Using machine learning to predict five-year transplant-free survival among infants with hypoplastic left heart syndrome.

Hypoplastic left heart syndrome (HLHS) is a congenital malformation commonly treated with palliative surgery and is associated with significant morbidity and mortality. Risk stratification models have often relied upon traditional survival analyses or outcomes data failing to extend beyond infancy. Individualized prediction of transplant-free survival (TFS) employing machine learning (ML) based analyses of outcomes beyond infancy may provide further valuable insight for families and healthcare providers along the course of a staged palliation. Data from both the Pediatric Heart Network (PHN) Single Ventricle Reconstruction (SVR) trial and Extension study (SVR II), which extended cohort follow up for five years was used to develop ML-driven models predicting TFS. Models incrementally incorporated features corresponding to successive phases of care, from pre-Stage 1 palliation (S1P) through the stage 2 palliation (S2P) hospitalization. Models trained with features from Pre-S1P, S1P operation, and S1P hospitalization all demonstrated time-dependent area under the curves (td-AUC) beyond 0.70 through 5 years following S1P, with a model incorporating features through S1P hospitalization demonstrating particularly robust performance (td-AUC 0.838 (95% CI 0.836-0.840)). Machine learning may offer a clinically useful alternative means of providing individualized survival probability predictions, years following the staged surgical palliation of hypoplastic left heart syndrome.

Precision Anesthesia in 2050.

Over the last few decades, the field of anesthesia has advanced far beyond its humble beginnings. Today's anesthetics are better and safer than ever, thanks to innovations in drugs, monitors, equipment, and patient safety.1-4 At the same time, we remain limited by our herd approach to medicine. Each of our patients is unique, but health care today is based on a one-size-fits-all approach, while our patients grow older and more medically complex every year. By 2050, we believe that precision medicine will play a central role across all medical specialties, including anesthesia. In addition, we expect that health care and consumer technology will continually evolve to improve and simplify the interactions between patients, providers, and the health care system. As demonstrated by 2 hypothetical patient experiences, these advancements will enable more efficient and safe care, earlier and more accurate diagnoses, and truly personalized treatment plans.

Application of natural language processing to identify social needs from patient medical notes: development and assessment of a scalable, performant, and rule-based model in an integrated healthcare delivery system.

Objectives: To develop and test a scalable, performant, and rule-based model for identifying 3 major domains of social needs (residential instability, food insecurity, and transportation issues) from the unstructured data in electronic health records (EHRs). Materials and Methods: We included patients aged 18 years or older who received care at the Johns Hopkins Health System (JHHS) between July 2016 and June 2021 and had at least 1 unstructured (free-text) note in their EHR during the study period. We used a combination of manual lexicon curation and semiautomated lexicon creation for feature development. We developed an initial rules-based pipeline (Match Pipeline) using 2 keyword sets for each social needs domain. We performed rule-based keyword matching for distinct lexicons and tested the algorithm using an annotated dataset comprising 192 patients. Starting with a set of expert-identified keywords, we tested the adjustments by evaluating false positives and negatives identified in the labeled dataset. We assessed the performance of the algorithm using measures of precision, recall, and F1 score. Results: The algorithm for identifying residential instability had the best overall performance, with a weighted average for precision, recall, and F1 score of 0.92, 0.84, and 0.92 for identifying patients with homelessness and 0.84, 0.82, and 0.79 for identifying patients with housing insecurity. Metrics for the food insecurity algorithm were high but the transportation issues algorithm was the lowest overall performing metric. Discussion: The NLP algorithm in identifying social needs at JHHS performed relatively well and would provide the opportunity for implementation in a healthcare system. Conclusion: The NLP approach developed in this project could be adapted and potentially operationalized in the routine data processes of a healthcare system.

Machine Vision and Image Analysis in Anesthesia: Narrative Review and Future Prospects.

Machine vision describes the use of artificial intelligence to interpret, analyze, and derive predictions from image or video data. Machine vision-based techniques are already in clinical use in radiology, ophthalmology, and dermatology, where some applications currently equal or exceed the performance of specialty physicians in areas of image interpretation. While machine vision in anesthesia has many potential applications, its development remains in its infancy in our specialty. Early research for machine vision in anesthesia has focused on automated recognition of anatomical structures during ultrasound-guided regional anesthesia or line insertion; recognition of the glottic opening and vocal cords during video laryngoscopy; prediction of the difficult airway using facial images; and clinical alerts for endobronchial intubation detected on chest radiograph. Current machine vision applications measuring the distance between endotracheal tube tip and carina have demonstrated noninferior performance compared to board-certified physicians. The performance and potential uses of machine vision for anesthesia will only grow with the advancement of underlying machine vision algorithm technical performance developed outside of medicine, such as convolutional neural networks and transfer learning. This article summarizes recently published works of interest, provides a brief overview of techniques used to create machine vision applications, explains frequently used terms, and discusses challenges the specialty will encounter as we embrace the advantages that this technology may bring to future clinical practice and patient care. As machine vision emerges onto the clinical stage, it is critically important that anesthesiologists are prepared to confidently assess which of these devices are safe, appropriate, and bring added value to patient care.

Survival analysis for pediatric heart transplant patients using a novel machine learning algorithm: A UNOS analysis.

BACKGROUND: Impact of pretransplantation risk factors on mortality in the first year after heart transplantation remains largely unknown. Using machine learning algorithms, we selected clinically relevant identifiers that could predict 1-year mortality after pediatric heart transplantation. METHODS: Data were obtained from the United Network for Organ Sharing Database for years 2010-2020 for patients 0-17 years receiving their first heart transplant (N = 4150). Features were selected using subject experts and literature review. Scikit-Learn, Scikit-Survival, and Tensorflow were used. A train:test split of 70:30 was used. N-repeated k-fold validation was performed (N = 5, k = 5). Seven models were tested, Hyperparameter tuning performed using Bayesian optimization and the concordance index (C-index) was used for model assessment. RESULTS: A C-index above 0.6 for test data was considered acceptable for survival analysis models. C-indices obtained were 0.60 (Cox proportional hazards), 0.61 (Cox with elastic net), 0.64 (gradient boosting), 0.64 (support vector machine), 0.68 (random forest), 0.66 (component gradient boosting), and 0.54 (survival trees). Machine learning models show an improvement over the traditional Cox proportional hazards model, with random forest performing the best on the test set. Analysis of the feature importance for the gradient boosted model found that the top 5 features were the most recent serum total bilirubin, the travel distance from the transplant center, the patient body mass index, the deceased donor terminal Serum glutamic pyruvic transaminase/Alanine transaminase (SGPT/ALT), and the donor PCO2. CONCLUSIONS: Combination of machine learning and expert-based methodology of selecting predictors of survival for pediatric heart transplantation provides a reasonable prediction of 1- and 3-year survival outcomes. SHapley Additive exPlanations can be an effective tool for modeling and visualizing nonlinear interactions.

A machine-learning approach for decision support and risk stratification of pediatric perioperative patients based on the APRICOT dataset.

BACKGROUND: Pediatric anesthesia has evolved to a high level of patient safety, yet a small chance remains for serious perioperative complications, even in those traditionally considered at low risk. In practice, prediction of at-risk patients currently relies on the American Society of Anesthesiologists Physical Status (ASA-PS) score, despite reported inconsistencies with this method. AIMS: The goal of this study was to develop predictive models that can classify children as low risk for anesthesia at the time of surgical booking and after anesthetic assessment on the procedure day. METHODS: Our dataset was derived from APRICOT, a prospective observational cohort study conducted by 261 European institutions in 2014 and 2015. We included only the first procedure, ASA-PS classification I to III, and perioperative adverse events not classified as drug errors, reducing the total number of records to 30 325 with an adverse event rate of 4.43%. From this dataset, a stratified train:test split of 70:30 was used to develop predictive machine learning algorithms that could identify children in ASA-PS class I to III at low risk for severe perioperative critical events that included respiratory, cardiac, allergic, and neurological complications. RESULTS: Our selected models achieved accuracies of >0.9, areas under the receiver operating curve of 0.6-0.7, and negative predictive values >95%. Gradient boosting models were the best performing for both the booking phase and the day-of-surgery phase. CONCLUSIONS: This work demonstrates that prediction of patients at low risk of critical PAEs can be made on an individual, rather than population-based, level by using machine learning. Our approach yielded two models that accommodate wide clinical variability and, with further development, are potentially generalizable to many surgical centers.

Examining Implicit Bias Differences in Pediatric Surgical Fellowship Letters of Recommendation Using Natural Language Processing.

OBJECTIVE: We analyzed the prevalence and type of bias in letters of recommendation (LOR) for pediatric surgical fellowship applications from 2016-2021 using natural language processing (NLP) at a quaternary care academic hospital. DESIGN: Demographics were extracted from submitted applications. The Valence Aware Dictionary for sEntiment Reasoning (VADER) model was used to calculate polarity scores. The National Research Council dataset was used for emotion and intensity analysis.  The Kruskal-Wallis H-test was used to determine statistical significance.  SETTING: This study took place at a single, academic, free standing quaternary care children's hospital with an ACGME accredited pediatric surgery fellowship. PARTICIPANTS: Applicants to a single pediatric surgery fellowship were selected for this study from 2016 to 2021. A total of 182 individual applicants were included and 701 letters of recommendation were analyzed. RESULTS: Black applicants had the highest mean polarity (most positive), while Hispanic applicants had the lowest.  Overall differences between polarity distributions were not statistically significant.   The intensity of emotions showed that differences in "anger" were statistically significant (p=0.03).  Mean polarity was higher for applicants that successfully matched in pediatric surgery. DISCUSSION: This study identified differences in LORs based on racial and gender demographics submitted as part of pediatric surgical fellowship applications to a single training program. The presence of bias in letters of recommendation can lead to inequities in demographics to a given program. While difficult to detect for humans, natural language processing is able to detect bias as well as differences in polarity and emotional intensity. While the types of emotions identified in this study are highly similar among race and gender groups, the intensity of these emotions revealed differences, with "anger" being most significant. CONCLUSION: From this work, it can be concluded that bias in LORs, as reflected as differences in polarity, which is likely a result of the intensity of the emotions being used and not the types of emotions being expressed.   Natural language processing shows promise in identification of subtle areas of bias that may influence an individual's likelihood of successful matching.

Feasibility of Conducting Long-term Health and Behaviors Follow-up in Adolescents: Longitudinal Observational Study.

BACKGROUND: Machine learning uses algorithms that improve automatically through experience. This statistical learning approach is a natural extension of traditional statistical methods and can offer potential advantages for certain problems. The feasibility of using machine learning techniques in health care is predicated on access to a sufficient volume of data in a problem space. OBJECTIVE: This study aimed to assess the feasibility of data collection from an adolescent population before and after a posterior spine fusion operation. METHODS: Both physical and psychosocial data were collected. Adolescents scheduled for a posterior spine fusion operation were approached when they were scheduled for the surgery. The study collected repeated measures of patient data, including at least 2 weeks prior to the operation and 6 months after the patients were discharged from the hospital. Patients were provided with a Fitbit Charge 4 (consumer-grade health tracker) and instructed to wear it as often as possible. A third-party web-based portal was used to collect and store the Fitbit data, and patients were trained on how to download and sync their personal device data on step counts, sleep time, and heart rate onto the web-based portal. Demographic and physiologic data recorded in the electronic medical record were retrieved from the hospital data warehouse. We evaluated changes in the patients' psychological profile over time using several validated questionnaires (ie, Pain Catastrophizing Scale, Patient Health Questionnaire, Generalized Anxiety Disorder Scale, and Pediatric Quality of Life Inventory). Questionnaires were administered to patients using Qualtrics software. Patients received the questionnaire prior to and during the hospitalization and again at 3 and 6 months postsurgery. We administered paper-based questionnaires for the self-report of daily pain scores and the use of analgesic medications. RESULTS: There were several challenges to data collection from the study population. Only 38% (32/84) of the patients we approached met eligibility criteria, and 50% (16/32) of the enrolled patients dropped out during the follow-up period-on average 17.6 weeks into the study. Of those who completed the study, 69% (9/13) reliably wore the Fitbit and downloaded data into the web-based portal. These patients also had a high response rate to the psychosocial surveys. However, none of the patients who finished the study completed the paper-based pain diary. There were no difficulties accessing the demographic and clinical data stored in the hospital data warehouse. CONCLUSIONS: This study identifies several challenges to long-term medical follow-up in adolescents, including willingness to participate in these types of studies and compliance with the various data collection approaches. Several of these challenges-insufficient incentives and personal contact between researchers and patients-should be addressed in future studies.

Secondary structure of peptides mimicking the Gly-rich regions of major ampullate spidroin protein 1 and 2.

Spider dragline silk has highly desirable material properties, possessing high extensibility, strength, and biocompatibility. Before it is spun, the constituent proteins are stored in a concentrated dope that is void of fibrils. To investigate the structural properties of the amorphous fiber regions in the dope, computer simulations were performed on model peptides representing the N. clavipes Gly-rich regions. Analysis of the secondary structure found predominantly turns, bends and coils; a small 31-helical population decreased with increasing concentration. Interestingly, the population of 31-helices saw a large increase in octanol. These results indicate that the unusual 31-helical secondary structure of the Gly-rich region of the fiber is a consequence of the spinning process, and that the low dielectric environment of the fiber may assist in favoring this structure.

How do we measure organisational wellness? Development of a comprehensive patient-centred and employee-centred visual analytical solution.

BACKGROUND: Dashboards are visual information systems frequently employed by healthcare organisations to track key quality improvement and patient safety performance metrics. The typical healthcare dashboard focuses on specific metrics, disease processes or units within a larger healthcare organisation. Here, we describe the development of a visual analytical solution (keystone dashboard) for monitoring an entire healthcare organisation. METHODS: The improvement team reviewed and assessed various data sources across the organisation and selected a group of patient and employee related metrics that afforded a broad overview of the organisation's well-being. Metrics spanned the organisation and included data from patient safety, quality improvement, human resources, risk management and medical staff affairs. Each metric was assigned a numeric weight that correlated with its impact. A visual model incorporating the various data fields was then constructed. RESULTS: The keystone dashboard incorporates a data heatmap and density visualisation to emphasis areas of higher density and/or weighted values. The heatmap is used to indicate the weight/magnitude of each metric within a data range in two dimensions: location and time. The visualisation 'heats up' depending on the combination of counts events and their assigned impact for the reporting month. Most data sources update in near real time. SUMMARY: The keystone dashboard serves as a comprehensive and collaborative integration of data from patient safety, quality improvement, human resources, risk management and medical staff affairs. This visual analytical solution incorporates and analyses metrics into a single view with the intent of providing valuable insight into the health of an entire organisation. This dashboard is unique as it provides a broad overview of a healthcare organisation by incorporating key metrics that span the organisation.

Impact of Coronavirus-2019 On Pediatric and Adult Heart Transplantation Waitlist Activity and Mortality in The United States: A Descriptive Approach.

BACKGROUND: Transplant centers saw a substantial reduction in deceased donor solid organ transplantation since the beginning of the coronavirus 2019 (COVID-19) pandemic in the United States. There is limited data on the impact of COVID-19 on adult and pediatric heart transplant volume and variation in transplant practices. We hypothesized that heart transplant activity decreased during COVID-19 with associated increased waitlist mortality. METHODS: The United Network for Organ Sharing (UNOS) database was used to identify patients at the time of listing for heart transplant from 2017-2020. Patients were categorized as pediatric (<18 years) or adult (≥18 years) and as pre-COVID (2017-2019) or post-COVID (2020). Regional and statewide data were taken from United States Census Bureau. CovidActNow project was used to obtain COVID-19 mortality rates. FINDINGS: Among pediatric patients, average time on the waiting list decreased by 28 days. Even though the average number of pediatric transplants (n=39 per month) did not change significantly during 2020, there was a temporal decline in the first quarter of 2020 followed by a sharp increase. Overall absolute pediatric waitlist mortality decreased from 5•31 to 4•73, however female mortality increased by 2%. Regional differences in pediatric mortality were observed: Northeast, decreased by 7•5%; Midwest, decreased by 9%; West, increased by 3•5%; and South, increased by 13%. North Dakota (0•55), Oklahoma (0•21) and Hawaii (0•33) showed higher mortality than other states per 100,000. In adults, average time on waiting list increased by 40 days and there was an increase in the number of transplants from 242 to 266. Adult waitlist mortality had a larger decrease, 18•44 to 15•70, with an increase in female mortality of 7%. Regional differences in adult mortality were also observed: Northeast, decreased by 3%; Midwest, increased by 5•5%; West, increased by 4•5% and South, decreased by 5%. Iowa (0•37), Wyoming (0•22), Arkansas (0•18) and Vermont (0•19) had the highest mortality per 100,000 compared to the other states. INTERPRETATION: Pediatric heart transplant volume declined in early 2020 followed by a later increase, while adult transplant volume increased all year round. Although, overall pediatric waitlist mortality decreased, female waitlist mortality increased for both adults and pediatrics. Regional differences in waitlist mortality were observed for both pediatrics and adults. Future studies are needed to understand this initial correlation and to determine the impact of COVID-19 on heart transplant recipients. FUNDING: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Orthogonal Halogen-Bonding-Driven 3D Supramolecular Assembly of Right-Handed Synthetic Helical Peptides.

Peptide-mediated self-assembly is a prevalent method for creating highly ordered supramolecular architectures. Herein, we report the first example of orthogonal C-X⋅⋅⋅X-C/C-X⋅⋅⋅π halogen bonding and hydrogen bonding driven crystalline architectures based on synthetic helical peptides bearing hybrids of l-sulfono-γ-AApeptides and natural amino acids. The combination of halogen bonding, intra-/intermolecular hydrogen bonding, and intermolecular hydrophobic interactions enabled novel 3D supramolecular assembly. The orthogonal halogen bonding in the supramolecular architecture exerts a novel mechanism for the self-assembly of synthetic peptide foldamers and gives new insights into molecular recognition, supramolecular design, and rational design of biomimetic structures.

Hydrogen-Bonding-Driven 3D Supramolecular Assembly of Peptidomimetic Zipper.

Hydrogen-bonding-driven three-dimensional (3D) assembly of a peptidomimetic zipper has been established for the first time by using an α/AApeptide zipper that assembles into a de novo lattice arrangement through two layers of hydrogen-bonded linker-directed interactions. Via a covalently bridged 1D 413-helix, drastic enhancement in stability has been achieved in the formed 3D crystalline supramolecular architecture as evidenced by gas-sorption studies. As the first example of an unnatural peptidic zipper, the dimensional augmentation of the zipper differs from metal-coordinated strategies, and may have general implications for the preparation of peptidic functional materials for a variety of future applications.

Phosphorylation induced cochaperone unfolding promotes kinase recruitment and client class-specific Hsp90 phosphorylation.

During the Hsp90-mediated chaperoning of protein kinases, the core components of the machinery, Hsp90 and the cochaperone Cdc37, recycle between different phosphorylation states that regulate progression of the chaperone cycle. We show that Cdc37 phosphorylation at Y298 results in partial unfolding of the C-terminal domain and the population of folding intermediates. Unfolding facilitates Hsp90 phosphorylation at Y197 by unmasking a phosphopeptide sequence, which serves as a docking site to recruit non-receptor tyrosine kinases to the chaperone complex via their SH2 domains. In turn, Hsp90 phosphorylation at Y197 specifically regulates its interaction with Cdc37 and thus affects the chaperoning of only protein kinase clients. In summary, we find that by providing client class specificity, Hsp90 cochaperones such as Cdc37 do not merely assist in client recruitment but also shape the post-translational modification landscape of Hsp90 in a client class-specific manner.

One-Bead-Two-Compound Thioether Bridged Macrocyclic γ-AApeptide Screening Library against EphA2.

Identification of molecular ligands that recognize peptides or proteins is significant but poses a fundamental challenge in chemical biology and biomedical sciences. Development of cyclic peptidomimetic library is scarce, and thus discovery of cyclic peptidomimetic ligands for protein targets is rare. Herein we report the unprecedented one-bead-two-compound (OBTC) combinatorial library based on a novel class of the macrocyclic peptidomimetics γ-AApeptides. In the library, we utilized the coding peptide tags synthesized with Dde-protected α-amino acids, which were orthogonal to solid phase synthesis of γ-AApeptides. Employing the thioether linkage, the desired macrocyclic γ-AApeptides were found to be effective for ligand identification. Screening the library against the receptor tyrosine kinase EphA2 led to the discovery of one lead compound that tightly bound to EphA2 (Kd = 81 nM) and potently antagonized EphA2-mediated signaling. This new approach of macrocyclic peptidomimetic library may lead to a novel platform for biomacromolecular surface recognition and function modulation.

Molecular dynamics simulations and molecular flooding studies of the retinoid X-receptor ligand binding domain.

Bexarotene is an FDA approved retinoid X-receptor (RXR) agonist for the treatment of cutaneous T-cell lymphoma, and its use in other cancers and Alzheimer's disease is being investigated. The drug causes serious side effects, which might be reduced by chemical modifications of the molecule. To rationalize known agonists and to help identify sites for potential substitutions we present molecular simulations in which the RXR ligand-binding domain was flooded with a large number of drug-like molecules, and molecular dynamics simulations of a series of bexarotene-like ligands bound to the RXR ligand-binding domain. Based on the flooding simulations, two regions of interest for ligand modifications were identified: a hydrophobic area near the bridgehead and another near the fused ring. In addition, positional fluctuations of the phenyl ring were generally smaller than fluctuations of the fused ring of the ligands. Together, these observations suggest that the fused ring might be a good target for the design of higher affinity bexarotene-like ligands, while the phenyl ring is already optimized. In addition, notable differences in ligand position and interactions between the RXRα and RXRß were observed, as well as differences in hydrogen bonding and solvation, which might be exploited in the development of subspecies-specific ligands.

Computer Simulations of the Retinoid X Receptor: Conformational Dynamics and Allosteric Networks.

As the heterodimerization partner for a large number of nuclear receptors, the retinoid X receptor (RXR) is important for a large and diverse set of biochemical pathways. Activation and regulation of RXR heterodimers is achieved by complex allosteric mechanisms, which involve the binding of ligands, DNA, coactivators and corepressors, and entail large and subtle conformational motions. Complementing experiments, computer simulations have provided detailed insights into the origins of the allostery by investigating the changes in structure, motion, and interactions upon dimerization, ligand and cofactor binding. This review will summarize a number of simulation studies that have furthered the understanding of the conformational dynamics and the allosteric activation and control of RXR complexes. While the review focuses on the RXR and RXR heterodimers, relevant simulation studies of other nuclear receptors will be discussed as well.

Secondary Structure Adopted by the Gly-Gly-X Repetitive Regions of Dragline Spider Silk.

Solid-state NMR and molecular dynamics (MD) simulations are presented to help elucidate the molecular secondary structure of poly(Gly-Gly-X), which is one of the most common structural repetitive motifs found in orb-weaving dragline spider silk proteins. The combination of NMR and computational experiments provides insight into the molecular secondary structure of poly(Gly-Gly-X) segments and provides further support that these regions are disordered and primarily non-ß-sheet. Furthermore, the combination of NMR and MD simulations illustrate the possibility for several secondary structural elements in the poly(Gly-Gly-X) regions of dragline silks, including ß-turns, 310-helicies, and coil structures with a negligible population of α-helix observed.

Importance of ß2-ß3 Loop Motion for the Increased Binding and Decreased Selectivity of the ΔLL Mutant of the Human Papillomavirus Type 6 E2 Protein.

The binding affinity of the human papillomavirus type 6 E2 protein is strongly mediated by the sequence of the DNA linker region, with high affinity for the AATT linker and low affinity for the CCGG linker. When two terminal leucine residues are removed from the protein, the level of binding to both strands increases, but unequally, resulting in a significant decrease in selectivity for the AATT linker strand. To rationalize this behavior, we performed molecular dynamics simulations of the wild-type and mutant protein in the apo state and bound to DNA with high-affinity AATT and low-affinity CCGG linker strands. While no stable contacts were made between the ß2-ß3 loop and DNA in the wild type, this loop was repositioned in the mutant complexes and formed electrostatic contacts with the DNA backbone. More contacts were formed when the mutant was bound to the CCGG linker strand than to the AATT linker strand, resulting in a more favorable change in interaction energy for the CCGG strand. In addition, significant differences in correlated motions were found, which further explained the differences in binding. The simulations suggest that ß2-ß3 loop motions are responsible for the increased affinity and decreased selectivity of the mutant protein.