Machine Learning-Based Critical Congenital Heart Disease Screening Using Dual-Site Pulse Oximetry Measurements.

BACKGROUND: Oxygen saturation (Spo2) screening has not led to earlier detection of critical congenital heart disease (CCHD). Adding pulse oximetry features (ie, perfusion data and radiofemoral pulse delay) may improve CCHD detection, especially coarctation of the aorta (CoA). We developed and tested a machine learning (ML) pulse oximetry algorithm to enhance CCHD detection. METHODS AND RESULTS: Six sites prospectively enrolled newborns with and without CCHD and recorded simultaneous pre- and postductal pulse oximetry. We focused on models at 1 versus 2 time points and with/without pulse delay for our ML algorithms. The sensitivity, specificity, and area under the receiver operating characteristic curve were compared between the Spo2-alone and ML algorithms. A total of 523 newborns were enrolled (no CHD, 317; CHD, 74; CCHD, 132, of whom 21 had isolated CoA). When applying the Spo2-alone algorithm to all patients, 26.2% of CCHD would be missed. We narrowed the sample to patients with both 2 time point measurements and pulse-delay data (no CHD, 65; CCHD, 14) to compare ML performance. Among these patients, sensitivity for CCHD detection increased with both the addition of pulse delay and a second time point. All ML models had 100% specificity. With a 2-time-points+pulse-delay model, CCHD sensitivity increased to 92.86% (P=0.25) compared with Spo2 alone (71.43%), and CoA increased to 66.67% (P=0.5) from 0. The area under the receiver operating characteristic curve for CCHD and CoA detection significantly improved (0.96 versus 0.83 for CCHD, 0.83 versus 0.48 for CoA; both P=0.03) using the 2-time-points+pulse-delay model compared with Spo2 alone. CONCLUSIONS: ML pulse oximetry that combines oxygenation, perfusion data, and pulse delay at 2 time points may improve detection of CCHD and CoA within 48 hours after birth. REGISTRATION: URL: https://www.clinicaltrials.gov/study/NCT04056104?term=NCT04056104&rank=1; Unique identifier: NCT04056104.

Prospective Evaluation of High Titer Autoantibodies and Fetal Home Monitoring in the Detection of Atrioventricular Block Among Anti-SSA/Ro Pregnancies.

OBJECTIVE: This prospective study of pregnant patients, Surveillance To Prevent AV Block Likely to Occur Quickly (STOP BLOQ), addresses the impact of anti-SSA/Ro titers and utility of ambulatory monitoring in the detection of fetal second-degree atrioventricular block (AVB). METHODS: Women with anti-SSA/Ro autoantibodies by commercial testing were stratified into high and low anti-52-kD and/or 60-kD SSA/Ro titers applying at-risk thresholds defined by previous evaluation of AVB pregnancies. The high-titer group performed fetal heart rate and rhythm monitoring (FHRM) thrice daily and weekly/biweekly echocardiography from 17-26 weeks. Abnormal FHRM prompted urgent echocardiography to identify AVB. RESULTS: Anti-52-kD and/or 60-kD SSA/Ro met thresholds for monitoring in 261 of 413 participants (63%); for those, AVB frequency was 3.8%. No cases occurred with low titers. The incidence of AVB increased with higher levels, reaching 7.7% for those in the top quartile for anti-60-kD SSA/Ro, which increased to 27.3% in those with a previous child who had AVB. Based on levels from 15 participants with paired samples from both an AVB and a non-AVB pregnancy, healthy pregnancies were not explained by decreased titers. FHRM was considered abnormal in 45 of 30,920 recordings, 10 confirmed AVB by urgent echocardiogram, 7 being second-degree AVB, all <12 hours from normal FHRM and within another 0.75 to 4 hours to echocardiogram. The one participant with second/third-degree and two participants with third-degree AVB were diagnosed by urgent echocardiogram >17 to 72 hours from an FHRM. Surveillance echocardiograms detected no AVB when the preceding interval FHRM recordings were normal. CONCLUSION: High-titer antibodies are associated with an increased incidence of AVB. Anti-SSA/Ro titers remain stable over time and do not explain the discordant recurrence rates, suggesting that other factors are required. Fetal heart rate and rhythm (FHRM) with results confirmed by a pediatric cardiologist reliably detects conduction abnormalities, which may reduce the need for serial echocardiograms.

Fetal Echocardiographic Z Score Pilot Project: Study Design and Impact of Gestational Age and Variable Type on Reproducibility of Measurements Within and Across Investigators.

BACKGROUND: Fetal echocardiography is widely available, but normative data are not robust. In this pilot study, the authors evaluated (1) the feasibility of prespecified measurements in a normal fetal echocardiogram to inform study design and (2) measurement variability to assign thresholds of clinical significance and guide analyses in larger fetal echocardiographic Z score initiatives. METHODS: Images from predefined gestational age groups (16-20, >20-24, >24-28, and >28-32 weeks) were retrospectively analyzed. Fetal echocardiography expert raters attended online group training and then independently analyzed 73 fetal studies (18 per age group) in a fully crossed design of 53 variables; each observer repeated measures for 12 fetuses. Kruskal-Wallis tests were used to compare measurements across centers and age groups. Coefficients of variation (CoVs) were calculated at the subject level for each measurement as the ratio of SD to mean. Intraclass correlation coefficients were used to show inter- and intrarater reliabilities. Cohen's d > 0.8 was used to define clinically important differences. Measurements were plotted against gestational age, biparietal diameter, and femur length. RESULTS: Expert raters completed each set of measurements in a mean of 23 ± 9 min/fetus. Missingness ranged from 0% to 29%. CoVs were similar across age groups for all variables (P < .05) except ductus arteriosus mean velocity and left ventricular ejection time, which were both higher at older gestational age. CoVs were >15% for right ventricular systolic and diastolic widths despite fair to good repeatability (intraclass correlation coefficient > 0.5); ductal velocities and two-dimensional measures, left ventricular short-axis dimensions, and isovolumic times all had high CoVs and high interobserver variability despite good to excellent intraobserver agreement (intraclass correlation coefficient > 0.6). CoVs did not improve when ratios (e.g., tricuspid/mitral annulus) were used instead of linear measurements. Overall, 27 variables had acceptable inter- and intraobserver repeatability, while 14 had excessive variability between readers despite good intraobserver agreement. CONCLUSIONS: There is considerable variability in fetal echocardiographic quantification in clinical practice that may affect the design of multicenter fetal echocardiographic Z score studies, and not all measurements may be feasible for standard normalization. As missingness was substantial, a prospective design will be needed. Data from this pilot study may aid in the calculation of sample sizes and inform thresholds for distinguishing clinically significant from statistically significant effects.

Congenitally Corrected Transposition of the Great Arteries: Fetal Diagnosis, Associations, and Postnatal Outcome: A Fetal Heart Society Research Collaborative Study.

Background Fetal diagnosis of congenitally corrected transposition of the great arteries (ccTGA) has been increasingly reported; however, predictors of clinical outcomes remain underexplored. We undertook a multicenter, retrospective study to investigate natural history, associated anomalies, and outcomes of fetal ccTGA. Methods and Results Fetuses with ccTGA diagnosed from January 2004 to July 2020 within 20 North American programs were included. Fetuses with severe ventricular hypoplasia thought to definitively preclude biventricular repair were excluded. We included 205 fetuses diagnosed with ccTGA at a median gestational age of 23 (interquartile range, 21-27) weeks. Genetic abnormalities were found in 5.9% tested, with extracardiac anomalies in 6.3%. Associated cardiac defects were diagnosed in 161 (78.5%), with atrioventricular block in 23 (11.3%). On serial fetal echocardiogram, 39% demonstrated a functional or anatomic change, most commonly increased tricuspid regurgitation (6.7%) or pulmonary outflow obstruction (11.1%). Of 194 fetuses with follow-up, 26 were terminated, 3 experienced fetal death (2 with atrioventricular block), and 165 were live-born. Of 158 with postnatal data (median follow-up 3.7 years), 10 (6.6%) had death/transplant before 1 year. On univariable analysis, fetal factors associated with fetal death or death/transplant by 1 year included ≥ mild tricuspid regurgitation, pulmonary atresia, aortic obstruction, fetal arrhythmia, and worsening hemodynamics on serial fetal echocardiogram (defined as worse right ventricular function, tricuspid regurgitation, or effusion). Conclusions Associated cardiac lesions and arrhythmias are common in fetal ccTGA, and functional changes commonly occur through gestation. Worse outcomes are associated with fetal tricuspid regurgitation (≥mild), any arrhythmia, pulmonary atresia, aortic obstruction, and worsening hemodynamics on serial echocardiograms. These findings can inform prenatal counseling and perinatal management planning.

A Measurement-Based Protocol Improves Interrater Agreement and Accuracy of Right Ventricular Systolic Pressure Assessment by Echocardiography in Children: A Call for Quality Improvement.

BACKGROUND: Echocardiographic measurements carry the promise of improving interrater (IR) agreement over subjective assessment. The aim of this study was to assess the effect of implementing a measurement-based protocol on IR agreement and accuracy in reporting of right ventricular (RV) systolic pressure in children. The effect of this reporting protocol on IR agreement in reporting RV dilation, hypertrophy, and systolic function was also evaluated. METHODS: Five echocardiography readers reported their assessment of RV systolic pressure, dilation, hypertrophy, and systolic function on 40 deidentified echocardiograms using their individual accustomed methods and then using an agreed-upon protocol based solely on RV measurements. IR agreement was assessed using κ statistics. Accuracy of RV systolic pressure ratings was assessed using the McNemar test in comparison with hemodynamic data obtained by cardiac catheterization. The reliability of the RV measurements was assessed using intraclass correlation coefficient (ICC) and coefficient of variation. RESULTS: IR agreement and accuracy of RV systolic pressure assessment improved after using the measurement-based protocol (agreement from 0.39 [95% CI, 0.27-0.5] to 0.62 [95% CI, 0.48-0.76] and accuracy from 18 of 40 to 29 of 40 [P = .03]). IR agreement of RV dilation improved (from 0.36 [95% CI, 0.25-0.48] to 0.63 [95% CI, 0.48-0.79]), while IR agreement of RV hypertrophy (from 0.29 [95% CI, 0.17-0.42] to 0.35 [95% CI, 0.15-0.55]) and RV systolic function (from 0.57 [95% CI, 0.45-0.69] to 0.53 [95% CI, 0.41-0.66]) did not improve. The reliability of the measurements was good (ICC > 0.8), except for RV free wall thickness (ICC = 0.62, coefficient of variation = 24%) and RV fractional area change (ICC = 0.47, coefficient of variation = 22%), suggesting a possible reason for the lack of improvement in IR agreement of RV hypertrophy and RV systolic function. Heteroscedasticity was observed in the reliability of RV measurements, with the ICC being significantly lower at larger magnitudes for all RV measurements. CONCLUSIONS: Standardization of reporting protocols using RV measurements in place of subjective assessment improved IR agreement and accuracy of RV systolic pressure assessment. Reliable measurements (RV systolic pressure and dilation) resulted in improvement in IR agreement while unreliable measurements (RV hypertrophy and systolic function) did not. Special attention to measurements' reliability and heteroscedasticity of reliability is required when designing clinical protocols to decrease IR disagreement as a source of error.

Enhanced Critical Congenital Cardiac Disease Screening by Combining Interpretable Machine Learning Algorithms.

Critical Congenital Heart Disease (CCHD) screening that only uses oxygen saturation (SpO2), measured by pulse oximetry, fails to detect an estimated 900 US newborns annually. The addition of other pulse oximetry features such as perfusion index (PIx), heart rate, pulse delay and photoplethysmography characteristics may improve detection of CCHD, especially those with systemic blood flow obstruction such as Coarctation of the Aorta (CoA). To comprehensively study the most relevant features associated with CCHD, we investigated interpretable machine learning (ML) algorithms by using Recursive Feature Elimination (RFE) to identify an optimal subset of features. We then incorporated the trained ML models into the current SpO2-alone screening algorithm. Our proposed enhanced CCHD screening system, which adds the ML model, improved sensitivity by approximately 10 percentage points compared to the current standard SpO2-alone method with minimal to no impact on specificity.Clinical relevance- This establishes proof of concept for a ML algorithm that combines pulse oximetry features to improve detection of CCHD with little impact on false positive rate.

Multi-Institutional Practice-Patterns in Fetal Congenital Heart Disease Following Implementation of a Standardized Clinical Assessment and Management Plan.

Background Prenatal diagnosis of congenital heart disease has been associated with early-term delivery and cesarean delivery (CD). We implemented a multi-institutional standardized clinical assessment and management plan (SCAMP) through the University of California Fetal-Maternal Consortium. Our objective was to decrease early-term (37-39 weeks) delivery and CD in pregnancies complicated by fetal congenital heart disease using a SCAMP methodology to improve practice in a high-risk and clinically complex setting. Methods and Results University of California Fetal-Maternal Consortium site-specific management decisions were queried following SCAMP implementation. This contemporary intervention group was compared with a University of California Fetal-Maternal Consortium historical cohort. Primary outcomes were early-term delivery and CD. A total of 496 maternal-fetal dyads with prenatally diagnosed congenital heart disease were identified, 185 and 311 in the historical and intervention cohorts, respectively. Recommendation for later delivery resulted in a later gestational age at delivery (38.9 versus 38.1 weeks, P=0.01). After adjusting for maternal age and site, historical controls were more likely to have a CD (odds ratio [OR],1.8; 95% CI, 2.1-2.8; P=0.004) and more likely (OR, 2.1; 95% CI, 1.4-3.3) to have an early-term delivery than the intervention group. Vaginal delivery was recommended in 77% of the cohort, resulting in 61% vaginal deliveries versus 50% in the control cohort (P=0.03). Among pregnancies with major cardiac lesions (n=373), vaginal birth increased from 51% to 64% (P=0.008) and deliveries ≥39 weeks increased from 33% to 48% (P=0.004). Conclusions Implementation of a SCAMP decreased the rate of early-term deliveries and CD for prenatal congenital heart disease. Development of clinical pathways may help standardize care, decrease maternal risk secondary to CD, improve neonatal outcomes, and reduce healthcare costs.

Fetal Cerebral Oxygenation Is Impaired in Congenital Heart Disease and Shows Variable Response to Maternal Hyperoxia.

Background Impairments in fetal oxygen delivery have been implicated in brain dysmaturation seen in congenital heart disease (CHD), suggesting a role for in utero transplacental oxygen therapy. We applied a novel imaging tool to quantify fetal cerebral oxygenation by measuring T2* decay. We compared T2* in fetuses with CHD with controls with a focus on cardiovascular physiologies (transposition or left-sided obstruction) and described the effect of brief administration of maternal hyperoxia on T2* decay. Methods and Results This is a prospective study performed on pregnant mothers with a prenatal diagnosis of CHD compared with controls in the third trimester. Participants underwent a fetal brain magnetic resonance imaging scan including a T2* sequence before and after maternal hyperoxia. Comparisons were made between control and CHD fetuses including subgroup analyses by cardiac physiology. Forty-four mothers (CHD=24, control=20) participated. Fetuses with CHD had lower total brain volume (238.2 mm3, 95% CI, 224.6-251.9) compared with controls (262.4 mm3, 95% CI, 245.0-279.8, P=0.04). T2* decay time was faster in CHD compared with controls (beta=-14.4, 95% CI, -23.3 to -5.6, P=0.002). The magnitude of change in T2* with maternal hyperoxia was higher in fetuses with transposition compared with controls (increase of 8.4 ms, 95% CI, 0.5-14.3, P=0.01), though between-subject variability was noted. Conclusions Cerebral tissue oxygenation is lower in fetuses with complex CHD. There was variability in the response to maternal hyperoxia by CHD subgroup that can be tested in future larger studies. Cardiovascular physiology is critical when designing neuroprotective clinical trials in the fetus with CHD.

Fetal Cardiac Intervention for Pulmonary Atresia with Intact Ventricular Septum: International Fetal Cardiac Intervention Registry.

INTRODUCTION: Invasive fetal cardiac intervention (FCI) for pulmonary atresia with intact ventricular septum (PAIVS) and critical pulmonary stenosis (PS) has been performed with small single-institution series reporting technical and physiological success. We present the first multicenter experience. OBJECTIVES: Describe fetal and maternal characteristics of those being evaluated for FCI, including pregnancy/neonatal outcome data using the International Fetal Cardiac Intervention Registry (IFCIR). METHODS: We queried the IFCIR for PAIVS/PS cases evaluated from January 2001 to April 2018 and reviewed maternal/fetal characteristics, procedural details, pregnancy and neonatal outcomes. Data were analyzed using standard descriptive statistics. RESULTS: Of the 84 maternal/fetal dyads in the registry, 58 underwent pulmonary valvuloplasty at a median gestational age of 26.1 (21.9-31.0) weeks. Characteristics of fetuses undergoing FCI varied in terms of tricuspid valve (TV) size, TV regurgitation, and pulmonary valve patency. There were fetal complications in 55% of cases, including 7 deaths and 2 delayed fetal losses. Among those who underwent successful FCI, the absolute measurement of the TV increased by 0.32 (±0.17) mm/week from intervention to birth. Among 60 liveborn with known outcome, there was a higher percentage having a biventricular circulation following successful FCI (87 vs. 43%). CONCLUSIONS: Our data suggest a possible benefit to fetal therapy for PAIVS/PS, though rates of technically unsuccessful procedures and procedure-related complications, including fetal loss were substantial. FCI criteria are extremely variable, making direct comparison to nonintervention patients challenging and potentially biased. More uniform FCI criteria for fetuses with PAIVS/PS are needed to avoid unnecessary procedures, expose only fetuses most likely to sustain a benefit, and to enable comparisons to be made with nonintervention patients.

Neurobehavioral evaluation of neonates with congenital heart disease: a cohort study.

AIM: To describe neurobehavioral patterns in neonates with congenital heart disease (CHD). METHOD: A cohort study describing neurobehavioral performance of neonates with CHD requiring cardiac surgery. The neonates were evaluated preoperatively and postoperatively with the Neonatal Intensive Care Unit Network Neurobehavioral Scale (NNNS) and scores were compared with published normative values. Clinical factors were obtained by chart review to assess their association with behavior. The CHD NNNS score pattern was compared with previously reported profiles in other high-risk populations. RESULTS: NNNS evaluations were completed on 67 neonates with CHD, resulting in 97 evaluations (50 preoperative, 47 postoperative). Compared with normative values, the cohort with CHD demonstrated decreased attention, regulation, asymmetry, stress, arousal, and excitability, along with increased non-optimal reflexes, lethargy, and need for handling (p<0.05 for all). Additional clinical factors had a minimal effect on the neurobehavioral pattern. Compared with previously published patterns in high-risk neonates without CHD, the cohort with CHD demonstrated a unique pattern of behavior. INTERPRETATION: Neonates with CHD demonstrate different neurobehavioral performance compared with typically developing neonates born at term as well as other high-risk neonates. Our experience suggests there is a unique neonatal neurobehavioral pattern in the hospitalized population with CHD. Targeted neonatal neurobehavioral evaluations may be useful in developing specific therapies to improve neurodevelopmental outcomes in neonates with CHD. WHAT THIS PAPER ADDS: Neonates with congenital heart disease demonstrate different neurobehavioral performance than typically developing neonates. Evaluation of neonatal neurobehavioral performance provides an opportunity to identify neurodevelopmental variability early. Identification of neurobehavioral performance variability allows targeted interactions and therapy.

Impact of educational video on critical congenital heart disease screening.

OBJECTIVE: To assess the status of pulse oximetry screening and barriers to implementing screening programs. METHODS: This was a prospective pre-post intervention survey of nurse managers and medical directors of hospital-based birthing centers in Oregon, Idaho, and Southern Washington. The intervention was a 7-minute video demonstrating and discussing pulse oximetry screening for critical congenital heart disease. RESULTS: Analysis of matched pairs showed a significant increase in the use of pulse oximetry screening during the study period from 52% to 73% (P < .0001). Following implementation of the video, the perception of all queried potential barriers decreased significantly among individuals from hospitals self-identified as nonscreening at baseline. Viewing the educational video was associated with an increase in the percentage of individuals from nonscreening hospitals that rated screening as "very beneficial" (45% vs 90%, P = .0001). CONCLUSIONS: An educational video was associated with improved opinions of pulse oximetry screening among hospitals not currently screening.