Real-Time Data Collection to Examine Relations between Physical Activity and Affect in Adults with Mental Illness

Abstract

Adults with serious mental illness (SMI) engage in limited physical activity, which contributes to significant health disparities. This study explored the use of both ecological momentary assessments (EMAs) and activity trackers with adults with SMI to examine the bidirectional relationship between activity and affect with multilevel modeling. Affective states were assessed up to 7 times per day using EMA across four days. Participants (n=20) were equipped with a waist-worn accelerometer to measure moderate-to-vigorous physical activity (MVPA). Participants had a mean EMA compliance rate of 88.3%, and over 90% of completed EMAs were matched with 30-minute windows of accelerometer wear. Participants who reported more positive affect than others had a higher probability in engaging in MVPA. Engaging in more MVPA than one’s usual was associated with more negative affect. This study begins to address the effect of momentary mood on physical activity in a population of adults that is typically difficult-to-reach.

Engaging in regular moderate-to-vigorous physical activity (MVPA) can greatly reduce the risk of certain health issues including cancer, diabetes, and heart disease (Physical Activity Guidelines Advisory Committee, 2008). Nonetheless, lack of MVPA has been declared a global public health problem (World Health Organization, 2009). In the general population, more than 80% of U.S. adults fail to meet national guidelines for physical activity (Office of Disease Prevention and Health Promotion, 2019). In order to explore differences in physical activity among individuals, there is increasing attention in research to the role of one’s affective state and how affect may be related to the propensity to engage in or avoid physical activity. One’s current feeling may trigger, and be triggered by, behavior (Leone et al., 2005; Updegraff et al., 2004). Furthermore, affective responses either before, during or after engaging in a specific behavior may influence one’s decision about whether to repeat the behavior (Kahneman et al., 1993). According to longstanding behavioral theories, individuals are generally more likely to engage in any behavior that derives pleasure (Bentham, 1962; Mellers, 2000). On the other hand, negative emotional states (e.g., stress) may result in avoidance of a behavior (Leone et al., 2005). In research on the relationship between affect and physical activity in the general population, this notion holds: if positive feelings are felt in relation to physical activity, individuals are more likely to engage in activity but negative affective states predict less engagement (Liao et al., 2015; Niermann et al., 2016).

Adults with serious mental illness (SMI), such as individuals diagnosed with schizophrenia, bipolar disorder, or major depressive disorder, tend to be even less active than the general population (Daumit et al., 2005; Richardson et al., 2005). There are several reasons to explain why adults with SMI are less likely to meet recommended guidelines of at least 150 minutes of MVPA per week than healthy adults (World Health Organization, 2009). Adults with SMI have reported difficulty engaging in MVPA due to the presence of symptoms associated with their mental health issue (e.g., apathy, anhedonia). Furthermore, adults with SMI have identified psychiatric medication side effects (e.g., lack of energy) and other existing physical comorbidities as additional barriers (e.g., cardiovascular disease, arthritis) (Glover et al., 2013). Adults with SMI often report low self-efficacy for exercise and generally have limited knowledge of healthy behaviors that may prevent obesity (Daumit et al., 2005; Dickerson et al., 2006; Druss et al., 2011; Fagiolini & Goracci, 2009; Viron & Stern, 2010; Williams et al., 2010). In qualitative studies addressing barriers to activity, adults with SMI mentioned functional issues correlated to mental health including poverty and unemployment. Although walking was a favored exercise, participants felt their surrounding neighborhoods were too unsafe to support this activity (Glover et al., 2013). In conjunction, adults with SMI have high rates of cardiovascular disease (Druss et al., 2011), and obesity is nearly twice as prevalent in this population compared to healthy adults (Allison et al., 2009). Thus, it is increasingly important to determine what may impact a lack of activity in order to help prevent diseases linked to sedentary behavior.

In research that explores the relationship between affect and activity in the general population using real-time data collection strategies, such as ecological momentary assessments (EMA) and activity trackers (e.g., accelerometers) (Fisher, 2008; Haskell, 2012), findings point to the importance of within-person variability. Essentially, one’s feeling states (e.g., mood or affect) fluctuate over time and this intraindividual variability in affect or mood is a predictor of physical activity behavior. For instance, in one particular EMA study of adults, greater positive mood compared to one’s own average was positively related to more time spent in subsequent MVPA and an inverse relationship was found for negative mood (Niermann et al., 2016). These results suggest the importance of observing momentary states with methods that reduce biases and improve ecological validity. To date, research on the relationship between affect or mood and activity among adults with SMI has examined between person differences and relied on self-report methods based on long periods of recall that are prone to errors (Loney et al., 2011; Nicaise et al., 2011). Advantages of EMA sampling may be particularly suited for adults with SMI who may have executive functioning impairments associated with their diagnoses (Trivedi, 2006) and thus issues accurately recalling past behaviors. Although momentary data collection methods have been employed in studies with non-clinical samples (Dunton et al., 2012; Liao et al., 2015), to the best of our knowledge there have been no attempts to date to explore the bidirectional association between acute affective states and physical activity among adults with SMI. Furthermore, since adults with SMI self-report lack of energy as a barriers to activity, it is also important to view the relationship between intraindividual variation in feelings of energy and how this may relate to physical activity.

This pilot study examined patterns of activity and affect or physical feeling states (i.e., energy or tiredness) in a population of adults with SMI using both smartphone-based EMA and activity measurement. In particular, we aimed to examine the association between momentary affective or physical feeling states and engaging in any MVPA or the amount of time spent in MVPA in those with SMI. Although adults with SMI may have difficulties with emotional regulation and affect that may fluctuate more throughout the day than those without serious mental health issues (Aldao et al., 2010), we hypothesized a similar relationship between affect and activity that has been found in research with a general populations of adults (Dunton et al., 2015; Niermann et al., 2016). We hypothesized the following: 1) individuals with SMI who are generally more positive or energetic than their peers will participate in more MVPA and spend more time in MVPA than those who are less positive or energetic and inversely, those who are generally more negative or tired than their peers, will be less inclined to participate in activity and spend less time in MVPA (between-person associations); 2) greater than average within-person positive affect or feeling energetic will be subsequently related to both engaging in MVPA and more minutes of MVPA than one’s own average. In addition, greater than average within-person negative affect or feeling tired will be subsequently related to less likelihood of engaging in MVPA and minutes spent in MVPA (within-person associations); and 3) more minutes spent in MVPA will be subsequently associated to more positive affect and feelings of energy and less negative affect or feelings of tiredness (within-person associations).

Methods

Participants and Recruitment

Twenty-one participants were recruited from an integrated healthcare clinic in a large southwestern city in the United States that serves a predominantly Latinx community. Recruitment flyers were distributed by behavioral health providers to clients enrolled in a program that provides physical and behavioral health services to low-income individuals with co-morbid SMI and chronic health conditions. Participants were eligible to participate in this study if they were aged 21 or older and able to speak and respond to surveys in either English or Spanish. Eligible individuals indicated their interest to participate to their providers who assisted in scheduling an on-site data collection appointment with a research team member. This research was reviewed and approved by the Institutional Review Board of the corresponding author’s institution. Accelerometer data were available for all 21 participants; however, EMA data were not available for one participant due to technical problems with the smartphone application and thus this participant was dropped from analyses.

Procedure

Protocols for this pilot project were taken from previous studies utilizing EMA with populations without serious mental health conditions (Dunton et al., 2012, 2015), though others have shown smartphone-based survey methods to be feasible and acceptable with youth with mental illness (Morgan et al., 2017) and other types of vulnerable adults (Moore et al., 2017). In this study, mobile phone technology was utilized to assess momentary affective and physical feeling states for four consecutive days while an accelerometer was simultaneously worn to provide a measure of movement. Participants initially attended a baseline appointment to complete a survey, set up an EMA application on loaned smartphones, and receive a study-provided ambulatory monitor to measure activity. Baseline appointments were approximately 60 minutes in length. Participants were loaned a Motorolla XT1032 phone for the study period in order to maintain consistency in the type of smartphone that was utilized by subjects and to ensure both data availability and app functioning. All participants were provided with a waist-worn accelerometer (Actigraph wGT3x-BT) and instructed to wear this monitor at all times except when sleeping, showering, or swimming. Participants were also instructed to carry the phone with them at all times during the study period.

After the in-person meeting, monitoring occurred across four days between 8:00am and 10:00pm. EMA data were collected using a custom software application compatible for smartphones with an Android operating system (Google Inc.). Seven EMA surveys were prompted at random each day within seven pre-programmed windows (i.e., randomly within two-hour periods) in order to ensure adequate spacing of prompts across a day (28 total prompts). EMA surveys were prompted either with an audible signal or vibration. At each prompt, participants were instructed to stop their current activity and complete a short EMA survey. The process required 2–3 minutes. If a signal occurred during an incompatible activity (e.g. sleeping or bathing), participants were instructed to ignore it. Participants received up to three reminder signals at 3-minute intervals and after 10 minutes, the survey was no longer accessible. EMA items measured momentary affective states and were available in English or Spanish. EMA data from smartphones were locally encrypted and wirelessly uploaded after each entry and stored on a cloud server. After the 4-day collection period, participants returned equipment. Participants were compensated $20 for the baseline portion of this study and up to $60 for compliance on EMAs ($60 if responded to 80% of the 28 total prompts and $2 per completed survey if responded to less than 80%).

Measures

Baseline survey.

This short survey captured demographic characteristics such as participants’ age, sex, race/ethnicity, educational attainment, income, mental health diagnosis, and self-rated general health status (ranging on a 4-point scale from ‘poor’ to ‘very good’).

Momentary measures.

The EMA question sequence measured current affective and feeling states, other mental health symptoms (e.g., lack of focus, crying spells, or poor appetite), and type of physical activity in the past hour. Participants were only able to select one response option for each item. Each prompt contained a minimum of 19 items and as many as 25 items if participants endorsed certain items. Table 1 includes EMA items relevant to this inquiry; questions not included in the table covered activity behavior other than exercise (e.g., reading, hanging out), location (e.g., home, work), social context (e.g., with friends, coworker), type of food intake (e.g., fast food, fruits or vegetables) and use of tobacco or alcohol. The two items addressing positive affect were averaged to create a composite score (Cronbach’s α = 0.63), and four items addressing negative affect were also combined for a composite scale (Cronbach’s α = 0.89). All EMA items were pilot tested with older adults for comprehension and applicability (Dunton et al., 2012).

Table 1.

Selected Ecological Momentary Assessment [EMA] Items

Variable	Item	Response options
[Q1–2] Positive Affect	1: How HAPPY were you feeling just before the beep went off? 2: How CALM or RELAXED were you feeling just before the beep went off?	1. Not at all  2. A little  3. Moderately  4. Quite a bit  5. Extremely
[Q3–6] Negative Affect	3: How STRESSED were you feeling just before the beep went off? 4: How FRUSTRATED or ANGRY were you feeling just before the beep went off? 5: How TENSE or ANXIOUS were you feeling just before the beep went off? 6: How SAD or DEPRESSED were you feeling just before the beep went off?
[Q7] Physical Feeling – Fatigue	7. How TIRED or FATIGUED were you feeling just before the beep went off?
[Q8] Physical Feeling - Energy	8. How ENERGETIC or FULL OF PEP were you feeling just before the beep went off?
[Q9] Paranoia or hallucinations	Within the past hour, have you had the impression that someone was spying on you or plotting against you, felt that someone was reading your thoughts, or heard voices, had visions, or experienced things that others could not see or hear?	• Yes • No
[Q10] Crying spells, focus, or poor appetite	In the past hour, have you had crying spells, trouble keeping your mind on what you were doing, or poor appetite?
[Q11] Physical Activity	In the past hour, have you done any physical activity/sports/exercise?
[Q11.1] Physical Activity Type	[If YES is selected for Q11] What type of physical activity/ sports/ exercise?	• Running/ jogging • Walking • Weightlifting/ strength training • Using cardiovascular equipment [such as Stairmaster] • Bicycling • Other
[Q11.1a]	[If OTHER is selected for Q11.1] Enter the type of physical activity/ sports/ exercise	[TEXT ENTRY]

Physical activity (MVPA).

The Actigraph, Inc. GT3X accelerometer was worn on the right hip attached to an adjustable belt. A 30-second epoch was utilized to collect movement. MVPA was defined using thresholds described in other work (Troiano et al., 2008). The outcome variable was defined as the number of minutes that were above the MVPA threshold. Accelerometer data were linked to EMA survey responses using electronic time stamps. The number of MVPA minutes occurring within the 30-minute windows before and after each EMA survey response was recorded. If there were zero minutes of wear time in the 30-minute windows before or after an EMA survey prompt, the entries were excluded from analyses.

Data Analyses

To examine the temporal effects of affective and physical feeling states on MVPA and vice versa, we conducted a series of multilevel regression models using Stata 13. The multilevel models adjust the standard errors for clustering of observations within people (Raudenbush, 2002). For the main predictor of each model, we generated both between-subject (BS) and within-subject (WS) versions (i.e., partitioning the variances) (Hedeker et al., 2008, 2012; Neuhaus & Kalbfleisch, 1998). The BS version represents a participant’s average level relative to the group mean. The WS version represents the level at any given EMA prompt relative to his or her own average level across all EMA prompts. In each model we controlled for potential confounders determined a priori including the person-level variables of age and gender, and the prompt-level variables of day of week and time of day. We also controlled for affect at the prior prompt.

First, we tested whether affective and physical feeling states predicted subsequent MVPA. Total MVPA minutes within the 30-minute windows after the EMA prompts were not normally distributed and some observations had zero MVPA minutes. We used a two-piece model to address this problem (Baldwin et al., 2016). For the Piece 1 Model, we used multilevel logistic regression to predict the probability of engaging in some MVPA (non-zero MVPA minutes) versus no MVPA (i.e. zero MVPA minutes). The Piece 2 Model used multilevel linear regression to predict the number of non-zero MVPA minutes. Stata GSEM (Stata 13) was utilized to run the two-piece model [36]. Second, we used multilevel linear regression models to test whether MVPA predicted subsequent affective and physical feeling states using Stata MIXED (Stata 13). Negative affect was not normally distributed and we log-transformed before fitting the models. Affective states were lagged to the prior prompt within that day, therefore, if affective data was missing from the first prompt, we excluded that from the analyses. Predictors (i.e., MVPA and affect) were time-lagged to the prior prompt within each day, resulting in exclusion of the first prompt of each day from analyses.

Results

We enrolled a total of 21 participants in this study. Demographic characteristics are available in Table 2. Participants (N=21) were on average 49 years old and included a majority of individuals identifying as Hispanic/Latino (n=20) and female (76.2%). Approximately half of the sample did not complete high school (47.6%), one third of the sample experienced homelessness (33.3%), and almost half reported no source of monthly income (42.9%). All participants had a diagnosis of major depressive disorder, though 47% of the sample also had a diagnosis of generalized anxiety disorder, and 14% were dually diagnosed with bipolar disorder. Over 65% of the sample self-reported poor or fair general health status. For mental health state, participants reported having crying spells, poor focus or poor appetite in the past hour in 14.1% of the EMA prompts (n=43/304 responses). In addition, participants reported having paranoia or hallucinations in the past hour in 6.3% of the EMA prompts (n=19/304 responses). When asked about the activities participants engaged in the hour prior to the prompt, participants self-reported engaging in planned physical activity (though not necessarily an indicator of MVPA) in only 16.4% of EMA prompts (n=11/412 responses). If physical activity was self-reported during the past hour, walking was the most commonly reported activity (83.3% of physical activity prompts).

Table 2.

Participant Characteristics [N=21]

Variable	n [%] x̅ [SD]
Female	16 [76.2%]
Age	49 [13.1]
Race [n=10 missing]
White	5 [23.8%]
Black	1 [4.8%]
Multi-Race	5 [23.8%]
Ethnicity
Hispanic/ Latino	20 [95.2%]
Non-Hispanic/ Latino	1 [4.8%]
Education
< High School	10 [47.6%]
High School	3 [14.3%]
Some College	4 [19.1%]
College Degree	4 [19.1%]
Mental Health Diagnosis
Depression	7 [33.3%]
Bipolar Disorder and Depression	3 [14.3%]
Anxiety and Depression	10 [47.6%]
General Health
Very good	1 [4.8%]
Good	6 [28.6%]
Fair	13 [61.9%]
Poor	1 [4.8%]
Ever Experienced Homelessness	7 [33.3%]
No Source of Monthly Income	9 [42.9%]

Compliance and Accelerometer-Matched EMA

The remainder of data is based on the 20 participants who had matched EMA and accelerometer data. Participants responded to an average of 23 prompts and the mean compliance rate was 88.3% with a range of 40.7% to 100%. Of the 460 completed EMA responses, 428 (93%) were matched with a period of accelerometer wear (32 responses were excluded because of non-wear in the 30-minute window before or after the EMA prompt). All participants had four valid days of accelerometer wear (i.e., at least ten hours each day). The percent of missed EMA responses per participant did not differ by demographics (i.e., sex, age, race/ethnicity, education) or time-varying covariates (i.e., day of the week and time of the day). Available accelerometer-matched EMA data on affect and physical feeling states were as follows: Positive Affect (n= 412 responses; M=2.9, SD=0.9, ICC=0.24), Negative Affect (n=395 responses; M=1.9, SD=1.0, ICC=0.50), Energy (n=408 responses; M=2.5, SD=1.0, ICC=0.17), and Fatigue (n=406 responses; M=2.2, SD=1.1, ICC=0.42).

Relation between Affective, Physical Feeling States and Activity

Average MVPA within the 30-minute window before and after the EMA prompt (see Table 3) was 1.2 minutes (SD=3.0) and 1.0 minute (SD=2.5), respectively (ICC=0.03). Approximately 10% of 30-minute windows before or after an EMA prompt did not correspond to accelerometer wear. The results of the two-part models testing the effects of affective and physical feeling ratings on the number of MVPA minutes occurring in the 30 minutes after the EMA prompts are available in Table 4. Participants who, on average, reported more positive affect (PA) than others (i.e., between-subjects) in the study had a higher probability of engaging in some MVPA versus no MVPA in the next 30 minutes (Part 1: β = 0.60, SE = 0.30, p < 0.05). However, among those participants who engaged in some MVPA in the next 30 minutes, reporting more positive affect, on average, than others (i.e., between-subjects) was associated with less total time engaged in MVPA (Part 2: β = −0.80, SE = 0.26, p < 0.01). Also, among participants who engaged in some MVPA in the next 30 minutes, reporting feeling more energetic than others on average (i.e., between-subjects) was associated with less MVPA (Part 2: β = −0.67, SE = 0.30, p < 0.05). Negative affect (NA) and feeling tired ratings at any given prompt were unrelated to subsequent levels of MVPA. There were also no significant within-subject effects.

Table 3.

Physical Activity and Accelerometer-Matched EMA Prompt Responses

Variable	n [%] x̅ [SD]
Minutes of physical activity from accelerometer data [N=428 1 ]
MVPA2 30 minutes before prompt	1.2 [3.0]
MVPA 30 minutes after prompt	1.0 [2.5]
Affective state right before prompt
Positive affect [n=412]	2.9 [0.9]
Negative affect [n=395]	1.9 [1.0]
Feeling energetic [n=408]	2.5 [1.0]
Feeling tired [n=406]	2.2 [1.1]
Mental health state during past hour [n=304]
Crying/ poor focus/ poor appetite	43 [14.1%]
Paranoia/ hallucinations	19 [6.3%]
None reported	242 [79.6%]
Types of physical activities during past hour [n=398]
Running/ jogging	5 [1.3%]
Walking	55 [13.8%]
Weightlifting/ strength training	2 [0.5%]
Using cardiovascular equipment	1 [0.2%]
Bicycling	0 [0.0%]
Other activity	3 [0.8%]
No activity	332 [83.4%]

¹Represents matched EMA and wear cases [verified wear time at 30-min windows before or after prompt];

²Moderate-to-vigorous physical activity [MVPA]

Table 4.

The Association of Affective and Physical Feeling States with MVPA¹ Minutes Occurring in the 30 Minutes After EMA² Prompt³

	MVPA 30 minutes after EMA prompt4
	Within-Subject		Between-Subject
	β [SE]	p-value	β [SE]	p-value
Positive Affect
Piece 1 Model [some vs. zero MVPA minutes]	−0.13 [0.15]	0.383	0.60 [0.30]	0.047
Piece 2 Model [MVPA minutes]	0.02 [0.11]	0.837	−0.80 [0.26]	0.002
Negative Affect
Piece 1 Model [some vs. zero MVPA minutes]	0.18 [0.18]	0.328	−0.04 [0.18]	0.826
Piece 2 Model [MVPA minutes]	−0.12 [0.14]	0.548	−0.18 [0.20]	0.390
Feeling Energetic
Piece 1 Model [some vs. zero MVPA minutes]	−0.23 [0.12]	0.064	0.55 [0.29]	0.039
Piece 2 Model [MVPA minutes]	−0.09 [0.09]	0.325	−0.67 [0.30]	0.022
Feeling Tired
Piece 1 Model [some vs. zero MVPA minutes]	0.17 [0.13]	0.216	−0.15 [0.21]	0.468
Piece 2 Model [MVPA minutes]	0.04 [0.11]	0.720	−0.11 [0.20]	0.581

¹Moderate to vigorous physical activity [MVPA];

²Ecologically momentary assessments [EMA];

³Each factor and affective state represents a separate model. All models adjusted for day of week [weekday vs. weekend day], time of day [morning, afternoon, evening], age, sex, and prior affects;

⁴Positive Affect (n= 412 responses), Negative Affect (n=395 responses), Energy (n=408 responses), and Fatigue (n=406 responses).

The results of the mixed models testing the effects of MVPA in the 30 minutes leading up to the EMA prompts on ratings of affective and physical feeling states are displayed in Table 5. Participants with more MVPA minutes before the prompts on average, compared with other participants (i.e., between-subjects), reported significantly lower levels of positive affect (BS: β = −0.25, SE = 0.10, p < 0.05) and feeling energetic (BS: β = −0.21, SE = 0.10, p < 0.05). Engaging in more MVPA than one’s usual in the 30-minute window before an EMA prompt (i.e., within-subject) was associated with more negative affect at that prompt (WS: β = 0.02, SE = 0.01, p < 0.05). MVPA was unrelated to feeling tired at the subsequent EMA prompt.

Table 5.

The Association of MVPA¹ Minutes Occurring in the 30 Minutes Before the EMA² Prompt with Affective and Physical Feeling States^3,7

	Positive Affect		Negative Affect4		Feeling Energetic		Feeling Tired
	β [SE]	p	β [SE]	p	β [SE]	p-value	β [SE]	p
MVPA
Intercept	2.71 [0.41]	<0.001	0.22 [0.20]	0.283	2.16 [0.40]	<0.001	1.87 [0.64]	0.004
WS5	−0.01 [0.02]	0.449	0.02 [0.01]	0.030	< −0.01 [0.02]	0.940	<−0.01 [0.02]	0.940
BS6	−0.25 [0.10]	0.014	0.04 [0.06]	0.541	−0.21 [0.10]	0.037	0.14 [0.19]	0.454

¹Moderate to vigorous physical activity [MVPA];

²Ecological momentary assessment [EMA];

³Each factor and affective state represents a separate model. All models adjusted for day of week [weekday vs. weekend day], time of day [morning, afternoon, evening], age, sex, and prior affects;

⁴Log of negative affect;

⁵Within-subjects [WS];

⁶Between-subjects [BS];

⁷Positive Affect (n= 412 responses), Negative Affect (n=395 responses), Energy (n=408 responses), and Fatigue (n=406 responses).

Discussion

This study blends both repeated momentary surveys and activity monitors to study affect, physical feeling states and physical activity in adults with SMI and illustrates the dynamic relationship between mood and behavior in a vulnerable adult population. Though other studies recruiting adults with SMI have utilized wearable devices to measure activity (Chapman et al., 2016), none to our knowledge have included ecological momentary assessments to address affect or other feeling states. Our protocol achieved high compliance and illustrated the benefit of EMA data collection with this population.

As evidenced by the preliminary findings of this study, the relationship between energy, positive affect, and time spent in activity is particularly nuanced for adults with SMI. Similar to research with the general population, participants were more likely to engage in any MVPA if they were generally more positive than others (i.e., between-persons assocations). However, for those who engaged in any MVPA during the study period (i.e., between-persons assocations), more positive affect and energy than others was associated with less time engaging in the activity (i.e., less MVPA minutes). Contrary to our expectations, there were no significant within-person associations between affect or feeling states and subsequent MVPA. On the other hand, if participants engaged in more minutes of MVPA than their usual, they were significantly more likely to report negative affect in subsequent prompts (within-person assocations).

It is a bit surprising that additional significant within-person associations were not found similar to other studies with a general population (Niermann et al., 2016). Among the individuals in this study, generally being more positive and energetic related to any physical activity, but fluctuations in momentary affect did not appear to modify one’s own behavior. Feeling energetic or feeling more positive than one’s peers may result in less active minutes while exercising, though it is unclear why this may be the case. Similar to results of other EMA studies with healthy adults (Dunton et al., 2015), participants in this study were also likely to report negative feelings higher than their average when engaged in MVPA for a longer period of time.

Generally, planned physical activity was not commonly reported in this study (only 17% of EMA) and if planned activity was self-reported, 82% of the time the participant was walking. Walking tends to be the most common physical activity reported in surveys by adults with SMI (Chapman et al., 2016; Daumit et al., 2005) and might be more the byproduct of transportation rather than a leisure activity (Daumit et al., 2005). It is not possible in this study to infer the motivations behind a specific physical activity, and it is possible participants were simply reporting their daily commutes. Perhaps heightened negative feelings following sustained exercise is a byproduct of some aspect of the activity that we are not measuring. Energy that is expended for general activity (e.g., walking) is referred to as non-exercise activity thermogenesis (NEAT) (Levine, 2002). NEAT can be a critical source of activity for individuals. Ergo, future studies should focus on how NEAT can be increased among individuals with SMI, as well as ways in which increased time spent engaging in NEAT can result in positive emotions rather than increased negative affect. In cross-sectional studies, adults with SMI report motivation to be active to improve mood, reduce stress, or lose weight (Chapman et al., 2016), but it is difficult to infer how one’s motivation may impact activity momentarily. Future studies should consider including a momentary measure of motivation to engage in physical activity in the EMA question sequence in hopes this may shed some additional light. Moreover, adults with SMI are often socially isolated and the propensity to engage in activity may be especially impacted by their sparse social networks (Suetani et al., 2016; Vancampfort et al., 2012). Similar to other work (Dunton et al., 2015), future endeavors may consider examining the influence of social and physical contexts on MVPA of this unique population. Lastly, future studies are warranted to determine if exercise may have a different neurobiological impact on adults with SMI. For instance, adults with SMI may be more sensitive to fatigue post MVPA similar to individuals with chronic fatigue syndrome (Staud et al., 2015) and thus feel more negative after exercising for sustained periods.

Though this study is strengthened by its inclusion of momentary collection methods and device measured activity, conclusions are limited primarily due to a small sample size of adults with a wide mix of mental health issues. Symptoms associated with different types of mental health diagnoses may differentially impact an individual and thus their behavior. Many of the participants also reported a variety of physical health issues that may impact the ability to exercise (e.g., back problems, asthma, arthritis, Fibromyalgia). Furthermore, the participants in this study were only monitored for four days and it is possible the observation period occurred during a time of atypical behavior. Though 4 days is a typical length of time in EMA studies with other populations, this small observation window may not provide an adequate time frame to represent the physical activity of adults with SMI. Longer monitoring periods with a larger sample would greatly improve the power to further test assumptions. Though we observed good compliance with prompts, generous incentives may have impacted participation and thus future studies may experience feasibility issues when payment is less. Due to random allocation of the EMA prompts, it is also feasible that activity was missed in this study. Even in longer observation periods, it may be beneficial to administer end of the day surveys (i.e., daily diaries) in conjunction with time-varying EMAs. It is also possible our method to exclude periods of non-wear was ineffective and resulted in exclusion of long bouts of sedentary time.

Despite the small size of this pilot, this work begins to identify potentially modifiable correlates of MVPA and begins to point to the utility of momentary interventions though larger studies are warranted. Adults with SMI often develop diseases that are largely avoidable and represent a population that would especially benefit from lifestyle interventions (Firth et al., 2016). Nonetheless, most physical activity interventions are complex and require ongoing self-regulation (Richardson, 2005). Many interventions do not consider the additional needs of individuals with SMI or attend to context or intraindividual variations in mood (Cabassa et al., 2010). Adults with SMI are motivated to engage in MVPA to improve mood and reduce stress, but both mood and stress are cited as the most common barriers to activity (Firth et al., 2016). Lifestyle interventions in this population might be enhanced with more attention towards avenues to increase NEAT and recreational walking (Chapman et al., 2016). Recent advances in mobile technology have created additional opportunities to creatively intervene on MVPA in naturalistic settings (Dunton & Atienza, 2009; Patrick et al., 2008). Ecological momentary interventions (EMI) or just-in-time adaptive interventions (Nahum-Shani et al., 2018) may be particularly attractive to a population that is often hard to reach. Mobile phone use is prolific among adults and also among those with SMI (Ben-Zeev et al., 2013) and may present an avenue for intervention that mitigates the need for travel and is attuned to one’s context (Firth et al., 2016; Richardson et al., 2005).