서브메뉴
검색
Modern Modeling Techniques for Adverse Events in Chronic Kidney Disease Patients
Modern Modeling Techniques for Adverse Events in Chronic Kidney Disease Patients
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103631
- ISBN
- 9798315791829
- DDC
- 574
- 저자명
- Qian, Qi.
- 서명/저자
- Modern Modeling Techniques for Adverse Events in Chronic Kidney Disease Patients
- 발행사항
- [Sl] : University of California, Los Angeles, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 179 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-12, Section: B.
- 주기사항
- Advisor: Senturk, Damla.
- 학위논문주기
- Thesis (Ph.D.)--University of California, Los Angeles, 2025.
- 초록/해제
- 요약Chronic kidney disease (CKD) affects nearly 15% (37 million) of adults in the United States. End-stage kidney disease (ESKD), the final stage of CKD, impacted over 807,920 individuals in 2020, with approximately 70% receiving dialysis - a life-sustaining treatment. Among CKD patients, the progressive decline in kidney function is closely linked to hospitalizations due to cardiovascular (CV) disease and terminal events such as kidney failure and death. Therefore, jointly modeling the correlated multivariate outcomes of hospitalization and mortality, along with their spatiotemporal patterns and associated risk factors, is essential for improving targeted patient monitoring and intervention.In the first chapter, using national data from the United States Renal Data System (USRDS), we propose a novel multivariate spatiotemporal functional principal component analysis (MST-FPCA) model to study joint spatiotemporal patterns in hospitalization and mortality rates among dialysis patients. This model leverages a multivariate Karhunen-Loeve expansion to capture leading directions of variation across time while inducing spatial correlation among region-specific scores. We develop an efficient estimation procedure that combines only univariate principal component decompositions with a Markov chain Monte Carlo (MCMC) framework to target spatial dependencies. The finite sample performance of the proposed method is studied through simulation studies. Application to the USRDS data identifies geographic "hot spots" in the U.S. with elevated hospitalization and/or mortality rates and highlights time periods of heightened clinical risk.Building on these findings, the second chapter explores the time-varying effects of regional-level covariates on hospitalization and mortality. We propose a multivariate varying coefficient spatiotemporal model (MV-VCSTM) to study the dynamic effects of risk factors (e.g., urbanicity, area deprivation index) on hospitalization and mortality rates as functions of time on dialysis. The proposed model flexibly accommodates time-varying covariate effects on the mean structure while modeling the spatiotemporal correlations of the residuals for efficient inference. Estimation is achieved through a fusion of functional principal component analysis and MCMC techniques, following basis expansions for varying coefficient functions and a multivariate Karhunen-Loeve expansion for region-specific random deviations. The finite sample performance of the proposed method is studied through extensive simulations. Application to the USRDS data uncovers significant regional risk factors and identifies critical time periods on dialysis and geographic areas with increased hospitalization and mortality risks.In the third chapter, we shift focus to individual-level disease progression. We introduce a novel Bayesian multivariate joint model (BM-JM) for the correlated outcomes of longitudinal kidney function (measured by estimated glomerular filtration rate, eGFR), recurrent CV events, and competing-risk terminal events (ESKD and death). The proposed modeling framework facilitates joint estimation of risk factors for each outcome and enables dynamic, personalized prediction of cumulative incidence probabilities for competing risks. These predictions are informed by a subject's baseline characteristics and their history of longitudinal eGFR and recurrent CV events. Estimation and prediction are conducted within a flexible Bayesian framework using MCMC methods. We assess predictive performance via the dynamic area under the receiver operating characteristic curve (AUC) and expected Brier score (BS). Simulation studies demonstrate the model's robustness. Application to data from the Chronic Renal Insufficiency Cohort (CRIC) study illustrates its practical utility and further emphasizes the value of integrating both longitudinal and recurrent event data to improve prognostic accuracy in CKD populations.
- 일반주제명
- Biostatistics
- 일반주제명
- Medicine
- 일반주제명
- Bioinformatics
- 키워드
- CKD population
- 기타저자
- University of California, Los Angeles Biostatistics 0132
- 기본자료저록
- Dissertations Abstracts International. 86-12B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017358014
■00520260202103631
■006m o d
■007cr#unu||||||||
■020 ▼a9798315791829
■035 ▼a(MiAaPQ)AAI32046910
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aQian, Qi.
■24510▼aModern Modeling Techniques for Adverse Events in Chronic Kidney Disease Patients
■260 ▼a[Sl]▼bUniversity of California, Los Angeles▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a179 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-12, Section: B.
■500 ▼aAdvisor: Senturk, Damla.
■5021 ▼aThesis (Ph.D.)--University of California, Los Angeles, 2025.
■520 ▼aChronic kidney disease (CKD) affects nearly 15% (37 million) of adults in the United States. End-stage kidney disease (ESKD), the final stage of CKD, impacted over 807,920 individuals in 2020, with approximately 70% receiving dialysis - a life-sustaining treatment. Among CKD patients, the progressive decline in kidney function is closely linked to hospitalizations due to cardiovascular (CV) disease and terminal events such as kidney failure and death. Therefore, jointly modeling the correlated multivariate outcomes of hospitalization and mortality, along with their spatiotemporal patterns and associated risk factors, is essential for improving targeted patient monitoring and intervention.In the first chapter, using national data from the United States Renal Data System (USRDS), we propose a novel multivariate spatiotemporal functional principal component analysis (MST-FPCA) model to study joint spatiotemporal patterns in hospitalization and mortality rates among dialysis patients. This model leverages a multivariate Karhunen-Loeve expansion to capture leading directions of variation across time while inducing spatial correlation among region-specific scores. We develop an efficient estimation procedure that combines only univariate principal component decompositions with a Markov chain Monte Carlo (MCMC) framework to target spatial dependencies. The finite sample performance of the proposed method is studied through simulation studies. Application to the USRDS data identifies geographic "hot spots" in the U.S. with elevated hospitalization and/or mortality rates and highlights time periods of heightened clinical risk.Building on these findings, the second chapter explores the time-varying effects of regional-level covariates on hospitalization and mortality. We propose a multivariate varying coefficient spatiotemporal model (MV-VCSTM) to study the dynamic effects of risk factors (e.g., urbanicity, area deprivation index) on hospitalization and mortality rates as functions of time on dialysis. The proposed model flexibly accommodates time-varying covariate effects on the mean structure while modeling the spatiotemporal correlations of the residuals for efficient inference. Estimation is achieved through a fusion of functional principal component analysis and MCMC techniques, following basis expansions for varying coefficient functions and a multivariate Karhunen-Loeve expansion for region-specific random deviations. The finite sample performance of the proposed method is studied through extensive simulations. Application to the USRDS data uncovers significant regional risk factors and identifies critical time periods on dialysis and geographic areas with increased hospitalization and mortality risks.In the third chapter, we shift focus to individual-level disease progression. We introduce a novel Bayesian multivariate joint model (BM-JM) for the correlated outcomes of longitudinal kidney function (measured by estimated glomerular filtration rate, eGFR), recurrent CV events, and competing-risk terminal events (ESKD and death). The proposed modeling framework facilitates joint estimation of risk factors for each outcome and enables dynamic, personalized prediction of cumulative incidence probabilities for competing risks. These predictions are informed by a subject's baseline characteristics and their history of longitudinal eGFR and recurrent CV events. Estimation and prediction are conducted within a flexible Bayesian framework using MCMC methods. We assess predictive performance via the dynamic area under the receiver operating characteristic curve (AUC) and expected Brier score (BS). Simulation studies demonstrate the model's robustness. Application to data from the Chronic Renal Insufficiency Cohort (CRIC) study illustrates its practical utility and further emphasizes the value of integrating both longitudinal and recurrent event data to improve prognostic accuracy in CKD populations.
■590 ▼aSchool code: 0031.
■650 4▼aBiostatistics
■650 4▼aMedicine
■650 4▼aBioinformatics
■653 ▼aChronic kidney disease
■653 ▼aEnd-stage kidney disease
■653 ▼aUnited States Renal Data System
■653 ▼aMarkov chain Monte Carlo
■653 ▼aCKD population
■690 ▼a0308
■690 ▼a0564
■690 ▼a0715
■71020▼aUniversity of California, Los Angeles▼bBiostatistics 0132.
■7730 ▼tDissertations Abstracts International▼g86-12B.
■790 ▼a0031
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17358014▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


