PEDIATRICS Vol. 118 No. 6 December 2006, pp. 2322-2325 (doi:10.1542/10.1542/peds.2006-1557)
ARTICLE |
Assessing Children’s Heart Sounds at a Distance With Digital Recordings
a Children’s Heart Centre, Izaak Walton Killam Health Centre, Halifax, Nova Scotia, Canada
b Department of Pediatrics, Dalhousie University, Halifax, Nova Scotia, Canada
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
OBJECTIVE. The objective of this study was to assess whether computer-stored digital sound recordings can be used to distinguish innocent from pathologic systolic murmurs.
METHODS. Recordings of 55 children aged 1 month to 19 years were made remotely with the use of a digital stethoscope and were e-mailed to a computer in our center for later assessment. Eight-second recordings were made by a physician in 2 to 4 locations on the chest. Three cardiologists who were blinded to the diagnosis reviewed the recordings independently using stethophones to assess the splitting of the second heart sound and whether murmurs were innocent or pathologic. Diagnoses were confirmed with echocardiography.
RESULTS. Seventeen children had innocent murmurs and 38 had pathologic murmurs. For the 3 cardiologists, sensitivity was 0.87 to 1.0, specificity was 0.82 to 0.88, negative predictive value was 0.75 to 1.0, and positive predictive value was 0.93 to 0.95. Assessment of splitting of second heart sound was highly accurate.
CONCLUSIONS. Digital recordings of children’s heart sounds allow reliable differentiation between innocent and pathologic murmurs. Use of this technology may allow remote diagnosis of childhood murmurs and avoid the expense and stress of travel to pediatric cardiology centers for some children. Cardiologists who use recordings should assess their diagnostic accuracy before clinical application.
Key Words: pediatrics • murmurs • auscultation • stethoscope • telemedicine
Abbreviations: NPV—negative predictive value • PPV—positive predictive value
Advances in technology now permit accurate assessment of children with suspected heart disease at a distance from pediatric cardiology centers. Echocardiography and electrocardiography commonly are performed at remote sites, and the images are transmitted to a cardiac center for diagnosis.1 Development of stethoscopes also has permitted remote acoustic assessment of heart sounds using analog2 and digital systems.3 This offers the potential for remote diagnosis of heart disease and also the diagnosis of the innocent murmur.
Fifty-six percent of new referrals to our pediatric cardiac center from family physicians and pediatricians have innocent murmurs.4 Travel to our center may take several hours, with the attendant expenses and loss of income for parents. Because the diagnosis of an innocent murmur in most cardiac centers is based primarily on auscultation by a pediatric cardiologist, a heart sound recording and transmission service between regional hospitals and our cardiac center could allow remote diagnosis, thereby eliminating the expense and stress of travel.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
The patients were 55 children who were aged 1 month to 19 years and attended a pediatrician’s (M.A.) office (7 children) or our Cardiac Clinic (48 children). All were known to have a murmur and were considered sufficiently cooperative for the recording. The only exclusions were infants who were younger than 1 month. Patients were not randomly assigned and were a convenience sample. Informed consent was obtained from the parent or the patient.
Recordings of heart sounds were made with a Littmann 4000 (3M Health Care, St Paul, MN) digital stethoscope by the attending pediatrician or pediatric cardiologist in 2 to 4 standard positions on the chest: the right and left upper sternal border, the left lower sternal border, and the apex. The patients usually were supine. Recordings lasted between 4 and 8 seconds and were downloaded from the stethoscope to a computer via an infrared connection. The recordings that were done by the pediatrician in her office were e-mailed to the same computer. All recordings were labeled using a code, thereby blinding the pediatric cardiologists to patient identification. All patients had an echocardiogram, and all were seen by a pediatric cardiologist. The echocardiographic diagnosis and patient name and date of birth all were recorded separately from the sound files.
The pediatric cardiologist who recorded the sounds described the findings on a score sheet. This included the second heart sound (split/single/indeterminate), the timing of the murmur (systolic/diastolic/continuous), and the murmur diagnosis (innocent/pathologic/uncertain). When the murmur was pathologic the diagnosis was recorded.
All recordings were assessed by 3 pediatric cardiologists who were blinded to each other and to the patient diagnosis and clinical information. The chest location of the recordings was provided to the cardiologists. With the use of the software that was provided with the digital stethoscope (Littman Sound Analysis Software for Electronic Stethoscope Model 4000, Version 1.0 for Heart Sounds; 3M Health Care), 2 to 4 recordings per patient were replayed and listened to using a stethophone. A stethophone has the body of a stethoscope with the head replaced by a miniature speaker that is attached to the audio output of the computer. The sounds were listened to unfiltered and also using low- and high-pass filters to mimic the bell and the diaphragm of the standard stethoscope. The same score sheet, described in the previous paragraph, was used for assessment of the heart sounds. The clarity of the recorded heart sounds also was assessed as very clear/moderate/unclear. The correlation of the score sheet with the echocardiographic assessment was performed only after all recordings had been reviewed by the 3 pediatric cardiologists.
Data analysis included calculation of sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) for murmur diagnosis. Two diagnostic categories were used for comparison: innocent and pathologic/uncertain. The standard for comparison was the echocardiographic diagnosis. A generalized estimating equation logistic regression model was used to assess the influence of age and heart sound clarity on diagnostic accuracy. The 
statistic was calculated as a measure of agreement among all 3 pediatric cardiologists considered together. The study was approved by our institutional Research Ethics Board.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
There were a total of 55 patients, 29 boys and 26 girls. The age range was 1 month to 19 years (median: 5 years). The diagnoses on the basis of echocardiography and clinical findings were as follows: 17 patients with innocent murmurs, and 38 patients with cardiac pathology. The pathologic diagnoses are listed in Table 1.
|
Average sensitivity, specificity, PPV, and NPV were 0.93, 0.86, 0.86, and 0.94, respectively (Table 2). The
statistic that was calculated for agreement among the 3 pediatric cardiologists was .62.
|
To check that the cardiologists truly were blinded to the patient data, we repeated the sensitivity and specificity analysis, showing the performance of each cardiologist on the patients whom he had never seen in person. There was no difference in these scores from those in Table 2.
There were 12 cases in which the murmur assessment of at least 1 pediatric cardiologist disagreed with the echocardiographic findings. For the 17 children with innocent murmurs, there were 51 evaluations (3 cardiologists x 17 patients) by the 3 pediatric cardiologists and 5 (10%) disagreements. For the 38 children with pathologic murmurs, there were 114 evaluations and 7 (6%) disagreements. The murmurs with disagreements were caused by small muscular ventricular septal defects (n = 3), mitral valve prolapse (n = 1), peripheral pulmonary stenoses (n = 2), and mild aortic stenosis (n = 1). In all cases except 1, only 1 pediatric cardiologist was at variance with the echocardiographic diagnosis.
The clarity of heart sounds and murmurs on the recordings was assessed by each pediatric cardiologist. For the 55 patients, there were a total of 165 evaluations, and in 22 (14%), the recording was judged unclear. The impact of recording clarity on diagnostic accuracy was assessed, and the relative risk for clear versus unclear/moderate recording in accurate diagnosis of the murmurs was 1.06 (0.96–1.09; P = .18, not significant). For 4 of the 7 pathologic cases for which a cardiologist disagreed with the echocardiographic diagnosis, 1 or more cardiologists believed that the recording was unclear. The patients’ age, whether < or >5 years, also was not a factor in diagnostic accuracy, with relative risk 1.03 (0.91–1.08; P = .46).
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
This study demonstrates that digital recordings of children’s heart sounds generally are of diagnostic quality. Our recordings of children from age 1 month to 19 years have allowed accurate distinction between innocent and pathologic murmurs in >90% of cases. The sensitivities and specificities of all 3 cardiologists were indicative of a high degree of accuracy and would be improved by eliminating any recording that was believed to be unclear.
It is difficult to compare our sensitivity and specificity results with in-person evaluations because those evaluations include other clinical findings. Smythe et al5 reported 96% and 95%, respectively, in 161 cases. Geva et al6 reported a comparison of clinical examination by 1 cardiologist (including electrocardiogram and chest radiograph) with echocardiography in 100 children. Recalculation of their results combining pathologic and uncertain diagnoses gives a sensitivity of 90%, specificity of 89%, NPV of 96%, and PPV of 88%, which are similar to our results with recorded heart sounds. We could not assess our in-person assessments of pathologic cases because the diagnoses were known to the attending cardiologist. With respect to the innocent cases, the echocardiographic diagnosis generally was not known at the time of examination, and there was 100% agreement between the in-person diagnosis and the echocardiographic diagnosis. In-person assessment with full knowledge of clinical information has the potential to improve the assessment of recorded heart sounds. In practice, an electrocardiogram also might improve diagnostic accuracy,5 and current research suggests a potential for computerized frequency analysis to improve further the accuracy of murmur assessment.7
The 2 previous studies of remote assessment of murmurs in children used different technology from that available to us: in 1 case, live analog transmission; in the other, a European digital recording stethoscope. We used stethophones for playback of recordings. These differences justified this pilot study with our own technology before clinical application.
Despite the technologic differences, the accuracy of assessment of heart sounds at a distance in our study was similar to the previous studies. A study of 76 children by Belmont and Mattioli2 using live analog transmission of heart sounds indicated similar overall levels of sensitivity and specificity but interestingly found much worse performance for children who were younger than 5 years. We did not find such a difference, although 28 of our patients were 5 years or younger. Dahl et al3 studied 87 children using a digital stethoscope, and 4 observers found levels of specificity and sensitivity similar to ours, with a mean for pairs of observers of .81.
The assessment of diagnostic performance of a cardiology service is described best by an average performance of the cardiologists who provide that service. In practice, assessment of patients (or recorded murmurs) would be done by various cardiologists in rotation, and although performance may vary among cardiologists, it is important to describe the service as a whole. Accordingly, we calculated a and average sensitivity and specificity as being characteristic of our service as a whole. This compares favorably with the in-person results of Smythe et al,5 although in their study, only 1 of 3 cardiologists examined the children. Unfortunately, performance of in-person clinical evaluation generally is not quantified in cardiology departments; therefore, comparison with diagnosis from recordings is uncertain. Nonetheless, the performance characteristics of any cardiology service using a diagnostic test such as digital recordings of heart sounds should be determined before the test is placed in clinical service.
Assessment of children’s heart sounds at a distance using e-mailed digital recording offers a potential method for triage of patients with murmurs. This is particularly useful for rural patients, and a heart sound recording service in regional hospitals should allow a significant number of patients to receive a diagnosis reliably without the need for travel. Fewer in-person consultations should shorten waiting lists and save some families the cost of travel and our public health system the cost of unnecessary consultations. Children with pathologic murmurs still would require in-person evaluation by a pediatric cardiologist. The disadvantage of distance diagnosis, of course, is the lack of personal contact between the family and a cardiologist and the lack of a complete history and physical examination by a cardiac specialist. Careful selection of referrals should minimize these potential disadvantages.
Limitations of our study include the deliberate exclusion of patients who were younger than 1 month, as well as those who could not cooperate. In-person examination often can overcome this difficulty, whereas cooperation truly is essential for adequate recording. Careful instruction and experience are necessary before reliable recordings can be made with a digital stethoscope. Those who use a recording stethoscope should have the ability to assess heart sounds and determine the adequacy of a recording. Various models of digital stethoscopes may use different filtering systems; therefore, the results of this study may not be applicable to all commercial digital stethoscopes. The stethoscope that was used in our study did not have a pediatric head, which made studies on infants awkward. Finally, our study had a preponderance of pathologic murmurs. Our results may not reflect the diagnostic accuracy of recordings in a typical referred population with primarily innocent murmurs, in whom PPV might be expected to be lower but NPV higher.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
This study shows that remote digital recordings of children’s heart sounds can be of diagnostic quality when carefully recorded and can be used for reliable remote assessment of murmurs in children. The use of this technology by pediatric cardiologists could have economic advantages for families and for health care providers. We continue to believe, however, that auscultation education for referring physicians remains crucial to allow appropriate selection of patients for referral, whether in-person or using remote recordings.
|
ACKNOWLEDGMENTS |
|---|
We thank Mary Fotheringham, MSc, for coordinating the study; Linda Dodds, PhD, and Colleen O’Connell, PhD, for statistical consultation; and Donna King, RN, Sharon McIntyre, RN, Brian Hoyt, MS, and Douglas Roy, MD, for help and advice.
|
FOOTNOTES |
|---|
Accepted Aug 21, 2006.
Address correspondence to John Finley, MD, IWK Children’s Heart Centre, 5850/5950 University Ave, Halifax, NS, Canada B3K6R8. E-mail: john.finley{at}iwk.nshealth.ca
The authors have indicated they have no financial relationships relevant to this article to disclose.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
- Finley JP, Justo R. Telecardiology in Canada and Australia. In: Wootton R, Batch J, eds. Telepediatrics: Telemedicine and Child Health. London, United Kingdom: Royal Society of Medicine Press; 2005:41–52
- Belmont JM, Mattioli LF. Accuracy of analog telephonic stethoscopy for pediatric telecardiology.
Pediatrics. 2003;112
:780
–786
[Abstract/Free Full Text] - Dahl LB, Hasvold P, Arild E, Hasvold T. Heart murmurs recorded by a sensor based electronic Stethoscope and e-mailed for remote assessment.
Arch Dis Child. 2002;87
:297
–301
[Abstract/Free Full Text] - Wong KK, Barker AP, Warren AE. Paediatricians’ validation of learning objectives in paediatric cardiology. Paediatr Child Health. 2005;10 :95 –99
- Smythe JF, Teixeira OHP, Vlad P, Demers PP, Feldman W. Initial evaluation of heart murmurs: are laboratory tests necessary?
Pediatrics. 1990;86
:497
–500
[Abstract/Free Full Text] - Geva T, Hegesh J, Frand M. Reappraisal of the approach to the child with a heart murmur: is echocardiography mandatory? Int J Cardiol. 1988;19 :107 –113[CrossRef][Web of Science][Medline]
- DeGroff CG, Bhatikar S, Herzburg J, Shandas R, Valdes-Cruz L, Mahajan RL. Artificial neural network-based method of screening heart murmurs in children.
Circulation. 2001;103
:2711
–2716
[Abstract/Free Full Text]
PEDIATRICS (ISSN 1098-4275). ©2006 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
This article has been cited by other articles:
|
|
 |
|
  Other articles noted Evid. Based Med., August 1, 2007; 12(4): 127 - 128. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
