Seeing is not always believing

The proponents of replacing the stethoscope with HHU proclaim seeing is believing,12 and understandably so, since instead of listening to the sounds created by aortic stenosis (AS), why not just look at an image created by ultrasound reflection from the diseased valve itself? Instead of listening for a murmur indicating the presence of a ventricular septal defect (VSD), why not just look at the septum during the course of your physical examination to see if there is a defect or not? At face value this makes good sense, and certainly most cardiologists when suspicious of AS or a VSD will obtain an echo as an essential part of the evaluation to confirm their suspicions and provide important morphologic and hemodynamic information to appropriately manage the patient. But likewise, cardiologists know from experience how challenging and potentially frustrating it can be at times to obtain, review, measure, document and archive cardiac images, especially when a simple listen can mean no imaging is necessary. Seeing while performing an echo is sometimes initially believing, until on further review, when less preoccupied with obtaining the images, the initial impression is revised. This takes time and focus, skill and experience. This is not easily taught in a short course, and proficiency for performing this task is only gained after much experience, similar to auscultation. Thus, while the workflow of routine use of HHU may initially appear comparable to the traditional physical examination when measured without additional time spent re-examining images, mistakes of omission and commission will be made following that paradigm. Although consensus guidelines have attempted to differentiate focused cardiac ultrasound (FCU) used as part of the physical examination from traditional echocardiography, important practical concerns are raised by the expert writing group.13 ‘The implications of following-up on abnormalities detected by routine use of FCU at the time of physical examination, many of which would be false positives, needs to be considered. In addition, the potential impact of failure to refer symptomatic patients for complete echocardiographic evaluation, because of a “normal” FCU physical examination, needs to be considered. The infrastructure to educate and train all physicians who perform physical examination in FCU would be a massive undertaking’.

There are of course potential appropriate uses for HHU, such as in the emergency room or intensive care unit where the pretest probability of heart disease is increased and the additional training required to develop the skills needed to perform and interpret these studies is therefore justifiable. Likewise, in areas with high risk of rheumatic heart disease, incremental benefit of screening with echocardiography over auscultation has been demonstrated.14 However, even for those settings, simply solving the problems of miniaturisation, cost and initial training is only a helpful start towards creating a useful tool that works well as a supplement to existing echo resources. Images contain information that can definitively diagnose diseases. As such, the tolerance for underdiagnosis or overdiagnosis even in the hands of non-cardiologists should be low, and despite consensus guidelines to the contrary,13 archiving these images may ultimately be desirable and required for optimising patient care through review and comparison to prior studies and to justify and review actions or inactions taken as a result, with all the associated costs and requirements of video data management and storage. In fact, a significant shortcoming of traditional auscultation has been the lack of a saved record of the heart sounds for documentation and comparison, and advocating imaging without archiving would create similar problems.

Can hearing be believing too?

Hearing, in the appropriate situations, is also believing and can be sufficient, even if not as sensitive, specific or fully informative as ultrasound. In practice, if the goal is to quickly screen for an abnormal valve or septum among patients at low risk for having disease, the incremental value of imaging does not stand up quite as well against auscultation. Additionally, once a definitive diagnosis is made by echo, in many cases enough can be surmised by a quick follow-up auscultation to make repeat imaging at every subsequent encounter unnecessary and of no incremental benefit. Cardiologists, as the traditional gatekeepers of echocardiography, may have in the past been more likely to perform an echo on patients even when they had low suspicion of heart disease, as overuse of technology was not as big a concern as possibly missing pathology, no matter how unlikely, and additional testing was incentivised. But times are changing and current mandates are to optimise outcomes with the dual goal of reducing diagnostic errors and unnecessary testing. The American Society of Echocardiography has recently developed consensus guidelines for appropriate use of echo in adult and paediatric patients.15 ,16 Among the appropriate uses of echo is to look for heart disease in adults and children with a pathologic murmur. Pathologic murmurs are only known to be present if someone has listened to the patient with a stethoscope. If recently convened groups of experts in both adult and paediatric cardiology have concluded that echocardiography needs to be justified, and a gateway justification involves the use of auscultation, how can we abandon the stethoscope? Also, even though these guidelines were meant to inform referral decisions for echocardiography as a diagnostic test and not as part of the physical examination as in FCU, images are the product of both the complete and the focused echo, whether or not the ultrasound device is large or hand-held. Thus, if imaging is not deemed an appropriate diagnostic test for some referral reasons by knowledgeable experts, is it logical to think it should be considered necessary and appropriate for every routine physical examination?

In the paediatric echocardiography laboratory at Johns Hopkins, we perform over 10 000 echocardiograms annually. In 2014, we did 1383 first-time echocardiograms on outpatients aged under 18 years, 183 (13%) of which were ordered because of a heart murmur with no other signs or symptoms of heart disease (M. Cartoski, personal communication). Of these, the murmur description had features of an innocent murmur in 86, all but one of which had either no heart disease or minor abnormalities not requiring follow-up. Of the 97 with a pathologic murmur, 55 had significant findings requiring either surgery, cardiac catheterisation or cardiology follow-up in the next 2 years. This group constituted by far the largest percentage yield of abnormal findings among all the reasons for first-time outpatient echo, highlighting the value of auscultation as an important screening tool in paediatrics to discover heart disease before onset of symptoms. These data also show that distinguishing innocent from pathologic murmurs is needed in order to avoid unnecessary echocardiograms, especially in paediatrics where 60%–70% of patients have innocent murmurs yet <1% have congenital heart disease. Improving auscultation skills and developing alternative ways to analyse heart sounds would help the busy clinician to screen for patients with pathologic murmurs.

What should we be doing with the stethoscope instead of trading it in for a hand-held ultrasound device?

First, a consensus statement with expert review and opinion of the current status, merits, pros and cons of cardiac auscultation with recommendations regarding appropriate use of auscultation would be helpful in guiding training decisions, manpower allocation and industry poised to develop strategies to assist in the transformation of auscultation.

Next, training should be re-evaluated using these guidelines with the goal of developing new paradigms indicating what specifically needs to be taught and how to best evaluate competency for recognising those findings. Finley17 has published an excellent review of current methods for improving cardiac auscultation. Recent studies show specific difficulties for learning certain findings,18 the confirmation of which could lead to a realistic, goal-oriented statement of the specific aims of auscultation training. We are beginning to understand the biologic mechanisms involved in learning to recognise heart sounds, which may eventually lead to technological innovation to enhance auscultation training given inherent limitations of the human ear and brain.19 Training and proficiency testing modules are now available, including websites with large collections of heart sounds (https://murmurQuiz.org), recordings demonstrating key findings (Heart Songs—American College of Cardiology) and multimedia training programmes (http://www.Blaufuss.org), but studies addressing the clinical impact of these efforts are currently lacking.

Algorithms have been developed over the past 20 years using advanced signal processing and artificial intelligence methods, borrowing knowledge from submarine detection and speech recognition technologies, with sensitivities and specificities approaching that of the experienced clinician.20–23 These advances may in the near future allow the practitioner to listen to heart sounds with an electronic stethoscope while simultaneously analysing the sounds for feature content predictive of pathology. Similar to automated ECG diagnosis, the result will be an automated ‘reading’ of the heart sounds with creation of an archived recording and visual representation for future retrieval and comparison. However, in order for these methods to become clinically useful, a standardised, rigorous testing method must first be developed. A large testbed of heart sounds should be assembled with careful attention to recording technique, anonymised and thoroughly vetted clinical information associated with each set of sound files and ground truth echocardiography data along with expert description of the recorded sounds. Regulatory agencies such as the Food and Drug Administration as well as industry and clinicians would then be able to compare new algorithms to existing ones to determine incremental improvements and validate performance claims without the time and expense of first having to assemble large cohorts of patients or gather new data, which may not be comparable to datasets used to validate prior art.

Once algorithms have acceptable accuracy in clinical practice, they will reduce variability inherent in unassisted auscultation making it possible to avoid unnecessary and costly referrals and diagnostic testing. Enhanced auscultation will play an important role in routine clinical practice, pre participation evaluation of athletes to detect signs of hypertrophic cardiomyopathy and screening for patients with valvar or congenital heart disease in resource-limited areas.