Artificial intelligence–based screening for cardiomyopathy in an obstetric population: A pilot study

Background Cardiomyopathy is a leading cause of pregnancy-related mortality and represents the number one cause of death in the late postpartum period. Purpose The goal of this study was to prospectively evaluate the effectiveness of an artificial intelligence-based electrocardiogram (AI-ECG) and a point-of-care artificial intelligence-enhanced digital stethoscope for cardiomyopathy screening among pregnant and postpartum women. Methods We conducted a single-arm prospective study of pregnant and postpartum women (up to 12 months following delivery) enrolled at three sites between October 2021 and October 2022. Study participants completed a standard 12-lead ECG, ECG + phonocardiogram recordings with a digital stethoscope, a demographic questionnaire, and a transthoracic echocardiogram (TTE) within a 24-hour period. Diagnostic performance of the AI-ECG and digital stethoscope for cardiomyopathy detection, defined as left ventricular ejection fraction (LVEF) < 45%, was evaluated using the area under the curve (AUC) and standard measures of diagnostic performance. Results We included 100 consecutive pregnant and postpartum women who provided informed consent and had at least one ECG recorded, completed the questionnaire, and had an echocardiogram performed. All participants self-identified as female, and 41% were recruited from a federally qualified health center. The median age of participants was 31 years (Q1: 27, Q3: 34). Thirty-eight percent identified as non-Hispanic White, 32% as non-Hispanic Black, 21% as Hispanic or Latino, 6% as Asian, and 3% as Other (1 Native Hawaiian and 2 multiracial). Five percent (5/100) had LVEF <45%. The 12-lead AI-ECG model demonstrated close to perfect discrimination (accuracy 99%; AUC 1.00, sensitivity 100%) for detection of cardiomyopathy with 1 false positive screen (LVEF 49%), Figure 1A. The digital stethoscope (angled position) had an AUC of 0.99 (95% CI: 0.97, 1.00) for detection of LVEF <45%. Sensitivity, specificity, positive predictive value, and negative predictive value for the digital stethoscope were 60%, 98%, 60% and 98%, respectively for detection of LVEF <45% (Figure 1B & 1C). Conclusions In this small prospective study, we found that 5% of pregnant and postpartum women had cardiomyopathy, a potentially life-threatening condition, and that a rapid, inexpensive, ubiquitous, massively scalable point-of-care test - the ECG - showed promising results, identifying it with an AUC > 0.98 with the application of artificial intelligence.Figure 1

Background: Prior risk models in patients with heart failure (HF) have focused on hospitalizations for worsening HF (WHF) and have not evaluated for differences in predictors by left ventricular ejection fraction (LVEF). We used natural language processing (NLP) and machine learning methods with access to longitudinal electronic health record (EHR) data to develop risk prediction models for WHF events across practice settings and by LVEF category. Methods: We identified all adults with HF and known LVEF on January 1 st of each year from 2011-2019 in an integrated health care system. WHF events within 1 year were defined as any hospitalization, emergency department, or outpatient encounter with ≥1 symptom, ≥2 objective findings including ≥1 sign, and ≥1 change in HF-related therapy. Signs and symptoms were ascertained using rule-based NLP. We conducted boosted decision tree-based ensemble models for any WHF event within each LVEF category: HF with reduced EF (HFrEF; LVEF ≤40%), HF with mildly reduced EF (HFmrEF; LVEF 41-49%), and HF with preserved EF (HFpEF; LVEF ≥50%). We evaluated model discrimination using area under the curve (AUC) and model calibration using Brier scores. Results: Among 359,298 patients from 2011-2019, 65,838 (18%) had HFrEF, 52,491 (15%) had HFmrEF, and 240,969 (67%) had HFpEF. Mean age was 75±12, 47% were women, and 37% were minorities including 10% Black, 11% Asian/Pacific Islander, and 12% of Hispanic ethnicity. WHF events occurred in 22% of patients with HFrEF, 17% with HFmrEF, and 16% with HFpEF. The models displayed an AUC of 0.75 and Brier score of 0.15 for HFrEF and an AUC of 0.77 and Brier scores of 0.12 for both HFmrEF and HFpEF. Clinical predictors were similar across LVEF categories ( Table ). Conclusions: Longitudinal EHR data can be leveraged using NLP and machine learning for accurate risk estimation that reliably identifies clinical predictors across a range of LVEF. These findings may provide novel insight into the natural history of HF.

Background: The regulatory effect of the left ventricular ejection fraction (LVEF) categories on the association of malnutrition and all-cause mortality in patients undergoing coronary angiography (CAG) have not been adequately addressed.Methods: Forty-five thousand eight hundred and twenty-six patients consecutively enrolled in the Cardiorenal ImprovemeNt (CIN) study (ClinicalTrials.gov NCT04407936) from January 2008 to July 2018 who underwent coronary angiography (CAG). The Controlling Nutritional Status (CONUT) score was applied to 45,826 CAG patients. The hazard ratios of mortality across combined LVEF and/or malnutrition categories were estimated by Cox regression models. Variables adjusted for in the Cox regression models included: age, gender, hypertension (HT), DM, PCI, coronary artery disease (CAD), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglyceride (TRIG), chronic kidney disease (CKD), statins, atrial fibrillation (AF), anemia, and stroke. Population attributable risk (PAR) was estimated for eight groups stratified by nutritional status and LVEF categories.Results: In our study, 42,181(92%) of patients were LVEF ≥ 40%, of whom, 41.55 and 9.34% were in mild and moderate or severe malnutrition status, respectively, while 46.53 and 22.28% in mild and moderate or severe malnutritional status among patients with LVEF < 40%. During a median follow-up time of 4.5 years (percentile 2.8–7.1), 5,350 (11.7%) patients died. After fully adjustment, there is no difference of mortality on malnutrition in LVEF < 40% group (mild, moderate and severe vs. normal, HR (95%CI): [1.00 (0.83–0.98)], [1.20 (0.95–1.51)], [1.41 (0.87–2.29)], respectively, p for trend =0.068), but malnutrition was related to markedly increased risk of mortality in LVEF ≥ 40% group (mild, moderate, and severe vs. normal, HR (95%CI): [1.21 (1.12–1.31)], [1.56 (1.40–1.74)], and [2.20(1.67–2.90)], respectively, p for trend < 0.001, and p for interaction < 0.001). Patients with LVEF ≥ 40% had a higher malnutrition-associated risk of mortality and a higher PAR than those with LVEF < 40%.Conclusions: Malnutrition is common in CAG patients and it has a greater effect on all-cause mortality and a higher PAR in patients with LVEF ≥ 40% than LVEF < 40%.

Objectives: Emergency department (ED) bedside echocardiography may offer useful information on cardiac function and volume status. The authors evaluated the accuracy of emergency physician (EP) performance of echocardiography in the assessment of left ventricular ejection fraction (LVEF) and central venous pressure (CVP). Methods: The authors conducted a cross‐sectional observational study at an urban teaching ED, involving a convenience sample of patients presenting to the ED between September 2000 and February 2001. Level III–credentialed EP sonographers who had undergone a three‐hour training session in limited echocardiography, focusing on LVEF and CVP measurement, performed echocardiograms. Vital signs and indication for echocardiography were documented on a study data sheet. LVEF was rated as poor (&lt;30%), moderate (30%–55%), or normal (&gt;55%) and an absolute percentage. Central venous pressure categories included low (&lt;5 cm), moderate (5–10 cm), and high (&gt;10 cm). Formal echocardiograms were obtained within a four‐hour window on all patients and interpreted by a staff cardiologist. Correlation analysis was performed using the κ correlation coefficient for LVEF and CVP categories and a Pearson correlation coefficient for LVEF measurement. Results: A total of 115 patients were assessed for LVEF, and 94 patients had complete information for CVP. Indications for echocardiography included chest pain (45.1%), congestive heart failure (38.1%), dyspnea (5.7%), and endocarditis (10.6%). Results showed a LVEF correlation of r2= 0.712 with 86.1% overall agreement. Subgroup analysis revealed the highest agreement (92.3%) between EP and formal echocardiograms within the normal LVEF category, followed by 70.4% agreement in the poor LVEF category and 47.8% in the moderate LVEF category. Central venous pressure measurements resulted in 70.2% overall raw agreement between EP and formal echocardiograms. Subgroup analysis revealed the highest agreement (83.3%) within the high CVP category followed by 66.6% in the moderate and 20% in the low categories. Conclusions: Experienced EP sonographers with a small amount of focused additional training in limited bedside echocardiography can assess LVEF accurately in the ED.

Emergency department (ED) bedside echocardiography may offer useful information on cardiac function and volume status. The authors evaluated the accuracy of emergency physician (EP) performance of echocardiography in the assessment of left ventricular ejection fraction (LVEF) and central venous pressure (CVP). The authors conducted a cross-sectional observational study at an urban teaching ED, involving a convenience sample of patients presenting to the ED between September 2000 and February 2001. Level III-credentialed EP sonographers who had undergone a three-hour training session in limited echocardiography, focusing on LVEF and CVP measurement, performed echocardiograms. Vital signs and indication for echocardiography were documented on a study data sheet. LVEF was rated as poor (<30%), moderate (30%-55%), or normal (>55%) and an absolute percentage. Central venous pressure categories included low (<5 cm), moderate (5-10 cm), and high (>10 cm). Formal echocardiograms were obtained within a four-hour window on all patients and interpreted by a staff cardiologist. Correlation analysis was performed using the kappa correlation coefficient for LVEF and CVP categories and a Pearson correlation coefficient for LVEF measurement. A total of 115 patients were assessed for LVEF, and 94 patients had complete information for CVP. Indications for echocardiography included chest pain (45.1%), congestive heart failure (38.1%), dyspnea (5.7%), and endocarditis (10.6%). Results showed a LVEF correlation of r(2) = 0.712 with 86.1% overall agreement. Subgroup analysis revealed the highest agreement (92.3%) between EP and formal echocardiograms within the normal LVEF category, followed by 70.4% agreement in the poor LVEF category and 47.8% in the moderate LVEF category. Central venous pressure measurements resulted in 70.2% overall raw agreement between EP and formal echocardiograms. Subgroup analysis revealed the highest agreement (83.3%) within the high CVP category followed by 66.6% in the moderate and 20% in the low categories. Experienced EP sonographers with a small amount of focused additional training in limited bedside echocardiography can assess LVEF accurately in the ED.

Incremental value of high-risk CMR attributes to predict adverse remodeling after ST-segment elevation myocardial infarction across LVEF categories

Background There is a paucity of data on the clinical characteristics, management, and outcomes of women compared with men with heart failure, especially in low-and middle-income countries. Purpose In this study from the Global Congestive Heart Failure registry, we examined sex differences in risk factors, clinical characteristics, treatments, and the risk of heart failure hospitalization and mortality by country economic status and by left ventricular ejection fraction (LVEF) category. Methods In the prospective Global-Congestive Heart Failure (G-CHF) study, participants with established heart failure were considered for inclusion from 40 high-, middle-, and low-income countries. We recorded information on the demographic characteristics, medical history, and treatments of participants. We report data on heart failure hospitalization and mortality by sex overall, by country economic status, and by LVEF category. Results From December 2016 to July 2022, 23,000 participants were recruited and followed up. The average age of women in the study was 62 years compared to 64 years in men. Fewer women than men had a LVEF ≤40 (51.7% women vs 66.3% men). By contrast, more women than men had an LVEF≥ 50 (33.2% women vs 18.6% men). Hypertensive heart failure was the most common etiology in women (25.5% women vs 16.8% men), and ischemic heart failure was the most common etiology in men (45.6% men vs 26.6% women). Signs and symptoms of heart failure were more common in women than men: 42.6% of women were NYHA functional class III/IV compared to 37.9% of men. Among participants with a LVEF &amp;lt;35, use of implantable cardiac defibrillator (ICD) was lower in women than men in the overall study (8.7% women vs 17.2% men), and within countries categorized by economic status. However, the use of heart failure medicines and cardiac tests did not differ systematically by sex. Differences by sex were not observed in the adjusted risk of heart failure hospitalization overall (women-to-men adjusted hazard ratio 1.0 (95% CI 0.94 to 1.07)). This pattern was consistent within countries categorized by economic status, geographic region, and by LVEF category. Women had a lower adjusted risk of mortality overall (women-to-men adjusted hazard ratio 0.82 (95% CI 0.77 to 0.87), which was consistently observed within countries categories by economic status, geographic region, and LVEF levels. Conclusions There are differences between women and men in heart failure symptoms, underlying causes, ejection fraction categories, and use of an ICD. However, use heart failure medications, cardiac tests, and hospitalization were similar in women and men; but women had a lower risk of death.

Prognosis of Low Normal Left Ventricular Ejection Fraction in an Asymptomatic Population-Based Adult Cohort: The Multiethnic Study of Atherosclerosis

Background The multiorgan involvement of heart failure is similar to the spread observed in cancer. We proposed a new score, named HLM, analogous to TNM classification used in oncology. HLM refers to (i) H: heart damage, instead of “T” for tumor; (ii) L: lung involvement, instead of “N” for lymph nodes; (iii) M: systemic multiorgan involvement, instead of “M” for metastasis. Objectives to compare HLM score to NYHA class, ACC/AHA stages and left ventricular ejection fraction (LVEF) classification, to assess the most accurate prognostic tool for HF patients. Methods We performed a multicentric, observational, prospective study of consecutive patients admitted for HF, or at risk for HF. Parameters for heart, lungs and organs’ function were collected. Each patient was classified according to HLM, NYHA, ACC/AHA stages and LVEF classification. Patients were followed up for 12 months. The primary composite endpoint was all-cause death and rehospitalization due to HF. Results We enrolled 1720 patients who completed the 12-month follow-up. As shown by Kaplan Meier curves, HLM was the most accurate score to predict primary endpoint at 12- month. The area under the ROC curve (AUC) was greater for HLM score than NYHA, ACC/AHA stages and LVEF classification, regarding the composite endpoint (HLM=0.645; NYHA=0.580; ACC/AHA=0.589; LVEF=0.572) (Figure 1). HLM score related AUC showed statistically significant differences compared to LVEF (p=0.002), ACC/AHA (p=0.029) and NYHA (p=0.009). Conclusions HLM score has a greater prognostic power compared to other nosologies, in terms of the composite endpoint of all-cause death and rehospitalization due to HF, at 12-month follow-up. Key words: heart failure; HLM score; prognosis; mortality; rehospitalization. Figure 1.

Background: Delays in follow-up after breast cancer screening are thought to contribute to disparities in breast cancer outcomes. The primary objective of this study is to determine the impact of race/ethnicity and type of health insurance on the diagnostic delay time, defined as the number of days from abnormal screening to definitive diagnosis. Methods: This is a retrospective study of 976 women examined for breast cancer between 1998 and 2009 at six hospitals and clinics located in the District of Columbia. We used a full-factorial ANOVA model to test for significant differences in diagnostic delay time among non-Hispanic white (NHW), non-Hispanic black (NHB), and Hispanic women with private, government, or no health insurance. A log transformation was taken on the diagnostic delay time to normalize our data, and geometric means were estimated and compared. Results: The average geometric mean (95% CI) diagnostic delay times were as follows: among those with private insurance, 15.9 (12.2,20.6) days for NHW, 27.0 (22.4,32.6) days for NHB, and 51.4 (34.8,76.0) days for Hispanic women; among those with government insurance, 11.9 (7.3,19.3) days for NHW, 39.5 (32.2,48.6) days for NHB, and 71.6 (47.8,107.1) days for Hispanic women; and among those without insurance, 44.5 (16.4,120.6) days for NHW, 59.7 (38.8,91.8) days for NHB, and 66.4 (55.8,79.1) days for Hispanic women. In fitting a full-factorial ANOVA model, we found that NHW women with government insurance had a significantly shorter delay in diagnosis than NHB (p=0.0003) and Hispanic (p&amp;lt;0.0001) women with government insurance. We also found that NHW women with private insurance had a significantly shorter delay in diagnosis than NHB (p=0.03) and Hispanic (p&amp;lt;0.0001) women with private insurance. However, there were no significant differences within the uninsured women (p&amp;gt;0.05). Finally, we found that NHB women with private insurance had a significantly shorter delay in diagnosis than uninsured NHB women (p=0.03). Conclusions: NHB and Hispanic women with government or private insurance waited more than twice as long to reach their definitive diagnosis than NHW women with government or private insurance. Uninsured NHB women waited more than twice as long to reach their definitive diagnosis than NHB women with private insurance. Having private health insurance markedly increased the speed of diagnostic resolution in NHB women; however, the speed of diagnostic resolution remained significantly longer for NHB women with private insurance than for NHW women with private insurance. These results suggest that while both insurance and race/ethnicity affect diagnostic resolution, health insurance may not be the primary barrier to optimal diagnostic resolution in NHB women. It will be important to determine what other factors serve as the primary barriers, as well as if these delays affect the final breast cancer outcome for the patients. Funding Mechanism: Grant Number 1 U01 CA116937; Patient Navigation Research Program (PNRP), Center for Research on Cancer Health Disparities (CRCHD), National Cancer Institute (NCI). Citation Information: Cancer Epidemiol Biomarkers Prev 2010;19(10 Suppl):PR-10.

Outcome of Percutaneous Coronary Intervention Utilizing Drug-Eluting Stents in Patients With Reduced Left Ventricular Ejection Fraction

Background The current guidelines recommend different medical treatment strategies for heart failure (HF) patients according to category of left ventricular ejection fraction (LVEF). Angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACE-I/ARB) is an established medical treatment for heart failure with reduced ejection fraction (HFrEF), whereas its usefulness remains to be elucidated for non-HFrEF, especially for heart failure with mid-range ejection fraction (HFmrEF). Purpose This study aimed to assess the difference in association between ACE-I/ARB and clinical outcomes depending on LVEF category. Methods The Kyoto Congestive Heart Failure (KCHF) Registry is a multicentre registry without any exclusion criteria which included consecutive patients hospitalized for congestive HF in Japan. In each LVEF group (HFrEF, HFmrEF and heart failure with preserved ejection fraction [HFpEF]), we compared those who were prescribed ACE-I/ARB as discharge medication and those not, and assessed their 1-year clinical outcomes. We defined the primary outcome measure as a composite of all-cause death and HF hospitalization. We constructed a multivariable Cox regression model incorporating 24 clinically relevant factors. We assessed adjusted hazard ratios (HRs) of those with ACE-I/ARB relative to those not, and also interaction between ACE-I/ARB prescription at discharge and LVEF category. Results A total of 3717 patients were included in this study, where the number of patients in each LVEF group were as follows; 1383 patients with HFrEF, 703 with HFmrEF and 1631 with HFpEF, respectively (Figure). As shown in the table, the HRs for the primary outcome measure were significant in the HFrEF and HFmrEF groups, whereas the HR in the HFpEF group was insignificant. The interaction between ACE-I/ARB prescription and LVEF category for the primary outcome measure was statistically significant. Hazard ratios by LVEF category Outcome measures HFrEF HFmrEF HFpEF P interaction HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value All-cause death + HF hospitalization 0.66 (0.54–0.79) &lt;0.001 0.61 (0.45–0.82) 0.001 0.95 (0.80–1.14) 0.61 0.01 All-cause death 0.62 (0.48–0.81) &lt;0.001 0.52 (0.35–0.77) 0.001 0.73 (0.58–0.93) 0.01 0.10 HF hospitalization 0.73 (0.57–0.92) 0.009 0.59 (0.40–0.87) 0.007 1.14 (0.90–1.44) 0.28 0.07 Hazard ratios of ACE-I/ARB relative to non-ACE-I/ARB for primary outcome measures in each LVEF category. Study flowchart Conclusions The risk ratios of those who were prescribed ACE-I/ARB relative to those not were significantly low in HFmrEF as well as HFrEF, whereas the risk ratios were insignificant in HFpEF. ACE-I/ARB could be a potential choice of treatment for HFmrEF patients.

Background: the role of echocardiography in septic shock remains controversial, since depressed cardiac afterload may overestimate left ventricular (LV) systolic performance and mask septic cardiomyopathy (SC). We hypothesized that afterload-adjusted LV ejection fraction (LVEF) and LV outflow tract velocity-time integral (VTI) values for given systemic vascular resistances (SVR) could provide novel insights into recognizing and stratifying the severity of SC. Methods: in this observational, monocentric study, we prospectively included 14 mechanically-ventilated patients under septic-shock who all had a Pulse index Continuous Cardiac Output (PiCCO) system in place for hemodynamic monitoring. Echocardiographic and PiCCO longitudinal examinations (71 measurements overall) were performed simultaneously at the onset of septic shock and every 12 h for 60 h overall. Results: VTI-derived stroke volume (SV) and cardiac output (CO) were significantly correlated with PiCCO measurements (r ≥ 0.993, both p < 0.001). LVEF and VTI showed linear and exponential inverse correlation to SVR (R2 = 0.183 vs. 0.507 and p < 0.001 vs. p < 0.001, respectively). The equations LVEF = 86.168 − 0.011 × SVR and VTI = 41.23 × e(−0.0005×SVR) were found to provide “predicted” values for given SVR. Measured to predicted LVEF ratios (for given SVR), the afterload-adjusted LVEF defined the severity of SC (mild ≥ 90%, 80% ≤ moderate < 90% and severe < 80%). Mild SC demonstrated normal/supra-normal LVEF, normal VTI and SVR. Moderate SC showed lower LVEF and SVR, yet increased LV end-diastolic volume (LVEDV), VTI, SV and CO compared with mild SC (all p < 0.05). Severe SC was distinguished from moderate SC by markedly reduced LVEF, LVEDV, VTI, SV, CO and significantly increased SVR (all p < 0.05). LVEF and VTI decreased over time in mild SC, LVEF decreased in moderate SC, and LVEF and VTI increased over time in severe SC (p ≤ 0.038). LVEF and VTI demonstrated significant performance in identifying severe SC [cut-off < 61.5%, area under the curve (AUC) = 1 ± 0.0, sensitivity/specificity = 100/100, p < 0.001 vs. cut-off < 17.9 cm, AUC = 0.882 ± 0.042, sensitivity/specificity = 80/77, p < 0.001, respectively]. VTI but not LVEF demonstrated significant diagnostic performance in identifying both SVR < 800 dynes·s·cm−5 and SVR > 1500 dynes·s·cm−5 (cut-off > 24.46 cm, AUC = 0.889 ± 0.049, sensitivity/specificity = 75/100, p < 0.001; cut-off < 16.8, AUC = 0.0.857 ± 0.082, sensitivity/specificity = 83/86, p = 0.002, respectively).Conclusions: our study suggests that ICU bedside echocardiographic assessment of LVEF, VTI and their adjusted to corresponding SVR values provides valuable insights for the comprehension of SC phenotypes, underlying vasoplegia and cardiac output fluctuations in septic shock.

Focused cardiac ultrasound (FoCUS) is becoming standard practice in a wide spectrum of clinical settings. There is limited data evaluating the real-world use of FoCUS with artificial intelligence (AI). Our objective was to determine the accuracy of FoCUS AI-assisted left ventricular ejection fraction (LVEF) assessment and compare its accuracy between novice and experienced users. In this prospective, multicentre study, participants requiring a transthoracic echocardiogram (TTE) were recruited to have a FoCUS done by a novice or experienced user. The AI-assisted device calculated LVEF at the bedside, which was subsequently compared to TTE. 449 participants were enrolled with 424 studies included in the final analysis. The overall intraclass coefficient was 0.904, and 0.921 in the novice (n = 208) and 0.845 in the experienced (n = 216) cohorts. There was a significant bias of 0.73% towards TTE (p = 0.005) with a level of agreement of 11.2%. Categorical grading of LVEF severity had excellent agreement to TTE (weighted kappa = 0.83). The area under the curve (AUC) was 0.98 for identifying an abnormal LVEF (<50%) with a sensitivity of 92.8%, specificity of 92.3%, negative predictive value (NPV) of 0.97 and a positive predictive value (PPV) of 0.83. In identifying severe dysfunction (<30%) the AUC was 0.99 with a sensitivity of 78.1%, specificity of 98.0%, NPV of 0.98 and PPV of 0.76. Here we report that FoCUS AI-assisted LVEF assessments provide highly reproducible LVEF estimations in comparison to formal TTE. This finding was consistent among senior and novice echocardiographers suggesting applicability in a variety of clinical settings.

Prognostic value of echocardiography for heart failure and death in adults with chronic kidney disease