<div>Dynamic treatment regimes (DTRs) are often considered for the medical care of chronic diseases and complex conditions. They consist of multistage treatment decisions, each based on the individual's health information and their treatment and response history. In this dissertation, we consider this setting with binary responses (i.e., either respond favorably or unfavorably to a treatment) and highlight one type of heterogeneity, specifically the existence of subgroups of patients who respond favorably to only a distinct subset of study treatments. </div><div>Currently, most works employ model-free approaches to find the optimal DTR. In contrast, we propose a model-based approach, which focuses more on describing heterogeneity in treatment responses. We first consider the scenario when baseline covariates are not included. A mixture of mixed logit models is proposed along with an EM alogorithm to estimate these subgroup proportions and the probabilities of a favorable response. We describe how an optimal dynamic treatment regime can be determined given the model information. We also discuss the necessary identifiability conditions (i.e., what sets of parameters are necessary for DTR determination). </div><div><div>Then, we extend the proposed model to incorporate baseline covariates. Specifically, we include certain baseline covariates in the logistic model for the probabilities of a favorable response and develop a multivariate Bernoulli model to incorporate the remaining covariates in the determination of subgroup proportions. Furthermore, time effects are considered in the model to allow for a potential overall decline in response effectiveness over time. </div><div>In each setting, simulation studies are performed to demonstrate the effectiveness of the proposed method in both parameter and DTR estimation. We also compare our approach with another competing method, Q-learning, and provide the scenarios when our mixture model outperforms Q-learning in terms of finding the optimal DTR.</div></div>

Most studies examining predictors of treatment outcomes among problem drinkers have used a traditional statistical approach that examines group outcomes (e.g. analysis of variance, multiple regression analysis). Contrary to traditional methods, a person-centered approach identifies commonalities among clusters of individuals and provides the opportunity to examine the relationship between multiple individual differences and outcomes in a longitudinal manner. Specifically, the person-centered approach makes it possible to cluster individuals into subgroups based on their change patterns, and to examine the relationship between those subgroups and other variables of interest (e.g., drinking problem severity). This approach allows the inclusion of a relatively large number of variables to test complex hypotheses. The present study is a secondary data analysis of early (first three-month) Timeline Followback (TLFB) post-treatment drinking data from 200 problem drinkers who completed a short outpatient intervention. Using a growth mixture modeling (GMM) analysis, the goal was to identify different outcome drinking trajectories and examine the relationship between problem severity levels, treatment modality (i.e. individual versus group format), and goal choice (i.e. low-risk drinking versus abstinence) to those trajectories. Results demonstrated the existence of different outcome subgroups among problem drinkers. In addition, problem severity level was associated with outcomes and class membership. Observed significant differences in the relationships between predictor variables and specific outcome subgroups, and evidence of different drinking fluctuation patterns in the outcomes suggest that using a person-centered approach adds value beyond traditional statistical outcome analyses. The person-centered approach can facilitate the identification of relevant variables for patient-treatment matching hypotheses for problem drinkers.

No description available.

The current body of individual risk behavior research has been mainly driven by two streams of literature. Stable risk researchers propose individuals tend to display similar risk behaviors across time and situations due to individuals' underlying propensity to either engage in risk averse or risk seeking behaviors. Changeable risk researchers have sought out to examine variability in risk behaviors due to factors such as personality and contextual characteristics. However, might it be the case that there are subgroups of individuals who may be more prone to display static risk behaviors and other subgroup whose risk behaviors are more amendable? To better address this question, the current study integrates literature on risk behavior, psychological safety, and personality to investigate the existence of these potential risk behavior classes. Furthermore, supervisor rated performance is captured in order to better understand the potential organizational value of these various risk behavioral classes. Latent class growth modeling revealed five different risk classes: stable risk seeking, stable risk averse, highly adaptive, moderately adaptive, and mixed risk behaviors. Individual personality traits were shown to be significant predictors of class membership, although the pattern of results was somewhat in contrast to predicted relationships. Furthermore, in support of proposed hypothesis, individuals in the highly adaptive risk behavior class received the highest ratings of supervisor rated performance. I discuss results in terms of risk research and organizational practice, and specifically advocate for the increased examination of risk behavior within organizational research.

No description available.

No description available.

No description available.

No description available.

No description available.

Heterogeneity in the profile of HIV-associated neuropsychological disorder (HAND) may obscure understanding of its etiology and prognosis. Despite longstanding acknowledgement of this heterogeneity, HAND diagnostic approaches such as the Frascati criteria characterize neuropsychological function based on the level of impairment, without regard to the pattern of strengths and weaknesses. Attention to these patterns may enhance etiologic and prognostic specificity. We used latent class analysis (LCA) to identify relatively homogeneous subtypes of neurocognitive function in adults with well-treated HIV infection. We compared the diagnostic agreement of latent classes and Frascati categories, as well as their associations with demographics, HIV markers and antiretroviral factors, comorbid medical and psychiatric conditions, and everyday functioning. LCA identified four classes, whose cognitive profiles are depicted in Figure 1: cognitively intact, mild-to-moderate motor/speed impairment, mild-to-moderate memory/visuoconstruction impairment, and moderate mixed impairment. Latent classes and Frascati categories demonstrated good agreement in the overall classification of impaired cognition but more disagreement regarding subtypes of impairment. Both latent classes and Frascati categories demonstrated unique associations with etiologic factors and significant associations with functional outcomes. However, only latent classes, not Frascati categories, were associated with HIV variables. Additionally, functional difficulties were significantly elevated in the motor impairment class but not the memory impairment class despite similar levels of cognitive impairment in the two groups. Findings support the utility of a diagnostic approach that accounts for both the level and pattern of neurocognitive impairment. Future research should examine the neuropathological mechanisms, longitudinal trajectories, and treatments of empirically identified HAND subtypes. / Psychology