Estimation of systolic blood pressure by Random Forest using heart sounds and a ballistocardiogram.

Cuffless blood pressure measurement enables unobtrusive and continuous monitoring that can be integrated with wearable devices to extend healthcare to non-hospital settings. Most of the current research has focused on the estimation of blood pressure based on pulse transit time or pulse arrival time using ECG or peripheral cardiac pulse signals as proximal time references. This study proposed the use of a phonocardiogram (PCG) and ballistocardiogram (BCG), two signals detected noninvasively, to estimate systolic blood pressure (SBP). For this, the PCG and the BCG were simultaneously measured in 21 volunteers in the rest, activity, and post-activity conditions. Different features were considered based on the relationships between these signals. The intervals between S1 and S2 of the PCG and the I, J, and K waves of the BCG were statistically analyzed. The IJ and JK slopes were also estimated as additional features to train the machine-learning algorithm. The intervals S1-J, S1-K, S1-I, J-S2, and I-S2 were negatively correlated with changes in SBP (p-val < 0.01). The features were used as explanatory variables for a regressor based on the Random Forest. It was possible to estimate the systolic blood pressure with a mean error of 3.3 mmHg with a standard deviation of ± 5 mmHg. Therefore, we foresee that this proposal has potential applications for wearable devices that use low-cost embedded systems.

Design, assembly and validation of a low-cost, high-fidelity simulator for heart exploration.

INTRODUCTION: Heart exploration is an essential clinical competence that requires continuous training and exposure. Low availability and accessibility to patients with heart disease constitutes a barrier to acquiring this competence. Inadequate cardiac auscultation skills in medical students, residents, and graduate physicians have been documented. OBJECTIVE: To develop and validate a low-cost, high-fidelity simulator for heart exploration. METHODS: A low-cost, high-fidelity heart examination simulator capable of reproducing normal cardiac sounds was designed and developed. Subsequently, the simulator was validated by a group of experts who gave their opinion according to a Likert scale. RESULTS: Ninety-four percent agreed that the simulator motivates the learning of heart exploration, and 92 % considered it to be a realistic model; 91 % considered that the simulator is an attractive tool to reinforce learning and 98 % recommended its further use. CONCLUSIONS: The use of the simulator facilitates the acquisition of skills and stimulates learning in the student, which can be attributed to repeated practice, longer exposure time and cognitive interaction.

INTRODUCCIÓN: La exploración cardiaca es una competencia clínica fundamental que requiere exposición o entrenamiento continuo. La baja disponibilidad y accesibilidad de pacientes con patología cardiaca constituye una barrera para adquirir esta competencia. Se han documentado inadecuadas habilidades de auscultación cardiaca en estudiantes de medicina, residentes y médicos graduados. OBJETIVO: Elaborar y validar un simulador de alta fidelidad y bajo costo para exploración cardiaca. MÉTODOS: Se diseñó y elaboró un simulador para exploración cardiaca, realista y de bajo costo capaz de reproducir ruidos cardiacos normales. Posteriormente se realizó la validación del simulador por un grupo de expertos que emitieron su opinión de acuerdo con una escala tipo Likert. RESULTADOS: El 94 % afirmó que el simulador motiva el aprendizaje de la exploración cardiaca y 92 % lo consideró un modelo realista; 91 % consideró que el simulador es una herramienta atractiva para fortalecer el aprendizaje y 98 % recomendó seguir utilizándolo. CONCLUSIONES: El uso del simulador facilita la adquisición de competencias y estimula el aprendizaje en el estudiante, lo cual puede ser atribuido a la práctica deliberada, a un mayor tiempo de exposición y a la interacción cognitiva.

Diseño, elaboración y validación de un simulador realista y de bajo costo para exploración cardiaca / Design, assembly and validation of a low-cost, high-fidelity simulator for heart exploration

Resumen Introducción: La exploración cardiaca es una competencia clínica fundamental que requiere exposición o entrenamiento continuo. La baja disponibilidad y accesibilidad de pacientes con patología cardiaca constituye una barrera para adquirir esta competencia. Se han documentado inadecuadas habilidades de auscultación cardiaca en estudiantes de medicina, residentes y médicos graduados. Objetivo: Elaborar y validar un simulador de alta fidelidad y bajo costo para exploración cardiaca. Métodos: Se diseñó y elaboró un simulador para exploración cardiaca, realista y de bajo costo capaz de reproducir ruidos cardiacos normales. Posteriormente se realizó la validación del simulador por un grupo de expertos que emitieron su opinión de acuerdo con una escala tipo Likert. Resultados: El 94 % afirmó que el simulador motiva el aprendizaje de la exploración cardiaca y 92 % lo consideró un modelo realista; 91 % consideró que el simulador es una herramienta atractiva para fortalecer el aprendizaje y 98 % recomendó seguir utilizándolo. Conclusiones: El uso del simulador facilita la adquisición de competencias y estimula el aprendizaje en el estudiante, lo cual puede ser atribuido a la práctica deliberada, a un mayor tiempo de exposición y a la interacción cognitiva.

Abstract Introduction: Heart exploration is an essential clinical competence that requires continuous training and exposure. Low availability and accessibility to patients with heart disease constitutes a barrier to acquiring this competence. Inadequate cardiac auscultation skills in medical students, residents, and graduate physicians have been documented. Objective: To develop and validate a low-cost, high-fidelity simulator for heart exploration. Methods: A low-cost, high-fidelity heart examination simulator capable of reproducing normal cardiac sounds was designed and developed. Subsequently, the simulator was validated by a group of experts who gave their opinion according to a Likert scale. Results: Ninety-four percent agreed that the simulator motivates the learning of heart exploration, and 92 % considered it to be a realistic model; 91 % considered that the simulator is an attractive tool to reinforce learning and 98 % recommended its further use. Conclusions: The use of the simulator facilitates the acquisition of skills and stimulates learning in the student, which can be attributed to repeated practice, longer exposure time and cognitive interaction.

Tele-Clinic Visits in Pediatric Patients with Marfan Syndrome Using Parentally Acquired Echocardiography.

OBJECTIVE: To test feasibility of tele-clinic visits using parentally acquired vital signs and focused echocardiographic images in patients with Marfan syndrome. STUDY DESIGN: We included patients with Marfan syndrome aged 5-19 years followed in our clinic. We excluded patients with Marfan syndrome and history of previous aortic root (AoR) surgery, cardiomyopathy, arrhythmia, or AoR ≥4.5 cm. We trained parents in-person to acquire focused echocardiographic images on their children using a hand-held device as well as how to use a stadiometer, scale, blood pressure (BP) machine, and a digital stethoscope. Before tele-clinic visits, parents obtained the echocardiographic images and vital signs. We compared tele-clinic and on-site clinic visit data. Parental and clinic echocardiograms were independently analyzed. RESULTS: Fifteen patient/parent pairs completed tele-clinic visits, conducted at a median of 7.0 (IQR 3.0-9.9) months from the in-person training session. Parents took a median of 70 (IQR 60-150) minutes to obtain the height, weight, heart rate, BP, cardiac sounds, and echocardiographic images before tele-clinic visits. Systolic BP was greater on-site than at home (median +13 mm Hg, P = .014). Height, weight, diastolic BP, heart rate, and AoR measurements were similar. CONCLUSIONS: This study provides information for implementing tele-clinic visits using parentally acquired vital signs and echocardiographic images in patients with Marfan syndrome. The results show that tele-clinic visits are feasible and that parents were able to obtain focused echocardiographic images on their children. TRIAL REGISTRATION: ClinicalTrials.gov: NCT03581682.

Phonocardiogram classification using deep neural networks and weighted probability comparisons.

Cardiac auscultation is one of the most conventional approaches for the initial assessment of heart disease, however the technique is highly user-dependent and with low repeatability. Several computational approaches based on the analysis of the phonocardiograms (PCG) have been proposed to classify heart sounds into normal or abnormal, but most often do not achieve acceptable levels of sensitivity (Se) and specificity (Sp) or require the use of special hardware. We propose a novel approach for classification of PCG. First, the system makes use of deep neural networks for computing individual cardiac cycle probabilities, followed by classification using weighted probability comparisons. The system was tested on an extended dataset consisting of a balanced sample of 18179 normal and abnormal cycles, achieving Se and Sp values of 91.3% and 93.8% respectively. In addition, the system overcomes previous limitations since it was trained with a balanced sample; also, the decision factor used during the classification stage allows to control the trade-off between Se and Sp, making the proposed system suitable for clinical applications.

Ensemble methods with outliers for phonocardiogram classification.

OBJECTIVE: Heart sound classification and analysis play an important role in the early diagnosis and prevention of cardiovascular disease. To this end, this paper introduces a novel method for automatic classification of normal and abnormal heart sound recordings. APPROACH: Signals are first preprocessed to extract a total of 131 features in the time, frequency, wavelet and statistical domains from the entire signal and from the timings of the states. Outlier signals are then detected and separated from those with a standard range using an interquartile range algorithm. After that, feature extreme values are given special consideration, and finally features are reduced to the most significant ones using a feature reduction technique. In the classification stage, the selected features either for standard or outlier signals are fed separately into an ensemble of 20 two-step classifiers for the classification task. The first step of the classifier is represented by a nested set of ensemble algorithms which was cross-validated on the training dataset provided by PhysioNet Challenge 2016, while the second one uses a voting rule of the class label. MAIN RESULTS: The results show that this method is able to recognize heart sound recordings efficiently, achieving an overall score of 96.30% for standard signals and 90.18% for outlier signals on a cross-validated experiment using the available training data. SIGNIFICANCE: The approach of our proposed method helped reduce overfitting and improved classification performance, achieving an overall score on the hidden test set of 80.1% (79.6% sensitivity and 80.6% specificity).

An mHealth monitoring system for traditional birth attendant-led antenatal risk assessment in rural Guatemala.

Limited funding for medical technology, low levels of education and poor infrastructure for delivering and maintaining technology severely limit medical decision support in low- and middle-income countries. Perinatal and maternal mortality is of particular concern with millions dying every year from potentially treatable conditions. Guatemala has one of the worst maternal mortality ratios, the highest incidence of intra-uterine growth restriction (IUGR), and one of the lowest gross national incomes per capita within Latin America. To address the lack of decision support in rural Guatemala, a smartphone-based system is proposed including peripheral sensors, such as a handheld Doppler for the identification of foetal compromise. Designed for use by illiterate birth attendants, the system uses pictograms, audio guidance, local and cloud processing, SMS alerts and voice calling. The initial prototype was evaluated on 22 women in highland Guatemala. Results were fed back into the refinement of the system, currently undergoing RCT evaluation.

Usefulness of the second heart sound for predicting pulmonary hypertension in patients with interstitial lung disease / Utilidade da segunda bulha cardíaca na predição de hipertensão pulmonar em portadores de doenças intersticiais pulmonares

CONTEXT AND OBJECTIVE: P2 hyperphonesis is considered to be a valuable finding in semiological diagnoses of pulmonary hypertension (PH). The aim here was to evaluate the accuracy of the pulmonary component of second heart sounds for predicting PH in patients with interstitial lung disease. DESIGN AND SETTING: Cross-sectional study at the University of Brasilia and Hospital de Base do Distrito Federal. METHODS: Heart sounds were acquired using an electronic stethoscope and were analyzed using phonocardiography. Clinical signs suggestive of PH, such as second heart sound (S2) in pulmonary area louder than in aortic area; P2 > A2 in pulmonary area and P2 present in mitral area, were compared with Doppler echocardiographic parameters suggestive of PH. Sensitivity (S), specificity (Sp) and positive (LR+) and negative (LR-) likelihood ratios were evaluated. RESULTS: There was no significant correlation between S2 or P2 amplitude and PASP (pulmonary artery systolic pressure) (P = 0.185 and 0.115; P= 0.13 and 0.34, respectively). Higher S2 in pulmonary area than in aortic area, compared with all the criteria suggestive of PH, showed S = 60%, Sp= 22%; LR+ = 0.7; LR- = 1.7; while P2> A2 showed S= 57%, Sp = 39%; LR+ = 0.9; LR- = 1.1; and P2 in mitral area showed: S= 68%, Sp = 41%; LR+ = 1.1; LR- = 0.7. All these signals together showed: S= 50%, Sp = 56%. CONCLUSIONS: The semiological signs indicative of PH presented low sensitivity and specificity levels for clinically diagnosing this comorbidity.

RESUMO CONTEXTO E OBJETIVO: Hiperfonese de P2 tem sido considerada como achado valoroso no diagnóstico semiológico de hipertensão pulmonar (HP). O objetivo foi de avaliar a acurácia do componente pulmonar da segunda bulha cardíaca em predizer HP nos pacientes portadores de doenças intersticiais pulmonares. TIPO DE ESTUDO E LOCAL: Estudo transversal na Universidade de Brasília e Hospital de Base do Distrito Federal. MÉTODOS: Os sons cardíacos foram adquiridos com estetoscópio eletrônico e analisados por fonocardiografia. Os sinais clínicos sugestivos de HP, como B2 mais intensamente audível em área pulmonar que aórtica, P2 > A2 na área pulmonar e P2 presente em área mitral foram confrontados com parâmetros cardiográficos no exame de Doppler sugestivos de HP. Sensibilidade (S), especificidade (E), razões de verossimilhança positiva (RV+) e negativa (RV-) foram avaliados. RESULTADOS: Não houve correlação significativa entre amplitude de B2 e P2 e a PSAP (pressão sistólica arterial pulmonar) (P = 0,185 e 0,115; P = 0,13 e 0,34; respectivamente). A análise da presença de B2 mais intensa na área pulmonar que aórtica, quando comparada a todos os critérios sugestivos de HP, mostrou S = 60%; E = 22%; RV+ = 0,7; RV- = 1,7; enquanto P2 > A2 mostrou: S = 57%; E = 39%; RV+ = 0,9; RV- = 1,1; e P2 no foco mitral mostrou: S = 68%; E = 41%; RV+ = 1,1; RV- = 0,7. Todos os sinais juntos mostraram S = 50%; E = 56%. CONCLUSÃO: Os sinais semiológicos indicativos de HP apresentam baixos valores de especificidade e sensibilidade para diagnóstico clínico dessa comorbidade.

Usefulness of the second heart sound for predicting pulmonary hypertension in patients with interstitial lung disease.

CONTEXT AND OBJECTIVE: P2 hyperphonesis is considered to be a valuable finding in semiological diagnoses of pulmonary hypertension (PH). The aim here was to evaluate the accuracy of the pulmonary component of second heart sounds for predicting PH in patients with interstitial lung disease. DESIGN AND SETTING: Cross-sectional study at the University of Brasilia and Hospital de Base do Distrito Federal. METHODS: Heart sounds were acquired using an electronic stethoscope and were analyzed using phonocardiography. Clinical signs suggestive of PH, such as second heart sound (S2) in pulmonary area louder than in aortic area; P2 > A2 in pulmonary area and P2 present in mitral area, were compared with Doppler echocardiographic parameters suggestive of PH. Sensitivity (S), specificity (Sp) and positive (LR+) and negative (LR-) likelihood ratios were evaluated. RESULTS: There was no significant correlation between S2 or P2 amplitude and PASP (pulmonary artery systolic pressure) (P = 0.185 and 0.115; P= 0.13 and 0.34, respectively). Higher S2 in pulmonary area than in aortic area, compared with all the criteria suggestive of PH, showed S = 60%, Sp= 22%; LR+ = 0.7; LR- = 1.7; while P2> A2 showed S= 57%, Sp = 39%; LR+ = 0.9; LR- = 1.1; and P2 in mitral area showed: S= 68%, Sp = 41%; LR+ = 1.1; LR- = 0.7. All these signals together showed: S= 50%, Sp = 56%. CONCLUSIONS: The semiological signs indicative of PH presented low sensitivity and specificity levels for clinically diagnosing this comorbidity.

Comparison between users of a new methodology for heart sound auscultation.

Auscultation is a routine exam and the first line of screening in heart pathologies. The objective of this study was to assess if using a new data collection system, the DigiScope Collector, with a guided and automatic annotation of heart auscultation, different levels of expertise/experience users could collect similar digital auscultations. Data were collected within the Heart Caravan Initiative (Paraíba, Brasil). Patients were divided into two study groups: Group 1 evaluated by a third year medical student (User 1), and an experienced nurse (User 2); Group 2 evaluated by User 2 and an Information Technology professional (User 3). Patients were auscultated sequentially by the two users, according to the randomization. Features extracted from each data set included the length (HR) of the audio files, the number of repetitions per auscultation area, heart rate, first (S1) and second (S2) heart sound amplitudes, S2/S1, and aortic (A2) and pulmonary (P2) components of the second heart sound and relative amplitudes (P2/A2). Features extracted were compared between users using paired-sample test Wilcoxon test, and Spearman correlations (P<;0.05 considered significant). Twenty-seven patients were included in the study (13 Group 1, and 14 Group 2). No statistical significant differences were found between groups, except in the time of auscultation (User 2 consistently presented longer auscultation time). Correlation analysis showed significant correlations between extracted features from both groups: S2/S1 in Group 1, and S1, S2, A2, P2, P2/A2 amplitudes, and HR in Group 2. Using the DigiScope Collector, we were able to collect similar digital auscultations, according to the features evaluated. This may indicate that in sites with limited access to specialized clinical care, auscultation files may be acquired and used in telemedicine for an expert evaluation.

The effect of data exchange protocols on decision support systems for heart sounds.

Heart auscultation is one of the basic exams performed during a patient physical examination, but it is also one that has a high skill ceiling. Decision support systems can provide physicians with a tool that can help to reduce the demanding skill requirements of this exam. Nevertheless, this second opinion needs to be delivered in a timely interval in order to be truly useful for a physician. To do this we need not only optimized algorithms, but also a well designed system. In this paper, we have studied how two different data exchange protocols, that define how data should be transferred from an acquisition to a process module, can impact the celerity of delivering second opinion to a physician. With data collected from real exams, acquired in a field hospital initiative in Brazil, we recreated two use cases that allowed us to measure performance in the form of time and resources spent, as well as power consumption. Results have shown that different data exchange protocols can have a significant impact on a decision support system response time.

Segmentação automática para classificação digital de sinais de fonocardiograma / Automatic segmentation for signal classification of digital phonocardiogram

Com o avanço tecnológico surgem novas ferramentas que auxiliam os médicos no diagnóstico de diversas doenças. Na área cardiovascular, após permanecer por um longo período em segundo plano, a ausculta cardíaca voltou a ser muito utilizada devido ao surgimento, no mercado, de estetoscópios digitais. Tais aparelhos contam com novos recursos tecnológicos que permitem a captação e a análise de dados de forma automática, oferecendo mais informações ao profissional da área. Levando em conta essa nova ascensão da área de Fonocardiografia,o presente trabalho se dedicou à separação das bulhas S1 e S2 por meio de ferramentas computacionais, com o propósito de auxiliar médicos não especialistas em Cardiologia a verificar a existência de possíveis anormalidades no som cardíaco. Acreditando na possibilidade de este procedimento vir a ser utilizado posteriormente para auxiliar no reconhecimento de padrões dos sons cardíacos, este trabalho se propôs a criar um algoritmo para detecção automática de anormalidades que afetam as bulhas S1 e S2. Assim, aplicou-se a técnica de Wavelet sobre uma base de dados de sons cardíacos constituída de 1209 bulhas...

Cardiac auscultation simulator embedded in virtual learning environment to support medical teaching.

Heart auscultation remains as an important clinical tool in heart disease diagnostics. A prototype was developed, using an existing database of clinical and cardiac sounds of pregnant women [1]. The context for virtual learning is cardiac disease during pregnancy (Partum and Post Partum situations). Based on real histories, cardiac sounds, and complementary exams, this pilot presents the most frequent diagnosis of cardiac diseases during pregnancy.

Assessment of time-frequency representation techniques for thoracic sounds analysis.

A step forward in the knowledge about the underlying physiological phenomena of thoracic sounds requires a reliable estimate of their time-frequency behavior that overcomes the disadvantages of the conventional spectrogram. A more detailed time-frequency representation could lead to a better feature extraction for diseases classification and stratification purposes, among others. In this respect, the aim of this study was to look for an omnibus technique to obtain the time-frequency representation (TFR) of thoracic sounds by comparing generic goodness-of-fit criteria in different simulated thoracic sounds scenarios. The performance of ten TFRs for heart, normal tracheal and adventitious lung sounds was assessed using time-frequency patterns obtained by mathematical functions of the thoracic sounds. To find the best TFR performance measures, such as the 2D local (ρ(mean)) and global (ρ) central correlation, the normalized root-mean-square error (NRMSE), the cross-correlation coefficient (ρ(IF)) and the time-frequency resolution (res(TF)) were used. Simulation results pointed out that the Hilbert-Huang spectrum (HHS) had a superior performance as compared with other techniques and then, it can be considered as a reliable TFR for thoracic sounds. Furthermore, the goodness of HHS was assessed using noisy simulated signals. Additionally, HHS was applied to first and second heart sounds taken from a young healthy male subject, to tracheal sound from a middle-age healthy male subject, and to abnormal lung sounds acquired from a male patient with diffuse interstitial pneumonia. It is expected that the results of this research could be used to obtain a better signature of thoracic sounds for pattern recognition purpose, among other tasks.

Time-Frequency Representations for second heart sound analysis.

Several researches have tried to provide a means to analyze the second heart sound (S2) in an attempt to understand the functional mechanisms in its genesis and for diagnosis purposes. In this work we tested Time-Frequency Representation (TFR) for simulated S2 selecting and applying classical and modern TFRs such as the Spectrogram, the Wigner-Ville Distribution, the Time Varying Autoregressive (TVAR) model, the Scalogram, and the Hilbert-Huang Spectrum (HHS) by Empirical Mode Decomposition. Two performance measures are proposed, the first one based on local 2D correlations (rho) between the ideal and the estimated TFRs images, while the second one based on time moments of the TFR images to provide the normalized root-mean-square error (NRMSE). Under no noise conditions, the TFRs by HHS and the TVAR modeling, by the Burg algorithm, resulted in a rho(average) of 0.788 and 0.812, and NRMSE of 0.172 and 0.195, respectively. Therefore, based on the lowest NRMSE, HHS was considered the TFR with the best performance. Afterward, HHS was applied to real S2 acquired at the aortic and pulmonary focal points.

Development of a tele-stethoscope and its application in pediatric cardiology.

Over the years, many attempts have been made to develop special stethoscopes for the teaching of auscultation. The objective of this article is to report on the experience with the development and implementation of an electronic stethoscope and a virtual library of cardiac sounds. There were four stages to this project: (1) the building of the prototype to acquire, filter and amplify the cardiac sounds, (2) the development of a software program to record, reproduce and visualize them, (3) the testing of the prototype in a clinical scenario, and (4) the development of an internet site, to store and display the sounds collected. The first two stages are now complete. The prototype underwent an initial evaluation in a clinical scenario within the Unit and during virtual out-patient clinical sessions. One hundred auscultations were recorded during these tests. They were reviewed and discussed on-line by a panel of experience cardiologists during the sessions. Although the sounds were considered "satisfactory" for diagnostic purposes by the cardiology team, they identified some qualitative differences in the electronic recorded auscultations, such as a higher pitch of the recorded sounds. Prospective clinical studies are now being conducted to further evaluate the interference of the electronic device in the physicians' capability to diagnose different cardiac conditions. An internet site (www.caduceusvirtual.com.br/ auscultaped) was developed to host these cardiac auscultations. It is set as a library of cardiac sounds, catalogued by pathologies and already contains examples from auscultations of the majority of common congenital heart lesions, such as septal defects and valvar lesions.

Competency in cardiac examination skills in medical students, trainees, physicians, and faculty: a multicenter study.

BACKGROUND: Cardiac examination is an essential aspect of the physical examination. Previous studies have shown poor diagnostic accuracy, but most used audio recordings, precluding correlation with visible observations. The training spectrum from medical students (MSs) to faculty has not been tested, to our knowledge. METHODS: A validated 50-question, computer-based test was used to assess 4 aspects of cardiac examination competency: (1) cardiac physiology knowledge, (2) auditory skills, (3) visual skills, and (4) integration of auditory and visual skills using computer graphic animations and virtual patient examinations (actual patients filmed at the bedside). We tested 860 participants: 318 MSs, 289 residents (225 internal medicine and 64 family medicine), 85 cardiology fellows, 131 physicians (50 full-time faculty, 12 volunteer clinical faculty, and 69 private practitioners), and 37 others. RESULTS: Mean scores improved from MS1-2 to MS3-4 (P = .003) but did not improve or differ significantly among MS3, MS4, internal medicine residents, family medicine residents, full-time faculty, volunteer clinical faculty, and private practitioners. Only cardiology fellows tested significantly better (P<.001), and they were the best in all 4 subcategories of competency, whereas MS1-2 were the worst in the auditory and visual subcategories. Participants demonstrated low specificity for systolic murmurs (0.35) and low sensitivity for diastolic murmurs (0.49). CONCLUSIONS: Cardiac examination skills do not improve after MS3 and may decline after years in practice, which has important implications for medical decision making, patient safety, cost-effective care, and continuing medical education. Improvement in cardiac examination competency will require training in simultaneous audio and visual examination in faculty and trainees.