Despite advances in therapies, heart failure (HF) remains a chronic, highly prevalent condition and a leading cause of morbidity and mortality in the US.1 Heart failure has long been described as a dynamic condition with fluctuating symptomatology that can vary over short periods of time (e.g., day-to-day), but empirical evidence to support this claim is surprisingly sparse. Physical activity is a safe and effective intervention for patients with HF and has been shown to reduce HF symptoms and improve quality of life.2–5 Despite these benefits, few patients with HF are sufficiently active, and HF patients engage in lower levels of physical activity then age-matched and other cardiovascular disease populations.6–8 Psychosocial factors (e.g., mood, self-efficacy, social support) may influence physical activity levels amongst HF patients9–13 though this relationship is poorly understood and the extent to which these factors vary – particularly over shorter periods of time – is unknown.This observational study was designed to explore both stable psychosocial factors and within day temporal changes in symptoms and situational psychosocial factors in HF patients, and their association with physical activity levels – a topic with limited evidence to date. To accomplish this, we employed smartphone-delivered ecological momentary assessments (EMAs), an intensive data-capture methodology designed to assess patients in naturalistic environments. The theoretical basis for this study was self-determination theory (SDT), a successful approach for understanding and modifying human behavior that centers on the importance of values and goals in behavioral self-regulation and posits that three basic psychological needs (competence, autonomy, relatedness) contribute to an individual’s adaptive functioning and motivational quality.14,15 We chose SDT as the basis for this study as prior literature has demonstrated its association with physical activity and exercise, primarily amongst healthy adults,16 and therapeutic approaches that build upon SDT have been shown to increase physical activity and improve self-care amongst patients with HF.17–19 We hypothesized that 1) stable psychological traits of a person (e.g., motivation) would be linked to daily physical activity; 2) there would be substantial variability in HF symptoms and situational psychosocial states (e.g., vitality) over short periods of time; and 3) variability in HF symptoms and psychosocial states would be linked to immediate changes in physical activity. Our findings have important implications for designing interventions to increase physical activity levels and reduce symptoms amongst patients with HF.
Methods
The MObile health interVention to INcrease activity in Heart Failure (MOVIN-HF) study was a prospective, observational study designed to understand factors influencing physical activity levels amongst HF patients. Participants answered baseline and 4-week surveys and completed twice-daily EMAs for 28 days. Participants were provided with a smartwatch (Fitbit Versa 2) to objectively measure physical activity levels (step count). The study was approved by the University of Michigan Health IRB (HUM00230476). The deidentified data that support the findings of this study are openly available through Github (https://github.com/WIRED-L-um/MOVIN-HF).
The MOVIN-HF study launched in February 2024 at an academic medical center in the state of Michigan with enrollment completed in June 2024. Patients were eligible for the study if they were 18 or older at the time of enrollment, had a diagnosis of HF with New York Heart Association (NYHA) class 2 or 3 symptoms, owned a compatible smartphone, and were safe for home physical activity, as determined by their treating clinician. Exclusion criteria included a HF hospitalization within the prior 6 weeks or the presence of potentially high-risk cardiovascular conditions which would make physical activity potentially unsafe. Full inclusion and exclusion criteria are listed in the supplemental methods.
Participants underwent remote recruitment and consent processes with the consent form signed within the mobile study application MyDataHelps (CareEvolution, LLC) and assigned baseline surveys within the mobile application. Participants were subsequently mailed a smartwatch and underwent remote enrollment appointments, during which they were assisted with pairing their smartwatches and phones as needed. Visits were conducted by telephone with the option to use video teleconferencing as needed. Participants were asked to wear their smartwatch daily, with an emphasis on the importance of wearing it during bouts of physical activity. Participants then completed a training EMA with the clinical research coordinator to familiarize themselves with accessing and completing EMA prompts.
For the next 28 days, participants received 2 EMAs/day in the mobile study application, with alerts by text message and push notification when EMAs became available. Participants had two hours to complete each EMA and received a reminder by push notification 1 hour after the EMA was delivered if not yet complete. EMAs were delivered over an 11-hour period at different times each day (between 8am and 7pm), once in the morning and once in the afternoon, and separated by at least 4 hours. Participants were contacted for non-adherence with completing EMAs or smartwatch wear. After 28 days, participants were reassigned surveys. Participants were considered to have completed the study at 6-weeks or completion of all surveys, whichever came first.
Demographic and clinical information was collected using the mobile study application following informed consent and included participants self-reported sex, race/ethnicity, and HF symptoms as measured by the 12-item Kansas City Cardiomyopathy Questionnaire (KCCQ-12).20 Heart failure phenotype (e.g., HF with preserved ejection fraction, HF with reduced ejection fraction), NYHA functional classification (based on the most recent clinic note), and ejection fraction were extracted from the electronic health record (EHR).
Physical activity was measured using a Fitbit Versa 2 that was provided to the participant. Heart rate was collected from the Fitbit to determine wear time. Valid days included days in which participants wore their smartwatches for at least 8 hours during the monitoring period.
Surveys and EMAs were completed in the mobile study application and designed to assess constructs predicted to influence physical activity levels in the published literature and from SDT. Briefly, SDT centers on the importance of values and goals in behavioral self-regulation and ascribes three innate human needs: competence, autonomy, and relatedness.14,15 Self-determination theory postulates that positive growth, in the case of health, is more likely to occur when these needs are met.14 Furthermore, SDT differentiates between low and high quality motivation, referred to as controlled and autonomous regulation. Autonomous motivation occurs when a behavior change is driven by internal personal meaning. Conversely, controlled motivation occurs when a person is coaxed by external factors (e.g., spouse) or internalized guilt or shame. We hypothesize that HF patients will be more likely to change their behaviors when their psychological needs are met and motivation is more autonomous than controlled (Figure S1).
Surveys were taken from the published literature and assessed stable psychological factors (Table S1). Preference was given to surveys with strong psychometric properties and which were specific to physical activity and exercise, when available. Surveys included measures of depression (8-item Patient Health Questionnaire [PHQ-8]),21,22 autonomy (Behavioral Regulation in Exercise Questionnaire version 2 [BREQ-2]),23 social support for physical activity,24–26 and outcome expectations for exercise (Multidimensional outcome expectations for exercise scale [MOEES])27. A 15-item physical activity barriers scale24–26 was administered once at baseline. Surveys were scored according to published methods, when available. For the BREQ-2, external regulation and introjected regulation were combined to form a single subscale termed controlled motivation to generate a total of 4 subscales (amotivation, controlled motivation, identified regulation, intrinsic regulation).
The EMA was designed to assess temporary states predicted to be influenced by current contextual factors and consisted of 12 questions intended to be completed over a 3–5 minute period. Questions were adapted from the published literature, when available, and both question stems and response items modified to allow for short form delivery using a smartphone. Questions focused on (1) HF symptoms,28 (2) behavior-specific psychological determinants (i.e., fear of physical activity, competence for physical activity, vitality), and (3) general psychological determinants (i.e., positive and negative affect).29 Heart failure symptoms were quantified using an indexed scoring system in which participants received 1 point for moderate or greater symptoms. In a sensitivity analysis, HF symptoms were quantified using a linear scoring system ranging from 1–5 points per question based on symptom severity with lower scores indicative of greater HF symptoms. The 12-item EMA and a description of the scoring criteria are presented in the supplemental methods.
Baseline clinical characteristics are described as means with standard deviations (SD) for continuous symmetric variables and median with interquartile range for skewed continuous variables. Categorical variables are presented as counts and percentages. Patients with NYHA class 2–3 HF symptoms based on last clinic note were assumed to have NYHA class 3 symptoms given unbalanced data. Variation in EMA scores was calculated at 7 and 28 days with 50% or greater change considered significant.
Mixed effects models were used to assess the relationship between surveys and mean daily step count. Model covariates were selected a priori based on clinical expertise and included age and sex in addition to baseline survey scores. Separate mixed effects models were constructed for each baseline survey measure using a random intercept. Due to the skewed nature of the data, the primary outcome was natural log transformed daily step count on valid days, defined by at least 8 hours of watch wear time. We elected to use 8-hour wear time to define a valid day because we found that overall step counts were similar when wear time thresholds were set at 8, 10, and 12 hours (Table S2).
Mixed effects models were constructed to evaluate the association between each EMA construct and step count. Features within models were determined a priori and included age, sex, time in study, time of day (AM versus PM delivery of the EMA), and EMA scores. Models included a random intercept with an unstructured correlation matrix to account for intra-individual variability. The primary outcome was natural log-transformed step count 12 hours after an EMA and, secondarily, natural log-transformed step count 4 hours after an EMA. Log transformation was used due to the skewed nature of the data. Model estimates were subsequently exponentiated to give a percent change in step count. All analyses were conducted using SAS version 14.2 (SAS Institute Inc., Cary, NC, USA). For all analyses, the level of significance was set at p <0.05.
Results
Between February 15, 2024 and June 25, 2024, we recruited 30 HF patients to participate in the study (Figure 1). One participant experienced technical difficulties when pairing the smartwatch and therefore provided only EMA data. Mean age of participants was 59.7 (SD 13.6) years, 53% were women, and most were White (73%) (Table 1). New York Heart Association functional classification was 2 for 63% of participants, 2–3 for 10% of participants, and 3 for 27% of participants. Mean KCCQ overall summary score was 63.3 (SD 22.5) and symptom frequency score was 65.9 (SD 23.1). Most participants had HF with improved ejection fraction (40%) or HF with reduced ejection fraction (27%).
Over the duration of the study, participants provided smartwatch data for a median of 27 days (IQR 26, 27), nearly of all of which were valid days (median 27 [IQR 26, 27]) and wore their smartwatches for a median of 20.5 (IQR 14, 24) hours/day. Median daily step count was 3702 (IQR 2011, 5949) steps, which was similar when accounting only for valid days (median 3837 [IQR 2157, 6216] steps).
Participants had high baseline levels of identified and intrinsic motivation (i.e., feeling internally motivated and not coerced) for exercise as assessed by the BREQ-2, mild depression symptoms as assessed by the PHQ-8 (7 participants [23.3%] with PHQ-8 scores >10), and few significant barriers to physical activity (Table 2). Over the 4-week study duration, survey scores did not significantly change.
Mixed effects models explored the association between baseline survey scores and mean daily step count (Table 3; Tables S3–S12). Amotivation and controlled motivation, external forms of motivation, were associated with significantly lower daily step counts, with a 36% and 46% relative decrease in step count per 1 point increase, respectively (Amotivation: 95% CI −57% – −5%, p=0.03; Controlled motivation: 95% CI −62% – −23%, p<0.001). In contrast, intrinsic regulation (i.e., an internally driven form of motivation) was associated with significantly higher daily steps counts with a 39% increase in step count per 1 point increase (95% CI 3% – 87%, p=0.03). Higher physical outcome expectations as assessed by the MOEES were also significantly associated with higher levels of physical activity (Physical outcome expectations: Estimate 12%, 95% CI 1% – 24%, p=0.03). Finally, each additional barrier to physical activity (rated “somewhat difficult” or “very difficult” on the physical activity barriers scale) was associated with a 25% decrease in mean daily step count (95% CI −33% – −17%, p<0.001). Depression and social support were not significantly associated with daily step count.
Participants initiated 84% of EMA prompts (1412 of a potential 1680 prompts; range 83.9–84.0% for individual EMA constructs). Over the study period, on average, most participants had moderate or greater of at least 1 HF symptom (mean 0.7 [SD 1.2]), high positive affect (mean 78.4 [SD 23.8]), moderate competence for physical activity (mean 4.0 [SD 2.0]), and moderate levels of vitality (mean 5.3 [SD 3.0]) (Table S13).
EMA constructs varied within and across days for most individuals (Figure S2) though remained stable on a population level over the course of the study (Table S13). For example, over the first 7 days of the study, 13 (43%) participants experienced a 50% or greater change in vitality and 12 (40%) a 50% or greater change in competence for physical activity (Figure S3). These results were even more pronounced over the 28-day study period, with 10 (33%) participants experiencing a 50% or greater change in HF symptoms, 18 (60%) a 50% or greater change in vitality, and 19 (63%) a 50% or greater change in competence for physical activity (Figure 2).
A series of mixed effects models explored the association between EMA constructs and step count 12 hours after an EMA. Heart failure symptoms, competence for physical activity, vitality, and positive affect were all significantly associated with 12-hour step count (Table 4; Table S14–S19). Specifically, each 1 point increased in HF symptoms (on a 5-point scale) was associated with a 14% decrease in 12-hour step count (95% CI −20% – −7%, p<0.0001). In contrast, each 1-point increase in competence for physical activity (on a 7-point scale) and vitality (on a 10-point scale) was associated with a 14% (95% CI 8% – 21%, p<0.0001) and 11% (95% CI 7% – 15%, p<0.0001) increase in 12-hour step count, respectively (Table 4). Results were similar in a sensitivity analysis which used a linear scoring system for HF symptoms (Table S20) and when evaluating the association between EMA constructs and step count 4-hours after an EMA (Table S21).
Discussion
In this observational study of symptomatic HF patients, we found 3 important results that linked psychosocial factors with physical activity levels. First, we found that stable psychosocial factors like motivational quality and extent of autonomous motivation for exercise were significantly associated with physical activity levels, with higher levels of intrinsic motivation (i.e., self-motivation) for exercise associated with higher mean daily step counts. Second, we found that temporary states which may be context-dependent such as vitality, competence, HF symptoms, and positive affect all varied within and across days for a given individual even over this relatively short study period. Finally, we found that these psychosocial states were significantly associated with immediate physical activity levels over the next 12 hours. Overall, these finding support that both stable and temporary, time-varying psychological factors are associated with physical activity. This suggests that future interventions to augment physical activity may require a combination of strategies to target both long-term change in stable psychological factors as well as more in-the-moment support for temporary states.
Despite overwhelming evidence to support the benefits of physical activity amongst patients with HF, physical activity levels remain grossly inadequate,6–8 and HF patients participate in less physical activity than age-matched and other cardiovascular disease populations.6,7 Efforts to increase physical activity amongst HF patients have relied predominantly on resource-intensive programs, such as cardiac rehabilitation. Such programs, however, are underutilized with less then 5% of eligible HF patients enrolling in cardiac rehabilitation, are not approved by most insurance companies for patients with an ejection fraction greater than 35%, and have insufficient capacity to accommodate all patients.30,31 Mobile health (mHealth) interventions represent a promising, scalable approach for increasing physical activity levels though the effects from these interventions often decrement over time.32–34 Dynamically tailored interventions have the potential to be more effective than static interventions by accounting for users changing contexts and circumstances (e.g., just-in-time adaptive interventions). For example, in patients with alcohol use disorders leaving a residential treatment facility, a smartphone-intervention grounded in SDT and tailored on participants’ location reduced the number of risky drinking days at 4 and 12 months.35 Herein, we identified that more stable traits, such as source of motivation (i.e., internal versus external forms of motivation), were associated with physical activity as were time-varying states such as competence, vitality, HF symptoms, and positive affect, with the latter traits varying within and across days for a given individual. Taken together, these findings suggest that a combination of approaches may be necessary to change physical activity behavior. An effective approach may consist of a more resource-intensive strategy at lower doses (e.g., counseling, cardiac rehabilitation) paired with a dynamically tailored mHealth intervention to address time-varying states influenced by contextual factors (e.g., just-in-time adaptive intervention).
One of the major findings of our study was the association between key constructs from SDT and physical activity levels amongst patients with symptomatic HF. Self-determination theory is a key foundational model of behavior change that centers on the importance of values and goals in behavioral self-regulation and ascribes three innate human needs: competence, feeling related to others, and feeling internally motivated (i.e., autonomy).9–13 SDT postulates that these needs vary within individuals over time, contexts, and social interactions.14 Furthermore, SDT asserts that sustained behavior change can be achieved only when an extrinsic behavior (e.g., stopping smoking, exercising) is brought into congruence with ones intrinsic goals and values so that the behavior assumes meaning. In contrast, when the behavior is incentivized by external rewards (e.g., financial incentives), the targeted behavior is only maintained so long as incentives are in place. Prior studies have demonstrated an association between key constructs from SDT and physical activity and exercise. In a meta-analysis of 66 studies, predominantly of healthy adults, motivational quality was strongly associated with moderate-vigorous exercise and, amongst the three psychological needs, perceived competence was most strongly associated with physical activity.16 Within HF patients specifically, however, the evidence has been more limited. In a cross-sectional study of 163 asymptomatic or mildly symptomatic HF patients, physical limitation and self-efficacy were both associated with physical activity levels though depression mediated this effect, at least in part. Additionally both intrinsic and extrinsic motivation for physical activity had a significant effect on physical activity levels independent of depression.36 Our results align with many of the findings from these studies but now extend them to a symptomatic HF population over time, demonstrating that motivational quality and perceived competence, but not depression, are strongly associated with physical activity levels.
Our study has multiple strengths. First, we enrolled patients with significant HF symptoms, as evidenced by their higher NYHA scores prior to enrollment, lower step count relative to other mHealth studies enrolling HF patients,37 and lower KCCQ scores on enrollment, a validated and sensitive measure of patient health status which has been shown to strongly correlate with clinical events over time.20,38 Prior studies have demonstrated the safety and benefits of physical activity even in frail, symptomatic HF patients2 though these individuals are often excluded from mHealth studies despite their potential to benefit from these more accessible interventions. Second, we assessed a range of factors with the potential to influence physical activity levels in participants’ naturalistic environments longitudinally through EMAs. While prior studies have evaluated factors influencing physical activity amongst HF patients, most have done so cross-sectionally and no studies have provided similar deep temporal resolution into changes in these factors over time.9–13 Finally, we had outstanding adherence to the protocol with participants wearing their smartwatches for over 20 hours/day and initiating 84% of EMAs.
Our study does have important limitations. First, participants were recruited from a single academic medical center, were predominantly White, and were required to be English speaking. Whether the observed relationships between psychological traits and states and physical activity apply to other HF populations or to non-English speaking populations is unknown. Second, we enrolled 30 participants thus limiting overall power as well as potential subgroup comparisons, for example by HF phenotype. Prior research suggests, however, that HF patients may share barriers to physical activity irrespective of their ejection fraction9,11 and that changes in physical activity are associated with changes in patient-reported quality-of-life (KCCQ scores), irrespective of HF phenotype.39 Third, the EMA assessed 6 constructs and, as such, some factors may have been inadequately captured or represented in the limited number of questions devoted to assessing that construct. Additionally, surveys were validated primarily in general populations. As these surveys largely target broader constructs, we have no reason to suspect these would be affected by an individual disease process like HF though that cannot be confirmed. Fourth, we only evaluated the association between psychosocial factors and step count but not other forms of physical activity. It is thus possible that participants engaged in forms of physical activity (e.g., swimming) which would not be reflected in their step count. Finally, as is the nature of any observational study, we demonstrated an association between HF symptoms and psychosocial factors, including foundational constructs from SDT, and physical activity levels. Whether this relationship is causal and whether intervening on these factors can improve activity levels remains an area of future research. Importantly, however, motivational interviewing, a therapeutic approach that builds upon SDT, has been shown to increase physical activity and functional capacity and to improve HF self-care.17–19
In summary, in this observational study of symptomatic HF patients, we identified significant temporal variation in psychosocial states, many from SDT. These, as well as other more stable psychological traits, such as motivation, were strongly associated with physical activity levels. Additional research is needed to understand the extent to which tailored interventions can target traits of patients as well as psychological states to improve physical activity levels.
