MPMichael L. Pennell

Michael L. Pennell

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Michael L. Pennell has a total of 142 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Bayesian variable selection for joint models of heterogeneous longitudinal variables and a binary outcome

    by Lingpeng Shan, Michelle J Naughton, Electra D Paskett, Michael L Pennell

    Biomedical studies often collect mixed-type longitudinal data (e.g., clinical, lifestyle) to identify predictors of a binary health outcome. A common analytical challenge is to link these irregularly collected variables to the outcome, particularly when the number of potential predictors is large. While joint models (JM) can handle this complex data structure, a critical gap exists, as they lack a formal framework for variable selection, limiting their utility for identifying relevant predictors. This article fills that methodological gap by introducing two novel, structured Bayesian variable selection strategies. Our first approach (JM1) selects predictors for the binary outcome, while our more advanced approach (JM2) performs simultaneous two-level selection for both the outcome and the longitudinal trajectories. Our framework leverages shrinkage priors to handle high-dimensional predictors and interactions, preventing overfitting. To guide final variable selection, we extend false discovery rate-based rules to our complex, multi-part joint model, which also accommodates grouped categorical predictors. We apply our method to Women’s Health Initiative and Life and Longevity After Cancer study data, identifying factors for post-treatment insomnia in breast cancer survivors. Our model identified a key predictor missed by conventional methods. This integrated approach provides a robust, interpretable, and computationally efficient framework, offering substantial gains in statistical power.

  • Open Access

    Effect of graphic warning label on acute changes in waterpipe tobacco smoking behavior, biomarkers of exposure and harm, and subjective effects in a randomized trial

    by Marielle C. Brinkman, Toral Mehta, Michael L. Pennell, David Angeles, Soliana Kahassai, Chieh-Ming Wu, Hayley Curran, Brittney Keller-Hamilton, Elizabeth G. Klein, Megan E. Roberts, et al.

  • Semi-parametric testing for ordinal treatment effects in time-to-event data via dynamic Dirichlet process mixtures of the inverse-Gaussian distribution

    by Jonathan A Race, Amy S Ruppert, Yvonne Efebera, Michael L Pennell

    Time-to-event data often violate the proportional hazards assumption under which the log-rank test is optimal. Such violations are especially common in the sphere of biological and medical data where heterogeneity due to unmeasured covariates or time varying effects are common. A variety of parametric survival models have been proposed in the literature which make more appropriate assumptions on the hazard function, at least for certain applications. One such model is derived from the first hitting time paradigm which assumes that a subject’s event time is determined by a latent stochastic process reaching a threshold value. Several random effects specifications of the first hitting time model have also been proposed which allow for better modeling of data with unmeasured covariates. We propose a Bayesian model which loosens assumptions on the mixing distribution inherent in the random effects first hitting time models currently in use and we do so in a manner which is ideally suited for testing for effects of ordinal treatment variables. We demonstrate via a simulation study that the proposed model has better power than log-rank based methods in detecting ordinal treatment effects in the presence of nonproportional hazards. Additionally, we show that the proposed model is almost as powerful as log-rank based methods when the proportional hazards assumption holds. We also apply the proposed methodology to two biomedical data sets: a toxicity study in rodents and an observational study of cancer patients.

  • Open Access

    A hierarchical constrained density regression model for predicting cluster‐level dose‐response

    by Michael L. Pennell, Matthew W. Wheeler, Scott S. Auerbach

    With the advent of new alternative methods for rapid toxicity screening of chemicals comes the need for new statistical methodologies which appropriately synthesize the large amount of data collected. For example, transcriptomic assays can be used to assess the impact of a chemical on thousands of genes, but current approaches to analyzing the data treat each gene separately and do not allow sharing of information among genes within pathways. Furthermore, the methods employed are fully parametric and do not account for changes in distribution shape that may occur at high exposure levels. To address the limitations of these methods, we propose Constrained Logistic Density Regression (COLDER) to model expression data from different genes simultaneously. Under COLDER, the dose‐response function for each gene is assigned a prior via a discrete logistic stick‐breaking process (LSBP) whose weights depend on gene‐level characteristics (e.g., pathway membership) and atoms consist of different dose‐response functions subject to a shape constraint that ensures biological plausibility. The posterior distribution for the benchmark dose among genes within the same pathways can be estimated directly from the model, which is another advantage over current methods. The ability of COLDER to predict gene‐level dose‐response is evaluated in a simulation study and the method is illustrated with data from a National Toxicology Program study of Aflatoxin B1.


  • Open Access

    Intolerance of Uncertainty and Cognition in Breast Cancer Survivors: The Mediating Role of Anxiety

    by Yesol Yang, Stephanie M. Gorka, Michael L. Pennell, Kellie Weinhold, Tonya Orchard

    Cancer-related cognitive impairment (CRCI) is one of the most prevalent symptoms that breast cancer survivors experience. While cancer treatments are established contributors to CRCI, inter-individual differences in CRCI are not well understood. Individual differences in sensitivity to uncertainty are potential contributors to CRCI; however, no prior studies have attempted to examine this link in the context of breast cancer. To address the gap, we used preliminary findings from an ongoing cross-sectional study. A total of 38 women with stage I–III breast cancer (1–4 years post-treatment) were included in this study. Intolerance of uncertainty (IU) was assessed using the Intolerance of Uncertainty Scale. Self-reported cognitive function was assessed with the Neuro-QoL questionnaire. Anxiety was assessed using the Patient-Reported Outcomes Measurement System Bank. From this study, we found that anxiety mediates the association between IU and cognitive function of survivors. In other words, among post-menopausal breast cancer survivors, those with higher IU showed higher anxiety and consequently had lower cognitive function. This finding suggests that assessing IU may help predict the risk of CRCI. This study expands the current knowledge that addresses the importance of IU as a factor associated with cognitive health.


  • Accelerating Colorectal Cancer Screening and Follow-up through Implementation Science (ACCSIS) in Appalachia: protocol for a group randomized, delayed intervention trial

    by Electra D Paskett, Aaron J Kruse-Diehr, Jill M Oliveri, Robin C Vanderpool, Darrell M Gray, Michael L Pennell, Bin Huang, Gregory S Young, Darla Fickle, Mark Cromo, et al.

    Appalachian regions of Kentucky and Ohio are hotspots for colorectal cancer (CRC) mortality in the USA. Screening reduces CRC incidence and mortality; however, screening uptake is needed, especially in these underserved geographic areas. Implementation science offers strategies to address this challenge. The aim of the current study was to conduct multi-site, transdisciplinary research to evaluate and improve CRC screening processes using implementation science strategies. The study consists of two phases (Planning and Implementation). In the Planning Phase, a multilevel assessment of 12 health centers (HC) (one HC from each of the 12 Appalachian counties) was conducted by interviewing key informants, creating community profiles, identifying HC and community champions, and performing HC data inventories. Two designated pilot HCs chose CRC evidence-based interventions to adapt and implement at each level (i.e., patient, provider, HC, and community) with evaluation relative to two matched control HCs. During the Implementation Phase, study staff will repeat the rollout process in HC and community settings in a randomized, staggered fashion in the remaining eight counties/HCs. Evaluation will include analyses of electronic health record data and provider and county surveys. Rural HCs have been reluctant to participate in research because of concerns about capacity; however, this project should demonstrate that research does not need to be burdensome and can adapt to local needs and HC abilities. If effective, this approach could be disseminated to HC and community partners throughout Appalachia to encourage the uptake of effective interventions to reduce the burden of CRC.


  • Bayesian semiparametric joint modeling of a count outcome and inconveniently timed longitudinal predictors

    by Woobeen Lim, Michael L Pennell, Michelle J Naughton, Electra D Paskett

    The Women's Health Initiative (WHI) Life and Longevity After Cancer (LILAC) study is an excellent resource for studying the quality of life following breast cancer treatment. At study entry, women were asked about new symptoms that appeared following their initial cancer treatment. In this article, we were interested in using regression modeling to estimate associations of clinical and lifestyle factors at cancer diagnosis (independent variables) with the number of new symptoms (dependent variable). Although clinical and lifestyle data were collected longitudinally, few measurements were obtained at diagnosis or at a consistent timepoint prior to diagnosis, which complicates the analysis. Furthermore, parametric count models, such as the Poisson and negative binomial, do not fit the symptom data well. Thus, motivated by the issues encountered in LILAC, we propose two Bayesian joint models for longitudinal data and a count outcome. Our two models differ according to the assumption on the outcome distribution: one uses a negative binomial (NB) distribution and the other a nonparametric rounded mixture of Gaussians (RMG). The mean of each count distribution is dependent on imputed values of continuous, binary, and ordinal variables at a time point of interest (e.g. diagnosis). To facilitate imputation, longitudinal variables are modeled jointly using a linear mixed model for a latent underlying normal random variable, and a Dirichlet process prior is assigned to the random subject-specific effects to relax distribution assumptions. In simulation studies, the RMG joint model exhibited superior power and predictive accuracy over the NB model when the data were not NB. The RMG joint model also outperformed an RMG model containing predictors imputed using the last value carried forward, which generated estimates that were biased toward the null. We used our models to examine the relationship between sleep health at diagnosis and the number of new symptoms following breast cancer treatment in LILAC.


  • Open Access

    Effectiveness of a Diabetes-Focused Electronic Discharge Order Set and Postdischarge Nursing Support Among Poorly Controlled Hospitalized Patients: Randomized Controlled Trial

    by Audrey White, David Bradley, Elizabeth Buschur, Cara Harris, Jacob LaFleur, Michael Pennell, Adam Soliman, Kathleen Wyne, Kathleen Dungan

    Background: Although the use of electronic order sets has become standard practice for inpatient diabetes management, there is limited decision support at discharge. Objective: In this study, we assessed whether an electronic discharge order set (DOS) plus nurse follow-up calls improve discharge orders and postdischarge outcomes among hospitalized patients with type 2 diabetes mellitus. Methods: This was a randomized, open-label, single center study that compared an electronic DOS and nurse phone calls to enhanced standard care (ESC) in hospitalized insulin-requiring patients with type 2 diabetes mellitus. The primary outcome was change in glycated hemoglobin (HbA1c) level at 24 weeks after discharge. The secondary outcomes included the completeness and accuracy of discharge prescriptions related to diabetes. Results: This study was stopped early because of feasibility concerns related to the long-term follow-up. However, 158 participants were enrolled (DOS: n=82; ESC: n=76), of whom 155 had discharge data. The DOS group had a greater frequency of prescriptions for bolus insulin (78% vs 44%; P=.01), needles or syringes (95% vs 63%; P=.03), and glucometers (86% vs 36%; P<.001). The clarity of the orders was similar. HbA1c data were available for 54 participants in each arm at 12 weeks and for 44 and 45 participants in the DOS and ESC arms, respectively, at 24 weeks. The unadjusted difference in change in HbA1c level (DOS – ESC) was −0.6% (SD 0.4%; P=.18) at 12 weeks and −1.1% (SD 0.4%; P=.01) at 24 weeks. The adjusted difference in change in HbA1c level was −0.5% (SD 0.4%; P=.20) at 12 weeks and −0.7% (SD 0.4%; P=.09) at 24 weeks. The achievement of the individualized HbA1c target was greater in the DOS group at 12 weeks but not at 24 weeks. Conclusions: An intervention that included a DOS plus a postdischarge nurse phone call resulted in more complete discharge prescriptions. The assessment of postdischarge outcomes was limited, owing to the loss of the long-term follow-up, but it suggested a possible benefit in glucose control. Trial Registration: ClinicalTrials.gov NCT03455985; https://clinicaltrials.gov/ct2/show/NCT03455985


  • Correlates of physical activity among older breast cancer survivors: Findings from the Women's Health Initiative LILAC study

    by Jessica L. Krok-Schoen, Michael L. Pennell, Nazmus Saquib, Michelle Naughton, Xiaochen Zhang, Aladdin H. Shadyab, Candyce H. Kroenke, Jennifer W. Bea, Lindsay L. Peterson, Tracy Crane, et al.


  • Visual attention to blu’s parody warnings and the FDA’s warning on e-cigarette advertisements

    by Brittney Keller-Hamilton, Makala Fioritto, Elizabeth G. Klein, Marielle C. Brinkman, Michael L. Pennell, Paul Nini, Joanne G. Patterson, Amy K. Ferketich