Faculty Bibliography

A novel functional additive model is proposed, which is uniquely modified and constrained to model nonlinear interactions between a treatment indicator and a potentially large number of functional and/or scalar pretreatment covariates. The primary motivation for this approach is to optimize individualized treatment rules based on data from a randomized clinical trial. We generalize functional additive regression models by incorporating treatment-specific components into additive effect components. A structural constraint is imposed on the treatment-specific components in order to provide a class of additive models with main effects and interaction effects that are orthogonal to each other. If primary interest is in the interaction between treatment and the covariates, as is generally the case when optimizing individualized treatment rules, we can thereby circumvent the need to estimate the main effects of the covariates, obviating the need to specify their form and thus avoiding the issue of model misspecification. The methods are illustrated with data from a depression clinical trial with electroencephalogram functional data as patients' pretreatment covariates.

Sparse additive modeling is a class of effective methods for performing high-dimensional nonparametric regression. This article develops a sparse additive model focused on estimation of treatment effect modification with simultaneous treatment effect-modifier selection. We propose a version of the sparse additive model uniquely constrained to estimate the interaction effects between treatment and pretreatment covariates, while leaving the main effects of the pretreatment covariates unspecified. The proposed regression model can effectively identify treatment effect-modifiers that exhibit possibly nonlinear interactions with the treatment variable that are relevant for making optimal treatment decisions. A set of simulation experiments and an application to a dataset from a randomized clinical trial are presented to demonstrate the method.

We consider a logistic regression model for a binary response where part of its covariates are subject-specific random intercepts and slopes from a large number of longitudinal covariates. These random effect covariates must be estimated from the observed data, and therefore, the model essentially involves errors in covariates. Because of high dimension and high correlation of the random effects, we employ longitudinal principal component analysis to reduce the total number of random effects to some manageable number of random effects. To deal with errors in covariates, we extend the conditional-score equation approach to this moderate dimensional logistic regression model with random effect covariates. To reliably solve the conditional-score equations in moderate/high dimension, we apply a majorization on the first derivative of the conditional-score functions and a penalized estimation by the smoothly clipped absolute deviation. The method was evaluated through a set of simulation studies and applied to a data set with longitudinal cortical thickness of 68 regions of interest to identify biomarkers that are related to dementia transition.

This paper presents a Bayesian reformulation of covariate-assisted principal regression for covariance matrix outcomes to identify low-dimensional components in the covariance associated with covariates. By introducing a geometric approach to the covariance matrices and leveraging Euclidean geometry, we estimate dimension reduction parameters and model covariance heterogeneity based on covariates. This method enables joint estimation and uncertainty quantification of relevant model parameters associated with heteroscedasticity. We demonstrate our approach through simulation studies and apply it to analyze associations between covariates and brain functional connectivity using data from the Human Connectome Project.

OBJECTIVE:To retrospectively evaluate the diagnostic utility of dual-energy computed tomography (DECT) virtual non-calcium bone marrow edema (BME) maps in the assessment of osseous pelvic trauma compared to conventional CT. In addition to fracture detection rates, we sought to determine the DECT effect on reader diagnostic confidence and follow-up recommendations. MATERIALS AND METHODS/METHODS:Three radiologists performed independent retrospective reviews of 92 consecutive pelvic CTs for the presence of fractures, without the use of DECT-generated BME maps. Following a washout period of ≥ 4 weeks, the cases were re-reviewed with the addition of DECT BME maps. Comparison of fracture detection rates, frequency of "indeterminate" reads, reader confidence ratings, and follow-up imaging recommendations were performed using McNemar's test and mixed-effects logistic regressions. Mixed-effect models were utilized to assess the effect of DEXA-derived patient bone mineral density, when available, on fracture detection rates. RESULTS:The use of DECT BME maps resulted in higher pooled reader diagnostic confidence, with a mean 0.19 increase on a 5-point Likert scale (p < 0.001). There was, however, no statistically significant difference in fracture detection rates for readers utilizing DECT versus conventional CT alone, with similar rates of "indeterminate" interpretations between the groups. For patients with recent DEXAs (n = 27), DECT was not associated with change in fracture detection rates among patients with osteoporosis or osteopenia. Readers recommended follow-up imaging more frequently in cases interpreted with DECT, but this effect was not significant (p = 0.266). CONCLUSION/CONCLUSIONS:When interpreting pelvic CT for trauma, DECT-generated BME maps increased diagnostic confidence without changing detection rates.

Infancy poses a spatial-coordinate challenge for developmental EEG: cortical geometry, head geometry, and EEG forward propagation change rapidly, so a cortical coordinate system that is meaningful at one age may not be directly comparable at another. We addressed this question in a template-based computational analysis using age-specific infant anatomical templates distributed through MNE-Python. Specifically, we computed cortical Laplace-Beltrami (LB) eigenmodes (cortical harmonics), EEG forward models, and forward-projected eigenmode dictionaries. We asked which eigenmode orders are preferentially expressed at the scalp and whether independently computed age-specific eigenmode coordinates remain stable for cross-age comparison. Forward-projected scalp gain was concentrated in lower eigenmode orders across infancy, indicating a stable coarse-to-fine transfer profile. However, neighboring-age LB bases were only locally comparable by nominal mode index. Same-index modes showed local reordering and mode-index drift, and the same cortical pattern produced coefficient leakage into nearby modes when re-expressed in a neighboring-age basis. These template-level coordinate differences affected simulated downstream analyses: neighboring-age dictionaries were not fully substitutable in low-dimensional sensor space, and a fixed adult-derived basis was suboptimal for recovering same-index coordinates, with larger penalties in early- to mid-infancy. Sequential Procrustes tracking improved same-index consistency, supporting local alignment as a practical step toward age-aware coordinates. Because all quantitative summaries are derived from population-average templates and simulations, they should be interpreted as mechanistic evidence about coordinate-system effects rather than direct estimates of empirical EEG error. These results motivate age-parameterized or explicitly tracked cortical harmonic coordinates for longitudinal developmental EEG and related lifespan analyses.

BACKGROUND/UNASSIGNED:The coronavirus disease 2019 (COVID-19) pandemic spurred a tremendous increase in the adoption and use of remote patient monitoring (RPM) for hypertension (HTN) management. However, limited evidence exists on the associations between frequency of utilization and uncontrolled blood pressure (BP). OBJECTIVES/UNASSIGNED:The present study comprehensively explores the associations between RPM use frequency and uncontrolled BP among a metropolitan-dwelling sample of hypertensive patients. METHODS/UNASSIGNED:Of 2,920 participants from a single urban health system, we employed a range of analytical perspectives to evaluate the RPM utilization-uncontrolled BP relationship across widely used engagement metrics: Frequency of BP transmission, digitally enabled clinician interactions, patient portal interactions, and a composite measure of utilization. Our dichotomized primary and secondary endpoints were BP >140/90 mm Hg and BP >130/80 mm Hg. RESULTS/UNASSIGNED:Fifty-nine percent of participants were females (59%), one-third (37%) were ≥65 years old, and Hispanic patients were most represented (39%). Our primary uncontrolled BP endpoint demonstrated strong adjusted associations with suboptimal RPM use across dichotomized measures: Low BP transmission (odds ratio [OR]: 2.02, 95% confidence interval [CI]: 1.41-2.96), low clinician interactions (OR: 1.83, 95% CI: 1.43-2.36), low patient portal interactions (OR: 1.83, 95% 1.46-2.30), and low overall engagement (OR: 3.50, 95% 2.77-4.46). Our causal evaluations mirrored these findings, showing moderate causal associations after comprehensive adjustment for confounding. Assessments using other data types, such as continuous and quartiles, showed significant associations and an apparent dose-response relationship, though not at a similar magnitude. CONCLUSION/UNASSIGNED:We observed strong associations between low RPM utilization and uncontrolled BP, with promising implications for patients with collectively high RPM use. These findings highlight the need to strengthen digital inclusion initiatives to improve RPM uptake and support existing efforts aimed at developing RPM clinical practice guidelines and expanding RPM reimbursement policies. Further research is warranted across diverse utilization components to better understand the linkages between engagement frequency and improved clinical outcomes.

BACKGROUND:Anxiety disorders may partly stem from altered neurodevelopment of attention-related networks. Neonatal alterations in resting-state functional connectivity (rsFC) among the dorsal attention (DAN); frontal parietal (FPN); salience (SN); and default mode networks (DMN)) relate to fearful temperament, a risk marker for anxiety. Nevertheless, little research examines development of these networks beyond the first months of life, particularly in fearful infants. This study examines how changes in these networks in the first two years of life relate to fearful temperament. METHODS:Using data from the Baby Connectome Project (from 180 infants across 396 sessions), we conducted independent components analysis to extract rsFC among the DMN, SN, DAN, and FPN. Longitudinal modeling characterized 1) age-related changes (slope) in rsFC through age two; 2) relations between rsFC change (slope) and fearfulness at age 2; 3) relations between rsFC and fearfulness trajectories (slope and intercept) over the first two years of life. RESULTS:Age-related decreases occurred in rsFC in DAN - FPN and DMN - SN. Smaller decreases in DAN - FPN rsFC over time related to greater fear at age 2, and to increases in fearfulness over time. High initial DAN-FPN rsFC and low initial DAN - SN rsFC also related to increasing fearfulness over time. CONCLUSION/CONCLUSIONS:This study provides the first evidence that changes in attention-related brain networks are related to early-life fearfulness, a robust early-life risk marker of anxiety.

An important goal of precision medicine is to personalize medical treatment by identifying individuals who are most likely to benefit from a specific treatment. The likely responder (LR) framework, which identifies a subpopulation where treatment response is expected to exceed a certain clinical threshold, plays a role in this effort. However, the LR framework, and more generally, data-driven subgroup analyses, often fail to account for uncertainty in the estimation of model-based data-driven subgrouping. We propose a simple two-stage approach that integrates subgroup identification with subsequent subgroup-specific inference on treatment effects. We incorporate model estimation uncertainty from the first stage into subgroup-specific treatment effect estimation in the second stage, by utilizing Bayesian posterior distributions from the first stage. We evaluate our method through simulations, demonstrating that the proposed Bayesian two-stage model produces better calibrated confidence intervals than naïve approaches. We apply our method to an international COVID-19 treatment trial, which shows substantial variation in treatment effects across data-driven subgroups.