서브메뉴
검색
Statistical Machine Learning Methodology in Precision Medicine for Multiple Survival Outcomes
Statistical Machine Learning Methodology in Precision Medicine for Multiple Survival Outcomes
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103049
- ISBN
- 9798315713661
- DDC
- 574
- 서명/저자
- Statistical Machine Learning Methodology in Precision Medicine for Multiple Survival Outcomes
- 발행사항
- [Sl] : The University of North Carolina at Chapel Hill, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 157 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
- 주기사항
- Advisor: Kosorok, Michael R.
- 학위논문주기
- Thesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2025.
- 초록/해제
- 요약Precision medicine leverages patient heterogeneity to estimate individualized treatment regimes (ITRs), which are formalized, data-driven approaches designed to match patients with tailored treatments. Currently, precision medicine methods do not simultaneously account for two endpoints. This dissertation addresses this gap by proposing three novel statistical methodologies that estimate optimal ITRs in novel multiple outcomes frameworks accounting for competing and recurrent events. With competing events, assessing cause-specific risks aids clinical decision-making. In Chapter 3, we present a nonparametric ITR estimator that identifies treatment regimes by maximizing survival from any event while minimizing the cumulative incidence of an event of interest. We also introduce a novel value function to identify the optimal ITR. To predict individualized overall survival and cause-specific cumulative incidence curves, we developed precision medicine random survival and cumulative incidence forests. Using empirical process theory and nonparametric kernel methods, we prove that the proposed forest estimators and optimal ITR estimator are each uniformly consistent over time and across patient information. We conduct extensive simulation studies to test finite-sample performance and analyze an observational cohort of peripheral artery disease patients at high risk for limb loss and mortality. In Chapter 4, we consider settings involving recurrent events with a terminal event in our optimization framework for multiple outcomes. We propose an optimal ITR estimator that prioritizes survival, but also reduces an unfavorable recurrence if mean survival time is similar across treatments. We developed and implemented forest algorithms to predict individualized curves. We establish uniform consistency for both the forests and optimal ITR estimator. Performance is examined using simulation studies and data from a bladder cancer clinical trial. Chapter 5 extends the work in Chapter 3 to the multi-stage treatment setting. We propose a reinforcement learning method that uses random stochastic draws to append to the data at each stage and jointly optimizes overall survival and cumulative incidence of an event of interest through backwards recursion. We rigorously prove the theoretical foundations of these joint forests. Additionally, we explore practical applications of our proposed method for competing risk survival data in scenarios with multiple decision points.
- 일반주제명
- Biostatistics
- 일반주제명
- Public health
- 일반주제명
- Bioinformatics
- 키워드
- Competing risks
- 키워드
- Random forests
- 키워드
- Recurrent events
- 기타저자
- The University of North Carolina at Chapel Hill Biostatistics
- 기본자료저록
- Dissertations Abstracts International. 86-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017356856
■00520260202103049
■006m o d
■007cr#unu||||||||
■020 ▼a9798315713661
■035 ▼a(MiAaPQ)AAI31931463
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aZhou, Christina W.
■24510▼aStatistical Machine Learning Methodology in Precision Medicine for Multiple Survival Outcomes
■260 ▼a[Sl]▼bThe University of North Carolina at Chapel Hill▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a157 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-11, Section: B.
■500 ▼aAdvisor: Kosorok, Michael R.
■5021 ▼aThesis (Ph.D.)--The University of North Carolina at Chapel Hill, 2025.
■520 ▼aPrecision medicine leverages patient heterogeneity to estimate individualized treatment regimes (ITRs), which are formalized, data-driven approaches designed to match patients with tailored treatments. Currently, precision medicine methods do not simultaneously account for two endpoints. This dissertation addresses this gap by proposing three novel statistical methodologies that estimate optimal ITRs in novel multiple outcomes frameworks accounting for competing and recurrent events. With competing events, assessing cause-specific risks aids clinical decision-making. In Chapter 3, we present a nonparametric ITR estimator that identifies treatment regimes by maximizing survival from any event while minimizing the cumulative incidence of an event of interest. We also introduce a novel value function to identify the optimal ITR. To predict individualized overall survival and cause-specific cumulative incidence curves, we developed precision medicine random survival and cumulative incidence forests. Using empirical process theory and nonparametric kernel methods, we prove that the proposed forest estimators and optimal ITR estimator are each uniformly consistent over time and across patient information. We conduct extensive simulation studies to test finite-sample performance and analyze an observational cohort of peripheral artery disease patients at high risk for limb loss and mortality. In Chapter 4, we consider settings involving recurrent events with a terminal event in our optimization framework for multiple outcomes. We propose an optimal ITR estimator that prioritizes survival, but also reduces an unfavorable recurrence if mean survival time is similar across treatments. We developed and implemented forest algorithms to predict individualized curves. We establish uniform consistency for both the forests and optimal ITR estimator. Performance is examined using simulation studies and data from a bladder cancer clinical trial. Chapter 5 extends the work in Chapter 3 to the multi-stage treatment setting. We propose a reinforcement learning method that uses random stochastic draws to append to the data at each stage and jointly optimizes overall survival and cumulative incidence of an event of interest through backwards recursion. We rigorously prove the theoretical foundations of these joint forests. Additionally, we explore practical applications of our proposed method for competing risk survival data in scenarios with multiple decision points.
■590 ▼aSchool code: 0153.
■650 4▼aBiostatistics
■650 4▼aPublic health
■650 4▼aBioinformatics
■653 ▼aCompeting risks
■653 ▼aEmpirical processes
■653 ▼aPrecision medicine
■653 ▼aRandom forests
■653 ▼aRecurrent events
■653 ▼aSurvival analysis
■690 ▼a0308
■690 ▼a0573
■690 ▼a0715
■71020▼aThe University of North Carolina at Chapel Hill▼bBiostatistics.
■7730 ▼tDissertations Abstracts International▼g86-11B.
■790 ▼a0153
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17356856▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.
