Yukang Zeng, MS

Zeng’s research focuses on causal inference for target populations and the use of statistical learning to answer questions in cardiovascular medicine. He develops methods for treatment-effect estimation when study samples differ from the populations of interest, and applies Bayesian modeling and machine learning to clinical outcomes. His work also examines how AI can support statistical study design and analysis.

He studies how survey sampling weights and propensity-score balancing weights can be combined to estimate treatment effects in defined target populations. His first-author Statistics in Medicine article examines estimation and uncertainty assessment in survey observational studies. Ongoing work develops survey-weighted double score matching using propensity and prognostic scores, and examines how matching dimension and individual weights affect estimation and inference. His related interests include population generalizability, transportability, and causal mediation under selection. He maintains PSweight and co-develops wdsmatch to make these methods available in R.

His master’s research developed hierarchical accelerated failure-time Bayesian tree models to integrate SPRINT MIND and NACC-UDS data for cognitive-impairment analyses. This work addressed data harmonization, censored outcomes, and subgroup differences, with attention to uncertainty and differences across data sources.

At CMAC, Zeng analyzes cardiovascular trials, observational cohorts, and electronic health record data. In TRANSFORM-HF, he applies accelerated failure-time Bayesian additive regression trees to study variation in survival treatment effects across patient profiles. He also develops XGBoost survival models to predict mortality after heart failure hospitalization, assessing discrimination and calibration. His win-ratio analyses of TRANSFORM-HF and pooled STICHES/REVIVED-BCIS2 data compare treatments using clinically prioritized outcomes. These analyses require clear estimands, appropriate handling of censoring and missing data, and careful assessment of model assumptions and uncertainty. Other projects examine antihypertensive treatment and cognitive outcomes, obesity and mortality, coronary vasomotor dysfunction, and cardiovascular imaging.

His broader clinical collaborations examine coronary thrombectomy and stroke risk, BMI-related treatment-effect modification in heart failure, and GLP-1 receptor agonists and hemodynamic trends in heart failure. Other projects study mental-stress coronary microvascular dysfunction, concordance between PET imaging and invasive thermodilution, and the design and reporting of pragmatic cardiovascular trials. He also contributes to research on reproductive health documentation in cardiovascular care, and breastfeeding patterns and socioeconomic disparities.

He develops Power Agent for analytical and simulation-based power and sample-size determination. His ongoing AI Statistician work examines reproducible analysis, statistical validation, and the documentation of analytical decisions. Related methodological research studies prioritized and sequential evaluation of AI agents, and adaptive decision rules for multi-step workflows. He is interested in building AI systems that support causal reasoning and scientific inquiry, and in evaluating them through clearly defined tasks and outcomes.

Conference contributions and forthcoming work. His work on propensity-score weighting was presented in an invited research seminar, “Propensity score weighting with complex survey data: Best practice,” at the Jacobs Center for Productive Youth Development, University of Zurich, on March 3, 2025. He presented weighted double score matching at the ENAR Spring Meeting in Indianapolis on March 16, 2026, and Power Agent at Joint Statistical Meetings in Boston on August 5, 2026. His survey-weighting presentations also included the ENAR Spring Meeting in New Orleans on March 26, 2025. He presented Power Agent posters at the Yale Medical AI Symposium on March 26, 2026, and the AI at Yale Symposium on April 28, 2026. His coauthored survey-weighting work was also presented at CFE–CMStatistics 2024 in London and Joint Statistical Meetings 2025 in Nashville. His coauthored clinical conference work includes “Does symptom profile modify revascularization benefit in ischemic LV dysfunction? A win ratio analysis pooling STICHES and REVIVED-BCIS2” (ESC Congress, August 2026) and “Obesity measures as predictors of mortality risk among US adults” (SGIM Annual Meeting, May 2026). His TRANSFORM Score abstract on mortality-risk prediction after heart failure hospitalization has been accepted for AHA Scientific Sessions in Chicago, November 6–9, 2026; the presentation is forthcoming.

Selected publications

Peer-reviewed journal articles

1. Zeng Y, Li F, Tong G. Moving toward best practice when using propensity score weighting in survey observational studies. Statistics in Medicine. 2026;45(10–12):e70555. https://doi.org/10.1002/sim.70555. https://pubmed.ncbi.nlm.nih.gov/42031002/. First published online April 24, 2026.

2. Tong G, Zeng Y, Greene SJ, Anstrom KJ, Testani J, Mentz RJ, Li F, Velazquez EJ. Comparative effectiveness of torsemide vs furosemide in the management of heart failure patients: Win-ratio reanalysis of the TRANSFORM-HF trial. American Heart Journal. 2026;301:107525. https://doi.org/10.1016/j.ahj.2026.107525. https://pubmed.ncbi.nlm.nih.gov/42419438/. Published online July 8, 2026.

3. Latif N, Zeng Y, Kostantinis S, Furman M, Feher A, Hinchcliff M, Cigarroa N, Tong G, Kunnirickal SJ, Odanovic N, Shah SM. Association between coronary vasomotor dysfunction and autoimmune disease in patients undergoing invasive coronary function testing. International Journal of Cardiology Heart & Vasculature. 2026;66:102012. https://doi.org/10.1016/j.ijcha.2026.102012. https://pubmed.ncbi.nlm.nih.gov/42774444/. Published online September 14, 2026.

4. Tong G, Li C, Wang H, Fang X, Liu R, Zeng Y, Sun Q, Ouyang Y, Baumann MR, Davis-Plourde K, Wang Z, Li F, Taljaard M. Subgroup analyses in stepped-wedge cluster randomized trials: A systematic review of statistical practice and reporting from 2016 to 2023. Journal of Clinical Epidemiology. 2026:112512. Published online September 17, 2026. https://doi.org/10.1016/j.jclinepi.2026.112512. https://pubmed.ncbi.nlm.nih.gov/42753990/.

5. Wright CX, Kiage JN, Kyriakoulis I, Csecs I, Ahmed AI, Tong G, Zeng Y, Liu YH, Bellumkonda L, Kokkinidis DG, Sinusas AJ, Miller EJ, Feher A. Coronary calcifications and myocardial flow reserve in heart transplant patients: a longitudinal study. European Heart Journal. 2026;47(24):3214–3216. https://doi.org/10.1093/eurheartj/ehag163. https://pubmed.ncbi.nlm.nih.gov/41780909/. First published online March 5, 2026.

Preprint (not peer reviewed)

Tong G, Li C, Li F, Zeng Y, Greene SJ, Anstrom KJ, Testani J, Mentz RJ, Velazquez EJ. Using artificial intelligence to assess treatment-effect heterogeneity in pragmatic cardiovascular trials: Insights from TRANSFORM-HF. medRxiv. Posted January 19, 2026. https://doi.org/10.64898/2026.01.16.26344310. https://www.medrxiv.org/content/10.64898/2026.01.16.26344310v1. This work has not yet been peer reviewed.

Published conference abstract

Akman Z, Zeng Y, Rossi R, Nouri A, Al Mouslmani M, Arya Nezhad S, Wang SY, Tong G, Damluji A, Nanna M. Calcium channel blockers versus beta-blockers and risk of cognitive decline in older adults: Insights from SPRINT MIND. Journal of the American College of Cardiology. 2026;87(13 Supplement):A1172. ACC.26 conference abstract. Published online March 27, 2026. https://www.jacc.org/doi/10.1016/j.jacc.2026.02.2882.