Precision Examination of Real-World Stress and Behavior Using Deep Digital Phenotyping

Abstract

Life stress has been consistently linked to poor mental health outcomes, but we know comparatively less about the more proximate impacts of life stress on daily affect and behavior. The current dissertation leverages recent advances in digital technology and analytical approaches to examine real-world life stress and its affective and behavioral correlates in a cohort of first-year college students naturally exposed to multifaceted stressors as they transitioned to college life. For a full academic year, participants provided continuous accelerometer data from an actigraphy wristband for the objective estimation of sleep duration and waketime activity. Additionally, participants used their phones to complete daily self-reports of perceived stress sources, affect, behavior, as well as the main events of their day, and completed periodic web-based assessments of global psychopathology symptoms. Similar data was collected in a timely 3-month follow-up study as the same participants underwent a new, unprecedented life transition: the beginning of the COVID-19 pandemic. Following a general review of relevant background literature in Chapter 1, novel insights into real-world life stress drawn from these datasets are offered across three papers presented in Chapters 2 through 4. Chapter 2 (Paper 1) examines fluctuations in stress, affect, and behavior over the course of the first year of college. Negative emotions, sleep patterns, and academic and social activity varied substantially over the course of the year as well as between individuals. Critically, while academics were a common source of stress for all students, a vulnerable subgroup reporting greater frequency of perceived social stress went on to report the highest global clinical symptoms at the end of the year, suggesting dissociable effects of different stress sources on mental health. Two years later, during the COVID-19 pandemic, the first-year subgroup with highest distress again stood out by frequent social stress and elevated clinical symptoms, suggesting that focus on sustained interpersonal stress, relative to academic stress, may be especially helpful to identify students at heightened risk for psychopathology. Chapter 3 (Paper 2) introduces a novel individual-level linear model (iLM) to estimate day-to-day associations between perceived stress levels and actigraphy-derived sleep duration within individuals and unbiased by the group. While stress and sleep duration were inversely related in most participants, the iLM revealed that the temporal direction of these associations is person-specific, identifying a variety of individual phenotypes that may account for the diverse group-level findings reported in prior literature: for some, elevated stress in the day was associated with shorter sleep later that night; for others, shorter sleep was associated with elevated stress the next day; others showed both directions of associations, and some showed no association. Paired with intensive longitudinal data, our individual-level model provides a precision framework for the estimation of stable real-world behavioral and psychological dynamics, and may support the personalized prioritization of intervention targets for health and wellbeing. Chapter 4 (Paper 3) investigates the proximate correlates of academic and social stressful events on daily affect and behavior within individuals. Experiences of stress were characterized from participants’ daily voice diaries narrating the main events of the day, using a combination of human labeling and large language models fine-tuned for sentiment analysis and topic modeling. Multilevel models assessing within-person associations found that days with academic stressful events were characterized by shorter sleep duration and decreased physical activity, as well as by reduced social interaction and increased time spent on schoolwork. Meanwhile, days with social stressful events were not systematically associated to changes in behavior, but they stood out by heightened negative affect, above and beyond the effect of academic events. Our results suggest that academic and social dimensions of life stress may have distinct signatures on daily affect and behavior, which in turn may go on to shape individuals’ long-term wellbeing. Chapter 5 discusses the main contributions and limitations of this body of work. In all, the current dissertation provides valuable insights into the effects of life stress on real-world emotions, behavior, and overall wellbeing at multiple timescales. Our approach offers a foundation for the characterization of life dynamics at both the individual and the group levels, with potential implications for the development of precision health and wellbeing interventions.