KCKelly Cho

Kelly Cho

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Kelly Cho has a total of 776 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Lifestyle factors and the incidence of premature major adverse cardiovascular events in the VA Million Veteran Program

    by Xuan-Mai T. Nguyen, Lindsay R. Salvati, Yanping Li, Serena C. Houghton, April R. Williams, Yuk-Lam Ho, Luc Djousse, John Michael Gaziano, Peter W.F. Wilson, Kelly Cho


  • Open Access

    Vitiligo-associated variants and genes beyond immune-related loci: A genome-wide association and Mendelian randomization meta-analysis

    by David Perez, Bryan L. Peacker, Jonathan C. Hwang, Liana N. Ly, Christy Zheng, Kai Gravel-Pucillo, Brian R. Ferolito, Kelly Cho, John M. Gaziano, Lee E. Wheless, et al.

  • Open Access

    Data-driven Prediction of Fifteen-Year All-Cause Mortality among 2.3 Million Individuals in the VA

    by Sayera Dhaubhadel, Judith D. Cohn, Tanmoy Bhattacharya, Ruy M. Ribeiro, Kumkum Ganguly, Nicolas Hengartner, Janet P. Tate, Lauren Costa, Yuk-Lam Ho, Kelly Cho, et al.

    We present a data-driven framework to predict 15-year all-cause mortality using outpatient administrative records for 2.3 million Veterans in the largest integrated U.S. healthcare system. Rather than relying on predefined clinical phenotypes, we used the 1,000 most common outpatient medical codes from each of three data types/modalities – ICD-9 (Dx), Current Procedural Terminology (CPT), and prescription drugs (Rx), encoded as binary features. Using these features, we trained three machine learning (ML) algorithms (logistic regression with lasso, random forest, and a 3-layered feed-forward neural network) to predict 15-year mortality risk. The features were also mapped to variables for the widely used Charlson Comorbidity Index (CCI), Elixhauser, and Veterans Aging Cohort Study (VACS) indices, refitted for 15-year mortality prediction, for baseline comparison. All our models significantly outperformed the widely used CCI, Elixhauser, and VACS indices, with C-statistics ranging from 0.82 to 0.84 versus 0.739–0.804 for the baselines. Relative improvements in C-statistics of our approach over the baselines were consistent across different subgroups (age groups of <65 years, those 65+years, Blacks, Hispanics, etc.) Our approach enabled the identification of high-impact predictors with clinical grounding, without requiring hand-curated phenotypes. Cardiovascular diseases and mental health diagnoses/treatments emerged as leading long-term mortality indicators. Using unsupervised ML techniques including PCA and K-means clustering, we associated interpretable patterns and complex interactions between diagnoses and treatments, highlighting comorbidities, disease trajectories, and healthcare utilization patterns. The ability to achieve the predictive performance and algorithmically detect such relationships purely from outpatient data supports the scalability and broad applicability of our framework. This framework not only improves mortality risk stratification over existing clinical indices, but also enables better understanding of how medical codes, regardless of category, interact to predict long-term outcomes.

  • Genome-Wide Association Study of Chronic Venous Insufficiency and Lymphedema in the Million Veteran Program

    by Gina M. Peloso, Nimish Adhikari, Melissa M. Young, Kelly Cho, Scott Kinlay, Sumitra Muralidhar, Jennifer Moser, Jennifer E. Deen, Philip S. Tsao, J. Michael Gaziano, et al.

    BACKGROUND: Lymphedema is a chronic condition characterized by the accumulation of fluid due to impaired lymphatic drainage, frequently occurring secondary to chronic venous insufficiency (CVI), malignancy, or obesity. Despite known environmental and clinical risk factors, the genetic contributors to lymphedema and CVI have been understudied. METHODS: We conducted a multipopulation genome-wide association study in participants of the Million Veteran Program without cancer to identify genetic variants associated with CVI, lymphedema, and their cooccurrence. Individuals were categorized into 3 case groups: CVI only (n=34 664), lymphedema only (n=2452), and lymphedema+CVI (n=3283) and were compared with 367 684 controls. RESULTS: We identified 11 genome-wide significant variants ( P −8 ) in the multipopulation analysis, including 8 for CVI only, 1 for lymphedema only, and 2 for lymphedema+CVI, and 2 population-specific genetic variants. Three independent variants replicated in the UK Biobank for CVI only near CASZ1 , SLC12A2 , and NDP . Polygenic risk scores derived from the Million Veteran Program were associated with CVI in the UK Biobank and phenome-wide associations of CVI-associated variants revealed pleiotropic associations with cardiometabolic traits. CONCLUSIONS: These findings enhance our understanding of the genetic architecture of lymphatic dysfunction.

  • Open Access

    Comparative Risk of Stroke Associated with GLP-1 Receptor Agonists and SGLT2 Inhibitors in Veterans with Type 2 Diabetes

    by Sophia C. Sun, Serena C. Houghton, Yanping Li, Xuan-Mai Nguyen, Luc Djousse, Kelly Cho, Hugo J. Aparicio, Peter W.F. Wilson

    Introduction: Stroke is a leading cause of disability and death in adults with type 2 diabetes (T2D). We evaluated the comparative stroke risk in Veterans with T2D initiated on either of two glucose-lowering medications: GLP-1 receptor agonists (GLP-1RA) or SGLT-2 inhibitors (SGLT2i). Patients and Methods: We conducted a retrospective cohort study on diabetic Veterans aged 40 and older with no prior history of stroke or transient ischemic attack, who started on a GLP-1RA or SGLT2i between 2014 and 2021. Patients with contraindications or prior exposure to medication were excluded. Using national Veteran health data, we identified 195,072 eligible individuals and followed them from treatment initiation until stroke, death, loss to follow up, or end of follow up, whichever came first. Primary outcome was incident stroke, and secondary outcomes included ischemic and hemorrhagic stroke. We applied Kaplan-Meier methods and Cox proportional hazards models. Adjusted associations were estimated using inverse probability weighting. Results: Both unadjusted and adjusted analyses suggest GLP-1RA users have reduced stroke incidence compared SGLT-2i users (HR = 0.67, 95%CI 0.64-0.69;HR = 0.72, 95%CI 0.69-0.75). Similar results were found in secondary outcome and stratified analyses, with GLP-1RA users having reduced stroke risk compared to SGLT2i users for all age groups, chronic kidney disease stages, and hemoglobin A1c levels. Discussion and Conclusion: GLP-1RA treatment was associated with a lower risk of stroke compared with SGLT2i treatment in Veterans with T2D. These findings were consistent for ischemic and hemorrhagic strokes, suggesting potential differences in stroke risk between the treatments.

  • Evaluating the feasibility of the CIPHER metadata framework towards building a conceptual phenotype standard

    by Jacqueline Honerlaw, Yuk-Lam Ho, Michael Murray, Francesca Fontin, Edward Zielinski, Ashley Galloway, Vasa Curcin, Martin Chapman, Wei-Qi Wei, Azza Shoaibi, et al.

    Objective: Phenotype libraries accelerate health research through the reuse of algorithms. However, multiple libraries exist with differing metadata standards. We describe the integration of libraries using shared conceptual standards to harmonize definitions. Materials and Methods: We identified a pilot set of definitions from the Observational Health Data Sciences and Informatics (OHDSI), Phenotype Knowledgebase (PheKB), and Health Data Research UK (HDR UK) phenotype libraries. Metadata were mapped to the Centralized Interactive Phenomics Resource (CIPHER) standard, integrated into the CIPHER library, and validated. Results: Twenty-seven phenotypes from OHDSI, 25 from PheKB, and 25 from HDR UK were mapped to the CIPHER standard and are available at https://phenomics.va.ornl.gov/. Discussion: We demonstrate the feasibility of integrating metadata across phenotype repositories and illustrate the CIPHER standard as a suitable initial framework for a global conceptual metadata model. Conclusion: Harmonizing phenotype metadata across libraries allows the reuse of definitions throughout the entire health data community.


  • Open Access

    AI in Clinical Decision Support Systems: Promising Applications and Strategies for Managing Data Challenges

    by Jennifer E Daly, Dursun Delen, Zheng Han, River Smith, Jacqueline Honerlaw, Kelly Cho, Bridget Bennett, Jennifer Sippel

    The translation of big data analytics and artificial intelligence (AI) into clinical decision support systems (CDSSs) has advanced from proof of concept to real-world clinical practice. AI-informed CDSSs show measurable improvements in diagnostic accuracy, risk stratification, resource use, and patient outcomes compared to traditional models, offering the potential to assist clinicians in managing symptom complexity and uncertainty in health care delivery. Despite this potential, access to large amounts of high-quality and granular data remains one of the most significant bottlenecks to AI-enabled CDSSs. We argue that as health care systems increasingly adopt data-driven decision support, addressing the challenges of data accessibility and protection is essential to realizing the full potential of AI in clinical medicine. We use selected case examples of AI-informed CDSSs in oncology, organ transplantation, diabetic retinopathy, epilepsy, spinal cord injury, rare disease diagnosis, and emergency medicine to illustrate opportunities and challenges related to AI’s potential to improve patient outcomes. We discuss public and semipublic, medical institutional and commercial, and government and national data sources that are currently available for the development of CDSSs and highlight the practical and ethical constraints associated with these data. We consider alternative data resources and ways in which health care systems can strengthen data ecosystems to increase AI-driven CDSS efficacy and implementation to improve patient outcomes.


  • Increasing value in the Veterans Affairs Healthcare System (VA) with precision health: a continuing landmark collaboration with the Department of Energy

    by Amy C Justice, Benjamin McMahon, Daniel A Jacobson, Kelly Cho, Anuj J Kapadia, Samuel M Aguayo, Zeynep H Gümüş, Ioana Danciu, Jean C Beckham, Nathan A Kimbrel, et al.

    Objective: Phase II of MVP-CHAMPION, a federal collaboration between the Veterans Affairs Healthcare System (VA) and the Department of Energy (DoE), leveraged large-scale clinical, geo-spatial, and genetic data with state-of-the-art artificial intelligence (AI), and high-performance computing (HPC) to improve value in healthcare. Materials and Methods: Eight clinical priority projects for which AI was a critical missing capability were initiated to address: lung cancer screening (MVP 061), suicide risk screening (MVP 062), cardiovascular risk in obstructive sleep apnea (MVP 063), checkpoint inhibitor toxicity (MVP 064), heart failure (MVP 065), renal complications in diabetes (MVP 066), post COVID-19 sequelae (MVP 067), and antipsychotic medication toxicity (MVP 068). Results: Building on a strong regulatory and administrative foundation, we developed multimorbidity-aware analytic frameworks, reusable computational tools, and analytic pipelines. These greatly facilitated identification of novel risk factors including genetic variants and specification of more discriminating prediction models. Novel genetic risk factors are informing development and repurposing of medications and discriminating prediction models promise to improve healthcare value. Discussion: The research foundation developed in Phase I and extended in Phase II of MVP CHAMPION has supported an unprecedented federal collaboration and yielded significant scientific advances. Our clinical findings are poised for near-term application, while advances in machine learning and high-performance computing may accelerate the broader adoption of artificial intelligence in healthcare. Conclusion: This maturing VA-DoE federal collaboration is poised to transform the future of Veterans’ healthcare and the broader national landscape of precision health.

  • Open Access

    X-Chromosome-Wide Association Study Identifies Novel Genetic Signals for Heart Failure and Subtypes

    by Junling Ren, Chang Liu, Qin Hui, Maryam Rahafrooz, Nicole M Kosik, Kevin Urak, Jennifer Moser, Sumitra Muralidhar, Alexandre C Pereira, Kelly Cho, et al.

    Background: Heart failure (HF) is a major and growing public health problem, and prior studies support a meaningful genetic contribution to HF susceptibility. Clinically, HF is commonly categorized into the major clinical sub-types of HF with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF), which differ in pathophysiology and clinical profiles. However, previous genome-wide association studies have focused on autosomal variation and have routinely excluded the X chromosome, leaving X-linked genetic contributions to HF and its subtypes under-characterized. Methods: We performed X-chromosome wide association study (XWAS) utilizing directly genotyped data from 590,568 Million Veteran Program participants, including 90,694 HF cases across European, African, Hispanic, and Asian Americans. Sex- and ancestry-stratified logistic regression was used with XWAS quality control measures, adjusting for age and population structure, followed by fixed-effects multi-ancestry meta-analysis. Functional annotation, gene-based testing, fine-mapping, and colocalization were performed. We replicated genetic associations with all-cause HF in the UK Biobank. Results: In the multi-ancestry meta-analysis, we identified five X-chromosome-wide significant loci for all-cause HF, five for HFrEF, and one locus for HFpEF in males. No loci reached significance in female-specific analyses. In sex-combined analyses, we identified six loci for all-cause HF and four for HFrEF. The strongest and most emphasized signals mapped to genes were BRWD3, FHL1 , and CHRDL1 . Ancestry-specific analyses revealed additional loci, including NDP and WDR44 in African ancestry and PHF8 in Hispanic ancestry. One locus, BRWD3 , was replicated in UK Biobank HF cohort. Integrated post-GWAS analyses (fine-mapping, colocalization and pleiotropy trait association studies) reinforced the biological plausibility of the X-linked signals. Conclusions: This multi-ancestry, sex-stratified XWAS identifies X-linked genetic contributions to HF and its subtypes and highlights the role of X-chromosome in heart failure pathogenesis.


  • Open Access

    Detecting Manuscripts Related to Computable Phenotypes Using a Transformer-based Language Model

    by Junghoon Chae, David Heise, Keith Connatser, Jacqueline Honerlaw, Monika Maripuri, Yuk-Lam Ho, Francesca Fontin, Vidisha Tanukonda, Kelly Cho

    Objective: The demand for a comprehensive phenomics library, which requires identifying computable phenotype definitions and associated metadata from an ever-expanding biomedical literature, presents a significant, labor-intensive, and unscalable challenge. To address this, we introduce a transformer-based language model specifically designed for identifying biomedical texts containing computable phenotypes and piloted its use in the Centralized Interactive Phenomics Resource (CIPHER) platform. Materials and Methods: We fine-tuned a BioBERT model using a labeled dataset of 396 manuscripts. The model incorporates our novel sliding-window approach to effectively overcome token-length limitations, thereby enabling accurate classification of full-length manuscripts. For scalable deployment and continuous refinement, we developed a cohesive framework that integrates a web-based user interface, a control server, and a classification module. Results: The staged approach for model development yielded a final model with 95% accuracy. The web-based user interface was deployed on the CIPHER platform and enables user feedback for model retraining. Discussion: We developed a model and user interface which are currently in use by data curators to identify computable phenotype definitions from the literature. Conclusion: Through this system, users can submit literature, assess classification results, and provide feedback directly influencing future model training, thereby offering an efficient and adaptive solution for accelerating phenotype-driven literature curation.