Danny Tzu-Yu Wu

Purpose: Visual analytics is a nascent and underexplored area in the medical education literature that can assist in synthesizing large volumes of data and information to provide actionable insight. This scoping review aims to identify: 1) the characteristics of the studies that have applied visual analytics in medical education, 2) how visual analytics are being used to support medical education efforts. Method: This study followed the Joanna Briggs Institute (JBI) scoping review guidelines to search four databases in mid-April 2026: PubMed, IEEE Xplore, Scopus, ACM Digital, along with Google Scholar. Key search terms included “medical education” and “visual analytics”; papers that did not implement visual analytics systems in the medical education context were excluded. Variables of the eligible papers were extracted and further analyzed in terms of study characteristics (publication year, study location, primary focus & level of trainees assessed) and application of visual analytics (type of visualization, framework guidance, hardware/software, evaluation technique). Results: Of the 39 eligible papers from 584 screened, 20 (51.3%) were published in the United States. Using visual analytics for learner competency assessment and quality improvement of training programs were the primary focuses. For trainees, nearly all papers focused on assessing residents and medical students. More than 28 (71.8%) studies utilized an interactive tool/dashboard and/or applied an evaluation technique. Only 14 (35.9%) involved a framework-guided design of the visualization. Conclusions: This review highlights that while visual analytics is maturing in medical education, systemic application requires adopting framework-guided visualization design, expertise-matched tools, and goal-oriented evaluation. Future research should prioritize interoperable frameworks for knowledge sharing, controlled field studies assessing impacts on learner competencies, and patient outcomes, and "cross-continuum" dashboards spanning pre-medical through faculty levels. These strategies allow educators and learners to make sense of complex educational data through visual analytics dashboards, ultimately enhancing assessment and patient care quality.

Unplanned cancer readmissions present a significant burden on patients and hospitals. Current predictive models often overlook socioeconomic factors such as social determinants of health (SDoH), which have the potential to improve prediction performance, as measured by the Area Under the Receiver Operating Characteristic (AUROC) and the Precision-Recall Curve (AUPRC). To investigate this, the present study developed predictive models using cancer readmission data from a large health system in Hamilton County, OH. The models incorporated geocoding-based SDoH along with clustering techniques and compared machine learning (ML) and deep learning (DL) algorithms. Overall, models, regardless of algorithm type, not using SDoH variables had higher AUROC and AUPRCs. The best-performing ML and DL models are comparable (AUROC = 0.7605 for ML; AUROC = 0.7585 for DL). However, when top-performing models were evaluated across certain organ and system cancers, using SDoH and clustering techniques significantly improved model performance. This was most notable for cancers of the skin, subcutaneous tissue, and breast with improvements of 8.20% in AUROC and 11.04% in AUPRC. For all cancer patient cases, utilizing individualized SDoH information extracted from clinical notes was recommended for future studies.

This study evaluated the readability of ClinicalTrials.gov trial information using traditional readability measures (TRMs) and compared it to summaries generated by large language models (LLMs), specifically ChatGPT and a fine-tuned BART-Large-CNN (FBLC). The study involved: 1) assessing required reading levels (RRL) with TRMs, 2) generating sample LLM-based summaries, and 3) evaluating summary quality based on scores provided by two independent reviewers. The results show that the original ClinicalTrials.gov trial descriptions were scored above the recommended readability level. In contrast, ChatGPT-generated summaries had significantly lower RRLs and higher quality scores. We conclude that ChatGPT shows great promise of creating readable, high-quality summaries. Future research is warranted to assess whether LLMs could be a viable solution to improve the readability of ClinicalTrials.gov to facilitate comprehension by laypersons.

BACKGROUND AND OBJECTIVES: Traumatic brain injury (TBI) affects over 69 million people worldwide, most of whom suffer a so-called “mild” injury. Patients with mild TBI, defined as a Glasgow Coma Scale (GCS) of 13-15 on presentation, often have a significant symptom burden as detected by the Rivermead Post-Concussion Questionnaire (RPQ). In this study, we aimed to determine whether social determinants of health (SDOH) may influence patient self-report of symptoms within a month of injury. METHODS: Patients presenting to an academic Level I trauma center with GCS 13-15 were included in the study with data collected as part of a prospectively maintained neurotrauma registry. Overall, 451 individuals completed the RPQ at a follow-up clinic visit. Demographic variables, injury characteristics, comorbidities, and geocoded SDOH information were captured from the electronic medical record. Multivariable regression analysis was performed. RESULTS: Variables contributing to increased symptom burden on the RPQ included sex, living in an area with a high fraction of poverty, history of depression or anxiety, initial GCS score, history of illicit drug use, obesity, and positive head CT. CONCLUSION: In addition to the expected patient and injury characteristics, fraction living in poverty was a significant variable contributing to scores on the RPQ-3 and RPQ-13. Efforts to incorporate screening for SDOH factors should be considered to identify patients at risk of poor recovery after mild TBI.

The above article published in Applied Clinical Informatics Open, Volume 9, Number 2, e1-e11, on July 9, 2025 (doi: ) was published with incorrect page numbers. The page numbers have been corrected subsequently with the publication of this erratum. Article published online: 04 August 2025 © 2025. The Author(s). This is an open access article published by Thieme under the terms of the Creative Commons Attribution License, permitting unrestricted use, distribution, and reproduction so long as the original work is properly cited. (https://creativecommons.org/licenses/by/4.0/) Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany

Project management is crucial in academic hospitals due to the intensive involvement in research and clinical trials. Designing an application in a user-centered and workflow-compatible manner can prevent issues arising from hospital's limited resources and various risks (e.g., financial risk and ethical misconduct).This study aimed to (1) develop a hybrid method combining user-centered design (UCD) and agile software development (ASD) principles, (2) apply the hybrid model to develop a Clinical Research Project Management System (CRPMS), (3) assess the usability of the CRPMS and iteratively refine the application.A CRPMS was developed following the UCD and ASD principles and supported by a research core in a pediatric heart institute. In Phase 0, semi-structured interviews were conducted to understand processes and bottlenecks. In Phase 1, the project management and budgeting sections of the CRPMS were simultaneously developed. In Phase 2, the Principal Investigator dashboard and the intake section of the CRPMS were developed. Usability evaluation metrics included the System Usability Scale (SUS), Single Ease Question, and severity scores.In Phase 1, the average SUS score was 88.65. There were 126 usability issues and 68 were considered high severity due to the SUS score cutoff (>1.5). In Phase 2, the average SUS score was 87.1. All of the 71 usability issues were addressed during the iterative refinements.A workflow-compatible and highly user-friendly CRPMS was developed by employing a hybrid Agile UCD model. Future work includes continuing development and expanding the CRPMS within and outside the institute.

Fundamental knowledge and skills in biomedical and health informatics (BMHI) are critical for the next generation of healthcare professionals. Implementing new programs, especially those targeted toward undergraduate students who are studying to become physicians, must be flexible and fit in an already busy schedule. Here, we share our experience and approach at the University of Cincinnati College of Medicine as we created an informatics experience for pre-medical and pre-healthcare students in the Medical Sciences Baccalaureate Program (MSBP).To create an informatics experience that supports development of students while allowing them to achieve their personalized educational goals, three strategies were considered to maximize benefit to students and ensure easy implementation of this program: (1) Synergy with an already busy schedule, (2) modularity of course and experiential offerings, and (3) measurable outcomes that are based on core competencies of medical informatics.By utilizing existing course offerings and experiential learning programs, our team created new informatics-focused research experiences and further created an undergraduate certificate that provides students with a structured yet personalized path to train on medical informatics. Specifically, the informatics experience included two research courses, three summer programs, 1 year-long program, and targeted three presentation venues.This article chronicles the 5-year effort of our team to create an informatics experience for pre-medical students in the MSBP. We hope that other pre-medical programs can learn from our achievement to create their own informatics experience and tailor it to the needs of their students and resources of their institution.

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: The note corpus consisted of internal medicine resident admission notes (retrospective set: July 2020-December 2021, n=700 NYU and 450 UC notes and prospective validation set: July 2023-December 2023, n=155 NYU and 92 UC notes). Clinicians rated CR documentation quality in each note using a previously validated tool (Revised-IDEA), on 3-point scales across 2 domains: differential diagnosis (D0, D1, and D2) and explanation of reasoning, (EA0, EA1, and EA2). At NYU, the retrospective set was annotated for NER for 5 entities (diagnosis, diagnostic category, prioritization of diagnosis language, data, and linkage terms). Models were developed using different artificial intelligence approaches, including NER, logic-based model: a large word vector model (scispaCy en_core_sci_lg) with model weights adjusted with backpropagation from annotations, developed at NYU with external validation at UC, NYUTron LLM: an NYU internal 110 million parameter LLM pretrained on 7.25 million clinical notes, only validated at NYU, and GatorTron LLM: an open source 345 million parameter LLM pretrained on 82 billion words of clinical text, fined tuned on NYU retrospective sets, then externally validated and further fine-tuned at UC. Model performance was assessed in the prospective sets with F1-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: At NYU, the NYUTron LLM performed best: the D0 and D2 models had AUROC/AUPRC 0.87/0.79 and 0.89/0.86, respectively. The D1, EA0, and EA1 models had insufficient performance for implementation (AUROC range 0.57-0.80, AUPRC range 0.33-0.63). For the D1 classification, the approach pivoted to a stepwise approach taking advantage of the more performant D0 and D2 models. For the EA model, the approach pivoted to a binary EA2 model (ie, EA2 vs not EA2) with excellent performance, AUROC/AUPRC 0.85/ 0.80. At UC, the NER, D-logic–based model was the best performing D model (F1-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: Continuous renal replacement therapy (CRRT) is the favored renal replacement therapy in critically ill patients. Predicting clinical outcomes for CRRT patients is difficult due to population heterogeneity, varying clinical practices, and limited sample sizes. OBJECTIVE: We aimed to predict survival to ICUs and hospital discharge in children and young adults receiving CRRT using machine learning (ML) techniques. DERIVATION COHORT: Patients less than 25 years of age receiving CRRT for acute kidney injury and/or volume overload from 2015 to 2021 (80%). VALIDATION COHORT: Internal validation occurred in a testing group of patients from the dataset (20%). PREDICTION MODEL: Retrospective international multicenter study utilizing an 80/20 training and testing cohort split, and logistic regression with L2 regularization (LR), decision tree, random forest (RF), gradient boosting machine, and support vector machine with linear kernel to predict ICU and hospital survival. Model performance was determined by the area under the receiver operating characteristic curve (AUROC) and the area under the precision-recall curve (AUPRC) due to the imbalance in the dataset. RESULTS: Of the 933 patients included in this study, 538 (54%) were male with a median age of 8.97 years and interquartile range (1.81–15.0 yr). The ICU mortality was 35% and hospital mortality was 37%. The RF had the best performance for predicting ICU mortality (AUROC, 0.791 and AUPRC, 0.878) and LR for hospital mortality (AUROC, 0.777 and AUPRC, 0.859). The top two predictors of ICU survival were Pediatric Logistic Organ Dysfunction-2 score at CRRT initiation and admission diagnosis of respiratory failure. CONCLUSIONS: These are the first ML models to predict survival at ICU and hospital discharge in children and young adults receiving CRRT. RF outperformed other models for predicting ICU mortality. Future studies should expand the input variables, conduct a more sophisticated feature selection, and use deep learning algorithms to generate more precise models.