Ying Cui
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
Research Subjects
Ying Cui has a total of 112 co-authors. Here are the top co-authors by number of shared publications:
- Lu Tian
Department of Biomedical Data Science, Stanford University School of Medicine, Stanford, California
3 shared publications - Jeong Hoon Jang
Department of Biostatistics and Data Science University of Texas Medical Branch Galveston Texas USA
2 shared publications Akaash Kolluri
Stanford University, Stanford, CA, USA
1 shared publication- Akash Chaurasia
Department of Computer Science, Stanford University, Stanford, CA, USA
1 shared publication - Akshay Swaminathan
Department of Biomedical Data Science, Stanford University School of Medicine, Stanford, CA, USA
1 shared publication - Alejandro Lozano
Department of Biomedical Data Science, Stanford University, Stanford, CA, USA
1 shared publication
Latest Publications
- 24 September 2026
- Published by Wiley in Biometrical Journal
- Vol. 68 (5) , e70184
Clustering Hybrid Functional and Vector Data
Modern biomedical studies increasingly collect hybrid data, consisting of functional observations (e.g., curves or images) and multiple scalar (vector‐valued) clinical variables. Integrating these modalities in an unsupervised learning framework can improve patient subgroup identification and interpretation, yet most existing clustering approaches handle only a single data type. We introduce a clustering framework for hybrid data combining multivariate functional and vector outcomes. The method first applies hybrid principal component analysis to capture joint mean structures and modes of variation across modalities, then iteratively reassigns subjects using a hybrid distance metric based on cluster‐specific Karhunen–Loéve expansions. This approach accommodates multiple functions of varying dimensions, integrates contributions from functional and vector components in mean and covariance structures, and avoids parametric distributional assumptions. Simulation studies show that the proposed method outperforms competing approaches when subgroup information, in terms of mean and/or variance differences, is either distributed across modalities or concentrated in a single modality. We illustrate the method using renal radionuclide imaging data, including baseline and diuretic renogram curves and pharmacokinetic variables from patients with suspected kidney obstruction. The method identifies clinically meaningful subgroups consistent with expert assessment, demonstrating its potential to support data‐driven decision‐making in multimodal biomedical studies.
- 24 September 2026
- Published by Wiley in Biometrical Journal
- Vol. 68 (5) , e70184
Clustering Hybrid Functional and Vector Data
Modern biomedical studies increasingly collect hybrid data, consisting of functional observations (e.g., curves or images) and multiple scalar (vector‐valued) clinical variables. Integrating these modalities in an unsupervised learning framework can improve patient subgroup identification and interpretation, yet most existing clustering approaches handle only a single data type. We introduce a clustering framework for hybrid data combining multivariate functional and vector outcomes. The method first applies hybrid principal component analysis to capture joint mean structures and modes of variation across modalities, then iteratively reassigns subjects using a hybrid distance metric based on cluster‐specific Karhunen–Loéve expansions. This approach accommodates multiple functions of varying dimensions, integrates contributions from functional and vector components in mean and covariance structures, and avoids parametric distributional assumptions. Simulation studies show that the proposed method outperforms competing approaches when subgroup information, in terms of mean and/or variance differences, is either distributed across modalities or concentrated in a single modality. We illustrate the method using renal radionuclide imaging data, including baseline and diuretic renogram curves and pharmacokinetic variables from patients with suspected kidney obstruction. The method identifies clinically meaningful subgroups consistent with expert assessment, demonstrating its potential to support data‐driven decision‐making in multimodal biomedical studies.
- Open Access
One Size Does Not Fit All: A Data-Driven Framework for Personalized Comorbidity Scoring
Comorbidity indices are widely used in clinical research to summarize disease burden. However, traditional indices were developed decades ago in limited populations using fixed weights that do not reflect the diversity of patients in modern healthcare. We present the Personalized Comorbidity Score (PCS), a data-driven framework for context-dependent comorbidity scoring designed to capture patient complexity while remaining accessible for broad research adoption. PCS was developed using Epic Cosmos, a large national EHR network encompassing over 8 million adult inpatient encounters from 2015 to 2020, with comorbidities defined using AHRQ Clinical Classifications Software Refined categories. Models were developed separately across eight age-sex subgroups using LASSO-penalized Cox regression for feature selection and restricted mean survival time for score derivation. PCS is available in two versions: PCS Core, incorporating age, sex, and comorbidities, and PCS Extended, which additionally incorporates socioeconomic and geographic variables. PCS Core and PCS Extended achieved AUROCs of 0.812 and 0.813 for one-year mortality, outperforming traditional indices (AUROC, 0.714-0.730). PCS demonstrated consistently lower subgroup calibration error across demographic and socioeconomic groups without including race or ethnicity as model features. PCS was further evaluated in two complementary external EHR datasets (Stanford Health Care and MIMIC-IV) with distinct patient populations and data structures, where it consistently outperformed traditional indices. Open-source R and Python packages are provided to support broad adoption. PCS provides an updatable framework for comorbidity measurement that is accurate, context-dependent, and designed to evolve alongside clinical practice.
Open AccessOne Size Does Not Fit All: A Data-Driven Framework for Personalized Comorbidity Scoring
Comorbidity indices are widely used in clinical research to summarize disease burden. However, traditional indices were developed decades ago in limited populations using fixed weights that do not reflect the diversity of patients in modern healthcare. We present the Personalized Comorbidity Score (PCS), a data-driven framework for context-dependent comorbidity scoring designed to capture patient complexity while remaining accessible for broad research adoption. PCS was developed using Epic Cosmos, a large national EHR network encompassing over 8 million adult inpatient encounters from 2015 to 2020, with comorbidities defined using AHRQ Clinical Classifications Software Refined categories. Models were developed separately across eight age-sex subgroups using LASSO-penalized Cox regression for feature selection and restricted mean survival time for score derivation. PCS is available in two versions: PCS Core, incorporating age, sex, and comorbidities, and PCS Extended, which additionally incorporates socioeconomic and geographic variables. PCS Core and PCS Extended achieved AUROCs of 0.812 and 0.813 for one-year mortality, outperforming traditional indices (AUROC, 0.714-0.730). PCS demonstrated consistently lower subgroup calibration error across demographic and socioeconomic groups without including race or ethnicity as model features. PCS was further evaluated in two complementary external EHR datasets (Stanford Health Care and MIMIC-IV) with distinct patient populations and data structures, where it consistently outperformed traditional indices. Open-source R and Python packages are provided to support broad adoption. PCS provides an updatable framework for comorbidity measurement that is accurate, context-dependent, and designed to evolve alongside clinical practice.
- 8 April 2026
- Published by SAGE Publications in Statistical Methods in Medical Research
- Vol. 35 (5) , 945-961
Model-based clustering of multiple images incorporating covariates
In this paper, we develop a novel method for clustering multiple images while adjusting for the effect of available covariates on cluster membership. The key strategy is to represent each image as two-dimensional functional data and formulate a functional latent class mixed model, which fully leverages the structural information of images while effectively addressing their high dimensionality and accounting for covariate effects. We apply the proposed method to color intensity matrices extracted from patient-sourced smartphone fingernail photos to identify distinct subgroups, while adjusting for the effect of image metadata, which may act as an effect modifier on cluster membership. Information on these subgroups can assist public health officials in low-resource settings by enabling rapid and non-invasive identification of high-risk subpopulations for anemia, thereby facilitating the timely delivery of targeted interventions. The results suggest that the three clusters identified by the proposed method correspond to varying levels of anemia risk, with 0%, 79%, and 86% of subjects in each cluster classified as anemic. These findings demonstrate the utility of the proposed method and highlight the potential of a smartphone application leveraging fingernail images for non-invasive and cost-effective anemia screening.
- 8 April 2026
- Published by SAGE Publications in Statistical Methods in Medical Research
- Vol. 35 (5) , 945-961
Model-based clustering of multiple images incorporating covariates
In this paper, we develop a novel method for clustering multiple images while adjusting for the effect of available covariates on cluster membership. The key strategy is to represent each image as two-dimensional functional data and formulate a functional latent class mixed model, which fully leverages the structural information of images while effectively addressing their high dimensionality and accounting for covariate effects. We apply the proposed method to color intensity matrices extracted from patient-sourced smartphone fingernail photos to identify distinct subgroups, while adjusting for the effect of image metadata, which may act as an effect modifier on cluster membership. Information on these subgroups can assist public health officials in low-resource settings by enabling rapid and non-invasive identification of high-risk subpopulations for anemia, thereby facilitating the timely delivery of targeted interventions. The results suggest that the three clusters identified by the proposed method correspond to varying levels of anemia risk, with 0%, 79%, and 86% of subjects in each cluster classified as anemic. These findings demonstrate the utility of the proposed method and highlight the potential of a smartphone application leveraging fingernail images for non-invasive and cost-effective anemia screening.
- 17 July 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (15-17) , e70180
An IPCW Adjusted Win Statistics Approach in Clinical Trials Incorporating Equivalence Margins to Define Ties
In clinical trials, multiple outcomes of different priorities commonly occur as the patient's response may not be adequately characterized by a single outcome. Win statistics are appealing summary measures for between‐group difference at more than one endpoint. When defining the result of pairwise comparisons of a time‐to‐event endpoint, it is desirable to allow ties to account for incomplete follow‐up and not clinically meaningful difference in endpoints of interest. In this article, we propose a class of win statistics for time‐to‐event endpoints with a user‐specified equivalence margin. These win statistics are identifiable in the presence of right censoring and do not depend on the censoring distribution. We then develop estimation and inference procedures for the proposed win statistics based on inverse‐probability‐of‐censoring weighting adjustment to handle right censoring. We conduct extensive simulations to investigate the operational characteristics of the proposed procedure in the finite sample setting. A real oncology trial is used to illustrate the proposed approach.
- 17 July 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (15-17) , e70180
An IPCW Adjusted Win Statistics Approach in Clinical Trials Incorporating Equivalence Margins to Define Ties
In clinical trials, multiple outcomes of different priorities commonly occur as the patient's response may not be adequately characterized by a single outcome. Win statistics are appealing summary measures for between‐group difference at more than one endpoint. When defining the result of pairwise comparisons of a time‐to‐event endpoint, it is desirable to allow ties to account for incomplete follow‐up and not clinically meaningful difference in endpoints of interest. In this article, we propose a class of win statistics for time‐to‐event endpoints with a user‐specified equivalence margin. These win statistics are identifiable in the presence of right censoring and do not depend on the censoring distribution. We then develop estimation and inference procedures for the proposed win statistics based on inverse‐probability‐of‐censoring weighting adjustment to handle right censoring. We conduct extensive simulations to investigate the operational characteristics of the proposed procedure in the finite sample setting. A real oncology trial is used to illustrate the proposed approach.
- Top Cited PapersOpen Access
- 7 March 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-10
Red teaming ChatGPT in medicine to yield real-world insights on model behavior
Red teaming, the practice of adversarially exposing unexpected or undesired model behaviors, is critical towards improving equity and accuracy of large language models, but non-model creator-affiliated red teaming is scant in healthcare. We convened teams of clinicians, medical and engineering students, and technical professionals (80 participants total) to stress-test models with real-world clinical cases and categorize inappropriate responses along axes of safety, privacy, hallucinations/accuracy, and bias. Six medically-trained reviewers re-analyzed prompt-response pairs and added qualitative annotations. Of 376 unique prompts (1504 responses), 20.1% were inappropriate (GPT-3.5: 25.8%; GPT-4.0: 16%; GPT-4.0 with Internet: 17.8%). Subsequently, we show the utility of our benchmark by testing GPT-4o, a model released after our event (20.4% inappropriate). 21.5% of responses appropriate with GPT-3.5 were inappropriate in updated models. We share insights for constructing red teaming prompts, and present our benchmark for iterative model assessments.
- 7 March 2025
- Published by Springer Nature in npj Digital Medicine
- Vol. 8 (1) , 1-10
Top Cited PapersOpen AccessRed teaming ChatGPT in medicine to yield real-world insights on model behavior
Red teaming, the practice of adversarially exposing unexpected or undesired model behaviors, is critical towards improving equity and accuracy of large language models, but non-model creator-affiliated red teaming is scant in healthcare. We convened teams of clinicians, medical and engineering students, and technical professionals (80 participants total) to stress-test models with real-world clinical cases and categorize inappropriate responses along axes of safety, privacy, hallucinations/accuracy, and bias. Six medically-trained reviewers re-analyzed prompt-response pairs and added qualitative annotations. Of 376 unique prompts (1504 responses), 20.1% were inappropriate (GPT-3.5: 25.8%; GPT-4.0: 16%; GPT-4.0 with Internet: 17.8%). Subsequently, we show the utility of our benchmark by testing GPT-4o, a model released after our event (20.4% inappropriate). 21.5% of responses appropriate with GPT-3.5 were inappropriate in updated models. We share insights for constructing red teaming prompts, and present our benchmark for iterative model assessments.
- 8 August 2024
- Published by SAGE Publications in Clinical Trials
- Vol. 21 (5) , 541-552
Inferences for the distribution of the duration of response in a comparative clinical study
Duration of response is an important endpoint used in drug development. Prolonged duration for response is often viewed as an early indication of treatment efficacy. However, there are numerous difficulties in studying the distribution of duration of response based on observed data subject to right censoring in practice. The most important obstacle is that the distribution of the duration of response is in general not identifiable in the presence of censoring due to the simple fact that there is no information on the joint distribution of time to response and time to progression beyond the largest follow-up time. In this article, we introduce the restricted duration of response as a replacement of the conventional duration of response. The distribution of restricted duration of response is estimable and we have proposed several nonparametric estimators in this article. The corresponding inference procedure and additional downstream analysis have been developed. Extensive numerical simulations have been conducted to examine the finite sample performance of the proposed estimators. It appears that a new regression-based two-step estimator for the survival function of the restricted duration of response tends to have a robust and superior performance, and we recommend its use in practice. A real data example from oncology has been used to illustrate the analysis for restricted duration of response.
- 8 August 2024
- Published by SAGE Publications in Clinical Trials
- Vol. 21 (5) , 541-552
Inferences for the distribution of the duration of response in a comparative clinical study
Duration of response is an important endpoint used in drug development. Prolonged duration for response is often viewed as an early indication of treatment efficacy. However, there are numerous difficulties in studying the distribution of duration of response based on observed data subject to right censoring in practice. The most important obstacle is that the distribution of the duration of response is in general not identifiable in the presence of censoring due to the simple fact that there is no information on the joint distribution of time to response and time to progression beyond the largest follow-up time. In this article, we introduce the restricted duration of response as a replacement of the conventional duration of response. The distribution of restricted duration of response is estimable and we have proposed several nonparametric estimators in this article. The corresponding inference procedure and additional downstream analysis have been developed. Extensive numerical simulations have been conducted to examine the finite sample performance of the proposed estimators. It appears that a new regression-based two-step estimator for the survival function of the restricted duration of response tends to have a robust and superior performance, and we recommend its use in practice. A real data example from oncology has been used to illustrate the analysis for restricted duration of response.
- Open Access
- 1 April 2021
- Published by Elsevier in Kidney International Reports
- Vol. 6 (4) , S241
POS-551 INTRAOPERATIVE PLASMA (1-84) PTH LEVELS ARE BETTER THAN INTACT PTH FOR ASSESSING THE SUCCESS OF PARATHYROIDECTOMY IN UREMIC HYPERPARATHYROIDISM PATIENTS
- 1 April 2021
- Published by Elsevier in Kidney International Reports
- Vol. 6 (4) , S241
Open AccessPOS-551 INTRAOPERATIVE PLASMA (1-84) PTH LEVELS ARE BETTER THAN INTACT PTH FOR ASSESSING THE SUCCESS OF PARATHYROIDECTOMY IN UREMIC HYPERPARATHYROIDISM PATIENTS