Feasibility and Accuracy of Tele‐Echocardiography, With Examinations by Nurses and Interpretation by an Expert via Telemedicine, in an Outpatient Heart Failure Clinic

To study the feasibility and accuracy of focused echocardiography by nurses supported by near–real‐time interpretation via telemedicine by an experienced cardiologist.

to use telemedicine is available in most places in the world through local area networks and wireless mobile telecommunication technology.
Heart failure (HF) is associated with a poor prognosis and a reduced quality of life, and the financial burden on the health care system is substantial. 1 Despite current treatment options, HF morbidity and mortality are still high, and 25% to 50% of all patients with HF are readmitted within 6 months of hospitalization after decompensated HF. [1][2][3] Guidelines advocate classification of HF by the left ventricular ejection fraction (LVEF), as this is decisive for both treatment and prognosis. 4 Patients with HF who have an EF of 50% or higher are classified as having heart failure with a preserved ejection fraction (HFpEF). Similarly, patients with HF who have an EF of 40% to 49% and an EF of 40% or lower are classified as having heart failure with a midrange ejection fraction (HFmrEF) and heart failure with a reduced ejection fraction (HFrEF), respectively. Furthermore, the size and function of the cardiac chambers are easily depictable by echocardiography. 5 Fluid retention is the major consequence of decompensation, which usually happens over time with lateonset symptoms and an unpredictable course. 4 Patients with HF could benefit from frequent volume status assessments and more aggressive therapy. 2,[5][6][7] The assessment of the volume status in patients with HF may be improved by evaluation of the presence of pleural effusion and the dimension and collapsibility of the inferior vena cava. 4,8 The classification of the HF category, estimation of filling pressures, and estimation of the volume status can be performed by echocardiography. 5,9 Echocardiography is usually performed by specialized sonographers or experienced cardiologists. Interpretation of the recordings is usually performed at the same location as the examination. Implementation of telemedicine for interpretation of images at remote locations is common in the field of radiology to reduce the workload of local radiologists. 10 The research regarding tele-echocardiography is scarce. The first studies on tele-echocardiography were conducted by pediatric cardiologists as support for physicians in rural areas. 11 So far, telemedicine has not found its way into routine follow-up of patients with HF. However, tele-echocardiography allows for the performance of echocardiographic recordings at one location and interpretation by an expert at another; thus, patients can benefit from the positive impact of more precise diagnostics.
The aim of the study was to examine the feasibility and accuracy of tele-echocardiography in an outpatient HF clinic. We combined echocardiographic recordings by trained nurses with transfer of echocardiographic data by local area networks and wireless mobile telecommunication for interpretation by a cardiologist at a remote location. The purpose of the study was not to implement limited ultrasound in the routine follow-up of patients with HF, but the aim was to explore the benefit in HF follow-up where teleechocardiography can overcome geographic challenges to improve HF diagnostics and care. Second, we aimed to evaluate the accuracy of tele-echocardiography for classification of HF and evaluation of filling pressures.

Study Population
Patients from an outpatient HF clinic were recruited at Levanger Hospital, Nord-Trøndelag Health Trust. All patients were followed for known HF and had previous echocardiographic examinations performed by a cardiologist. Only patients older than 18 years were eligible for inclusion between October 2016 and February 2017. All participants gave their informed written consent before inclusion. The study was conducted in conformity with the policy statement for the use of human subjects of the Declaration of Helsinki. The study was approved by the Regional Committee for Medical and Health Research Ethics (REK 2015/2312) and registered in the ClinicalTrials. gov database (NCT02936050).

Training and Education of Nurses
Three registered cardiac nurses with 6 to 12 years of clinical experience from a nurse-led outpatient HF clinic were trained in performing echocardiographic recordings by two cardiologists. The nurses had no previous experience in echocardiographic recordings or image analyses. However, they were familiar with the use of handheld ultrasound devices for evaluation of the size and respiratory variation of the inferior vena cava and assessment of pathologic fluid in the pericardium and pleural cavities to aid in their clinical work with patients with HF. Lung ultrasound was not included in the training. They underwent systematic training by cardiologists with experience in echocardiography and performed a mean of 67 (range, 47-97) lifetime examinations before patient inclusion. Initially, they performed approximately five echocardiographic examinations with hands-on training support.
Tele-Echocardiography With Recordings by Nurses and Interpretation by Telemedicine A Vivid 7 scanner (GE Healthcare AS, Horten, Norway) was used by the nurses to obtain goal-directed echocardiographic recordings of the following standard views: parasternal long-and short-axis (with and without color Doppler), 3 standard apical views (4-chamber, 2-chamber, and long-axis) with and without color Doppler focusing on left ventricular (LV), left atrial (LA), and right ventricular subcostal views for assessment of the inferior vena cava, pulsed wave tissue Doppler imaging with a sample volume in the basal part of the septal and lateral walls (4-chamber), pulsed wave Doppler recordings of mitral inflow, and continuous wave Doppler imaging of the blood flow through the aortic valve and tricuspid regurgitations. All recordings contained at least 3 cardiac cycles. The recording of the inferior vena cava included both maximum and minimum dimensions by including a quick inspiration (sniff). Both pleural cavities were assessed in the midclavicular and midaxillary line in a sitting position with the transducer in the intercostal spaces, as described earlier, 12 and in cases of pleural effusion, longitudinal and transverse images were recorded. The amount of fluid was quantified as the distance from the diaphragm to the basal part of the lung, annotated as 0 in cases of no pleural effusion. No ultrasound examinations of the lungs were included. The nurses had access to patient histories and key words of previous echocardiographic examinations, but importantly, they did not have access to previous echocardiographic recordings.
Transfer of the recordings was done immediately after the examination. A commercial software-and hardware-based system (PaCentric; Fimreite Software, Stavanger, Norway) was used. PaCentric allows for secure transmission, interpretation, and reporting of medical Digital Imaging and Communications in Medicine images per the Internet. The data were stored securely and depersonalized. PaCentric is accredited (International Organization for Standardization 13485:2003) and certified (Conformitè Europëenne 0434), as well as approved by the US Food and Drug Administration (510 k100837). Transmission of the recordings was allowed by installing the software on the Vivid 7 scanner connected to the hospital's local area network. After the end of the examination, data were exported to the PaCentric server and stored on an ordinary, password-protected laptop computer by the cardiologist who performed all of the analyses. Both local area networks and mobile telecommunication network solutions were used, depending on the availability at the interpreter's actual location.
The interpretation of the recordings by the outof-hospital cardiologist was performed in EchoPAC SWO (version BT12; GE Healthcare). The out-ofhospital cardiologist was blinded to all previous echocardiographic recordings and patient histories. All measurements reflect the average of at least 3 cardiac cycles. The LV endocardial borders were traced in end diastole and end systole in 4-and 2-chamber views. The LV internal length was measured from the traces and LV volumes (end-diastolic and endsystolic), and the EF was calculated by biplanar Simpson method. The LV internal diameter and wall thickness were measured at the level of the tip of the mitral leaflets in 2-dimensional grayscale recordings. The LA endocardial border was traced in end systole in 4-and 2-chamber views, and the volume was calculated by the area-length method and subsequently indexed per square meter body surface area (left atrial volume index [LAVI]). The pulmonary veins and the LA appendage were not included in the trace. Mitral inflow peak early (E) and late (A) velocities and the early filling mitral deceleration time were measured in pulsed wave Doppler recordings from the apical 4-chamber view. The peak velocity of the tricuspid regurgitation was measured by continuous wave Doppler imaging. Mitral annular peak systolic (S 0 ) and peak early diastolic (e 0 ) longitudinal velocities were assessed by pulsed-wave tissue Doppler imaging. The ratio of the early mitral inflow to the early diastolic mitral annular velocity (E/e 0 ) was calculated. The LV filling pressure was estimated according to recommendations by the European Association of Cardiovascular Imaging and the American Society of Echocardiography, based on the LAVI, e 0 , E/e 0 , and peak velocity of the tricuspid regurgitation for the HF subgroups as normal or elevated. 9

Reference Echocardiography
Reference echocardiography was performed immediately after the nurses' recordings by 1 of 4 in-house physicians experienced in echocardiography (3 cardiologists and 1 experienced resident in cardiology). They were not blinded to medical histories or previous echocardiographic recordings from the patients. However, they were blinded to the echocardiographic examinations performed by the nurses and analyzed by the out-of-hospital cardiologist by telemedicine. The reference examinations were performed at the same department but in another room. High-end echocardiographic scanners (Vivid E9 and Vivid E95) were used. The reference imaging included the same recordings, and in addition, all other chambers and valves were assessed. The echocardiographic measurements were performed as indicated above.

Patient Flow
The patients were first examined by 1 of 3 registered cardiac nurses, who immediately transferred the echocardiographic data for further analyses by telemedicine. The recordings obtained by the nurses were interpreted in near real time via the tele-echocardiographic approach by an out-of-hospital cardiologist ( Figure 1). No further follow-up or ultrasound examinations of the participants were performed during the study.
Before echocardiography, blood samples were drawn the same day and analyzed at the in-hospital accredited laboratory. Serum N-terminal pro-brain natriuretic peptide, serum creatinine, and estimated glomerular filtration rate (calculated by the Cockcroft-Gault equation) values were measured for characterization of Figure 1. Patient flow throughout the study. Patients with HF were examined by 1 of 3 specialist nurses and immediately after by 1 of 3 expert echocardiographers. The images acquired by the nurses were interpreted and analyzed by an out-of-hospital cardiologist via telemedicine. the population. The New York Heart Association functional classification was scored by the nurses, and the body weight (kilograms), body height (centimeters), and blood pressure (millimeters of mercury) were measured. Anthropometric measurements were rounded up to the nearest multiple of 1.

Statistical Analyses
Descriptive statistics were used for describing the study population. Data are presented as mean AE SD, but data not following a normal distribution are presented as median (range). Categorical data are reported as numbers and percentages. The agreement of the measurements by the telemedical approach and reference was tested by Bland-Altman statistics, the coefficient of variation, and the Pearson or Spearman correlation coefficient. Proportions were analyzed by the χ 2 test. Agreement with respect to the correct HF classification was analyzed by the weighted κ statistic.
Comparisons of means were tested by paired t tests or the related-sample Wilcoxon signed rank test. Two-sided P < .05 was considered statistically significant. Semiquantitative data were further assessed by calculations of sensitivity and specificity and negative and positive predictive values. The association of the type of HF with the correct classification of LV filling pressures was analyzed by logistic regression analyses.

Population
Baseline data for the 50 participants are shown in Table 1. The mean age was 77 AE 12 years; 46% were women; and the mean body mass index was 25.6 AE 5.9 kg/m 2 . The mean estimated glomerular  Left and right pleural cavities were treated separately. The numbers relate to the millimeter distance from the diaphragm to the basal part of the lung. In case of no effusion, the measurement is annotated as 0. Values are presented as mean (range).   Table 3.
filtration rate was 44 AE 29 mL/min. Diuretics, beta blockers, and angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers were used by 84%, 80%, and 36% of the population, respectively. The low proportion of patients treated with angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers was related to the high prevalence of renal failure in the population and the fact that optimal HF therapy was not yet achieved. Table 2 shows the time used for the telemedical approach. The mean duration of the examination by the nurses from the start of echocardiography until the report was finalized and reported back electronically by the out-of- Comparison of the tele-echocardiographic data with the reference measurements is shown in Table 3. Feasibility was high for all indices, except for the mitral E/A ratio and peak velocity of the tricuspid regurgitation. Only the LV internal end-diastolic diameter, LV internal end-diastolic length, LV posterior wall enddiastolic thickness, and tricuspid regurgitation peak velocity were significantly different between the methods (P < .02). By tele-echocardiography, LV endocardial borders in end diastole were underestimated by 3 mm (49 versus 52 mm), and LV length was overestimated by 5 mm (84 versus 79 mm).

Feasibility of Tele-Echocardiography
The agreements of measurements by telemedicine and the reference for the EF, LAVI, maximal tricuspid regurgitation gradient, and E/e 0 ratio are illustrated in Figure 2 and Table 4. The biases for the different measurements were close to 0, and there was no significant relationship between the errors and the magnitudes of the measurements. Coefficients of variation for the above-mentioned central indices were all in the range of 6% to 15% ( Table 4). The correlations were high for all echocardiographic indices (r ≥ 0.71; P ≤ .007), except for measurements of wall thickness, for which the correlations were moderate for both the interventricular septum and the posterior wall (both r ≥ 0.60; P ≤ .03). Pleural effusion was revealed in a total of 9 pleural cavities by either the reference or the telemedical approach. The latter detected pleural effusion in 7 of 8 cavities in which reference imaging results were positive.
Tele-echocardiography showed substantial agreement with the reference for classification of the category of HF, with a weighted κ of 0.73 (P < .001). Importantly, no participants were misclassified between rEF and pEF, but 17 cases were misclassified between mEF and pEF or rEF. The LV filling pressure was determined by the telemedical approach and reference in 39 and 41 of the 50 cases, respectively, and in 31 cases, the filling pressure was determined by both approaches (Figure 3). Among the misclassified cases, HFpEF and HFmrEF were numerically more prevalent, including all cases by the telemedical approach and 5 (71%) by reference.
Valvular disease was evaluated semiquantitatively. The sensitivity and specificity for tele-echocardiography to detect at least moderate mitral stenosis, mitral regurgitation, and tricuspid regurgitation were 100% and Figure 3. Agreement of grading of LV filling pressure between the telemedical approach and the reference method in 31 available cases. Left ventricular filling pressures were assessed as high or normal. "Under," "equal," and "over" refer to the comparison of the LV filling pressure classification described by the telemedical approach compared to the reference echocardiography. 95% or higher, respectively. For detection of at least moderate aortic stenosis (n = 8, but only 7 cases available for the analyses) the sensitivity was lower (43%) but still with excellent specificity (97%).

Discussion
We are currently unaware of other studies evaluating tele-echocardiography with recordings by nonphysician personnel at a single geographic location combined with near-real-time interpretation by a cardiologist at another. The telemedical approach was feasible and reliable in this HF population. The most important finding of this study is that by using expert support by telemedicine, more patients with HF can gain the benefit of diagnostic ultrasound. Such an approach may improve diagnostics and care when distance and available resources matter.
The limited echocardiographic approach presented here may be used during the initial evaluation of a patient with suspected HF, making information available. Thus, the time delay to diagnosis can be reduced by improving the basis for decisions on the right workup. Importantly, the aim was not to replace comprehensive echocardiography by this approach but to evaluate whether telemedicine could support clinical decision making when the echocardiographic recordings were in the hands of inexperienced users.
The time spent on the echocardiographic recordings by the nurses was, on average, 0.5 hour and within range of what is acceptable and feasible in the everyday clinical practice for a nurse-lead outpatient clinic. Similarly, the time spent for transfer, analyses, and reporting was short and allows for implementation. Even though the time used for transfer of the recordings depended on the local area network available, the approach was feasible for near-real-time interpretation, also when a 3G/4G mobile network was used. This was in line with previous studies. 13 The out-of-hospital cardiologist's categorization of the type of HF via the telemedical approach was comparable to the in-hospital cardiologist's reference echocardiography. Thus, geographic challenges can be overcome with the use of telemedicine for support of dedicated health care personnel in remote areas where traveling is a burden. 14,15 The telemedical software used is approved and complies with the regulations set by the Norwegian Data Protection Authority. With this software, sensitive data can be transferred and directly imported into EchoPAC (GE Healthcare) software for analyses, which presents a great advantage in simplifying the work flow. Today, several vendors provide similar software for transfer of imaging data, in accordance with the European Union general data protection regulations.
The three nurses performing the recordings had undergone dedicated, but limited training. Their training exceeded the recommendations for focused cardiac ultrasound examinations but did not reach the level recommended for comprehensive echocardiography. [16][17][18] In line with others, our results may add knowledge, and consequently, more patients can benefit from the diagnostic yield of echocardiography. As shown both in HF and other populations, diagnostic ultrasound may add important information, even when those performing the examinations have limited experience. 8,19,20 In this study, all patients were additionally examined by a cardiologist, and the results indicate that echocardiographic recordings by nurses combined with interpretation by a cardiologist add important information to the clinical decision making. To evaluate the clinical benefit of this approach, larger clinical studies are warranted.
As shown by the semiquantitative assessment of valvular disease, more training of the users may be needed to safely exclude valvular disease. Other studies evaluating handheld ultrasound devices by inexperienced users have also shown that evaluating valvular disease may be challenging, 15 and our group has previously shown that inexperienced users of diagnostic ultrasound perform better in assessments of global LV function than valvular assessments. 21 However, a valvular assessment was not the purpose of this study; thus, the results presented are in line with what was expected. The results highlight the need for dedicated training in any given task for operators of diagnostic ultrasound.
The telemedical calculations of dimensions, volumes, and flow measurements were well in line with reference measurements. Most differences were nonsignificant. There were only small but significant differences for the LV internal end-diastolic diameter, end-diastolic length, and posterior wall end-diastolic thickness (mean differences of 3, 5, and 0.9 mm, respectively) and tricuspid regurgitation peak velocity (mean difference of 0.14 m/s). The data presented are quite similar to what has been presented from other studies evaluating test-retest variation of echocardiographic indices. [22][23][24] The agreement for correct classification of the HF category was substantial, with a κ of 0.73, and all of the inconsistencies between HF classification by telemedicine and the reference related to HFmrEF versus HFrEF or HFpEF. The recently introduced new HF class HFmrEF has caused intense debate, as it is based on the idea that the EF can correctly classify HF into different categories. Furthermore, on the basis of the published repeatability data from the Atherosclerosis Risk in Communities study, approximately 40% of patients will move from HFmrEF to HFrEF or HFpEF in repeated analyses. 25 The prevalence of HFpEF in the study population was similar to what has been shown other HF populations. 4 The agreement with respect to estimation of the LV filling pressure was good, with misclassification by the telemedical approach in only 17%. Studies performing similar tasks with repeated echocardiograms are scarce, but our data were similar to a study evaluating agreement between echocardiographers in 105 single echocardiograms. 26 The views and parameters included in the limited echocardiographic examination were based on the need for a proper assessment of a patient with HF, allowing for classification of subtypes and an assessment of the LV filling pressure. These are less than what is recommended for a comprehensive echocardiogram but substantially more than what is included in focused cardiac ultrasound. 27,28 Very little research has been done in which personnel not previously skilled in echocardiography performed echocardiographic recordings with interpretation by specialists, but the results are promising when compared to, for instance, the robotic-arm approach. 27,29 This approach differs from the training of echocardiography technicians and sonographers, as the nurses did not perform quantitative analyses but only dedicated recordings without interpretations of the recordings. The nurses had years of clinical experience evaluating the volume status with handheld ultrasound devices but had not performed echocardiography until training for this study. 8,30 This confirms previously shown results aiming to implement telemedicine to overcome geographic challenges. Thus, a limited echocardiographic examination by nonexperts can achieve quality that allows for a reliable assessment. 10,11 Considering the efforts and costs to transport patients to a hospital with an available specialist versus the efforts and costs of training nurses, the study results can justify implementation into everyday clinical practice.

Limitations
The aim of the study was to evaluate the LV function, volume status, and indices important for classification of subpopulations of HF. Thus, the results cannot be generalized for other tasks. The agreement between the methods for determination of an elevated filling pressure was not validated invasively; thus, only the agreement of the telemedical approach with the reference echocardiography could be assessed.
Most of the patients had previous echocardiographic examinations at the same hospital. The reference cardiologist had access to previous echocardiographic recordings. The out-of-hospital cardiologist who performed the interpretation of the nurses' recordings was blinded to all data from previous echocardiograms and medical histories, and it is not likely that the reference echocardiographic examination was influenced by previous echocardiography. Thus, we find it unlikely that the knowledge of previous echocardiograms biased the results, as the complete blinding of the telemedical cardiologist interpreting the echocardiograms would, if having any influence, tend to introduce a negative bias, which was not observed. The limited echocardiograms recorded by nonexperts described here should not be considered equal to comprehensive echocardiography by experts. However, the goaldirected examinations presented in this study add valuable quantitative information for clinical decision making, which may guide therapy beyond what is achievable by semiquantitative cardiac ultrasound from handheld devices. 17,28 Conclusions Tele-echocardiography, in the form of image acquisition by registered cardiac nurses supported by interpretation by a cardiologist, is feasible and provides reliable results of central indices for quantification of left-sided cardiac size and function, HF classification, and LV filling pressures. Implementing tele-echocardiography at remote locations where echocardiography experts are not available may improve diagnostics and therapy.