Faculty Bibliography

OBJECTIVE:Reduced-channel wearable electroencephalography (EEG) may overcome the accessibility and patient comfort limitations of traditional ambulatory electrographic seizure monitoring during extended-duration use. Automated algorithms are necessary for review of extended-duration reduced-channel EEG, yet current clinical support software is designed only for full-montage recordings. METHODS:The performance of a novel automated seizure detection algorithm for reduced-channel EEG (Epitel) was evaluated in a clinical validation study involving 50 participants (31 with seizures) with diverse demographic and seizure representation. RESULTS:The algorithm demonstrated an event-level sensitivity of 86.2% (95% confidence interval [CI] = 79.5%-93.2%) and a false detection rate of .162 per hour (95% CI = .116-.221), which is comparable to the performance of current clinical software for full-montage EEG. Performance varied by electrographic seizure type, with 91.4% sensitivity for focal evolving to generalized seizures, 86.7% for generalized seizures, and 77.3% for focal seizures. The algorithm maintained robust performance in both pediatric participants aged 6-21 years (83% sensitivity) and adults aged 22+ years (90% sensitivity), as well as in ambulatory (80%) and epilepsy monitoring unit (EMU) monitoring environments (87.5%). The false detection rate in ambulatory monitoring environments (.290 false positive [FP] detections/h), all of which involved pediatric participants, was notably higher than in the EMU (.136 FP/h), indicating an area with clear need for improvement for unrestricted at-home monitoring. The algorithm's supplemental Confidence metric, designed to engender trust in the algorithm, showed a strong correlation with detection precision. SIGNIFICANCE/CONCLUSIONS:These results suggest that this algorithm can provide crucial support for review of extended-duration reduced-channel wearable EEG, enabling electrographic seizure monitoring with no restrictions on a person's daily life.

BACKGROUND:Clinical reasoning (CR) is an essential skill; yet, physicians often receive limited feedback. Artificial intelligence holds promise to fill this gap. OBJECTIVE:We report the development of named entity recognition (NER), logic-based and large language model (LLM)-based assessments of CR documentation in the electronic health record across 2 institutions (New York University Grossman School of Medicine [NYU] and University of Cincinnati College of Medicine [UC]). METHODS:-scores for the NER, logic-based model and area under the receiver operating characteristic curve (AUROC) and area under the precision-recall curve (AUPRC) for the LLMs. RESULTS:-scores 0.80, 0.74, and 0.80 for D0, D1, D2, respectively. The GatorTron LLM performed best for EA2 scores AUROC/AUPRC 0.75/ 0.69. CONCLUSIONS:This is the first multi-institutional study to apply LLMs for assessing CR documentation in the electronic health record. Such tools can enhance feedback on CR. Lessons learned by implementing these models at distinct institutions support the generalizability of this approach.

BACKGROUND/UNASSIGNED:Donation after circulatory death (DCD) with cardiopulmonary bypass for thoracoabdominal normothermic regional perfusion (TA-NRP) has led to increased use of donor hearts. Rejection rates and long-term survival outcomes are not known. METHODS/UNASSIGNED:A single-center retrospective cohort review of patients who underwent DCD heart transplantation from January 2020 to December 2023 was performed. Donor and recipient characteristics, operative characteristics, and posttransplantation outcomes were analyzed. Subgroup analysis comparing co-localized vs distant donors and recipients was performed. The primary end point was 1-year survival. Secondary end points included incidences of primary graft dysfunction (PGD), cardiac allograft vasculopathy (CAV), rejection rate, and overall mortality. Our TA-NRP protocol has remained the same, consisting of sternotomy, ligation of aortic arch vessels, establishment of cardiopulmonary bypass, reintubation, resuscitation of the heart, and cold static storage during transport. RESULTS/UNASSIGNED:< .005) ischemia times, without any other differences. CONCLUSIONS/UNASSIGNED:Outcomes after DCD heart transplantation using TA-NRP remain encouraging with acceptable rates of rejection, PGD, CAV, and survival at 1 year.

BACKGROUND/UNASSIGNED:We investigated racial and ethnic disparities in telehealth counseling among Medicaid-insured patients in outpatient substance use disorder (SUD) treatment clinics and assessed whether the clinic-level proportion of Medicaid-insured patients moderated these disparities. METHODS/UNASSIGNED:Using New York State (NYS) Medicaid and statewide treatment registry data, we analyzed 24,814 admission episodes across 399 outpatient SUD clinics during the first 6 months of COVID-19 (April-September 2020). Our outcome measure was the number of tele-counseling sessions within the first 90 days of treatment. Key independent variables included beneficiary race/ethnicity and the clinic-level proportion of Medicaid-insured patients, divided into four quartiles: lowest, second, third, and highest. Mixed effects negative binomial models assessed the associations between race/ethnicity, Medicaid proportions, and telehealth use, with interaction terms evaluating the moderating role of Medicaid proportions. RESULTS/UNASSIGNED:Black and Latinx patients received fewer telehealth sessions than non-Latinx White patients, with adjusted incidence rate ratios (aIRRs) of 0.86 (95% CI: 0.82, 0.91) for Black patients and 0.93 (95% CI: 0.88, 0.98) for Latinx patients. Black patients at clinics with the highest Medicaid proportions had higher telehealth usage rates compared to those at clinics with the lowest Medicaid proportions (aIRR, 1.20; 95% CI, 1.03-1.41). Patients in clinics with the highest Medicaid proportions were more likely to use individual telehealth counseling (aIRR, 1.02-1.88; 95% CI, 1.01-3.04). CONCLUSIONS/UNASSIGNED:Significant racial disparities in telehealth use exist, with variations persisting across clinics with different Medicaid proportions. Targeted interventions are needed to address these access gaps.

OBJECTIVE:Recent guidelines suggest that non-invasive prenatal screening (NIPS) should be offered to all patients with singleton and twin pregnancies. Accurate determination of fetal fraction in cell-free DNA (cfDNA) is vital for reliable NIPS outcomes. We propose a methylation-based approach using droplet digital PCR (ddPCR) and methylation-sensitive restriction enzyme (MSRE) digestion for fetal fraction quantification as an affordable and fast solution. METHOD/METHODS:Following biomarker discovery using early pregnancy placental genomic DNA (gDNA) and cfDNA from non-pregnant female individuals, we designed assays targeting MSRE-compatible regions based on contrasting methylation patterns between maternal and fetal cfDNA. We established a proof-of-concept ddPCR workflow on the Bio-Rad Droplet Digital PCR QX600 instrument. RESULTS:Testing the fetal fraction assay multiplex on 137 prospective clinical samples demonstrated high concordance with NGS results for both female and male pregnancies as well as with chromosome Y-based calculations for samples with a male fetus. Reproducibility analysis indicated lower variability compared to previously reported NGS performance. CONCLUSION/CONCLUSIONS:This study showcases the potential of this novel, 6-color, high-multiplex methylation ddPCR panel for accurate measurement of fetal fraction in cfDNA samples. It presents opportunities to integrate such methodology as a standalone measurement to assess the quality of samples undergoing NIPS.

INTRODUCTION/BACKGROUND:Despite longstanding status as a public health priority, preterm birth rates continue to be higher among non-Hispanic Black women compared with other racial and ethnic groups. A growing body of literature highlights the site of care as a key factor in pregnancy outcomes. Although research shows that many individuals do not receive guideline-based prenatal care, little is known about site-level variation in the use of recommended prenatal services and its potential relationship with Black-White preterm birth disparities. METHODS:In this cross-sectional cohort study, we analyzed variation in site-level use of 4 key prenatal services: tetanus, diphtheria, and pertussis (Tdap) vaccination, [per the CDC website] and screening for bacteriuria, diabetes, and group Beta streptococcus, using administrative data from New York State Medicaid and the American Community Survey. We used multivariable logistic regressions to estimate the odds of attending a low-use site (mean <2 services per patient) by race and ethnicity, controlling for age, high-poverty residential address, and low prenatal care attendance. We performed Fairlie decomposition analyses to quantify the contribution of individual and site-level factors to the observed difference in preterm birth rates among Black and White non-Hispanic women. RESULTS:Site-level use of recommended prenatal services ranged from an average of 1 to 3.6 services per patient. Non-Hispanic Black women had more than twice the odds (adjusted odds ratio, 2.42; 95% CI, 2.32-2.52) of attending a low-use site compared with non-Hispanic White women. Among factors in the decomposition analysis, site-level screening for bacteriuria and diabetes accounted for the highest proportion of the explained variance in the observed preterm birth rates for non-Hispanic Black (10.7%) and non-Hispanic White (6.7%) women. DISCUSSION/CONCLUSIONS:Results from this research support immediate improvement in guideline-based prenatal care to narrow the gap in preterm birth for non-Hispanic Black women. Research is needed to identify and correct site-level barriers to recommended prenatal services.

Cancer is a systemic disease with complications beyond the primary tumor site. Among them, thrombosis is the second leading cause of death in patients with certain cancers (e.g., pancreatic ductal adenocarcinoma [PDAC]) and advanced-stage disease. Here, we demonstrate that pro-thrombotic small extracellular vesicles (sEVs) are secreted by C-X-C motif chemokine 13 (CXCL13)-reprogrammed interstitial macrophages in the non-metastatic lung microenvironment of multiple cancers, a niche that we define as the pro-thrombotic niche (PTN). These sEVs package clustered integrin β₂ that dimerizes with integrin α_X and interacts with platelet-bound glycoprotein (GP)Ib to induce platelet aggregation. Blocking integrin β₂ decreases both sEV-induced thrombosis and lung metastasis. Importantly, sEV-β₂ levels are elevated in the plasma of PDAC patients prior to thrombotic events compared with patients with no history of thrombosis. We show that lung PTN establishment is a systemic consequence of cancer progression and identify sEV-β₂ as a prognostic biomarker of thrombosis risk as well as a target to prevent thrombosis and metastasis.

BackgroundAdvances in STN-DBS technology, among which directional stimulation, improved Parkinson's disease (PD) treatment efficacy, while increasing the clinical programming complexity. Lead localization software may aid the stimulation contact selection process.ObjectiveWe aimed to assess the concordance between imaging-suggested (IGP) and conventional-programming (CP) selected stimulation contacts one year after surgery and its impact on motor outcomes.MethodsSixty-four PD patients with bilateral STN-DBS were enrolled. Lead localization was reconstructed with Brainlab^TM software. For each electrode, the vertical contact level and, when applicable, the directionality predicted by the lead reconstruction software to be the most effective were established and compared to the stimulation parameters clinically activated one-year post-surgery. IGP/CP concordance ratio was calculated for both stimulation level and directional contacts. Post-operative modifications of PD motor symptoms severity were compared among groups of concordant and discordant IGP/CP programming.ResultsOne-year post-surgery, IGP/CP concordance was 80% for active stimulation vertical contact level and 51% for directionality. No significant difference in motor outcomes was found between IGP/CP concordant and discordant patients for contact level activation, whereas patients with concordant IGP/CP active directional stimulation (c-Direction) showed superior motor outcomes at one-year follow-up than those discordant (d-Direction) (UPDRS-III stimulation-induced improvement: c-Direction = -25.66 ± 13.74 vs. d-Direction = -12.54 ± 11.86; p = 0.011).ConclusionsVisual reconstruction software correctly predicted the most clinically effective stimulation contact levels in most patients. Imaging therefore facilitates classic STN-DBS clinical programming while assuring similar outcomes. Moreover, better motor outcomes were reached by patients with concordant IGP/CP directional parameters, suggesting that visualization can represent an added value in particular for directional stimulation programming.

Existing deep learning algorithms for electrocardiogram (ECG) classification rely on supervised training approaches requiring large volumes of reliably labeled data. This limits their applicability to rare cardiac diseases like Brugada syndrome (BrS), often lacking accurately labeled ECG examples. To address labeled data constraints and the resulting limitations of supervised training approaches, we developed a novel deep learning model for BrS ECG classification using the Variance-Invariance-Covariance Regularization (VICReg) architecture for self-supervised pre-training. The VICReg model outperformed a state-of-the-art neural network in all calculated metrics, achieving an area under the receiver operating and precision-recall curves of 0.88 and 0.82, respectively. We used the VICReg model to identify missed BrS cases and hence refine the previously underestimated institutional BrS prevalence and patient outcomes. Our results provide a novel approach to rare cardiac disease identification and challenge existing BrS prevalence estimates offering a framework for other rare cardiac conditions.