Statistics in Medicine最新文献

We reexamine GOOOGLE, the group-regularized zero-inflated negative binomial (ZINB) approach of Chatterjee et al. We show that in the released implementation, the tuning parameter is selected using a Bayesian information criterion (BIC) computed on a Gaussian surrogate. Because the unpenalized model fits this surrogate exactly (with zero residual sum of squares), the BIC always favors essentially unpenalized solutions and fails to induce group sparsity. This results in zero group specificity in simulations mirroring the original paper's design. We demonstrate that this issue is resolved by selecting the tuning parameter using the true ZINB log-likelihood. Furthermore, we propose the fully iterative group broken adaptive ridge (grBAR) estimator as a more robust alternative. Our open-source R package, group regularization algorithms for zero-inflated models (GRAZIMs), provides these tools to enable reliable group selection in ZINB models.

We propose a random-effects approach to missing values for generalized linear mixed model (GLMM) analysis of longitudinal data. The method converts a GLMM with missing covariates to another GLMM without missing covariates. The standard GLMM analysis tools then apply. The method applies, in particular, to the cases of linear mixed models (LMMs) and logistic regression. Performance of the method is evaluated empirically, and compared with alternative approaches, including the popular MICE procedure of multiple imputation (MI). Theoretical justification of the method is given, and explained, for the patterns observed in the simulation studies. Two real-data examples from healthcare studies are discussed.

Despite advancements in healthcare data management, missing data in electronic health records (EHR) and patient-reported outcomes remain a persistent challenge, limiting their usability in healthcare analytics. Conventional imputation methods often struggle to capture complex nonlinear relationships, require extensive computation time, and are limited in addressing various types of missing data mechanisms. To overcome these challenges, we propose the clustering-informed shared-structure variational autoencoder (CISS-VAE), which utilizes the strengths of Bayesian neural networks. This model can effectively capture complex associations and accommodate various missing data mechanisms, including missing not at random (MNAR). We also develop iterative learning algorithms that further enhance missing data imputation accuracy while preventing overfitting. Comprehensive simulations demonstrate the superior accuracy of our model compared to traditional and contemporary methods. We apply our method to EHR data from early-stage breast cancer patients at Memorial Sloan Kettering Cancer Center, aiming to mitigate the impact of missing data and enhance health monitoring and analyses.

In diagnostic studies, researchers frequently encounter imperfect reference standards with some misclassified labels. Treating these as gold standards can bias receiver operating characteristic (ROC) curve analysis. To address this issue, we propose a novel likelihood-based method under a non-parametric density ratio model. This approach enables the reliable estimation of the ROC curve, area under the curve (AUC), partial AUC, and Youden's index with favorable statistical properties. To implement the method, we develop an efficient expectation-maximization (EM) algorithm. Extensive simulations evaluate its finite-sample performance, showing smaller mean squared errors in estimating the ROC curve, partial AUC, and Youden's index compared to existing methods. We apply the proposed approach to a malaria study.

Violations of the positivity assumption can render conventional causal estimands unidentifiable, including the average treatment effect (ATE), the average treatment effect on the treated (ATT), and the average treatment effect on the controls (ATC). Shifting the inferential focus to their alternative counterparts-the weighted ATE (WATE), the weighted ATT (WATT), and the weighted ATC (WATC)-offers valuable insights into treatment effects while preserving internal validity. In this tutorial, we provide a comprehensive review of recent advances in propensity score (PS) weighting methods, along with practical guidance on how to select a primary target estimand (while other estimands serve as supplementary analyses), implement the corresponding PS-weighted estimators, and conduct post-weighting diagnostic assessments. The tutorial is accompanied by a user-friendly R package, ChiPS. We demonstrate the pertinence of various estimators through extensive simulation studies. We illustrate the flow of the tutorial on two real-world case studies: (i) Effect of smoking on blood lead level using data from the 2007-2008 National Health and Nutrition Examination Survey (NHANES); and (ii) Impact of history of sex work on HIV status among transgender women in South Africa.

Assessing treatment effect heterogeneity (TEH) in clinical trials is crucial, as it provides insights into the variability of treatment responses among patients, influencing key decisions related to drug development. Furthermore, it can lead to personalized medicine by tailoring treatments to individual patient characteristics. This paper introduces novel methodologies for assessing treatment effects using the individualized treatment effect as a basis. To estimate this effect, we use a doubly robust (DR) learner to infer a pseudo-outcome that reflects the causal contrast. This pseudo-outcome is then used to perform three objectives: (1) a global test for heterogeneity, (2) ranking covariates based on their influence on effect modification, and (3) providing estimates of the individualized treatment effect. We compare the DR-learner with various alternatives and competing methods in a simulation study, and also use it to assess heterogeneity in a pooled analysis of five Phase III trials in psoriatic arthritis (PsA). By integrating these methods with the recently proposed Workflow to Assess Treatment Effect Heterogeneity in Drug Development for Clinical Trial Sponsors (WATCH) workflow, we provide a robust framework for analyzing TEH, offering insights that enable more informed decision-making in this challenging area.

Estimating the optimal individualized treatment regimes based on patients' characteristic information has become an increasingly important topic in personalized medicine study. These characteristic data can range from simple scalar values to complex functional data such as curves or images, which might be taken into account to recommend more beneficial treatment decisions for patients. In this paper, we propose a novel semiparametric partial functional regression model for estimating the optimal individualized treatment regimes with scalar and functional covariates. One advantage of this model is that it involves a fully nonparametric main effect of the covariates and a flexible interaction effect between the covariates and the treatment, and greatly reduces the risk of model misspecification. The form of single index interaction effect with a monotone link function ensures the estimated optimal individualized treatment regime preserving good interpretability. We estimate this model via B-spline and establish the convergence rate of the estimated optimal individualized treatment regime. Sufficient simulation studies and a real data analysis are conducted to assess the performance of the proposed method.

In streaming longitudinal data, status prediction becomes challenging when input variables are block-sparse, autocorrelated, and irregular in both dimension and distribution. General methods cannot model such data directly, especially when the classes are extremely imbalanced. This research proposes a K-Nearest Neighbor (KNN) algorithm where distance is measured by Kullback-Leibler (KL) divergence. The algorithm uses features extracted from metric conditional density, both with and without first-order lag. The developed streaming KNN algorithm is further applied to simulation data. Results show that when differences originate from the location hyperparameters of the Gaussian distribution or both the shape and scale hyperparameters of the inverse gamma distribution, the method performs quite well, as expected, with an AUC close to 1. Additionally, a numerical method is proposed for general distributions that lack an analytical expression in real data. This method is applied to a big medical streaming dataset with similar properties. Results indicate that the AUC value gradually increases to 0.913, with a sensitivity of 0.851 and a specificity of 0.816.

With the rapid advancements in wearable device technologies, there is a growing interest in learning useful digital biomarkers from wearable device data as objective, low-cost, real-time alternatives to use in healthcare settings. They have the potential to facilitate disease progression monitoring, medication tailoring, and supplementing clinical trial endpoints. For example, triaxial accelerometer sensor data is promising for monitoring symptoms of movement-related diseases, such as tremors in Parkinson's disease (PD). However, existing methods for accelerometer studies based on hidden Markov models (HMM) often analyze each individual's activity data separately, leading to inefficiency and limited generalizability. This paper proposes a joint nonparametric Bayesian method that extends the hierarchical Dirichlet process autoregressive HMM (HDP-AR-HMM) to incorporate subject-specific transition parameters. This approach allows for simultaneous estimation across multiple subjects and repeated measurements, accounts for between-subject variability, and provides consistent hidden state estimation without pre-specifying the number of states. We validate our method on simulated data and show that it can achieve higher accuracy in detecting the true hidden states compared to alternative methods. We apply the method to a free-living study, the Biomarker & Endpoint Assessment to Track Parkinson's disease (BEAT-PD) DREAM Challenge CIS-PD study, to demonstrate its utility in monitoring disease symptoms in PD patients.

Empirical substantive research, such as in the life or social sciences, is commonly categorized into the two modes exploratory and confirmatory, both of which are essential to scientific progress. The former is also referred to as hypothesis-generating or data-contingent research, while the latter is also called hypothesis-testing research. In the context of empirical methodological research in statistics, however, the exploratory-confirmatory distinction has received very little attention so far. Our paper aims to fill this gap. First, we revisit the concept of empirical methodological research through the lens of the exploratory-confirmatory distinction. Second, we examine current practice with respect to this distinction through a literature survey including 115 articles from the field of biostatistics. Third, we provide practical recommendations toward a more appropriate design, interpretation, and reporting of empirical methodological research in light of this distinction. In particular, we argue that both modes of research are crucial to methodological progress, but that most published studies-even if sometimes disguised as confirmatory-are essentially exploratory in nature. We emphasize that it may be adequate to consider empirical methodological research as a continuum between "pure" exploration and "strict" confirmation, recommend transparently reporting the mode of conducted research within the spectrum between exploratory and confirmatory, and stress the importance of study protocols written before conducting the study, especially in confirmatory methodological research.