Background
The rapid detection of left ventricular systolic dysfunction (LVSD) is an important step in the clinical management of patients admitted with acute decompensated heart failure, because it allows the initiation of treatment specific to LVSD and avoidance of contraindicated therapies. The aim of this study was to determine whether internal medicine residents with limited ultrasound training could use hand-carried ultrasound (HCU) to identify LVSD.
Methods
Fifty patients admitted with acute decompensated heart failure were imaged from the parasternal window at the bedside with an HCU device by residents blinded to all clinical data, who had undergone limited cardiac ultrasound training (20 practice studies). Ejection fraction (EF) on HCU was graded as >40% or <40%. HCU EF and a number of physical exam findings and electrocardiographic and laboratory variables were compared for their ability to predict to formal echocardiographic left ventricular EF.
Results
The average formal EF was 32 ± 16% (range, 7%–70%), with 66% of patients having EFs < 40%. The residents’ ability to detect an EF < 40% with HCU was excellent (sensitivity, 94%; specificity, 94%; negative predictive value, 88%; positive predictive value, 97%). Binary logistic regression demonstrated that HCU EF was the most powerful predictor of EF < 40%, with minimal additional value from clinical, exam, lab, and electrocardiographic variables. The time interval between clinical assessment and availability of formal echocardiographic results was 22 ± 17 hours.
Conclusions
Residents with limited training in cardiac ultrasound were able to identify LVSD in patients with acute decompensated heart failure with superior accuracy compared with clinical, physical exam, lab, and electrocardiographic findings and an average of 22 hours before the results of formal echocardiography were available.
Patients presenting with acute decompensated heart failure (ADHF) may have left ventricular (LV) systolic dysfunction (LVSD) or heart failure with normal ejection fraction (HFNEF). Differentiation between the two is an important initial step in the clinical management of these patients, because diagnostic and therapeutic strategies differ for these two groups of patients. Clinicians have traditionally relied on history, physical exam, and electrocardiographic (ECG) findings at the point of care to aid in this differentiation. However, previous work clearly demonstrates the lack of both sensitivity and specificity for these findings. Transthoracic echocardiographic (TTE) imaging is a class I recommendation in all patients with ADHF, in large part because of its ability to readily distinguish LVSD and HFNEF. However, traditional full-featured TTE platforms may not be immediately available at the patient’s bedside at the time of admission. Hand-carried ultrasound (HCU), which circumvents the impracticality of needing a large cumbersome platform at the bedside, is a natural fit for this need and has been shown to have relatively high sensitivity and specificity for identifying LVSD when used by experienced imagers. However, although HCU devices get the modality to the bedside, trained imagers to perform and interpret the studies are not available at all times when patients are admitted with ADHF, delaying the differentiation between LVSD and HFNEF.
We sought to determine whether internal medicine house staff could use a pocket-sized ultrasound device at the bedside in patients with ADHF to accurately and more expeditiously differentiate LVSD and HFNEF.
Methods
Fifty consecutive adult patients with ADHF who were admitted to the cardiology inpatient ward service were enrolled after providing informed consent. Inclusion criteria were the clinical diagnosis of ADHF, made by clinicians not involved in the study. No formal protocol to define ADHF was used, but as patients with ADHF and other dyspneic disorders (chronic obstructive pulmonary disease, pulmonary embolism, pneumonia) are admitted to different services, careful triage is customary. After enrollment, all patients were assessed with the Framingham criteria for congestive heart failure (CHF). Patient exclusion criteria were an inability to provide consent, admission to the critical care unit, and the presence of a mechanical assist device. No patients were excluded for poor echocardiographic images from the parasternal window. Prespecified demographic data, admission laboratory data, and initial diagnostic examination findings (electrocardiography, chest x-ray) were recorded. Admission vital signs and the primary service’s cardiovascular physical exam findings were noted. Only exam findings that might prove useful to distinguish LVSD from HFNEF were recorded (presence of a third or fourth heart sound, presence of mitral regurgitation murmur, displaced point of maximal impulse). Other cardiac examination findings, although useful signs of CHF, are present in ADHF with reduced and preserved LV ejection fraction (EF).
All patients underwent formal TTE imaging in the hospital echocardiography laboratory, as part of the primary service’s diagnostic evaluation. The formal TTE studies were interpreted by a level 3 trained echocardiographer blinded to all clinical data (other than the admission diagnosis of CHF). LV EF was calculated from the apical four-chamber view using the single-chamber method of disks technique. The time interval between initial clinical assessment by the primary team and formal echocardiography was measured. Both the time from initial clinical assessment to first echocardiographic image acquisition and the time of report finalization were recorded.
Patients underwent a focused bedside HCU examination (Vscan; GE Healthcare, Waukesha, WI) by one of three second-year internal medicine residents blinded to the history, physical examination findings, and all clinical data ( Figure 1 ). The Vscan is a battery-powered device that weighs <1 lb and displays images from a 1.7-MHz to 3.8-MHz phased-array transducer on a 3.5-inch screen. The residents were not specifically selected based on any criteria other than their stated interest to participate in the trial after being told the scope and time commitment required. The internal medicine residents had no prior training in ultrasound before participating in this study. Cardiac ultrasound instruction was focused solely on the acquisition of parasternal long-axis and short-axis views for the determination of LV systolic function. Before imaging subjects for the study, residents completed 20 focused exams (on patients not enrolled in the study) supervised by experienced imagers to gain scanning experience. The residents were provided instruction on how to visually assess LV EF from parasternal views by observing endocardial motion, overall LV cavity size, and mitral valve excursion. To gain cardiac ultrasound interpretive skills, the residents were provided with a DVD containing parasternal images from 50 patients, selected to provide a wide range of LV systolic function (<20%–70%). The DVD was intended to review, in connection with an answer key of actual measured LV EF, to gain experience in visual estimation of LV EF.
On the basis of their bedside focused HCU examinations, residents were asked to visually estimate LV EF as >40% or <40%. An LV EF of 40% was selected because the acute therapies for ADHF are similar for patients with EFs above this value. In addition, prior data from large databases suggest that the prognoses of patients with EFs of 40% to 55% are more similar to those with LV EFs > 55% than patients with LV EFs < 40%. No measurements were made, and images were not stored from the residents’ HCU examinations. Formal echocardiographic results were used to classify patients into two groups according to whether their actual LV EFs was >40% or <40%, allowing the calculation of sensitivity, specificity, and negative and positive predictive value. Accuracies were also determined for the subgroups of patients with LV EFs < 30%, 30% to 50%, and >50%.
Continuous variable are summarized as mean ± SD and were compared between groups using unpaired t tests, with P values < .05 indicating statistical significance. The frequencies of categorical variables between groups were compared using χ 2 tests. All variables were entered into binary logistic regression analysis (forward stepwise conditional variable entry with constant) and tested for their ability to correctly predict formal LV EF < 40%.
Results
All patients met the Framingham criteria for the diagnosis of CHF. All 50 patients underwent HCU imaging. and all had parasternal images adequate to allow estimation of EF by the residents. The average age of the patients was 57 ± 17 years, and 58% were men. By formal TTE evaluation, the average EF was 32 ± 16% (range, 7%–70%), and 66% of patients had LVSD (EF < 40%). There were 27 (54%), 15 (30%), and eight (16%) patients in the LV EF subgroups of <30%, 30% to 50%, and >50%.
Table 1 shows the values and prevalences for variables in the patients subgrouped by whether their formal echocardiograms demonstrated LVSD. Although men were more common in the group of patients with reduced LV EFs, no other variable, including vital signs and admitting resident physical examination findings, proved able to distinguish the two groups of patients. Patients found to have LVSD were no more likely to have low systolic blood pressure, tachycardia, cardiac gallop, or audible mitral regurgitation. Table 2 summarizes the ECG, clinical laboratory, and chest x-ray findings. The prevalence of ECG Q waves, cardiomegaly on chest x-ray, and azotemia were similar between groups with preserved and abnormal LV EFs. Serum N-terminal pro–B-type natriuretic peptide (and log N-terminal pro–B-type natriuretic peptide) also failed to discriminate the two groups.
| Variable | All patients | EF > 40% | EF < 40% | EF < 40% vs EF > 40% |
|---|---|---|---|---|
| Age (y) | 57 ± 16 | 57 ± 18 | 57 ± 16 | — |
| Weight (kg) | 95 ± 32 | 92 ± 27 | 97 ± 34 | — |
| Men | 58% | 31% | 71% | P < .01 |
| SBP (mm Hg) | 127 ± 26 | 136 ± 21 | 123 ± 27 | — |
| DBP (mm Hg) | 77 ± 19 | 74 ± 23 | 79 ± 18 | — |
| HR (beats/min) | 90 ± 19 | 84 ± 19 | 93 ± 19 | — |
| S3 | 8% | 0% | 12% | — |
| S4 | 2% | 0% | 3% | — |
| MR | 20% | 19% | 21% | — |
| Displaced PMI | 12% | 13% | 12% | — |
| Variable | All patients | EF > 40% | EF < 40% | EF < 40% vs EF > 40% |
|---|---|---|---|---|
| NSR | 66% | 69% | 65% | — |
| AF | 16% | 6% | 21% | — |
| LVH | 24% | 25% | 24% | — |
| Q wave | 24% | 25% | 24% | — |
| LBBB | 2% | 6% | 3% | — |
| Cardiomegaly (CXR) | 82% | 75% | 85% | — |
| Sodium (mEq/L) | 138 ± 4 | 138 ± 3 | 137 ± 4 | — |
| BUN (mg/dL) | 32 ± 27 | 39 ± 36 | 29 ± 21 | — |
| Creatinine (mg/dL) | 1.6 ± 0.4 | 1.9 ± 2.5 | 1.7 ± 1.0 | — |
| NT-proBNP (pg/mL) | 11,336 ± 17,018 | 8,322 ± 9,531 | 12,503 ± 19,160 | — |
| log NT-proBNP | 3.73 ± 0.52 | 3.63 ± 0.55 | 3.78 ± 0.52 | — |
Residents’ prediction of LVSD (EF < 40%) was statistically different between the groups ( P < .001). Of patients with preserved LV EFs on formal echocardiography, only one (6%) was felt to have a reduced LV EF by resident HCU examination. Conversely, of the patients with confirmed LVSD, 32 (94%) had predicted low EFs by resident HCU examination ( P < .001). The sensitivity, specificity, positive predictive value, and negative predictive value for resident HCU examinations to predict true LVSD were 94%, 94%, 97%, and 88%. The formal LV EF of the one false-positive and two false-negatives were 46%, 32%, and 39%. The ability to correctly predict LV EF < 40% was best in subgroups of subjects with EFs < 30% and EFs > 50%, for which the sensitivities were both 100%. In patients with baseline LV EFs of 30% to 50%, it was more difficult for residents to be certain whether the LV EF was >40% or <40%, which is reflected in the sensitivity of 71% and specificity of 88%.
The average duration of time between admitting resident physical examination and finalization of the formal echocardiographic report was 22 ± 17 hours (range, 1–70 hours). Quartiles of time from clinical evaluation and report finalization were 11, 18, and 33 hours. Only 17% of finalized reports were available in <4 hours, and 29% were available to clinicians in <12 hours. The average time from first image to report finalization was 3.2 hours. The average duration of time between admitting resident clinical evaluation and initial formal echocardiographic image was 19 ± 17 hours. Quartiles of time from clinical evaluation and report finalization were 8, 14, and 27 hours. Of the 50 patients with ADHF admitted to the hospital, the on-call resident evaluated only 20% during weekday echocardiography laboratory hours. Sixteen percent of patients were admitted on weekends, when the echocardiography laboratory was not open.
Binary logistic regression demonstrated an overall model that was highly statistically significant ( P < .001). HCU EF was the most powerful predictor of EF < 40% (odds ratio, 154; P < .001). There was no statistically significant improvement in the model prediction with the addition of clinical, physical examination, laboratory, or ECG variables.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
