HDHao Dai

Hao Dai

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Hao Dai has a total of 64 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Learning health system frameworks and the integration of artificial intelligence and data science in United States health care: a scoping review

    by Rotana M. Radwan, Hao Dai, Yao An Lee, Ronald Ramon, Wei-Han Chen, Hsin-Yueh Lin, Ya-Yun Yeh, Wenxi Huang, Hongyu Chen, Tiancheng Zhou, et al.

  • Open Access

    Glucagon‐Like Peptide‐1 Receptor Agonists and Cardiovascular Events in Adults With Obesity and Autoimmune Disease: A Target Trial Emulation

    by Hao Dai, Yao An Lee, Austin Natalie, Whitney Jackson, Angela Pham, Jake Levine, Rotana M. Radwan, Jingchuan Guo, Jiang Bian, Amy J. Sheer

    Background: Obesity and autoimmune diseases (AIDs) are each associated with elevated risk of cardiovascular and thromboembolic events. Glucagon‐like peptide‐1 receptor agonists (GLP‐1RAs) have demonstrated cardiovascular and metabolic benefits in patients with type 2 diabetes and obesity, but their effects in patients with obesity and comorbid AID remain uncertain. This study evaluated the association between GLP‐1RA use and major adverse cardiovascular and thromboembolic event risk among adults with obesity and AID. Methods: This retrospective cohort study emulated a target trial using 2014 to 2024 electronic health record data from the OneFlorida+ network. Adults (aged ≥18 years) with AID and obesity eligible for antiobesity medication therapy were included. Exposure was defined as GLP‐1RA use versus nonuse. Participants were matched using 1:1 time‐dependent propensity scores. The primary outcomes were myocardial infarction, stroke or transient ischemic attack, pulmonary embolism, venous thromboembolism, and coronary revascularization. Secondary outcomes included hospitalization, emergency department visits, and all‐cause mortality. Results: We matched 13 204 GLP‐1RA users and 13 204 nonusers (mean ± SD age, 54.7 ± 14.5 years; 73.4% women; mean body mass index, 37 kg/m 2 ). GLP‐1RA use was associated with lower hazard of stroke/transient ischemic attack (hazard ratio [HR], 0.87 [95% CI, 0.76–0.99]; P =0.039), pulmonary embolism (HR, 0.69 [95% CI, 0.56–0.86]; P =0.001), venous thromboembolism (HR, 0.83 [95% CI, 0.72–0.95]; P =0.007), emergency department visits (HR, 0.79 [95% CI, 0.75–0.83]; P P <0.001). Conclusions: Among adults with obesity and AID, GLP‐1RA use was associated with reduced thromboembolic events, lower emergency department use, and decreased mortality, suggesting potential cardiovascular and survival benefits in this high‐risk population.


  • Open Access

    GLP-1 receptor agonists and acute diabetes complications in adults with type 1 diabetes

    by Hao Dai, Yao An Lee, Rotana M. Radwan, Amy J. Sheer, Tamara S. Hannon, Matthew R. Hayes, Ramon C. Sun, Jiang Bian, Jingchuan Guo

  • Open Access

    Domain-Specific Knowledge Graphs in RAG-Enhanced Healthcare LLMs

    by Sydney Anuyah, Mehedi Mahmud Kaushik, Hao Dai, Rakesh Shiradkar, Arjan Durresi, Sunandan Chakraborty

    Large Language Models (LLMs) generate fluent answers but can struggle with trustworthy, domain-specific reasoning. We evaluate whether domain knowledge graphs (KGs) improve Retrieval-Augmented Generation (RAG) for healthcare by constructing three PubMed-derived graphs: $\mathbb{G}_1$ (T2DM), $\mathbb{G}_2$ (Alzheimer's disease), and $\mathbb{G}_3$ (AD+T2DM). We design two probes: Probe 1 targets merged AD T2DM knowledge, while Probe 2 targets the intersection of $\mathbb{G}_1$ and $\mathbb{G}_2$. Seven instruction-tuned LLMs are tested across retrieval sources {No-RAG, $\mathbb{G}_1$, $\mathbb{G}_2$, $\mathbb{G}_1$ + $\mathbb{G}_2$, $\mathbb{G}_3$, $\mathbb{G}_1$+$\mathbb{G}_2$ + $\mathbb{G}_3$} and three decoding temperatures. Results show that scope alignment between probe and KG is decisive: precise, scope-matched retrieval (notably $\mathbb{G}_2$) yields the most consistent gains, whereas indiscriminate graph unions often introduce distractors that reduce accuracy. Larger models frequently match or exceed KG-RAG with a No-RAG baseline on Probe 1, indicating strong parametric priors, whereas smaller/mid-sized models benefit more from well-scoped retrieval. Temperature plays a secondary role; higher values rarely help. We conclude that precision-first, scope-matched KG-RAG is preferable to breadth-first unions, and we outline practical guidelines for graph selection, model sizing, and retrieval/reranking. Code and Data available here - https://github.com/sydneyanuyah/RAGComparison

  • Open Access

    Inferring high-fat dietary patterns from electronic health record data using machine learning

    by Ya-Yun Yeh, Hsin-Yueh Lin, Jingchuan Guo, Ramon C Sun, Sizun Jiang, Jiang Bian, Hao Dai

    Objectives: Electronic health records (EHRs) rarely capture dietary detail, limiting diet–disease research. We aimed to develop machine learning (ML) computable phenotypes to identify high-fat diet (HFD) using variables typically available in EHRs. Materials and Methods: We used National Health and Nutrition Examination Survey (NHANES) 1999-2020 data, where 24-h dietary recall served as ground truth. Dietary fat intake was summarized into a score (0-30) based on percent energy from fat, carbohydrate, and protein; lower scores indicated HFD. We defined HFD at cutoffs of 10, 15, and 20, and trained ML models (Extreme Gradient Boosting, logistic regression, random forest) using EHR-compatible variables (demographics, comorbidities, labs, anthropometrics). Model interpretability was assessed using Shapley Additive Explanations. To evaluate clinical relevance, we compared cancer associations using ML-predicted vs true diet labels. Results: Machine learning models classified HFD with good performance, strongest at broader definitions. Random forest achieved an F1-score of 0.79 (recall 0.74, precision 0.84) at cutoff 20. Key predictors included race/ethnicity, triglycerides, obesity metrics (body mass index and derived indices), and metabolic panel results. Discussion: These findings indicate that dietary patterns, though seldom recorded in EHRs, can be inferred from routinely available variables. The ability of ML-derived phenotypes to reproduce known diet–disease relationships underscore their epidemiologic validity. Top predictors also align with established biological pathways linking obesity, lipid metabolism, and cancer risk, supporting plausibility. Conclusion: A high-fat dietary pattern can be inferred from EHR-compatible variables using ML-based phenotyping. This approach offers a scalable tool to integrate diet into EHR-based research and precision medicine.


  • Open Access

    High-Throughput Safety Signal Detection for Sodium-Glucose Cotransporter-2 Inhibitors in Type 2 Diabetes

    by Hao Dai, Yao An Lee, Jiang Bian, Jingchuan Guo

    Objective To systematically identify potential safety signals associated with sodium-glucose cotransporter-2 inhibitors (SGLT2i) in patients with type 2 diabetes (T2D) using a high-throughput target trial emulation pipeline applied to real-world data. Methods We utilized electronic health record (EHR) data from the OneFlorida+ Data Trust (2012–2023). Using a retrospective new-user cohort design, we compared SGLT2i initiators to initiators of other second-line glucose-lowering drugs (GLDs). We evaluated risk across all ICD-10-CM 4-digit diagnosis codes. A semi-Bayesian shrinkage method was employed to adjust hazard ratios (HR) and p-values to account for multiple testing and stabilize estimates for rare outcomes. Results The analysis identified several statistically significant safety signals. Validating our method, we observed known adverse events such as candidiasis of the vulva and vagina (HR=1.64) and other urogenital fungal infections. We also detected potential signals for conditions such as viral conjunctivitis and melanocytic nevi. Conclusion This high-throughput screening effectively identified both known and potential new safety signals for SGLT2i. The use of semi-Bayesian shrinkage provides a robust framework for post-marketing surveillance in large healthcare databases.

  • Open Access

    Heterogeneous Effects of Sodium-Glucose Cotransporter-2 Inhibitors on Acute Kidney Injury: A Causal Learning Approach

    by Hao Dai, Yao An Lee, Jiang Bian, Jingchuan Guo

    Background: Sodium–glucose cotransporter-2 inhibitors (SGLT2is) have been associated with lower risk of acute kidney injury (AKI), but existing studies rarely explore heterogeneous treatment effects or underlying causal pathways. We applied a comprehensive causal-learning framework to evaluate both overall and subgroup-specific effects of SGLT2i therapy on AKI. Methods: Using a new-user, active-comparator target trial emulation in the OneFlorida+ data (2014–2023), we estimated individualized and average treatment effects with a doubly robust meta-learner, assessed heterogeneity via subgroup and decision-tree analyses, and used causal structure learning and mediation methods to identify mechanistic pathways linking treatment to AKI. Results: SGLT2 inhibitors were associated with a significant reduction in AKI compared with other second-line glucose-lowering drugs, with an average individual treatment effect of −0.0039 (95% CI: −0.0065 to −0.0014). Kaplan-Meier curves demonstrated consistently lower cumulative AKI incidence among SGLT2i users. Subgroup analyses revealed substantial heterogeneity: protective effects were strongest in younger adults, males, and non-Hispanic Black patients, whereas benefits were attenuated in older adults, females, and those with baseline CKD. Decision tree-based heterogeneity modeling further identified atrial fibrillation, anemia, and antiparkinson agent use as key effect modifiers. Causal structure learning highlighted atrial fibrillation, anemia, chronic kidney disease, and eGFR as central intermediating nodes. Mediation analyses showed that most of the benefit operated through direct pathways (ADE ≈ −0.0034 to −0.0035), while anemia and heart failure contributed small but statistically significant indirect effects. Conclusion: SGLT2 inhibitors reduce AKI risk, but effects vary meaningfully across clinical subgroups and are partially mediated through interconnected cardio-renal pathways. Causal-learning methods provide mechanistic insight beyond average associations and may support more individualized SGLT2i therapy for AKI prevention.


  • Develop and validate a computable phenotype for identifying alcohol-use disorder patients using structure and unstructured EHR data

    by Hao Dai, Elliot B Tapper, Lili Zhao, Grant D Scheiffele, Xiaohan Li, Ronald Ramon, Xing He, Yao An Lee, Jingchuan Guo, Jiang Bian

    Background: Alcohol Use Disorder (AUD) drives significant morbidity through alcohol-related liver disease. Accurate AUD identification in electronic health records is critical for research and care delivery, yet International Classification of Diseases (ICD) code-based algorithms miss many cases while manual review is impractical at scale. Computable phenotypes (CPs) integrating structured and unstructured EHR data offer a scalable solution. Methods: Using University of Florida Health's Integrated Data Repository covering two million patients, we developed AUD CPs through a two-step process. First, candidate cohorts were identified using AUD-related ICD codes, medications, and keyword searches across structured and unstructured data. Second, rule-based combinations were iteratively refined through manual chart review. Final algorithms were evaluated against gold-standard chart review, measuring sensitivity, positive predictive value (PPV), and F1-score, then validated in an independent testing set and an external dataset. Results: The F1-optimized CP achieved an F1-score of .87 (sensitivity: .98, PPV: .78) in the testing set, while the precision-optimized CP achieved PPV of .9 (sensitivity: .68, F1-score: .77). Minimal performance attenuation between training and testing sets demonstrated robustness and generalizability. Both CPs substantially outperformed restricted AUD-specific ICD code-based approaches. Conclusions: CPs integrating structured and unstructured EHR data enable accurate, reproducible AUD identification, surpassing traditional AUD-specific ICD-based methods. This approach facilitates efficient cohort construction for clinical research, public health surveillance, and quality improvement initiatives targeting AUD and its consequences, addressing a critical gap in identifying patients who may benefit from screening and intervention.

  • Open Access

    Effect of Injectable Dual and Single Agonist Glucagon-Like Peptide-1 Based Therapy on Inflammatory Bowel Disease Activity Among Patients with Obesity

    by Jake Levine, Yao An Lee, Angela Pham, Jingchuan Guo, Hao Dai, Rotana M Radwan, Jiang Bian, Aleksey Novikov, Amy Sheer

    Background and Aims: Obesity is a risk factor for relapsing inflammatory bowel disease. GLP-1 and dual GLP-1/GIP agonists may offer superior weight loss; however, their effect on inflammatory bowel disease remains unknown. In this study, we assessed outcomes in patients with IBD and obesity using these medications. Methods: A retrospective, propensity-matched analysis was conducted using data from The OneFlorida+ network. We evaluated patients with IBD and obesity prescribed GLP-1-based therapy (liraglutide, semaglutide, or tirzepatide) versus those without. 562 patients prescribed GLP-1-based therapies were matched 1:1 to controls based on demographics and comorbidities. Results: No differences in GLP-1 users versus non-users with respect to all-cause hospitalization, IBD-hospitalization, IBD-related surgery, or pancreatitis were seen. Use of semaglutide reduced the risk of IBD-related surgery (HR 0.33, CI 0.13-0.83). Among 89 tirzepatide users, none required IBD-related surgery versus 2 matched controls. Black GLP-1 users had an increased risk of all-cause hospitalizations (HR 1.59, CI 1.11–2.29) but not IBD-related hospitalization or IBD-related surgery. Steroid use was comparable between groups. GLP-1 use significantly reduced serum CRP. Conclusions: Semaglutide appeared to lower the risk of IBD-related surgery. Black patients using GLP-1 therapy had an increased risk of all-cause hospitalization. Tirzepatide showed no significant difference regarding primary endpoints; however, the sample size was small. Overall, all GLP-1-based therapies were well tolerated among obese patients with IBD. Larger studies are warranted to understand the therapeutic potential of GLP-1-based treatment in this patient population. Graphical

  • Open Access

    GLP-1 Receptor Agonists and Cardiovascular Events in Adults with Obesity and Autoimmune Disease: A Target Trial Emulation

    by Hao Dai, Yao An Lee, Austin Natalie, Whitney Jackson, Angela Pham, Jake Levine, Rotana Radwan, Jingchuan Guo, Jiang Bian, Amy Joy Sheer

    Importance: Obesity and autoimmune diseases (AID) are each associated with elevated risk of cardiovascular and thromboembolic events due to chronic systemic inflammation. Glucagon-like peptide-1 receptor agonists (GLP-1RAs) have demonstrated cardiovascular and metabolic benefits in patients with type 2 diabetes and obesity, but their effects in patients with obesity and comorbid AID remain uncertain. Objective: To evaluate the association between GLP-1RA use and the risk of major adverse cardiovascular and thromboembolic events among adults with obesity and AID eligible for anti-obesity medication (AOM) therapy. Design: This retrospective cohort study emulated a target trial using 2014-2024 electronic health record data from the OneFlorida+ network, which includes 21 million individuals across Florida, Georgia, and Alabama. Adults with obesity and AID who met AOM eligibility criteria were included. Propensity score matching (1:1) was applied using a time-dependent framework to balance baseline covariates between GLP-1RA users and non-users. Participants: AID Adults (≥18 years) who were eligible for AOM treatment. Exposure: GLP-1RA use versus non-use. Main Outcomes and Measures: The primary outcomes were myocardial infarction, stroke or transient ischemic attack (TIA), pulmonary embolism (PE), venous thromboembolism (VTE), and coronary revascularization. Secondary outcomes included hospitalization, emergency department (ED) visits, and all-cause mortality. Results: The matched cohort included 13,204 GLP-1RA users and 13,204 non-users (mean age, 54.7 ± 14.5 years; 73.4% female; mean BMI, 37 kg/m²). Compared with non-users, GLP-1RA users had lower incidence rates (per 1000 person-years) of PE (6.4 vs 9.5), VTE (16.6 vs 20.4), and mortality (9.5 vs 16.9). GLP-1RA use was associated with lower hazard of stroke/TIA (HR, 0.87 [95% CI, 0.76-0.99]; P = .039), PE (HR, 0.69 [95% CI, 0.56-0.86]; P = .001), VTE (HR, 0.83 [95% CI, 0.72-0.95]; P = .007), ED visits (HR, 0.79 [95% CI, 0.75-0.83]; P = .000), and mortality (HR, 0.56 [95% CI, 0.47-0.66]; P = .000). Conclusions and Relevance: Among adults with obesity and AID, GLP-1RA use was associated with reduced thromboembolic events, lower emergency department utilization, and decreased mortality. These findings suggest potential cardiovascular and survival benefits of GLP-1RAs in a high-risk, understudied population. Key Points: Question: Is the use of glucagon-like peptide-1 receptor agonists (GLP-1RAs) associated with cardiovascular and thromboembolic outcomes among adults with obesity and autoimmune disease (AID)? Findings: In this target trial emulation using 2014-2024 electronic health record data from the OneFlorida+ network, 13,204 GLP-1RA users were compared with 13,204 matched non-users. GLP-1RA use was associated with significantly lower risks of pulmonary embolism, venous thromboembolism, emergency department visits, and all-cause mortality, with marginal associations for stroke or transient ischemic attack. Meaning: Among adults with obesity and AID, GLP-1RA therapy may confer thromboembolic and survival benefits without increasing cardiovascular risk.