Xinmeng Zhang

Xinmeng Zhang

Auto-generated profile, based on publicly available information

Xinmeng Zhang

Department of Computer Science, Vanderbilt University, Nashville, TN, USA

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
3 August 2020 in Journal of Healthcare Informatics Research
Verified
Journal Article
Closed access
PDF file
XZXinmeng Zhang
Xinmeng Zhang
CYChao Yan
Chao Yan
CGCheng Gao
Cheng Gao
BMBradley A. Malin
Bradley A. Malin
You ChenYou Chen
You Chen

8 March 2025 in Journal of the American Medical Informatics Association
Verified
Journal Article
Open access
KBKatherine E Brown
Katherine E Brown
CYChao Yan
Chao Yan
ZLZhuohang Li
Zhuohang Li
XZXinmeng Zhang
Xinmeng Zhang
BCBenjamin X Collins
Benjamin X Collins
BMBradley A Malin
Bradley A Malin
Objectives: To determine the extent to which current large language models (LLMs) can serve as substitutes for traditional machine learning (ML) as clinical predictors using data from electronic health records (EHRs), we investigated various factors that can impact their adoption, including overall performance, calibration, fairness, and resilience to privacy protections that reduce data fidelity. Materials and Methods: We evaluated GPT-3.5, GPT-4, and traditional ML (as gradient-boosting trees) on clinical prediction tasks in EHR data from Vanderbilt University Medical Center (VUMC) and MIMIC IV. We measured predictive performance with area under the receiver operating characteristic (AUROC) and model calibration using Brier Score. To evaluate the impact of data privacy protections, we assessed AUROC when demographic variables are generalized. We evaluated algorithmic fairness using equalized odds and statistical parity across race, sex, and age of patients. We also considered the impact of using in-context learning by incorporating labeled examples within the prompt. Results: Traditional ML [AUROC: 0.847, 0.894 (VUMC, MIMIC)] substantially outperformed GPT-3.5 (AUROC: 0.537, 0.517) and GPT-4 (AUROC: 0.629, 0.602) (with and without in-context learning) in predictive performance and output probability calibration [Brier Score (ML vs GPT-3.5 vs GPT-4): 0.134 vs 0.384 vs 0.251, 0.042 vs 0.06 vs 0.219)]. Discussion: Traditional ML is more robust than GPT-3.5 and GPT-4 in generalizing demographic information to protect privacy. GPT-4 is the fairest model according to our selected metrics but at the cost of poor model performance. Conclusion: These findings suggest that non-fine-tuned LLMs are less effective and robust than locally trained ML for clinical prediction tasks, but they are improving across releases.

8 March 2021 in JMIR Human Factors
Verified
Journal Article
Open access
CYChao Yan
Chao Yan
XZXinmeng Zhang
Xinmeng Zhang
CGCheng Gao
Cheng Gao
EWErin Wilfong
Erin Wilfong
JCJonathan Casey
Jonathan Casey
You ChenYou Chen
You Chen
Background: Few intensive care unit (ICU) staffing studies have examined the collaboration structures of health care workers (HCWs). Knowledge about how HCWs are connected to the care of critically ill patients with COVID-19 is important for characterizing the relationships among team structures, care quality, and patient safety. Objective: We aimed to discover differences in the teamwork structures of COVID-19 critical care by comparing HCW collaborations in the management of critically ill patients with and without COVID-19. Methods: In this retrospective study, we used network analysis methods to analyze the electronic health records (EHRs) of 76 critically ill patients (with COVID-19: n=38; without COVID-19: n=38) who were admitted to a large academic medical center, and to learn about HCW collaboration. We used the EHRs of adult patients who were admitted to the COVID-19 ICU at the Vanderbilt University Medical Center (Nashville, Tennessee, United States) between March 17, 2020, and May 31, 2020. We matched each patient according to age, gender, and their length of stay. Patients without COVID-19 were admitted to the medical ICU between December 1, 2019, and February 29, 2020. We used two sociometrics—eigencentrality and betweenness—to quantify HCWs’ statuses in networks. Eigencentrality characterizes the degree to which an HCW is a core person in collaboration structures. Betweenness centrality refers to whether an HCW lies on the path of other HCWs who are not directly connected. This sociometric was used to characterize HCWs’ broad skill sets. We measured patient staffing intensity in terms of the number of HCWs who interacted with patients’ EHRs. We assessed the statistical differences in the core and betweenness statuses of HCWs and the patient staffing intensities of COVID-19 and non–COVID-19 critical care, by using Mann-Whitney U tests and reporting 95% CIs. Results: HCWs in COVID-19 critical care were more likely to frequently work with each other (eigencentrality: median 0.096) than those in non–COVID-19 critical care (eigencentrality: median 0.057; P<.001). Internal medicine physicians in COVID-19 critical care had higher core statuses than those in non–COVID-19 critical care (P=.001). Nurse practitioners in COVID-19 care had higher betweenness statuses than those in non–COVID-19 care (P<.001). Compared to HCWs in non–COVID-19 settings, the EHRs of critically ill patients with COVID-19 were used by a larger number of internal medicine nurse practitioners (P<.001), cardiovascular nurses (P<.001), and surgical ICU nurses (P=.002) and a smaller number of resident physicians (P<.001). Conclusions: Network analysis methodologies and data on EHR use provide a novel method for learning about differences in collaboration structures between COVID-19 and non–COVID-19 critical care. Health care organizations can use this information to learn about the novel changes that the COVID-19 pandemic has imposed on collaboration structures in urgent care.

7 March 2025 in Journal of the American Heart Association
Verified
Journal Article
Open access
LWLei Wang
Lei Wang
XZXinmeng Zhang
Xinmeng Zhang
You ChenYou Chen
You Chen
CFCharles R. Flynn
Charles R. Flynn
WEWayne J. English
Wayne J. English
DYDanxia Yu
Danxia Yu
Background: We applied the novel Predicting Risk of Cardiovascular Disease EVENTs equations to evaluate cardiovascular–kidney–metabolic (CKM) health and estimated cardiovascular disease (CVD) risk after bariatric surgery. Methods: Among 7804 patients (aged 20–79 years) undergoing bariatric surgery at Vanderbilt University Medical Center during 1999 to 2022, CVD risk factors from before surgery to 2 years after surgery were extracted from electronic health records. The 10‐ and 30‐year risks of total CVD, atherosclerotic CVD, coronary heart disease, stroke, and heart failure (HF) were estimated for patients without CVD history at each time point (n=124–2910), using the social deprivation index–enhanced Predicting Risk of Cardiovascular Disease EVENTs equations. Paired t tests or McNemar tests were used to compare pre‐ with postsurgery CKM health and CVD risk. Two‐sample t tests were used to compare CVD risk reduction between patient subgroups. Results: CKM health was significantly improved after surgery with lower systolic blood pressure, non–high‐density lipoprotein cholesterol, and diabetes prevalence, and higher high‐density lipoprotein and estimated glomerular filtration rate. The 10‐year risks of total CVD and its subtypes decreased by 21.7% to 56.3% at 1 year after surgery and by 14.6% to 46.5% at 2 years after surgery, with the largest reduction observed for HF. Younger age, White race, >30% weight loss, and diabetes history were associated with greater HF risk reductions. Similar results were found for the 30‐year CVD risk estimates. Conclusions: Bariatric surgery significantly improves CKM health and reduces estimated CVD risk, particularly HF, by 47% to 56% within 1 to 2 years after surgery. HF risk reduction may vary by patient's demographics, weight loss, and disease history, which warrants further research.

9 December 2024 in Journal of Medical Internet Research
Verified
Journal Article
Open access
XZXinmeng Zhang
Xinmeng Zhang
KKKaidi Kang
Kaidi Kang
CYChao Yan
Chao Yan
YFYubo Feng
Yubo Feng
SVSimon Vandekar
Simon Vandekar
You ChenYou Chen
You Chen
Background: Bariatric surgery is an effective intervention for obesity, but comprehensive postoperative self-management is essential for optimal outcomes. While patient portals are generally seen as beneficial in engaging patients in health management, the link between their use and post–bariatric surgery weight loss remains unclear. Objective: This study aimed to investigate the association between patient portal engagement and postoperative BMI reduction among patients after bariatric surgery. Methods: This retrospective longitudinal study included patients who underwent Roux-en-Y gastric bypass or sleeve gastrectomy at Vanderbilt University Medical Center between January 2018 and March 2021. Patient portal engagement was measured during 4 stages: early (within 3 months after surgery), early midterm (3-6 months), late midterm (6-9 months), and late (9-12 months). Using generalized estimating equations, we estimated the associations between patients’ portal engagements at these stages and the percentage of BMI reduction (%BMIR) at 3, 6, and 12 months after surgery. Covariates included duration since surgery, patient’s age at the time of surgery, sex, race and ethnicity, type of bariatric surgery, severity of comorbid conditions, and socioeconomic disadvantage. Results: The study included 1415 patients, predominantly female (n=1145, 80.9%), with a racial composition of 76.9% (n=1088) White and 19.9% (n=282) Black. Overall, 805 (56.9%) patients underwent Roux-en-Y gastric bypass and 610 (43.1%) underwent sleeve gastrectomy. By 1 year after surgery, the median %BMIR was 31.5% (IQR 25.2%-36.8%), and the median number of active days on the patient portal was 54 (IQR 33-80). Early portal engagement was significantly associated with %BMIR at various postoperative times. Specifically, each additional 10 days of early portal engagement was associated with a 0.37% (95% CI –0.55% to –0.18%; P<.001) lower expected %BMIR at 3 months, a 1.11% (95% CI 0.82%-1.41%; P<.001) higher expected %BMIR at 6 months, and a 0.78% (95% CI 0.25%-1.31%; P=.004) higher expected %BMIR at 12 months. Furthermore, early midterm portal engagement was associated with a 0.36% (95% CI –0.69 to –0.03; P=.03) lower expected %BMIR at 6 months, but it was not significant at 12 months (P=.88). Late midterm and late portal engagement were not significantly associated with %BMIR at 12 months (P=.27 and P=.12, respectively). Furthermore, early engagement in various portal functions, such as messaging and accessing medical records, was significantly associated with a lower %BMIR at 3 months and a higher %BMIR at both 6 and 12 months (all P<.05). Conclusions: Higher patient portal engagement within 3 months after surgery—suggestive of stronger adherence to postoperative instructions and improved communication with care teams—is associated with less favorable weight loss immediately after surgery but enhanced postoperative weight loss outcomes at 6 and 12 months. However, the limitations of retrospective data-driven studies highlight the need for future intervention-based studies to validate these associations and establish causality.

14 September 2022
Verified
Book Chapter
Closed access
PDF file
XZXinmeng Zhang
Xinmeng Zhang
ZWZhiyu Wan
Zhiyu Wan
CYChao Yan
Chao Yan
JBJ. Thomas Brown
J. Thomas Brown
WXWeiyi Xia
Weiyi Xia
BMBradley Malin
Bradley Malin