For patients grappling with complex medical conditions, seamless transitions of care (TOC) are crucial to maintaining the continuity of their treatment. The COVID-19 pandemic's effects might unexpectedly strain healthcare organizations' usual operations, potentially placing a load on the TOC system. A network analytical approach is used in this study to evaluate and quantify the evolution of TOC structures, contrasting the pre-COVID-19 and intra-COVID-19 contexts. Care transitions of 5674 inpatients (2699 prior to COVID-19 and 2975 during COVID-19) admitted to Vanderbilt University Medical Center (VUMC) between January 2019 and May 2021 were investigated using a trauma registry repository. By analyzing network metrics such as assortativity, homophily, and small-world properties, the structures of TOC and their changes were quantified. TOC structures, demonstrating disassortative tendencies, homophily, and small-world characteristics, persisted before and during the COVID-19 period. The pandemic's influence on these properties was minimal. Transferring patients between highly-connected triage centers and less-connected receivers is facilitated by the disassortative structure of the TOC; the homophily characteristic reveals that similar patient types are often served by linked care units; and the small-world nature of the network means that transfer routes to highly collaborative units are generally short.Over 78 million individuals are projected to experience dementia by 2030, emphasizing the critical need for early detection in at-risk patients with mild cognitive impairment (MCI) and personalized clinical evaluation steps to diagnose potentially reversible causes. Machine learning models, developed using real-world electronic health records from the observational medical outcomes partnership (OMOP) data model, are employed to forecast MCI one year prior to the documented diagnosis. More than 531,000 patient visits were used to train and evaluate logistic regression, random forest, and XGBoost models. The findings suggest the random forest model achieved an ROC-AUC of 0.68207 in anticipating MCI onset. Through an examination of clinician notes, we isolate the clinical elements most predictive of MCI diagnosis. Employing analogous association mining methods, we construct a data-driven inventory of clinical procedures frequently requested during MCI case investigations, which can serve as a foundation for creating guidelines and standardized clinical order templates.In the United States, the leading cause of blindness, especially impacting African Americans and Hispanics, is primary open-angle glaucoma (POAG). In the detection of POAG, deep learning utilizing fundus images demonstrates a performance that is equal to, or exceeding, that of medical professionals conducting diagnoses. Deep learning models, commonly employed, can unfortunately reproduce and magnify human bias prevalent in clinical diagnosis, which subsequently influences their performance metrics. Prejudices can result in (1) overlooking a condition's presence, thus raising the probability of delayed or inadequate treatment, and (2) wrongly identifying a condition, which might contribute to increased stress, fear, diminished well-being, and needless/costly intervention. This study investigated the issues of underdiagnosis and overdiagnosis when employing deep learning for POAG detection, drawing data from the Ocular Hypertension Treatment Study (OHTS) collected from 22 centers in 16 US states. Our results demonstrate that the widely utilized deep learning model can lead to inaccurate diagnoses, potentially underdiagnosing or overdiagnosing members of underserved communities. The female group below the age of 60 years is the group most prone to underdiagnosis, while the Black group of 60 years and above is the most overdiagnosed group. Disease identification and treatment in ophthalmology clinics using traditional deep learning models might suffer from biased diagnoses, creating an uneven impact on underserved populations and potentially delaying care, consequently raising ethical concerns.Man-made compounds known as per- and polyfluoroalkyl substances (PFAS) exhibit human toxicity and have been detected in US drinking water supplies. In order to classify PFAS exposure levels for our patients in New Jersey, our electronic health record was supplemented with geospatial information. Three popular PFAS exposure classification methods, as seen in the literature, were tested, resulting in the identification of differing boundary types: public water supplier service areas, municipal boundaries, and ZIP code delineations. We delved into the convergence zone of the three limits. In our effort to identify potential bias, we investigated well-established connections between PFAS exposure and specific medical conditions, encompassing hypertension, thyroid disease, and parathyroid disease. The classification method for PFAS exposure proved influential in determining the statistical significance and the effect size of the associations. Our study revealed a higher prevalence of PFAS exposure among Black/African-American patients in various cohorts, highlighting the critical connection between knowledge discovery and environmental justice.Cancer patients on chemotherapy frequently experience functional decline and physical impairments, significantly impacting their lives. While exercise is demonstrably helpful in addressing these symptoms, cancer patients' access to exercise-based rehabilitation is often constrained. Remote rehabilitation initiatives have yielded promising outcomes in reducing barriers to care. A deep understanding of how cancer patients perceive telerehabilitation is critical for developing user interfaces and features that truly center on the patient experience. This study investigated patient viewpoints and experiences with a mobile cancer telerehabilitation system, employing a step-by-step walkthrough for data collection. Semi-structured qualitative interviews with 29 chemotherapy-receiving cancer patients were conducted after the demonstration. Employing a deductive thematic analysis, the researchers identified patterns and themes within the interviews. atezolizumab inhibitor Patients' feedback on the telerehabilitation system was overwhelmingly positive, specifically citing its user-friendliness and ease of access as major benefits. Patients demonstrating a limited understanding of technology were able to adjust to the system with very few challenges. Patients' needs and preferences concerning the telerehabilitation system's interface and functionality were detailed in the thematic analysis results. The upcoming design and execution of a patient-centric cancer telerehabilitation system will take these valuable insights into account.On any given night of 2021, more than half a million people in America were without a home, however, only approximately 50 percent of them utilized the shelter options available to them. To tackle the homelessness crisis, we report the development of H4H, a centralized repository of emergency shelters and housing support, which allowed us to deploy natural language processing to extract data on admission, available services, stay duration, and eligibility criteria specific to the homeless population. Information extraction is formulated as a question-and-answer procedure. A two-step classification and question-answering Roberta model, enhanced by prompting, emerged as the top-performing system, achieving a macro-average F1 score of 7583 on the 2055 question-answer pairs used for training and evaluation. Publicly available as a benchmark dataset, H4H provides access to its annotated entries.Over the past eight years, the rise in deaths from stroke has compelled clinicians to seek data-supported tools for informed decision-making. Recently, prediction models built upon deep learning and refined with detailed electronic health record (EHR) data, have displayed superior promise in predicting health outcomes. Exploration of deep learning models, drawing from electronic health records, for predicting hemorrhagic stroke outcomes remains comparatively limited. An ensemble deep learning approach is proposed in this paper to forecast early mortality among ICU patients with hemorrhagic stroke. The ensemble model's performance, reflected in an accuracy of 83%, was substantially better than the fusion model and various baseline methods, encompassing logistic regression, decision trees, random forests, and XGBoost. The ensemble model's prediction mechanisms were elucidated by SHAP values, which helped identify influential features in the prediction process. Moreover, the paper employs the MINIMAR (Minimum Information for Medical Artificial Intelligence Reporting) standard, marking a crucial advance in establishing trust between artificial intelligence systems and medical professionals.The Healthy People 2030 initiative, the fifth iteration of U.S. national health goals, centers on health literacy. Difficulty in comprehending health information, adhering to post-visit care instructions, and utilizing prescribed medications are common issues faced by individuals with limited health literacy, which can have negative consequences on health outcomes and contribute to health disparities. Natural language processing methods are proposed in this study to improve health literacy in patient education materials, specifically by automatically translating illiterate languages in given sentences. The online health information websites MedlinePlus.gov and three others were the source of the patient education materials that were scraped. Consult Drugs.com for comprehensive drug information. Mayoclinic.org and Reddit.com provide diverse types of information, from health details to online discussions. We trained state-of-the-art neural machine translation (NMT) models using a silver standard training dataset and tested them using a gold standard testing dataset, respectively. BiLSTM NMT models, as shown by the experimental results, achieved better performance than BERT-based NMT models. The effectiveness of NMT models for translating health-illiterate languages was scrutinized by comparing the ratio of health-illiterate expressions in the sentences.