restricted health care resources. NT-Pro-BNP has a high NPV and improves the ability to rule out HF. In our study, NT-pro-BNP was not always present during the GP assessment because of laboratory delay but the number of false-positive cases in patients with NT-pro-BNP below the usual threshold of 125 ng/L varied from 1 to 6 depending on the diagnostic stages. As NT-pro-BNP has a lower PPV, its availability would not necessarily reduce the number of false-positive diagnoses [
The diagnostic accuracy of plasma BNP and NTproBNP in patients referred from primary care with suspected heart failure: results of the UK natriuretic peptide study.
]. The sensitivity and specificity of any test are influenced by the distribution of the studied population, and this also relates to the diagnostic performance of NT-pro-BNP, as well as the diagnostic decision-support software used in the present study. The results indicated that inclusion of HUD examinations improved the GPs´ diagnostic precision, whereas the GPs were not able to improve their practice by adding the automatic tools providing measurements of LV function. There was no significant difference between diagnostics after the clinical examination and after adding automatic quantification tools. Most of the patients were overweight or obese, and comorbidities such as atrial fibrillation and hypertension were common. Thus, both poor acoustics and atrial fibrillation (present in 24%) may have interfered with image acquisition and the accuracy of the automatic measurements. As the present population had characteristics similar to those of other studies evaluating HUDs and HF [
Characteristics and outcomes of patients with suspected heart failure referred in line with National Institute for Health and Care Excellence guidance.
]. After completing the program, the GPs were able to perform and interpret cardiac ultrasounds with improvement in diagnostic precision. However, they were not able to interpret and adjust to the false-positive (and false-negative) results provided by the algorithms. Future work is needed to determine if more training would improve image quality, the ability to correctly diagnose HF and the ability to use the advantages of HUD applications.
Clinical influence of HUD examinations and diagnostic supportive tools
The precision of HF diagnosis based on standard examination alone is suboptimal, with low accuracy and high false-positive rates [
]. In our study, clinical examination alone had a low PPV (0.47). Of the patients suspected of having HF, only 43% were diagnosed with HF by the reference.
The proportion of correct classification increased after HUD examinations with fewer false positive and uncertain cases. This confirms the diagnostic value of adding HUDs to a clinical examination. Mjølstad et al. [
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
Recommendations for the evaluation of left ventricular diastolic function by echocardiography: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
]. Automatic LV quantification tools implemented on HUDs have previously scarcely been evaluated by inexperienced users. There was no improvement in correctly diagnosed patients after automatic quantification compared with clinical examination alone. False-positive results increased after autoMAPSE, and the uncertainty was high for both applications. The difference between autoMAPSE and autoEF was not significant. After upgrading the autoEF software, there was no improvement in false-positive or false-negative cases. The COVs revealed a modest variation of autoEF between GPs and reference cardiologists (COV = 15.4%) and quite a large variation for autoMAPSE (COV = 24.3 %) [
Real-time automatic quantification of left ventricular function by hand-held ultrasound devices in patients with suspected heart failure: a feasibility study of a diagnostic test with data from general practitioners, nurses and cardiologists.
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
], which was not available in this study. In our study, both automatic quantification tools underestimated the LV measurements (mean MAPSE: 8 mm vs. 12 mm, and mean EF: 48% vs. 53%, respectively). The underestimation by autoMAPSE was larger than previously reported [
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
], and the potential importance of reduced image quality, patient characteristics and users must be addressed in future studies. The underestimation may have contributed to the GPs’ uncertainty and overdiagnosis. Both automatic application tools were fully automatic, and it was not possible for the operators to adjust the endocardial tracking or mitral points. Further, no automatic feedback to optimize the recordings was provided. It is not known if novel tools providing automatic feedback to optimize the recordings could have improved the results [
]. The decrease in correctly diagnosed patients after adding the automatic quantification tools to the HUD examination may indicate that the GPs were unable to distinguish between correct and incorrect measurements. It may be hypothesized that feedback regarding the robustness of tracking of regions of interest and the measurements could potentially improve the GPs’ interpretation. Further, atrial fibrillation (AF) poses a challenge to HF diagnostics. AF was overrepresented by 50% among the uncertain cases after autoEF (13 [33%]). AF was present in 54% of all HF patients and in 62% of patients with HFpEF. The combination of HFpEF and AF also represents a clinical challenge. Diagnostics are difficult because of overlapping symptoms [
]. After telemedical support, the numbers of false-positive and uncertain cases decreased. There was a significant difference in the proportion of correct reclassification (NRI 0.19) compared with clinical evaluation and automatic quantification. Telemedical support of the HUD recordings had the numerically highest PPV (0.71), and the total number of correctly diagnosed patients improved in comparison with HUD alone (123 vs. 118). Improved image quality could further facilitate the evaluation [
]. The time between inclusion days for the participating GPs varied and may have influenced the image quality and subsequently reduced the quality of the telemedical support. Further, the GPs needed to fit the feedback into a clinical context, which may have been challenging in some cases. Access to the clinical information from the present exam could have improved the telemedical interpretation and feedback to the GPs.
There were false-negative cases at each stage of the examinations. In all but one instance, either HF was identified after telemedical support and/or the patients were referred for an echocardiography for other reasons such as AF, suspicion of aortic stenosis, coronary artery disease or hypertrophic cardiomyopathy. Of the patients with HF, 46% had HFpEF, which is challenging to diagnose [
How to diagnose heart failure with preserved ejection fraction: the HFA-PEFF diagnostic algorithm: a consensus recommendation from the Heart Failure Association (HFA) of the European Society of Cardiology (ESC).
]. As outlined in Table 3, the combined use of HUD, autoEF, autoMAPSE and telemedicine was not sufficient for precise diagnosis of HFpEF.
Despite a low proportion of patients with suspected HF, the GPs would still refer a large proportion to a cardiologist. In most cases, there was a valid reason for the referral, such as AF, suspicion of valvular disorders, pericardial effusion, suspicion of other cardiac pathology or poor image quality.
Limitations
This was a single-center study with a limited number of observers. The participating GPs were randomly selected by the health care administration in two municipalities and did not join because of a special interest. Further, the five GPs originated from five different GP offices. Evaluating an even broader set of operators would improve the ability to make stronger conclusions regarding the generalizability of the results. Diagnostic ultrasound is user dependent, and enthusiasm and interest would likely motivate participants to improve their recordings. The modest sample size may limit the power for secondary analyses. For example, including a larger population would have improved the power to detect subtle inter-operator differences. Additionally, we do not know whether differences between the two external experts influenced the results. Because of laboratory delay, the GPs did not have access to all the blood sample results, and the level of NT-pro-BNP was not available in all initial referrals to the outpatient clinic. NT-Pro-BNP is an important factor in the evaluation of suspected HF [
2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
The diagnostic accuracy of plasma BNP and NTproBNP in patients referred from primary care with suspected heart failure: results of the UK natriuretic peptide study.
]. However, in this study, most patients had elevated NT-pro-BNP, and we do not expect a substantially changed outcome if all results were available at the time of examination. Lastly, even though the GPs were not able to provide automatic measurements of the left ventricle in a substantial proportion of the examinations, the finding of ≥43% misclassified decisions after automatic measurements indicates that we had adequate power to conclude on the pre-set 15% limit for accepted misclassifications.
Conclusions
Addition of handheld ultrasound to examinations by general practitioners improved diagnostic precision in patients with suspected heart failure. The highest NRI was found after the HUD recordings were supported by telemedical interpretation. In the future, this may allow for better selection of patients with suspected HF in need for cardiac follow-up. The applications for automatic quantification of LV function added no significant benefit. Further refinement of the methods and more specific training of personnel may be needed before these methods add benefit to the diagnostic process.
Conflict of interest
L.L. is a part-time consultant for GE Vingmed Ultrasound, unrelated to this work. M.I.M., O.C.M., L.L. and H.D. hold positions at CIUS, a Centre of Research-based Innovation that is funded by the Research Council of Norway, Oslo, Norway, and industry. One of the industry partners is GE Vingmed Ultrasound, Horten, Norway. The company (GE) provided hand-held ultrasound devices with dedicated software for free use in the study but had no role in the planning of the study, data acquisition or drafting or revision of the manuscript.
Acknowledgments
We thank the clinicians and other employees at Nord-Trøndelag Hospital Trust for their support and for contributing to data collection in this research project. The study is funded by the Research Council of Norway and the Norwegian University of Science and Technology.
Data availability statement
Data will be made available on reasonable request.
References
van Riet EE
Hoes AW
Limburg A
Landman MA
van der Hoeven H
Rutten FH.
Prevalence of unrecognized heart failure in older persons with shortness of breath on exertion.
2016 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure: the Task Force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC). Developed with the special contribution of the Heart Failure Association (HFA) of the ESC.
ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure 2012: the Task Force for the Diagnosis and Treatment of Acute and Chronic Heart Failure 2012 of the European Society of Cardiology. Developed in collaboration with the Heart Failure Association (HFA) of the ESC.
Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
Clinical validation of an artificial intelligence-assisted algorithm for automated quantification of left ventricular ejection fraction in real time by a novel handheld ultrasound device.
Accuracy of left ventricular ejection fraction determined by automated analysis of handheld echocardiograms: a comparison of experienced and novice examiners.
Feasibility and reliability of automatic quantitative analyses of mitral annular plane systolic excursion by handheld ultrasound devices: a pilot study.
in: Tayal VS Blaivas M Foster TR Ultrasound program management: a comprehensive resource for administrating point-of-care, emergency, and clinical ultrasound. Springer,
Cham2018: 281-299
Real-time automatic quantification of left ventricular function by hand-held ultrasound devices in patients with suspected heart failure: a feasibility study of a diagnostic test with data from general practitioners, nurses and cardiologists.
The diagnostic accuracy of plasma BNP and NTproBNP in patients referred from primary care with suspected heart failure: results of the UK natriuretic peptide study.
Characteristics and outcomes of patients with suspected heart failure referred in line with National Institute for Health and Care Excellence guidance.
Recommendations for the evaluation of left ventricular diastolic function by echocardiography: an update from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.
How to diagnose heart failure with preserved ejection fraction: the HFA-PEFF diagnostic algorithm: a consensus recommendation from the Heart Failure Association (HFA) of the European Society of Cardiology (ESC).
00680-9&id=gr1.jpg){kind=link}
00680-9&id=gr2.jpg){kind=link}
00680-9&id=gr3.jpg){kind=link}
00680-9&id=gr4.jpg){kind=link}
00680-9&id=gr5.jpg){kind=link}