Cardiac Auscultation: An Art Based on Science

Both the history and the physical examination are of paramount importance in the assessment of cardiovascular disease. Often a careful history is all that is required to make a diagnosis. The thoughtful use of physical examination can provide a great deal of accurate clinical information and obviate the need for more expensive testing. The focus of this brief primer is to outline some the key points of the cardiac examination with particular attention to cardiac auscultation. Data regarding the diagnostic proficiency of physicians in training and the teaching of cardiac auscultation will be discussed initially. Based on this data, the general examination and twelve clinically important cardiac events will be outlined. The remainder of the discussion will center on the accuracy of the cardiac examination compared to more expensive technologies. Tables and figures referred to in the text are found in the appendix. A list of defined abbreviations is also available.

Assessment studies of Internal Medicine and Family Practice residents have shown less than adequate proficiency in cardiac auscultation. The literature also indicates that the formal teaching of cardiac auscultation has waned considerably despite positive attitudes about its importance as a useful clinical skill.

St. Clair et al (Ann Intern Med. 1992) performed a cross-sectional assessment of 63 internal medicine residents at Duke University. The residents were tested for diagnostic accuracy using three preprogrammed simulations (MR, MS, and AI) on the "Harvey" cardiology patient simulator (standardized reproducible simulations). Two cardiologists blinded to the objectives of the study and without prior exposure to "Harvey" were used as reference standards (100% correct diagnoses). Residents were tested at the beginning and end of the working year with no specific intervention between sessions besides routine rounds. The final diagnoses were graded as correct, partially correct, or incorrect. Residents were also asked to indicate which "key" observations (assembled by the study investigators) were associated with the diagnosis in question. A total of 60 residents (95%) completed both sessions.

There was no significant difference in responses when comparing the level of training (PGY1-PGY3). The correct diagnosis rate for all residents was 54% for AI, 52% for MR, and 37% for MS. There was no significant improvement in accuracy between sessions. Identification of the holosystolic murmur of MR, and the OS and late peaking diastolic murmur of MS were "key" observations that significantly predicted a correct or partially correct diagnosis.

Limitations of this study include the fact that it was performed using a cardiology patient simulator, was made up of a relatively small group of participants at a single medical center, and only tested three conditions representing valvular abnormalities (AS, the most common valvular disease, was not included). Nonetheless, this study suggests that housestaff are not particularly proficient at cardiac auscultation, and more importantly, do not improve through the three years of their training. Given this and the fact that simple "key" observations (like MR is a holosystolic murmur) significantly improved diagnostic accuracy, there likely is a breakdown in the teaching of basic cardiac auscultation. Studies designed to address this issue followed.

Mangione et al (Ann Intern Med. 1993) explored the time and importance given to cardiac auscultation during training. They also assessed the auscultatory proficiency of medical students and housestaff. All 659 internal medicine and cardiology program directors were surveyed via mailed questionnaire. Specific objectives were to define the scope of structured teaching, gain insight into the attitudes regarding the relative importance of cardiac auscultation, and determine a list of clinically important cardiac events that "should be mastered by any practicing physician." Based on these twelve events, a test of auscultatory proficiency was given to 187 internal medicine residents, 16 cardiology fellows, and 49 third-year medical students (mainly from university-affiliated programs in Philadelphia).

The surveys from 498 (75.6%) program directors were analyzed. Only 27.1% of internal medicine programs and 37.1% of cardiology programs offered structured auscultatory training. However, all program directors felt cardiac auscultation was of "great clinical significance" (IM 5.3, cardiology 5.46 on scale from 1-6). The following cardiac events were deemed clinically important (4.5-5.8 on scale from 1-6): S3, S4, ESCLK, MSCLK, OS, RUB, MR, AS, MS, AI, and PDA.

Proficiency testing was performed using the above events (digitally recorded from patients) plus AS/AI. Trainees used stethophones, were told the chest area of recording, and could listen repeatedly as needed. Answers were tabulated as "non-adjusted" (partially correct/overcall) or "adjusted" (completely correct). Adjusted accuracy scores for the various events ranged from 0% to 56.2% for cardiology fellows and 2% to 36.8% for medical residents. Cardiology fellows were statistically more accurate than residents (except in the case of ESCLK, OS, MS, PDA, and AI), but residents were never more accurate than medical students (even when analyzing non-adjusted scores). In terms of improvement based on year of training, only the S3 gallop and OS/diastolic rumble of MS statistically improved for internal medicine residents (adjusted scores). Less than 10% of the residents identified AS/AI, OS, MS, ESCLK, and MSCLK. Interestingly, all residents and fellows felt that cardiac auscultation was extremely important (3.7 on scale from 1-4).

Although limited by the fact that the proficiency test had no reference standard (i.e. cardiology attending), this study once again suggests a definite inaccuracy in diagnosis (that does not improve appreciably with level of training) when cardiac auscultation is used by physicians in training. Despite a proclaimed interest in cardiac auscultation as a clinically useful tool, less than three fourths of internal medicine programs and two thirds of cardiology programs offer structured teaching of this skill. It is not difficult to see how these observations are related.

In a larger study, Mangione and Nieman (JAMA. 1997) again assessed the cardiac auscultatory proficiency of housestaff, this time concentrating on internal medicine and family practice residents. The larger question was whether these future "gatekeepers" of managed care could effectively screen patients using their examination skills. A total of 453 residents (198 internal medicine and 255 family practice) from mid-Atlantic programs as well as 88 medical students (third and fourth year) were tested on accuracy using the same twelve cardiac events discussed above. Answers were tabulated as "non-adjusted" (partially correct/overcall) or "adjusted" (completely correct). An attitude survey was also completed.

Data were analyzed by program type, year of training, and self-motivated teaching. All residents indicated the extreme importance of cardiac auscultation (3.8 on scale from 1-4). Internal medicine residents (1.2 adjusted events/12) were statistically more accurate (p=0.02) than family practice residents (0.9 adjusted events/12), but not more so than medical students (1.1 adjusted events/12). Adding in non-adjusted scores only led to an overall accuracy of 20% for all events tested. Although recognition of some cardiac events significantly improved within the three years of training (S3 and MS for internal medicine, MR and MSCLK for family practice) there was no overall improvement from medical students to residents. Family practice residents were found to have a significantly higher independent use of teaching audiotapes, but this did not translate into better accuracy as shown above. This study, despite no reference standard for the proficiency test, again raises serious questions about the adequacy of teaching and accurate use of cardiac auscultation.

Although the results of these studies cannot be generalized to all physicians in training, it is disturbing to see such a consistent lack of skill in those tested. If general internists are to be the initial screen of cardiac disease in a cost-effective medical environment, these skills must be preserved. To make at least some attempt at this, the next major part of this primer is a focused review of the cardiac examination with particular attention to auscultation. Specifically, the twelve clinically important cardiac events described in the above literature will be outlined.

Low frequency sounds (S3, S4, MS rumble) are best heard with the bell applied lightly, whereas high frequency sounds (clicks, OS, MR) are best heard with the diaphragm. One can appreciate a spectrum of frequencies by applying differential pressure with the bell.

Although authorities differ on the exact progression of auscultation, the important point to remember is that one should proceed in a structured, methodical manner with each examination so that the process is standardized and findings are less likely to be overlooked.

Braunwald suggests that the exam proceed from apex to LLSB, from LLSB to left base, and from left base to right base. This follows the path of blood flow (inflow to outflow) and allows for physiological correlation with findings.

The patient should be examined in the left lateral decubitus position to bring out findings at the apex (S3, OS, MS). The standard examination includes both the supine and sitting positions as well. Use of maneuvers such as standing to squatting is referred to as dynamic auscultation and is discussed below.

It is also important to remember to auscultate areas such as the left axilla and back (MR), subxiphoid (RV sounds), supraclavicular (venous hums), and even the top of the head (if AI is suspected).

Murmurs should be described by noting intensity, pitch, shape, quality, duration, and timing. The intensity of murmurs may be described using the Levine grading scale:

Noting "RRR with SEM" is not very useful and is certainly not diagnostic. "SEM" simply means that there is a murmur occurring in systole due to the forward flow of blood (functional, AS, HOCM, PS, aortic sclerosis, etc.). In a patient with effort syncope, the diagnosis of AS could be made based on the following description:

RRR with III/VI mid-peaking (<>) systolic murmur, heard best at the right base, and radiating to the right carotid. Intensity increases following pauses, but decreases with Valsalva and isometric handgrip.

While this may seem somewhat cumbersome, it could easily be abbreviated with shorthand or symbols. The important point is that the more descriptive a notation, the more diagnostic (which is the whole idea).

Interventions that alter cardiac dynamics are particularly useful in differentiating murmurs with similar patterns during the standard exam. Rather than memorize the particular effect of a maneuver on a given murmur, remember the dynamic alterations that occur during a maneuver, and apply this knowledge to the physiology of the murmur in question. Basically, the maneuvers of dynamic auscultation cause the following:

The above diagram shows one complete cardiac cycle with the first and second heart sounds indicated (S1 and S2, respectively). S1 is produced by the audible closure of the mitral and tricuspid valves, in that order. S2 is produced by the audible closure of the aortic and pulmonic valves, in that order. Remember "MTAP" for the order of valve closure during the cardiac cycle. This knowledge can be applied to the interpretation of heart sound splitting and recognition of the loud P2 of pulmonary hypertension.

It is sometimes difficult to distinguish S1 from S2, particularly during tachycardia. Remember that, in general, S1>S2 at the apex, and S2>S1 at the base. It is also useful to listen to the pattern of time intervals between sounds, because systole is the shorter of the two intervals. One can usually identify S1 and S2 using these simple observations and not relying on palpation of the carotid pulse, which can be misleading due to time delay.

Splitting of S1 may be normal or may be associated with RBBB, Ebstein’s anomaly (due to abnormality of the tricuspid valve), or ventricular arrhythmias. Splitting of S2 is more common and is classified as physiological, persistent, fixed, or paradoxical (Figure 1):

Auscultate at the left base to hear splitting of S2 because P2 is confined to this area. A2 is heard more diffusely. It makes physiologic sense that A2 should be heard more widely because the closure of the aortic valve is produced by a higher pressure (systemic) than that of the pulmonic valve (pulmonic).

Once S1 and S2 are identified with the recognition of systole and diastole, it is useful to divide these intervals into early, middle, and late periods. Though this may seem to be complicating matters unnecessarily, it helps to train the ear to listen selectively, and allows for more accurate diagnosis. Simply placing a cardiac event in systole or diastole narrows the diagnostic possibilities considerably. Actually placing the event at a particular point in systole or diastole can make the diagnosis.

The diagram of the cardiac cycle below shows the early, middle, and late periods within systole and diastole. Representative cardiac events are indicated in their respective positions.

These are by no means the only important findings with which to be familiar; however, the ability to reliably recognize these specific events provides the physician with broad diagnostic potential. Information regarding the genesis and recognition of the sounds is provided rather than the formal diagnosis of the associated condition. Obviously, listening to these sounds is far more important than reading about them, but descriptions of their physiology can help to solidify their place in the cardiac cycle.

This is a high frequency, early systolic sound occurring 0.03-0.07 second after S1. The sound is generated either by the sudden upward doming of an abnormal semilunar valve (aortic or pulmonic), or the sudden distention of the associated great artery (Table 2). AS, PS, and a bicuspid aortic valve can all produce this sound, with or without a murmur. Pulmonic ejection sounds often decrease in intensity with inspiration and are best heard at the LSB. With outflow obstruction the presence of a click implies mobility of the valve cusps (not heavily calcified) and identifies the level of obstruction as valvular.

This is a high frequency, mid systolic sound that is diagnostic of MVP. There may or may not be an associated late systolic murmur (heard best at the apex) which is due to mitral regurgitation caused by valve prolapse. The click is produced by systolic prolapse of the mitral valve leaflets into the left atrium with tensing of the chordae tendinae. Dynamic auscultation using the Valsalva or squatting to standing maneuver causes the click to move towards S1 with lengthening of the associated murmur, if present. This is due to a reduced ventricular volume that allows for more redundancy of the chordae with earlier valve prolapse. The opposite situation is true with the use of isometric handgrip or standing to squatting (click towards S2 with shortening of associated murmur, if present).

This is a high frequency, early diastolic sound that is associated with MS. It occurs 0.04-0.12 second after S2 and may or may not be associated with a late peaking or rumbling diastolic murmur. The sound is produced by rapid reversal of the superior bowing of the anterior mitral valve leaflet during diastole (high left atrial pressure). Generally the closer the OS to S2 (<0.07 second) and the longer the murmur, the more severe the stenosis. The OS must not be confused with P2 or S3 (Table 3).

This is a low frequency, mid diastolic sound occurring 0.14-0.22 second after S2. It is generated during the rapid ventricular filling phase in diastole prior to atrial contraction. It can be normal up to age 40, but is usually suggestive of pathology beyond that time. The left ventricular S3 is best heard at the apex and may be augmented by passive leg raise. The right ventricular S3 is best heard at the LLSB or subxiphoid area and is typically louder with inspiration. Although both the normal and abnormal S3 produce a characteristic triple cadence within the cardiac cycle, only in the abnormal circumstance is this referred to as the S3 gallop (Table 4).

This is a low frequency, late diastolic sound occurring 0.08-0.20 second prior to S1. It is generated during presystolic ventricular filling due to atrial contraction. S4 can be seen in conditions such as hypertension, significant AS, and diastolic dysfunction. The left ventricular S4 is also common in acute coronary syndromes when ischemia leads to decreased ventricular wall contractility and motion (less compliant). The squatting to standing maneuver tends to move S4 closer to S1 and in some cases can make it disappear.

The pericardial rub is a scratchy, leathery sound that is typically triple phased. The three components occur in mid systole, mid diastole, and late diastole. This characteristic sound is produced by apposition of abnormal visceral and parietal pericardial surfaces due to an inflammatory process (transmural MI, infectious, collagen vascular, etc.). One can augment the rub by having the patient lean forward, and with firm stethoscope pressure, auscultating during held expiration. Pericardial rubs can also be double or single phased. The loss of the three components typically occurs in the following order from first to last: mid diastolic, late diastolic, and mid systolic.

Aortic stenosis may be congenital, rheumatic, or degenerative-calcific. The murmur is a crescendo-decrescendo (<>), mid to late peaking, harsh systolic murmur. It is heard best at the right base and often radiates to the right carotid. Intensity of the murmur varies with cycle length, typically becoming louder after a pause (more blood in ventricle means more blood flow past the stenotic valve). The murmur of AS can have a more musical quality at the apex, termed the Gallavardin effect. The diagram below illustrates the murmur of AS.

Aortic insufficiency may be congenital (with AS), rheumatic, or associated with endocarditis, VSD, and collagen vascular disease. The murmur is a high frequency (blowing) decrescendo murmur beginning in early diastole. It is best heard at the third or fourth left ICS in the chronic state. Several signs are associated with this condition including Hill’s sign (>20mmHg increase in leg SBP compared to arm SBP), the Waterhammer pulse, de Musset’s sign (head bob), and Quincke’s pulse (capillary pulsations). As mentioned previously, this murmur can radiate to the top of the head. If the murmur radiates to the right sternal border, it suggests aortic root dilatation as in Marfan’s syndrome. The diagram below illustrates the murmur of AI.

The combination of AS and AI due to congenital valvular disease is illustrated below.

Mitral stenosis may be congenital, rheumatic, or due to endocarditis and amyloidosis. The murmur is typically a low frequency rumble or late peaking diastolic murmur heard best at the apex beat. It is often associated with an OS and a loud S1. Placing the patient in the left lateral decubitus position can be helpful in identifying the murmur. MS can be confused with an Austin Flint murmur which is a mid diastolic murmur caused by flow across a mitral valve that has been narrowed by increased ventricular pressure as a result of AI (Table 5). MS can also be associated with the Graham Steell murmur of pulmonic regurgitation. Sir William Osler said, "Mitral stenosis may be concealed under a quarter of a dollar," referring to how localized the findings can be. The following diagram illustrates the murmur of MS with an OS.

Mitral regurgitation is associated with endocarditis, ischemic heart disease, and MVP. The murmur is typically a high frequency, holosystolic, plateau murmur that is best heard at the apex. The murmur often radiates to the left axilla and back. There is no appreciable change in murmur intensity with cycle length (as with AS). MR may be associated with S3 in more severe cases. Isometric handgrip and standing to squatting maneuvers make the murmur louder. The following diagram illustrates the murmur of MR.

This condition represents approximately 15% of congenital heart disease (<5% at age 50). It is described as a "machinery" or continuous murmur that typically envelops S2 and is heard best at the left base. The PDA murmur may be confused with the more common continuous murmur of the venous hum (Table 6). If pulmonary hypertension develops over time, the murmur may systematically disappear (with reversal of blood flow). In this case one may only hear a systolic murmur, pulmonic ejection sound, and notice differential cyanosis (feet > hands). The following diagram illustrates the typical murmur of PDA.

There have not been a large number of studies that address the clinical accuracy of cardiac examination. Most studies use cardiologists as examiners in an attempt to optimize and standardize the skill level so that the actual maneuver or physical finding is evaluated. However, it has been shown that cardiologists differ in opinion with regards to examination findings. This is the art of the matter, where individual skill and acuity of senses cannot be completely standardized.

Lembo et al (NEJM. 1988) looked at the accuracy of dynamic auscultation in the bedside diagnosis of systolic murmurs. Fifty outpatients (age 6 to 85) with at least a I/VI systolic murmur were examined. The lesion responsible for the murmur in each patient was confirmed by either cardiac catheterization or echocardiography. Lesions included PS, TR, AS, HOCM, MR, MVP, and VSD. All patients were examined independently by two cardiologists blinded to the patients’ identity, diagnosis, and the maneuver being performed. Patients were first taught the maneuvers, separated from the examiner by a partition, and taken through the following maneuvers in no particular order: respiration, Valsalva, Müller, squatting to standing, standing to squatting, passive leg raise, isometric handgrip, transient arterial occlusion, and administration of amyl nitrite. Observations were made using an electronic stethoscope and included whether the murmur intensity increased, decreased, or stayed the same. These findings were compared to the expected findings of a particular maneuver in each cardiac lesion.

The two cardiologist observers were in agreement 87% of the time. Table 7 shows the calculated sensitivity, specificity, positive predictive value, and negative predictive value of each diagnostic maneuver in the case of right-sided murmurs, HOCM, MR, and VSD. It was found that inspiration increased the intensity of right-sided murmurs (PS, TR) thereby distinguishing them from left-sided murmurs with a sensitivity of 100% and a specificity of 88%. The murmur of HOCM decreased with standing to squatting (sensitivity 95%, specificity 85%) and increased with squatting to standing (sensitivity 95%, specificity 84%). MR and VSD murmurs had parallel responses to all maneuvers, but they could be distinguished from other murmurs with isometric handgrip (increased intensity, sensitivity 68%, specificity 92%) and transient arterial occlusion (increased intensity, sensitivity 78%, specificity 100%).

Limitations of this study include the fact that all patients had murmurs as a result of known lesions (no functional murmurs as controls), and electronic stethoscopes were used (not common in routine practice). We can still derive some useful information regarding the diagnostic accuracy of cardiac auscultation. In most cases either the sensitivity or specificity of a maneuver was very high allowing for an accurate diagnosis based either on the presence or absence of an expected change, depending on the lesion in question. Therefore, this study illustrates the fact that no single maneuver is perfect by itself, but when dynamic auscultation is performed systematically and correctly, an accurate diagnosis can be made.

Roldan et al (Am J Cardiol. 1996) went further to assess the value of cardiac examination in the detection of asymptomatic valvular heart disease. The study included 143 patients (68 healthy volunteers and 75 outpatients with connective tissue disease that did not have any cardiac symptoms) who underwent a TEE (interpreted independently) and a complete cardiac examination by an experienced cardiologist (blinded to TEE results). Valvular heart disease (MR, MVP, bicuspid aortic valve, PS, AI, TR, PI) was detected in 33 subjects (23%) by TEE and 25 subjects (19%) by physical examination. The cardiac examination had a sensitivity of 70% (CI 51-84%), specificity of 98% (CI 94-99%), negative predictive value of 92%, and positive predictive value of 92%. AS and MS were overcalled in two cases of physical examination (AV thickening with AI, and AI alone on TEE). Interestingly, only two of the ten patients with abnormal valves by TEE and not on physical examination had more than just mild valve regurgitation or other clinically significant lesion. Dynamic auscultation distinguished functional murmurs (31%) from pathologic murmurs (13%) with a specificity of 98% and a positive predictive value of 92%.

In what boils down to be a test of a single cardiologist’s skill, this study does provide support for the accuracy of a thorough cardiac examination. The high specificity of the cardiac examination, particularly in the case of distinguishing functional from pathologic murmurs, suggests that this is an excellent screening tool in asymptomatic patients. This is of course assuming that the skills of the examiner are sharp.

Finally, Etchells et al (JAMA. 1997) conducted a rigorous review of MEDLINE data regarding the precision and accuracy of the clinical examination in evaluating abnormal systolic murmurs. Data were mainly available from studies using cardiologist examiners who were found to be very accurate at detecting abnormal systolic murmurs (LR ¥ with CI 14-¥ ). However, precision in the detection of systolic murmurs varied widely (k=0.29-0.79). This article is very useful because it defines the likelihood ratios associated with particular events and maneuvers in terms of their diagnostic accuracy. Some of these numbers are presented in the appendix (Tables 8-10). To summarize these findings:

Once again the cardiac examination is shown to be an accurate diagnostic tool in experienced hands. The wide variation in precision likely represents different experience levels and reflects the art of this skill. It is difficult to understand why this accurate diagnostic skill has waned so dramatically given its proven clinical utility and ability to save medical dollars.

Sir William Osler said, "The practice of medicine is an art, based on science." At this point it should be clear that this is particularly true of cardiac auscultation. Internists should spend more time learning this basic skill, given their important task of screening the general population. We cannot afford to lose this powerful and rewarding tool. It is accurate and economical, it fosters a patient’s confidence and trust, and it is rewarding to the physician made able to make a diagnosis with the senses. It is hard to imagine that any technology could compete with that.

Table 8. Precision of the Cardiac Examination of Systolic Murmurs (Reference 3).

- Accuracy of the Clinical Examination for Detecting Mitral Regurgitation

Finding	Reference Standard (No. of Patients)		Positive Likelihood Ratio (95% CI)*		Negative Likelihood Ratio (95% CI)^	Quality Grade
Murmur in mitrial area Study 1³⁶		Echocardiogram: moderate to severe MR (394)		3.9 (3.0 - 5.1)	0.34 (0.23 – 0.47)	C
Study 2³⁵		Cardiac catheterization: moderate to severe MR (35)		3.6 (1.9 - 7.7)	0.12 (0.02 – 0.50)	C
Late or holosystolic murmur		Echocardiogram: moderate to severe MR (80)		1.8 (1.3 – 11.0)	0.0 (0.0 – 0.8)	C
Any murmur during acute MI³⁷		Cardiac catheterization: moderate to severe MR (206)		4.7 (1.3 – 11.0)	0.66 (0.25 – 1.00)	C
With transient arterial occlusion, murmur increases in intensity²⁹		Cardiac catheterization: severity not statedt		7.5 (2.5 – 23.0)	0.28 (0.13 – 0.60)	C

AI – Aortic Insufficiency	MSCLK – Mid Systolic Click
AS – Aortic Stenosis	MVP – Mitral Valve Prolapse
ASD – Atrial Septal Defect	OS – Opening Snap
ESCLK – Early Systolic Click	PDA – Patent Ductus Arteriosus
HOCM – Hypertrophic Obstructive Cardiomyopathy	PS – Pulmonic Stenosis
MR – Mitrial Regurgitation	RUB – Pericardial Rub
MS – Mitrial Stenosis	TR – Tricuspid Regurgitation
	VSD – Ventricular Septal Defect