Article
Mathematical & Computational Biology
Di Shu, Peisong Han, Rui Wang, Sengwee Toh
Summary: The inverse probability weighted Cox model is used to estimate the marginal hazard ratio, requiring correct specification of the propensity score model. To address misspecification, a weighted estimation method rooted in empirical likelihood theory is proposed. The method demonstrates satisfactory performance in terms of consistency and efficiency in simulation studies and application to comparing postoperative hospitalization risks between two surgical procedures. Extending the method to multisite studies allows for site-specific propensity score models.
STATISTICS IN MEDICINE
(2021)
Article
Public, Environmental & Occupational Health
Sarah A. Reifeis, Michael G. Hudgens
Summary: Inverse probability weighting (IPW) is commonly used in observational studies to estimate treatment effects. The traditional method of estimating variance using the robust sandwich estimator may not always be valid. Instead, stacked estimating equations that consider weight estimation can provide a consistent variance estimator.
AMERICAN JOURNAL OF EPIDEMIOLOGY
(2022)
Article
Public, Environmental & Occupational Health
Jacqueline E. Rudolph, Enrique F. Schisterman, Ashley Naimi
Summary: The authors compared the performance of inverse probability weighting (IPW) and g-computation in time-varying analyses. They found that IPW and Monte Carlo g-computation performed similarly, while ICE g-computation had the least bias but lowest precision.
AMERICAN JOURNAL OF EPIDEMIOLOGY
(2023)
Article
Mathematical & Computational Biology
Peter Z. Schochet
Summary: In clustered randomized controlled trials, sample recruitment occurring after cluster randomization can lead to recruitment bias. This article presents a potential outcomes framework that yields a causal estimand related to individuals always recruited into the research conditions. A consistent inverse probability weighting estimator is developed using data on recruits only, and a generalized estimating equations approach is used to obtain robust clustered standard error estimators that account for estimation error in the weighting. A simple data collection strategy is discussed to improve the predictive accuracy of the logit propensity score models.
STATISTICS IN MEDICINE
(2023)
Article
Biology
Bonnie E. Shook-Sa, Michael G. Hudgens
Summary: This paper investigates the impact of inverse probability of treatment weights (IPTWs) on sample size calculations when estimating causal effects. It presents a simplified design effect approximation method and discusses practical considerations.
Article
Mathematical & Computational Biology
Peter C. Austin
Summary: We conducted Monte Carlo simulations to compare asymptotic variance estimators with the bootstrap method in estimating standard errors using propensity score weighting. The results showed that the bootstrap method outperformed the asymptotic estimators in certain scenarios, but also had limitations in some cases.
STATISTICS IN MEDICINE
(2022)
Article
Public, Environmental & Occupational Health
Joan T. Price, Yuri Sebastiao, Bellington Vwalika, Stephen R. Cole, Felistas M. Mbewe, Winifreda M. Phiri, Bethany L. Freeman, Margaret P. Kasaro, Marc Peterson, Dwight J. Rouse, Elizabeth M. Stringer, Jeffrey S. A. Stringer
Summary: Comparing the risks of preterm birth and stillbirth between IPOP and ZAPPS studies showed a significantly lower risk in IPOP, even after adjusting for baseline characteristics. The possible benefits of additional monetary reimbursement, more frequent visits, and group-based care in IPOP warrant further investigation to understand the reasons for the lower risk observed.
Article
Economics
Huijuan Ma, Jing Qin, Yong Zhou
Summary: It is well known that conditioning on covariates can improve the estimation of the marginal outcome distribution. This article establishes a connection between marginal quantile and conditional quantile regression and proposes two novel estimation approaches using conditional quantile regression. The consistency and asymptotic normality of the estimators are proven, and the second approach achieves semiparametric efficiency.
JOURNAL OF BUSINESS & ECONOMIC STATISTICS
(2023)
Article
Health Care Sciences & Services
Eleanor M. Pullenayegum, Catherine Birken, Jonathon Maguire
Summary: Data collected over time in usual healthcare can be used for research, but the timing of measurements may affect outcomes. Failure to consider this can result in biased inferences. A new semi-parametric joint model is proposed to address this issue, which shows promising performance in simulation studies.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2021)
Article
Statistics & Probability
Trinetri Ghosh, Yanyuan Ma, Xavier de Luna
Summary: In this study, a semi-parametric locally efficient dimension-reduction approach was used to assess treatment effects in observational studies. Results were integrated using imputation, inverse probability weighting, and doubly robust augmentation estimators. A shrinkage estimator combining the advantages of doubly robust and imputation estimators was proposed, retaining double robustness while improving variance when response models are correct. The performance of these estimators was demonstrated using simulated experiments and real data on the impact of maternal smoking on baby birth weight.
Article
Public, Environmental & Occupational Health
Elizabeth A. Lancet, Luisa N. Borrell, Janet Holbrook, Alfredo Morabia
Summary: This study examines the validity of Intention to Treat (ITT) analyses in the presence of non-adherence and lost to follow up in an existing randomized clinical trial on asthma treatment using Marginal Structural Models (MSMs). Results showed that the adjusted relative risks from MSMs and ITT analyses were nearly identical, indicating that adherence and censoring may not invalidate ITT analysis.
ANNALS OF EPIDEMIOLOGY
(2021)
Article
Public, Environmental & Occupational Health
Paola Gilsanz, Jessica G. Young, M. Maria Glymour, Eric J. Tchetgen Tchetgen, Chloe W. Eng, Karestan C. Koenen, Laura D. Kubzansky
Summary: Social epidemiology aims to identify social structural risk factors and inform interventions. However, determining the most effective interventions and timing is challenging due to variations in social conditions and time-varying confounding. Marginal structural models (MSMs) are useful but face challenges in studying social epidemiologic exposures over the life course. A study using simulated data found that correctly specified MSMs accurately estimated causal effects while conventional regression models showed bias.
AMERICAN JOURNAL OF EPIDEMIOLOGY
(2022)
Article
Medicine, General & Internal
Ian Shrier, Annabelle Redelmeier, Mireille E. Schnitzer, Russell J. Steele
Summary: Properly interpreting research results is crucial in evidence-based medicine. Most studies use multiple regression to adjust for variables and report adjusted effects. However, interpreting the estimates from these analyses as population average causal effects can be incorrect if there is an interaction between treatment and other variables with respect to the outcome. Researchers should be cautious when considering excluding interaction terms based on p values.
BMJ EVIDENCE-BASED MEDICINE
(2021)
Article
Mathematics, Applied
Qiang Zhao, Chao Zhang, Jingjing Wu, Xiuli Wang
Summary: Two types of weighted quantile estimators were proposed for nonlinear models with missing covariates, and the asymptotic normality of the estimators was established. The optimal weights were calculated, and the asymptotic distributions of the resulting estimators were derived. Numerical simulations and real data analysis were conducted for performance comparison with other methods.
Article
Mathematical & Computational Biology
Shuxi Zeng, Fan Li, Rui Wang, Fan Li
Summary: This article discusses methods for addressing chance imbalance in baseline characteristics in randomized clinical trials, advocating the use of overlap weighting (OW) for covariate adjustment. Through simulations, OW consistently outperforms IPW in finite samples, showing better efficiency when the degree of treatment effect heterogeneity is moderate or when the outcome model is incorrectly specified.
STATISTICS IN MEDICINE
(2021)
Article
Mathematical & Computational Biology
Dongdong Li, Wenbin Lu, Di Shu, Sengwee Toh, Rui Wang
Summary: This article proposes a distributed methodology for fitting Cox proportional hazards models in multi-site studies without sharing individual-level data. The method uses summary-level statistics and can accommodate different types of models and adjust for multiple covariates. Through simulation studies and real-world data application, the feasibility and accuracy of the proposed method are verified.
Article
Statistics & Probability
Dongdong Li, X. Joan Hu, Rui Wang
Summary: This article focuses on evaluating the association between two event times without specifying the joint distribution parametrically. The presence of informative censoring due to a terminating event, such as death, makes this task particularly challenging. The authors propose a method that links the joint distribution of the two event times and the informative censoring time using a nested copula function. They also use flexible functional forms to specify the covariate effects on both the marginal and joint distributions. The proposed approach estimates the association parameters, marginal survival functions, and covariate effects in a semiparametric model for the bivariate event time.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Mathematical & Computational Biology
Fan Li, Xinyuan Chen, Zizhong Tian, Denise Esserman, Patrick J. Heagerty, Rui Wang
Summary: This article explores the heterogeneity of treatment effects in different patient subpopulations in cluster randomized trials and presents a novel analytical design formula that can be widely applied to evaluate effect modifiers at different levels. The effectiveness of this new method is validated through simulation studies and real-world trial examples.
Article
Mathematical & Computational Biology
Xueqi Wang, Elizabeth L. Turner, Fan Li
Summary: Cluster randomized trials (CRTs) often use a small number of clusters, requiring small-sample corrections for valid inference. This study proposes nine bias-corrected sandwich variance estimators for the marginal Cox model in CRTs analyzing time-to-event outcomes, evaluating their performance through simulation studies. The results show that the choice of variance estimator depends on cluster size variability and the evaluation metric used. In a real-world CRT, the use of small-sample bias corrections affects the conclusion about intervention effectiveness. The proposed estimators are implemented in the R package CoxBcv.
BIOMETRICAL JOURNAL
(2023)
Article
Biology
Lara Maleyeff, Fan Li, Sebastien Haneuse, Rui Wang
Summary: This article proposes a new model formulation for assessing treatment effect heterogeneity over exposure time in stepped-wedge CRTs, using random effects to represent varying treatment effects by exposure time, leading to more precise estimation of average and exposure-time-specific treatment effects. It also develops a permutation test for the variance component of the heterogeneous treatment effect parameters.
Article
Public, Environmental & Occupational Health
Dongdong Li, Jenna Wong, Xiaojuan Li, Sengwee Toh, Rui Wang
Summary: This study compared four imputation methods for missing data in distributed research networks (DRNs). The results showed that all imputation methods produced unbiased and more efficient estimates under small effect sizes and homogeneous missingness mechanisms. Random forest (RF) method had higher efficiency under heterogeneous missingness mechanisms. Estimates from distributed imputation combined by meta-analysis were similar to those from imputation using pooled data. Substantive model compatible imputation (SMC) performed best when effect sizes were large.
PHARMACOEPIDEMIOLOGY AND DRUG SAFETY
(2023)
Article
Public, Environmental & Occupational Health
Brett M. Biebelberg, Shangyuan Ye, Rui Wang, Michael Klompas, Chanu Rhee
Summary: Hospital-acquired Aspergillus rates among COVID-19 patients were initially higher at a hospital with high negative-pressure room utilization compared to a similar hospital with low utilization. After the index hospital decreased negative-pressure utilization, rates at the 2 hospitals converged.
INFECTION CONTROL AND HOSPITAL EPIDEMIOLOGY
(2023)
Article
Clinical Neurology
Keith A. Robinson, Zhuoran Wei, Jerilynn Radcliffe, H. Gerry Taylor, Cristina M. Baldassari, Ronald D. Chervin, Stacey Ishman, Ron B. Mitchell, Ignacio E. Tapia, Susan Garetz, Fauziya Hassan, Sally Ibrahim, Lisa M. Elden, Carolyn E. Ievers-Landis, Ariel A. Williamson, Michelle Hjelm, Erin Kirkham, Addy Tham, Kamal Naqvi, Michael Rueschman, Carol L. Rosen, Rui Wang, Susan Redline
Summary: Sleep duration and continuity are associated with neurocognitive functioning in children with mild sleep-disordered breathing, supporting efforts to target these sleep health parameters as part of interventions for reducing neurobehavioral morbidity.
JOURNAL OF CLINICAL SLEEP MEDICINE
(2023)
Article
Health Care Sciences & Services
Yue Song, Rui Wang
Summary: Nonlinear mixed effects models are widely used in analyzing data from biological, agricultural, and environmental sciences. The estimation and inference of parameters in these models are typically based on a likelihood function. However, the specification of the random effects distribution can complicate the process, especially with multiple random effects. This study proposes a smoothed simulated pseudo-maximum likelihood estimation approach to fit nonlinear mixed effects models with left-censored observations, and provides consistency, asymptotic normality, and testing procedures for random effects correlation and distributional assumptions.
STATISTICAL METHODS IN MEDICAL RESEARCH
(2023)
Article
Mathematical & Computational Biology
Linda J. Harrison, Rui Wang
Summary: Randomized trials are widely used to evaluate interventions, but missing outcome data is a common problem. It is unclear how to account for this in sample size calculations. We propose a method based on inverse probability of response weighting (IPRW) to adjust sample size calculations for missing outcome data. We derive sample size formulas for individually randomized and cluster randomized trials, and provide an R shiny app for easy implementation.
STATISTICS IN MEDICINE
(2023)
Article
Multidisciplinary Sciences
Shangyuan Ye, Daniel Li, Tingting Yu, Daniel A. Caroff, Jeffrey Guy, Russell E. Poland, Kenneth E. Sands, Edward J. Septimus, Susan S. Huang, Richard Platt, Rui Wang
Summary: The Centers for Medicare and Medicaid Services require hospitals to report on quality metrics and financially penalize those in the lowest performance quartile. Surgical site infections (SSIs) are a critical component of these metrics, but accurate profiling is hindered by small surgical volumes. The exclusion of hospitals with less than one expected SSI from rankings is currently used, but its effectiveness is uncertain. Therefore, reliable evaluation criteria based on surgical volumes and predicted events are needed.
SCIENTIFIC REPORTS
(2023)
Article
Mathematical & Computational Biology
Chia-Rui Chang, Yue Song, Fan Li, Rui Wang
Summary: Covariate adjustment is important in analyzing data from randomized clinical trials, but missing data can be a barrier. This study reviews different covariate adjustment methods with incomplete covariate data. The researchers propose a weighting approach that combines inverse probability weighting and overlap weighting to adjust for missing outcomes and covariates, and conduct comprehensive simulation studies to evaluate the performance of the methods.
STATISTICS IN MEDICINE
(2023)
Article
Mathematical & Computational Biology
Lara Maleyeff, Rui Wang, Sebastien Haneuse, Fan Li
Summary: This study proposes a method for testing treatment effect heterogeneity in cluster randomized trials. Through a generalized linear mixed model, we derive sample size expressions for binary effect modifiers and develop a computationally efficient Monte Carlo approach for continuous effect modifiers. Our findings contribute to filling the methodological gap in existing research.
STATISTICS IN MEDICINE
(2023)
Article
Mathematical & Computational Biology
Di Shu, Xiaojuan Li, Qoua Her, Jenna Wong, Dongdong Li, Rui Wang, Sengwee Toh
Summary: Missing data in multi-site studies pose challenges for statistical analyses. We proposed a one-step estimation method that combines meta-analysis and within-site multiple imputation to estimate the average causal effect of a target population without sharing individual-level data. We evaluated six different approaches and found that combining results across sites using sample-standardization weights or Rubin's rules yielded the best performance. Applying inverse-variance weighted meta-analysis without accounting for treatment effect heterogeneity can lead to biased results.
RESEARCH SYNTHESIS METHODS
(2023)