Explainable AI (XAI) is often viewed as a mechanism to promote the transparency and interpretability of AI recommendations in high-stakes domains, such as healthcare. This has led many studies to focus on designing and evaluating XAI to foster trust, calibrate reliance, enable algorithmic recourse, and support model understanding. However, this limited design scope restricts our understanding of how XAI can be used more broadly to support critical tasks in complex workflows, such as facilitating shared decision-making and supporting communication between stakeholders. Our work aims to address these critical gaps in the design for and understanding of XAI's emerging uses by iteratively prototyping XAI designs for an AI-powered clinical decision-support system with clinical stakeholders. We then created a high-fidelity prototype from those iterative sessions and used it as a design probe to uncover four emerging uses of XAI: collaboratively exploring treatment options, identifying and reflecting on treatment plans, communicating with stakeholders, and supporting health education. We reflect on the implications of designing XAI for emerging uses in healthcare.

Pulmonary arterial hypertension (PAH) is a complex progressive disease associated with high morbidity and mortality. Circulating serum biomarkers have the potential to optimize diagnosis and prognosis in PAH. The cellular communication network (CCN) protein family is a group of similarly structured matricellular proteins with many roles ranging from fibrosis to malignancy. Individual CCN proteins have been associated with PAH in previous studies, but no study has evaluated multiple CCN proteins as potentially relevant biomarkers in PAH. This study sought to establish associations using the circulating concentrations of measurable CCN proteins and PAH diagnosis, severity, outcomes, and other biomarkers. Serum levels of CCN1, CCN2, CCN3, and CCN6 were measured utilizing 225 patients from the PAH Biobank (PAHBiobank) with available hemodynamic data and 40 control samples. Serum levels of CCN1, 2, 3, and 6 proteins were significantly increased in PAH compared to controls. CCN1, CCN2, and CCN3 were associated with a lower 6-min walk distance. CCN2 and CCN3 were also associated with worse New York Heart Association Functional Class. Higher CCN2 and CCN3 levels correlated with higher levels of Endostatin and NT-proBNP. CCN6 was not significantly associated with any hemodynamic or clinical variables in the PAH cohort. Our results suggest that multiple CCN proteins are increased in PAH and that CCN2 and CCN3 have the most potential as novel biomarkers in PAH.

BACKGROUND:"Low-risk" pulmonary arterial hypertension (PAH) patients are defined as having a risk of death <5% within 1 year. As treatment improves, accounting for the negative impact of short-term morbidity and considering extending the duration of <5% mortality in this definition seem prudent. OBJECTIVES:This study sought to propose a more stringent definition of low risk that extends the low mortality risk to 3 years and specifies a low risk of morbidity at 1 year. METHODS:This study used patient-level data from a harmonized data set of 8 PAH randomized controlled trials that included patients with a REVEAL (Registry to Evaluate Early and Long-Term PAH Disease Management) 2.0 risk score ≤6 (ie, low risk using contemporary scores). Low risk was a priori defined as "risk of death <5% at 3 years and risk of clinical worsening <10% at 1 year." Clinical worsening was defined at the individual trial level and included any of the following: lung transplantation, hospitalization for PAH, initiation of long-term oxygen therapy, 15% decrease in 6-minute walking distance (6MWD) from baseline with a worsening of NYHA functional class, or the addition of a new PAH medication. RESULTS:A total of 1,925 PAH patients were included: median age, 46 years (Q1-Q3: 34-57 years); mean pulmonary artery pressure, 47 mm Hg (Q1-Q3: 37-58 mm Hg); pulmonary vascular resistance, 8 WU (Q1-Q3: 6-12 WU), 6MWD, 409 m (Q1-Q3: 363-423 m); and N-terminal pro-B-type natriuretic peptide, 227 pg/mL (Q1-Q3: 101-621 pg/mL). Compared with patients with a REVEAL 2.0 score equal to 5 or 6, patients with a score ≤4 had a better clinical and hemodynamic presentation and met the definition of "refined low risk" (3-year mortality of 2.6% and 1-year clinical worsening of 6.4%). Idiopathic PAH, 6MWD, N-terminal pro-B-type natriuretic peptide, NYHA functional class I to II, heart rate, and the absence of previous atrial fibrillation were associated with refined low risk. CONCLUSIONS:This study is the first to embed both morbidity and mortality within the definition of low -risk for PAH and suggests that patients with a REVEAL 2.0 risk score ≤4 define this condition.

The study of blood and cerebrospinal fluid biomarkers is a promising and rapidly advancing field in the research of disorders of consciousness (DoC). The use of advanced biochemical and analytic techniques in biomarker research has improved our ability to identify new biomarkers that can aid in the diagnosis, prognosis, and treatment of patients with brain injury. However, the use of biomarkers in clinical practice is limited by several challenges, including the lack of standardization in test and research methodologies. Despite this, identifying the most promising biomarkers and supporting their findings with strong evidence can improve their clinical utility. This chapter discusses the most promising biomarkers for DoC, which fall into four categories: neuronal, glial, inflammatory, and metabolic biomarkers. Understanding the role of each category in DoC can provide valuable insights into the mechanisms of brain injury and inform the development of more effective diagnostic and treatment strategies. By integrating biomarker research with clinical practice, we can improve our understanding of DoC and provide better care for these patients.

Background: There is significant interest in NO pathway modulators, specifically type 5 phosphodiesterase inhibitors (PDE5is), to treat patients with a Fontan circulation. Trials, however, have had mixed results. The relationship between the NO pathway and clinical status in patients with Fontan circulation is a significant knowledge gap. Methods and Results: We performed targeted metabolomic analysis using liquid chromatography coupled to mass spectrometry to quantify plasma NO pathway metabolite concentrations from 2 well-characterized populations of patients with Fontan circulation: the Boston Adult Congenital Heart Disease Biobank and Fontan Udenafil Exercise Longitudinal studies. We investigated associations between NO metabolite concentrations and clinical outcomes, exercise capacity, and response to PDE5is. Increased plasma concentration of asymmetric dimethyl arginine (ADMA), an inhibitor of NO production, was associated with risk for hospitalization or death. Increased ADMA and symmetric dimethyl arginine (another inhibitor of NO production) concentrations were associated with decreased baseline exercise capacity among patients with Fontan circulation with <90% predicted peak oxygen uptake, and change in ADMA and symmetric dimethyl arginine concentrations were predictive of change in exercise capacity over time. Treatment with the PDE5i udenafil uncoupled this association. Finally, baseline ADMA and symmetric dimethyl arginine concentrations predicted response to PDE5is among patients with subnormal peak oxygen uptake. Conclusions: Plasma concentrations of metabolites that inhibit NO flux are associated with negative clinical outcomes and worse exercise capacity. Moreover, metabolite shifts over time associated with increased NO flux are associated with improved exercise capacity. In patients with a Fontan circulation, the NO pathway modulators ADMA and symmetric dimethyl arginine may be useful as biomarkers of clinical status and predictive of response to PDE5is.

Background: Presently pediatric pulmonary hypertension (PH) lacks a widely accepted risk assessment algorithm and clear guidelines for treatment. Most targeted therapies are used off-label in pediatrics, and current registries do not capture the full breadth of pediatric PH in the U.S., limiting insight into treatment patterns and outcomes. Question: How is pediatric PH treated in the U.S. and what is the expected prognosis? Methods: Patient data were obtained through the TriNetX Research Network, which provides de-identified patient health information from electronic medical records in the US, and we included patients aged 1 to 21 years diagnosed with PH based on ICD10 code, including primary PH and secondary PH, who were treated with targeted PH therapy, and had a pre-treatment BNP/NT-proBNP (BNP) level. Patients were excluded if they were on iNO alone or had PH due to left heart disease. BNP stratification was made through an adaptation of REVEAL 2.0 and recent ESC/ERS guidelines (Table 1). Primary outcomes were death or hospitalizations 1 year after diagnosis. Results: We identified 793 patients who fit the criteria. Pre-treatment BNP levels were grouped “low” (n=303), “mid” (n=135), or “high” (n=355). The cohort was predominately White (n=419) and Black or African-American (BAA) (n=206), with 51% of BAA and 40.6% of White with a high pre-treatment BNP. Initial therapy was predominately a single drug no matter the initial BNP level (low 77.2%, mid 82.2% or high 69%) (Table 1). High BNP levels were a significant predictor of 1- and 3-year survival with a Cox proportional hazard of 2.0 (0.67-2.04, p<0.01) and 1.8 (1.21, 2.69, p<0.01), respectively. When looking at initial treatment intensity, there was no difference in mortality between those who were on single or multi-drug therapy at 1- and 3-years (Table 2). Using a Sankey plot for 1-year post diagnosis trends, roughly a third of each group transitioned to another level, and those who experienced mortality were restricted to mainly those patients who worsened or remained in the high level BNP group (Figure 1). Conclusion: Mortality was seen in those with initial high BNP levels at 1- and 3- years after diagnosis, but was not significantly different in those who were started on single vs multi-drug therapy. To our knowledge this is the largest cohort of pediatric PH patients looking at the current landscape of pediatric PH, prognostic value of BNP, and current therapy choices and expected outcomes.

Background: Patients with idiopathic pulmonary arterial hypertension (IPAH), a progressive pulmonary vasculopathy, may be clinically similar, yet have different responses to therapy and outcomes. Understanding proteomic profiles that distinguish amongst clinically similar patients could improve disease classification to better target treatment. Research Question: Can proteomic profiling identify molecular sub-phenotypes amongst otherwise clinically similar patients with IPAH? Methods: A cross-sectional, prevalent cohort of patients with IPAH (N=120, 60 survivors/60 non-survivors) was selected from the PAH Biobank to have similar clinical, genetic hemodynamic and risk profiles. Plasma protein levels of 11,000 proteins (SomaScan 11K Platform) were assayed. The top 10% of proteins with highest variance were used for spectral clustering. Survival by cluster was evaluated by Kaplan-Meier analysis and Cox proportional hazards model, adjusted for enrollment age and sex. Clinical differences between clusters were assessed with Chi-squared, Kruskal-Wallis and Fisher’s exact tests. Pathway analysis identified major proteomic pathways differentiating each cluster. Results: Spectral clustering identified 3 clusters, with 36%, 45% and 68% events over 5-years in clusters 1, 2 and 3 respectively (Figure 1, p=0.0025). The adjusted hazard ratio for death was 1.40 and 2.78 in clusters 2, 3 comparing to cluster 1. Subjects in cluster 3 were older than cluster 1 and 2. Medications, six-minute walk test and hemodynamics were similar across clusters and most subjects were intermediate risk (ERS criteria) and low or average risk (REVEAL 2.0)(Table 1).Differentially expressed proteins showed enrichment of Vascular Endothelial Growth Factor (VEGF) signaling (p=9.30E-04) between clusters 1 and 2, the Wnt signaling between clusters 1 and 3 (p=0.0014), and Neurotrophin signaling (p=4.82E-09) and VEGF signaling (p=5.93E-09) between clusters 2 and 3. Conclusions: This study identified differential expression of 3 pathways across clusters of similar IPAH patients with differences in outcomes not predicted by clinical risk assessment. All three of these pathways, Wnt, VEGF and Neurotrophin, have been implicated in pulmonary vascular and right ventricular dysfunction and may identify patients at higher risk or at different stages of the IPAH disease process. Understanding the proteome may identify molecular phenotypes of IPAH associated with outcomes but missed by clinical classification.

Cellular communication network 2 (CCN2) is a secreted matricellular protein associated with pulmonary arterial hypertension (PAH) but has not been studied relative to PAH severity, outcomes, or right ventricle (RV) structure and function in a large human cohort and preclinical animal model. This study assessed the associations between CCN2 and PAH severity, survival, hemodynamic measurements, and cardiovascular dysfunction. Serum CCN2 levels were compared in 2548 adults with PAH and 216 controls. CCN2 levels in PAH patients were compared to functional and hemodynamic measurements, and survival outcomes. RV-pulmonary artery coupling and RV morphology were also assessed in a small subset of patients via pressure-volume loops and cardiac magnetic resonance imaging. In a preclinical PAH model, plasma CCN2 levels were compared between ventricles with PAH progression. CCN2 mRNA levels in both ventricles in the preclinical model were measured to compare with morphologic histologic variables. CCN2 serum levels were significantly higher in PAH compared to controls (p < 0.0001). Higher CCN2 levels were associated with reduced RV contractility (p = 0.003). Higher CCN2 levels were associated with worse 6MWD (p = 0.035), and higher risk of mortality or transplant (p = 0.025). In the preclinical model, prepulmonary CCN2 plasma levels increased with the progression of disease. CCN2 mRNA levels in the RV were associated with decreased RV capillary density (p = 0.015) and increased RV fibrosis (p = 0.045). Though more investigation is needed, it appears that CCN2 plays a role in the development of PAH and potentially in RV maladaptation in PAH.

Background: Exposure to cyclohexanone, an industrial solvent used to soften medical plastics, has been associated with unfavorable outcomes for neonates undergoing congenital heart surgery. Yet, the underlying physiologic mechanisms remain unclear. Hypothesis: Cyclohexanone exposure induces a systemic inflammatory response following cardiopulmonary bypass (CPB) in neonates undergoing congenital heart surgery. Methods: We analyzed data from 70 neonates enrolled in the placebo group of the Corticosteroid Therapy in Neonates Undergoing Cardiopulmonary Bypass trial. Serum samples were collected at five time points: preoperatively, immediately after CPB cessation, and at 4, 12, and 24 hours postoperation. Serum cyclohexanone and cytokine concentrations were measured using gas chromatography tandem mass spectrometry and multiplex enzyme-linked immunoassays, respectively. We used covariate-adjusted longitudinal quantile regression to separately estimate the expected differences in 75 th percentile cytokine concentrations (and 95% CIs) per interquartile range (IQR) increase in preoperation serum cyclohexanone concentration as well as contemporaneous, perioperative cyclohexanone concentrations. Covariates included age and weight at surgery, STAT mortality risk category, cytokine assay batch, and total CPB duration. Results: In covariate-adjusted models, an IQR increase in preoperation cyclohexanone concentration was associated with a 342.3 pg/mL (-460.0, -224.5) lower IL-10 concentration at 4 hours postoperation, 330.3 pg/mL (185.8, 474.8) higher IL-6 concentration at 4 hours and 438.2 pg/mL (37.9, 838.4) higher concentration at 12 hours postoperation, and 64.9 pg/mL (13.4, 116.5) higher IL-8 at 12 hours postoperation. An IQR increase in contemporaneous cyclohexanone concentration was associated with lower IL-10 levels after CPB cessation (-997 pg/mL; 95%CI: -1545, -449), whereas IL-8 was significantly higher after CPB cessation (83 pg/mL; 95%CI: 4.64, 161) and 4 hours (250 pg/mL; 95%CI: 95.3, 405), 12 hours (35.6 pg/mL; 95%CI: 12.1, 59.1), and 24 hours (77.3 pg/mL; 95%CI: 24.9, 130) postoperation. Conclusions: Perioperative cyclohexanone exposure is associated with a proinflammatory response. These results suggest that cyclohexanone may contribute to postoperative inflammation following CPB and further potentiate adverse postoperative outcomes. Further studies are needed to develop strategies to reduce cyclohexanone exposure in the perioperative environment.

Introduction: Congenital heart disease repair is commonly associated with neuro-developmental abnormalities that affect quality of life. Cyclohexanone (CXO) is an industrial solvent used to soften medical plastics. Exposure to CXO during neonatal cardiac surgery is associated with poor neurodevelopmental outcomes. Hypothesis: Neonatal cyclohexanone exposure causes neurological damage and initiates peripheral inflammation, resulting in functional and behavioral deficits persisting to adulthood that mimic developmental abnormalities seen in human congenital heart disease. Methods: CXO Exposure: Neonatal rats from both sexes were given CXO (0.53 ul/g/day, i.p.) or saline from post-natal day 7 (P7) to P14. Peripheral Blood Mononuclear Cell (PBMC) and Serum: Whole blood was collected at P15, P28 and P70. Inflammatory markers in serum and the PBMC secretome were evaluated using a MesoScale Discovery platform. Open Field and Magnetic Resonance Imaging (MRI) : At P28, functional MRI (fMRI), diffusion tensor imaging (DTI), and open field behavior were assessed. Touchscreen : As adults, rats were trained on a touchscreen task of visual discrimination (VD) and reversal learning (RL). Statistics: Student’s T-test or Mann-Whitney U-test in n=8-16/group with p<0.05 considered significant. Results: One day after CXO exposure (P15), levels of IFN-γ, IL-10, IL-13, IL-4 (all p<0.05), and IL-6 (p<0.01) were high in serum compared to controls concomitant with sustained peripheral immune hyperreactivity (SPIHR). By P28 through P70, levels of CXCL1 remained elevated (p<0.05) and PBMCs were hyper-inflammatory. Inflammation coincides with diminished cerebral functional connectivity related to attention and reduced white matter microstructure in tracts responsible for cognition (p<0.05). CXO-exposed rats were hypermobile and hyperactive in an open field (p<0.05) with diminished performance on VD and RL tasks. Conclusion: CXO induces inflammation and SPIHR. Adult rats exposed to CXO as neonates have diminished executive function, functional and anatomical connectivity, and poor cognitive flexibility alongside hyperactivity and hypermobility. This supports that neonatal CXO exposure initiates SPIHR and neurological damage that results in altered function and behavior that completely recapitulates the human neuro-developmental delay phenotype after neonatal cardiac surgery. Future work will focus on the mechanism of CXO injury and reducing CXO exposure to improve neurodevelopment.

Exposures to endocrine disrupting chemicals (EDCs) in early life have demonstrable adverse implications on child health and development. Yet, there is a dearth of studies evaluating the potential exposures to EDCs, such as bisphenols, parabens, phthalates, and volatile organic compounds (VOCs), in hospital-based settings among children who are critically ill and/or particularly vulnerable. This narrative review seeks to provide up-to-date evidence on the sources and magnitude of exposure to EDCs in neonatal-, pediatric-, and cardiac intensive care units (NICUs/PICUs/CICUs) as well as resulting health impacts. Thirty-three studies were included in this review. The most frequently studied and characterized EDCs in NICUs/PICUs/CICUs were phthalates followed by terephthalates and alternative plasticizers. Evaluations of health outcomes resulting from such exposures were scarce, and few studies assessed health outcomes after hospital discharge. EDCs are pervasive in NICU/PICU/CICU settings and pediatric exposure levels are much higher than in other environments. However, the existing evidence has multiple limitations that should be addressed in future work. Specifically, studies evaluating a more expansive array of chemicals, including contemporary and emerging replacements for legacy compounds, are needed, as are studies that consider chemical mixtures. Additionally, few studies evaluated the health impacts of chemical exposures, and their mixtures, in NICU/PICU/CICU settings, especially long-term health outcomes observed after hospital discharge. Such studies could be invaluable in supporting policy as well as development of medical products without toxic chemicals.

ABSTRACTIntroductionScreening and treatment of seizures (Sz) in neonates suffering from hypoxic-ischemic encephalopathy (HIE) is routine. Understanding if Sz worsen brain injury and outcomes would optimize treatment decisions, recognizing risks and benefits. We hypothesized that serum central nervous system (CNS)-specific biomarkers would discriminate neonates with Sz and relate to outcomes.MethodsThis retrospective cohort study was conducted between April 28, 2009 and November 15, 2019, including neonates diagnosed with seizures (Sz) and/or HIE treated with therapeutic hypothermia (TH), who had sufficient remnant serum for biomarker analysis. Neonates were grouped in i) only Sz without HIE (Sz-no HIE), ii) HIE with Sz (Sz-HIE) and iii) HIE without Sz (no Sz-HIE). Levels of glial fibrillary acidic protein (GFAP, astrocytic reactivity), Tau (neuronal injury) and neurofilament light chain (NF-L, axonal degeneration) were studied at admission, <72h and 72-144h of life against time to full oral feeds and brain injury in MRI.ResultsAfter exclusions 145 neonates were included (61% male; 33% Black). Admission GFAP levels were higher in Sz-HIE than in no Sz-HIE neonates. During the first 72h of life, levels of all 3 biomarkers were similar between Sz groups (Sz-no HIE & Sz-HIE), but higher than in no Sz-HIE. After 72h, NF-L and Tau remained higher in both Sz groups vs. no Sz-HIE. Stroke diagnosed in 31% (Sz-no HIE), 8% (Sz-HIE) and 11% (no Sz-HIE) of neonates had no effect in biomarkers levels. In adjusted regression models higher Tau and NF-L percentiles related to longer time to reach full oral feeding and higher odds of significant brain injury in MRI in both Sz groups.ConclusionsIn this study, Tau and NF-L levels are higher in those neonates developing Sz, irrespective of their HIE diagnosis. These results provide additional support for the notion that Sz may worsen brain injury and outcomes in neonates with HIE even with TH.

BACKGROUND:Cyclohexanone is a volatile organic compound known to be toxic to humans and animals, used in the medical setting as a solvent sealer for intravenous (IV) fluid administration devices. We aimed to determine exposure sources as well as plasma and urine levels of cyclohexanone and metabolites in critically ill infants and children. METHODS:We prospectively enrolled children in a single center pediatric intensive care unit (ICU) (n = 66), and conducted a secondary analysis of a multicenter trial in premature neonates (n = 69). Cyclohexanone and its predominant metabolites, trans-1,2-cyclohexanediol and trans-1,4-cyclohexanediol, were measured serially in medical fluids, plasma, and urine. RESULTS:Cyclohexanone was detected in all IV solutions used in standard ICU care (IV fluids, medications, dialysate and red blood cell bags, n = 53 fluid samples). Cyclohexanone and metabolites were higher in urine versus plasma in both cohorts. In premature neonates, plasma and urine cyclohexanone concentrations were highest on day of randomization, while metabolite concentrations were highest on days 7-14. CONCLUSIONS:Currently, cyclohexanone may represent an inevitable exposure to children who require intensive care inclusive of IV fluid and medication administration devices. Further studies are needed to develop replacement or mitigation strategies for cyclohexanone exposure in the vulnerable neonatal and pediatric ICU populations. IMPACT:Direct bloodstream exposure to cyclohexanone in the hospital environment has been poorly described in the healthcare setting. Cyclohexanone was present in all tested types of intravenous solutions used in standard intensive care (intravenous fluids, medications, dialysate and stored red blood cell bags). In a single center pediatric intensive care unit cohort and a multicenter neonatal intensive care unit cohort, cyclohexanone and its metabolites were detected in every blood and urine sample tested. In a multicenter neonatal intensive care unit cohort, plasma and urine cyclohexanone concentrations were highest on day 1 of admission and metabolite concentrations were highest on days 7-14.

Background: Multiple studies of preclinical models and adults with pulmonary arterial hypertension (PAH) have shown new insights into PAH pathophysiology using high-throughput proteomics. However, application of similar techniques to pediatric PAH is lacking, limiting our understanding of the unique mechanisms and features of pediatric PAH and potentially improving outcomes. Question: Can high-throughput proteomic profiling identify novel biomarkers of pediatric PAH that reflect disease severity and predict overall outcomes? Methods: We used the Olink® Explore HT platform to analyze 86 serum samples from the NHLBI PAHbiobank (Table 1). This platform utilized Proximity Extension Assay (PEA) technology, which employs protein-specific antibodies linked to DNA tags in combination with Next Generation Sequencing (NGS), to quantify 5,440 unique proteins. Using pediatric PAH samples from the PAHBiobank (n=171), we partially validated our top protein candidates. Results: T-test with false discovery rate correction identified the 10 eligible proteins based on the composite outcome of death, lung transplantation, atrial septostomy, or Pott’s shunt, with 7 proteins upregulated and 3 downregulated (Table 2, top). Initial validation confirmed BDNF and SDC4 were significantly lower, and REN, IGFBP2, and NT-proBNP were higher in those who experienced the composite outcome (all, p<0.01). Similarly, CTSB and CCL24 were increased in patients who experienced the composite outcome, but not significant (p=0.07 and 0.08, respectively) (Table 2, bottom). Kaplan-Meier survival analysis demonstrated a significantly reduced event-free survival among patients with lower BDNF and SDC4 levels (Log-Rank p=0.04). Similarly, elevated REN, IGFBP2, and NT-proBNP was associated with reduced survival (Log-Rank p=0.02, <0.01, and 0.04, respectively). CTSB and CCL24 were also elevated, but not significant (Log-Rank p=0.19 and 0.30, respectively) (Figure 1). Conclusion: Using high-throughput proteomics, we identified several proteins that were significantly associated with survival in pediatric PAH that can represent new inflammatory, or heart failure pathways. Our early validation has identified new biomarkers such as BDNF and REN that could serve as potential new treatment targets, as inhibitors of the renin-angiotensin axis have shown efficacy in adult PAH and largely unused in children. Future studies will focus on analyzing other target proteins and further validate their clinical relevance.