Nearly 4% of the population is diagnosed with a bipolar spectrum disorder (BD) [1], a condition that results in significant functional impairment [2] and increased mortality risk [3]. Despite its public health impact, etiological and maintenance factors of BD remain unclear, contributing to ongoing challenges in developing effective interventions. Increasingly, research indicates that physical activity may impact BD symptoms [4]. Understanding the interplay between physical activity and mood in individuals with BDs may reveal patterns that could inform treatment interventions.
It is well established that activity levels impact both physical and mental health. Reducing sedentary behavior and increasing physical activity helps prevent cardiovascular disease, diabetes, and metabolic syndromes [5], all of which are comorbidities commonly reported in BDs [6]. Given individuals with BDs may be prone to lower activity levels [7, 8], it is especially important to study the relationships between activity and physical health in this population.
Research also finds an inverse relationship between daily activity level and mental health symptoms like depression and anxiety [9,10,11,12]. These associations are also present in BD, where cross-sectional and longitudinal evidence suggests that increased activity levels mitigate symptoms of bipolar depression [6, 13, 14]. Recent work in this area suggests that improvements in mood derived from physical activity may occur as soon as next-day. For example, Merikangas and colleagues (2019) found that, over two weeks, increased motor activity was associated with better mood at the next time period, particularly for individuals with BD I [15]. Similarly, increased vigorous activity was associated with reduced next-day depressive symptoms in 28 individuals with BD over 20 days [16]. Together, these studies show that physical activity may confer mood benefits at the day level in BD, while highlighting the need to replicate these findings in larger, longer-term studies.
Additionally, growing evidence suggests individuals with BDs may differ from non-psychiatric populations in the function of their autonomic nervous system (ANS) [17, 18], which regulates involuntary physiological processes, including heart rate [19]. While much research relies on heart rate variability (HRV) as a proxy of ANS function, another easily accessible index of ANS activity is resting heart rate (RHR). In both psychiatric and non-psychiatric populations, higher RHRs have been linked with all-cause mortality [20, 21] and higher risk of suicide, an outcome commonly linked with severe depressive symptoms [22,23,24]. Furthermore, there is evidence that for some individuals, particularly those with psychiatric symptoms, RHR associates with mood state [25]. Despite these associations, research investigating RHR in BD populations remains limited, particularly regarding how RHR influences more proximal outcomes (e.g., daily mood), despite the elevated suicide and depression risk in this group.
Although long-term associations between activity and mood in BDs are well-established, day-to-day dynamics remain unclear. Most research relies on short-term (e.g., 2-week) or cross-sectional studies with small samples, limiting analysis of temporal fluctuations typical in BDs. Since mood instability is central to BDs [26, 27], examining daily predictors and outcomes of mood shifts in larger longitudinal studies could offer valuable insights for treatment.
Digital phenotyping offers a promising approach to address this gap, allowing for day-to-day measurement of mood, RHR, and objective physical activity [28, 29]. While still relatively new in BD research, digital phenotyping holds promise for disentangling complex relationships between variables as they unfold in daily life. For example, Lipschitz and colleagues (2024) applied machine learning models to Fitbit data and found that RHR and "very active minutes" (i.e., exercise minutes classified as higher intensity based on HR) were among the variables with highest relative importance in the algorithms predicting depressive and (hypo)manic episodes, respectively [30]. However, machine learning algorithms are designed for prediction rather than inferential analysis. Thus, while these findings highlight the importance of metrics such as very active minutes and RHR in relation to mood, they leave open questions about the temporal relationships with mood dynamics over time, like how changes in steps, exercise intensity, and RHR dynamically impact mood at the day-level. These types of questions are better addressed by inferential methods like Dynamic Structural Equation Modeling (DSEM) [31].
The purpose of the present investigation is to examine the day-to-day dynamics between physical activity (i.e., daily steps, very active minutes), RHR, and mood in BDs using a large longitudinal dataset of active and passive ambulatory assessment data. We leveraged data from the University of Michigan's PROviding Mental Health Precision Treatment (PROMPT) [32] study to test the following hypotheses: 1) Within-person increases in daily steps will be associated with better (i.e., increased) mood and vice versa; 2) Within-person increases in minutes classified as "very active" (i.e., higher intensity based on metabolic equivalent of task or MET) will be associated with better mood and vice versa; 3) Within-person decreases in RHR would be associated with better mood.