Within-person examination of the exercise intention-behavior gap among women in midlife with elevated cardiovascular disease risk

Abstract

Engaging in moderate-to-vigorous intensity physical activity (MVPA) is important for protecting cardiovascular health among women in midlife (i.e., ages 40-60), particularly if they have already developed conditions that increase their risk for cardiovascular disease (e.g., hypertension). Although the gap between MVPA intentions and behavior is well documented in other populations, little is known about the intention-behavior gap in this at-risk group – particularly as it plays a role in daily life. The present study employed an ecological momentary assessment design to examine the relation between women’s MVPA intentions and behavior in the subsequent 3 hours, as well as momentary moderators of this relation (i.e., affective states and body satisfaction). Surveys sent to women’s smartphones 5 times per day for 10 days while they wore ActiGraph GT3X waistband accelerometers. Women achieved their exercise intentions at only 13% of occasions on which they set intentions. Although the most common intended exercise was walking, women engaged in more minutes of MVPA after setting intentions to do yoga or Pilates than any other type of exercise (sr = 0.25). Multilevel models showed a modest within-person relation between minutes of intended MVPA and observed MVPA in the next 3 hours (sr = 0.20). This relation was moderated within-person by the reported extent of positive affect (particularly contentment) and body satisfaction (srs = 0.35 and 0.28, respectively). Findings extend knowledge about the physical activity intention-behavior gap to an at-risk population of women and identify positive affect and body satisfaction as potential contextual influences for this group, which could inform improvements to existing interventions (e.g., delivering intervention content at times with lower-than-usual body satisfaction).

Physical activity has myriad mental and physical health benefits, including reductions in risk for cardiovascular disease (CVD) among those with risk conditions such as hypertension (Piercy et al., 2018). Accruing these benefits is particularly important for women in midlife with such conditions (i.e., those aged 40-60; Brim et al., 2019), as the menopause transition is uniquely associated with increased risk of CVD (Brim et al., 2004, Matthews et al., 2009). Currently, more than half of women in midlife do not engage in the recommended levels of physical activity (PA) needed for health benefits - particularly activity that reaches moderate-to-vigorous intensity (i.e., exercise; National Center for Health Statistics, 2019). Elucidating the psychosocial determinants of exercise engagement in this at-risk population will likely enhance health promotion efforts that target volitional behaviors.

Prominent health behavior theories postulate that intentions to exercise are a proximal determinant of exercise behavior (e.g., Theory of Planned Behavior, Ajzen, 1991). Specifically, that setting (vs. not setting) an intention to exercise and/or experiencing stronger (vs. weaker) intentions to exercise are positively associated with exercise behavior. Evidence appears to support this notion; one meta-analysis found that intentions to be physically active (including intentions for moderate-to-vigorous activity) had a medium-sized effect on engaging in future PA or exercise (McEachan et al., 2011). This effect was larger than other correlates of PA or exercise participation (e.g., attitudes, subjective norms), suggesting the potential utility of promoting intentions as a way to foster increased engagement.

Yet, there is also substantial evidence indicating that behavioral intentions do not always translate to engagement, known as the intention-behavior gap (Rhodes & Dickau, 2012). A meta-analysis by Rhodes and de Bruijn (2013) concluded that only slightly more than half of individuals who intend to be physically active or exercise ultimately follow through with those intentions, and only a small percentage are able to engage in PA or exercise despite having no intention to do so. These findings suggest that intentions are useful, but that intentions alone are not sufficient to engage in PA or exercise. Importantly, however, the majority of research exploring PA or exercise intention-behavior coupling has focused on these relations among college students. University students tend to be young, fit, and have fewer time- and responsibility-related barriers to PA compared to adults (e.g., El Ansari & Lovell, 2009). Research is needed to understand intention-behavior coupling in at-risk groups such as women in midlife, as there may be specific factors that moderate associations between intentions and behavior and increase intention-behavior concordance in these populations.

Intention-Behavior Relations: Level of Analysis

In addition, most existing work examining intention-behavior relations focuses on between-person relations, or person-level associations, across marco-timescales (e.g., weeks, months). While this time scale of analysis can provide valuable information, it does not align with the time scale on which individuals make choices to engage in PA or exercise (e.g., daily, within days) or account for PA/exercise as repeated behaviors across days (Dunton, 2017). To address this, recent work has employed daily and within-day assessments of PA/exercise intentions and behavior to provide a more nuanced understanding of intention-behavior relations at the within-person level. These studies have documented significant within-person variability in daily and within-day intentions and behavior as well as significant associations between these constructs (Conroy et al., 2013; Maher & Dunton, 2020 Pickering et al., 2016; Schumacher et al., 2021). For instance, Maher et al. (2017) examined the within-person association between adults’ current PA intentions and subsequent PA behavior in the next 3 hours. On more than 80% of occasions when adults intended to be active, they failed to translate their intentions into behavior in the next 3 hours (Maher et al., 2017). Examination of potential within-person moderators of intention-behavior relations indicated that, on occasions when individuals experienced greater positive affect than was typical for them, they were more likely to enact their PA intentions over the next 3 hours (Maher et al., 2017).

Ecological momentary assessment (EMA) is an intensive longitudinal data capture strategy aimed at repeatedly assessing phenomena of interest in everyday life (Shiffman et al., 2008). EMA methods can overcome several limitations in previous PA and exercise research, by examining both within-person associations between PA determinants and behavior (Dunton, 2017). EMA is able to capture intention-behavior coupling across and within days, while reducing recall bias inherent in traditional and more commonly used retrospective, between-person measures (Shiffman et al., 2008). Preliminary evidence shows that EMA of PA, as well as of proximal psychosocial determinants of these PA (e.g., self-efficacy), is feasible and acceptable to women in midlife (Ehlers et al., 2016; Arigo et al., 2021). However, no existing study has used EMA to assess the exercise intention-behavior relation in this population.

Within-Person Moderators of the Intention-Behavior Relation

EMA may also be useful for exploring other momentary, within-person factors that play unique roles in intention-behavior coupling among women in midlife, in addition to more established within-person factors such as affective state (see Liao et al., 2015). For instance, body satisfaction is known to play a role in regulating PA and exercise among women; however, body satisfaction often is conceptualized as a between-person factor (i.e., distinguishing between people who are more or less satisfied with their bodies; cf. Leahey et al., 2011). Yet, considerable evidence suggests that body satisfaction fluctuates within people and its influence on exercise can be more or less salient depending on contextual factors (Cash et al., 2002; Mills et al., 2014). It is possible that body satisfaction at a given moment may modify exercise intention-behavior relations. For instance, more positive body satisfaction at a particular moment may weaken the intention-behavior relation, because women who are already satisfied with their physical appearance and do not feel the need to follow through with intentions to improve their appearance (cf. Arigo et al., 2016). Alternatively, positive body satisfaction at a particular moment may strengthen intention-behavior relations because women see exercise as a means to maintain their appearance and positive self-perceptions, and thus, follow through with intentions to be active. To date, no study has investigated the moderating role of body satisfaction in momentary exercise intention-behavior relations, particularly among women in midlife.

Aims of the Present Study

Toward the goal of elucidating the within-person exercise intention-behavior relation among women in midlife with elevated CVD risk, the first aim of this study was to describe daily and momentary exercise intentions in this population, with respect to the frequency, duration, and types of exercise intended. The second and primary aim was to determine whether exercise intentions predicted exercise behavior in the following 3 hours, defined in minutes of intended and actual moderate-to-vigorous PA (MVPA). We selected the 3-hour period following the EMA prompt for two reasons: (1) to best align with the time frame presented in the EMA intention item (i.e., the next few hours) without overlapping with the following reporting period, and (2) to mirror previous EMA research examining momentary exercise intention-behavior coupling using a 3-hour time frame (Maher et al., 2017). The third aim was to examine momentary moderators of the relation between exercise intentions and behavior (i.e., affective states and body satisfaction), to identify the contexts associated with stronger versus weaker intention-behavior relations in daily life.

Methods

Recruitment and Participants

Eligible individuals were women in midlife (40-60 years old, inclusive) with one or more conditions that confer risk for CVD; this included diagnosis of type 2 diabetes or prediabetes, hypertension or prehypertension, hyperlipidemia or hypercholesterolemia, metabolic syndrome, or smoking (current or had quit in the past 3 months). Additionally, eligibility required that they were fluent in English, had no medical constraints to participate in PA, were not pregnant, had no comorbid medical conditions or psychiatric symptoms that would hinder involvement, and were able to complete electronic surveys on their personal mobile devices. Recruitment employed web and print advertisements (e.g., Facebook, flyers in local community centers) and direct referrals in family medicine clinics. A total of 197 women expressed interest, 135 completed the telephone screening, and 76 attended the setup session. One woman withdrew after this session but before beginning the EMA protocol; 100% of those who began the EMA protocol completed the full 10 days of observation. The sample included 75 women (M_Age = 51.61, M_BMI = 34.02 kg/m²); the majority identified as White (73%) and 69% had BMIs in the obese range. Further demographics can be found in Table 1.

Table 1.

Participant demographics (n = 75).

	M (SD)
Age	51.61 (5.43)
BMI	34.02 (7.13)
Number of CVD risk factors	1.63 (0.82)
Racial/Ethnic Identification	n (%)
Caucasian/White	55 (73%)
African American/Black	16 (22%)
Asian or Pacific Islander	1 (1%)
Hispanic/Latina	2 (3%)
Mixed/Other	1 (1%)
Marital Status	n (%)
Never married	12 (16%)
Widowed	4 (5%)
Divorced	11 (15%)
Separated	4 (5%)
Married	44 (59%)
Household Income
< $25,000	5 (7%)
$25,000-$50,000	12 (16%)
$50,000-$75,000	12 (16%)
> $75,000	45 (61%)
Highest educational level
High School or GED	7 (9%)
Associate’s degree, technical degree, or partial college	16 (21%)
Bachelor’s degree	24 (32%)
Graduate/professional degree	28 (37%)
Menopause Status
Pre-menopause	14 (20%)
Perimenopause	16 (23%)
Post-menopause	29 (39%)
Other (e.g., surgical intervention)	12 (17%)
CVD risk condition(s)
Hypercholesterolemia or hyperlipidemia	39 (52%)
Hypertension or prehyptertension	35 (47%)
Type 2 diabetes	30 (40%)
Metabolic syndrome	8 (11%)
Smoker (or quit in last 3 months)	10 (14%)

Measures

Demographics

Characteristics such as age, racial/ethnic identification, medical conditions, and menopause were assessed via electronic survey, completed before the start of the EMA period. Height and weight were measured by research staff at baseline in-person visits using a Seca® scale and stadiometer; these values were used to calculate BMI (kg/m²).

Exercise Intentions

Intentions were assessed with 1 item: “Do you have plans to do cardiovascular exercise in the next few hours (such as going on a brisk walk or doing a strength DVD routine)?” Participants could respond with either “No, no plans to exercise” or “Yes.” If participants responded “Yes”, they were asked “How many planned minutes?” and “What kind of exercise?” Both were open-ended questions that allowed participants to enter the number of minutes intended and the type of exercise. The latter were coded by research staff into the following categories: (1) (brisk) walking, (2) running/jogging/wogging (i.e., a combination of walking and jogging), (3) bike riding, (4) swimming, (5) other cardio (e.g., Zumba, dance class, basketball, boxing, or using an elliptical machine, stair stepper, or treadmill), (6) yoga or Pilates, and (7) weight lifting or other strength training that would elevate the heart rate. Activities that were not counted as a type of exercise (i.e., those unlikely to reach moderate-to-vigorous intensity) were house/yard work (gardening, cleaning) and casual activity (shopping).

Affect and Body Satisfaction

Recent affect was measured with 5 items referencing how much they experienced each emotion since waking up (first survey of the day) or in the last 3 hours (all other surveys). Participants responded on a 3-point scale ranging from 1 (not at all) to 3 (very much) for each affective state (cf. Dockray et al., 2010; Dunton et al., 2011). Responses to happy or excited and content were summed to create a positive affect composite score; responses to angry, stressed or anxious, and sad were summed to create a negative affect composite score (Scott et al., 2020). The associated 2-factor structure showed good model fit (χ² = 9.5, p = 0.01); factor loadings for positive affect and negative affect items showed that each subset of items loaded significantly together, as expected. Current level of satisfaction with one’s body was measured with 1 item, “How would you describe your body satisfaction right now?” (cf. Fuller-Tyszkiewicz, 2019; Thøgersen-Ntoumani et al., 2017), on a 4-point scale from 1 (very dissatisfied with my body) to 4 (very satisfied with my body). These items were pre-tested with the population of interest following procedures described by Arigo et al. (2021).

Exercise Behavior

MVPA was assessed with the ActiGraph GT3X triaxial accelerometer (ActiGraph Corporation; Pensacola, FL). This was worn on the waist, aligned with the participant's dominant hip, during waking hours (removed for activities such as showering or swimming). Minutes spent in moderate- and vigorous-intensity PA (MVPA) were summed to generate the number of minutes of MVPA in the 3 hours following completion of each survey. Accelerometer data were processed with the ActiPro package for R (Dzubur, 2020) using MVPA cut points proposed by Matthews et al. (2008; see Arigo et al., 2020). Periods with 60 or more continuous minutes without activity counts (i.e., activity counts = 0) were considered nonwear times and were excluded from analyses (cf. Pickering et al., 2016). For descriptive purposes, MVPA in the 3 hours after each survey was coded as achieving versus not achieving the intended number of exercise minutes (1/0), using thresholds of 100% and 80% of intended minutes.

Procedures

The full protocol for this study is described elsewhere (Arigo, Brown et al., 2020) ) and all procedures were approved by the Institutional Review Board at the supporting institutions. Participants expressed interest via phone or email and completed a telephone screening with research staff to verify eligibility, explain study procedures, and answer questions. Those who were eligible were scheduled for an in-person setup appointment and were asked to complete an initial electronic questionnaire before this appointment to collect demographics. Individual setup appointments were conducted by a trained staff member, who conducted informed consent procedures and measured the participant’s height and weight. Staff members also trained participants in the completion of EMA surveys and wear of the accelerometer; this included defining exercise as moderate or vigorous movement that gets the heart rate up (i.e., elevates the heart rate more so than leisurely walking), including moderate-intensity strength or flexibility exercises such as Pilates and some types of yoga. Participants were asked to complete surveys within 1 hour of receiving them. Clocks on the accelerometer initialization software and participants’ mobile devices were internally synced to associated satellite timekeeping methods; research staff viewed both during setup appointments and confirmed that there were no deviations of more than a few seconds between these tools.

The EMA period began the day after the setup session, with surveys delivered via text message with an embedded web link. Participants received EMA prompts at 5 semi-random times for 10 days with a consistent schedule across weekdays and weekends. Three different EMA prompting schedules were available based on differences in participants’ typical sleep/wake times. These included early rising (prompts delivered between 6:30 AM and 6:45 PM), standard work schedule (prompts delivered between 8:30 AM and 9:45 PM), or late rising (prompts delivered between 11:00 AM and 11:15 PM). Within each sleep/wake timing category, participants were randomized to one of 3 different schedules for receiving prompts. Further details on the EMA prompting schedule can be found in Arigo, Brown et al. (2020). Participants wore the accelerometer to measure PA over the same 10-day period. After 10 days, participants came back to the research center for an exit interview and to return accelerometers. Participants were able to receive up to US $55 as compensation: $15 for baseline assessment, $30 for completing EMA component, $10 bonus for EMA survey completion >80% (cf. Bernstein et al., 2018).

Statistical Analysis

Research staff coded survey responses as completed within the allotted 1-hour time window versus not (1/0); only those completed within the allotted time window were included in analyses. As noted, the type of exercise intended was coded into categories, and non-structured exercise (e.g., household tasks, gardening) were removed from analyses. Surveys completed at the end of the day also were removed from analyses (n = 600), as they described exercise intentions for the following day, rather than the following hours. Surveys indicating more than 180 minutes of intended exercise in the following 3 hours were also excluded (n = 1). Across the sample, 2967 surveys were completed in the allotted 1-hour time window (89%), and 2367 valid responses were for surveys 1-4 of the day (79%). As described below, a subset of participants never set exercise intentions and these individuals were removed from primary analyses. Valid accelerometer data were available for 1865 observations, affording adequate power for the within-person analyses described below (i.e., >0.80 based on simulations by Maas & Hox, 2005).

All analyses were conducted in SAS Version 9.4 (Cary, NC). Descriptive statistics (e.g., means, frequencies) were calculated for exercise intention reports (yes/no), type of exercise intended, MVPA in the following 3 hours, and achievement of intended exercise minutes (yes/no). Intraclass correlation coefficients were used to confirm that time-sensitive variables (i.e., intended minutes of exercise, observed minutes of MVPA, affective state, body satisfaction) showed meaningful within-person variability (see supplemental Table A; all ICCs ≤ 0.78, ps for within-person/residual variance all < 0.001). Predictors of the likelihood of setting intentions (binary, yes/no), including age, BMI, survey of the day, day of observation, and MVPA in the previous 3 hours were evaluated with PROC GLIMMIX using likelihood estimation (Laplace).

Minutes of MVPA met assumptions of normality: estimates of skewness and kurtosis were below thresholds that would necessitate use of a non-normal distribution (i.e., 2.0 and 8.0, respectively; Kline, 2015). Consequently, primary analyses were conducted in PROC MIXED with a continuous predictor and outcome (number of intended exercise minutes and number of MVPA minutes in the following 3 hours, respectively), using maximum likelihood estimation to address missingness. This included accelerometer non-wear in the periods following an EMA prompt, which were designated as missing data. As adding the day level did not significantly improve model fit (χ²[1] = 1.20, p = 0.25), all models used 2 levels (moment and person). BMI, age, weekday versus weekend day, and survey of the day were associated with missingness and were included as covariates in all analyses; BMI and age were centered at the grand mean. As described below, MVPA in the previous 3 hours was associated with the likelihood of setting intentions and also was included as a covariate.

The relation between exercise intentions and MVPA in the following 3 hours was examined using the number of intended minutes of exercise as a continuous predictor, for times when intentions were set. Subsequently, potential momentary moderators of the relation between the numbers of intended and actual minutes of MVPA (i.e., affective states, body satisfaction) were added in separate models and were person-mean centered to isolate within-person effects (Hoffman, 2015). Affect composite scores were tested first, with responses to individual affective states tested in follow-up analyses. As recommended by Rebar, Rhodes, and Gardner, (2019), we confirmed that the proposed moderators were not associated with the likelihood of reporting exercise intentions or with MVPA in the following 3 hours (ps > 0.10). Sensitivity analyses using MVPA in 1-, 2-, and 4-hour time frames post-survey completion showed patterns similar to those reported below. Associated effect sizes were smaller for 1- and 2-hour blocks (as women engaged in fewer minutes of MVPA over these shorter time frames) and slightly larger for the 4-hour block (as women engaged in more minutes of MVPA over this longer time frame). Statistical significance was set at p < 0.05 and effect sizes are expressed as semipartial correlation coefficients (sr).

Results

Frequency and Type of Exercise Intentions

Across valid surveys, participants reported setting exercise intentions in 18% of their responses (427/2367). At the participant level, the number of reports with indications of setting exercise intentions ranged from 0 to 38 out of 50 surveys, with 11 participants never reporting an intention to engage in exercise over the course of the study. The average participant reported setting exercise intentions at 9.37 of 50 surveys (SD = 8.53, 19%). Women with higher (vs. lower) BMIs were less likely to report setting intentions (OR 1.07, 95% CI 1.02-1.12, p < 0.01), though the likelihood of setting intentions did not differ by age, racial/ethnic identification, or menopause status (ps > 0.35). Participants set intentions to exercise approximately once per day, on average (B = 0.98, SE = 0.10), though the daily frequency of intentions to exercise varied significantly within-person (Z = 15.89, p = 0.001; range 0-5 times per day).

At the moment level, setting exercise intentions was least likely at the fourth survey of the day (F[1,2886] = 92.92, p < 0.001, OR = 5.05, 95% CI = 3.63-7.02), which typically was completed between 6:00 and 8:00 PM. The likelihood of setting exercise intentions also was related to exercise in the previous 3 hours, such that participants were less likely to set intentions when they achieved more minutes of MVPA in the 3 hours prior to survey completion (F[1,1961] = 25.09, p < 0.001, RR = 0.98, 95% CI = 0.97-0.99). The likelihood of setting intentions was not related to day of week or weekday versus weekend (ps > 0.21), but was associated with day in study: participants were less likely to set intentions on later (vs. earlier) days of data collection (OR = 1.08, 95% CI = 1.01-1.12, p = 0.01).

The type of exercise intended was most often walking, followed by activities such as a dance class or using an elliptical machine (see Table 2). Participants indicated more than one intended type of exercise at 1% of surveys (28/2367). Minutes of intended exercise ranged from 3 to 180 (M = 37.12, SD = 36.53). Minutes of MVPA in the 3-hour blocks following surveys 1-4 ranged from 0 to 98.17 (B = 12.65, SE = 0.61) and showed a linear decrease from the first to fourth surveys of the day (F[1,1255] = 350.58, p = 0.05, sr = 0.43). Of note, minutes of MVPA were highest in the 3 hours subsequent to an intention to do yoga or Pilates (and higher after these intentions than all other categories of exercise; F[1,32] = 4.03, p = 0.05; sr = 0.25), followed by intentions to do other cardiovascular activities and (brisk) walking (see Table 2).

Table 2.

Descriptive statistics for exercise intentions and behavior.

Exercise Intentions	n (%) YES of all valid surveys	n (%) YES Survey #1	n (%) YES Survey #2	n (%) YES Survey #3	n (%) YES Survey #4
Reported intention	427 (21%)	141 (24%)	121 (20%)	111 (18%)	54 (9%)
Type of Exercise Intended	n (%) YES of all valid surveys	n (%) YES Survey #1	n (%) YES Survey #2	n (%) YES Survey #3	n (%) YES Survey #4
(Brisk) Walking	273 (12%)	88 (62%)	80 (65%)	73 (69%)	32 (19%)
Running, jogging, wogging	8 (0.3%)	2 (1%)	4 (3%)	2 (2%)	0 (0%)
Bike riding	9 (0.4%)	2 (1%)	3 (2%)	3 (3%)	1 (2%)
Swimming	5 (0.2%)	1 (1%)	3 (2%)	1 (1%)	0 (0%)
Other cardio	57 (2%)	26 (18%)	13 (11%)	9 (8%)	9 (15%)
Yoga/Pilates	17 (0.7%)	5 (6%)	3 (2%)	5 (5%)	4 (7%)
Weight lifting/strength straining	19 (0.8%)	11 (8%)	3 (2%)	3 (3%)	2 (3%)
Exercise Behavior	B (SE) after all valid surveys	B (SE) after Survey #1	B (SE) after Survey #2	B (SE) after Survey #3	B (SE) after Survey #4
Minutes of MVPA	12.65 (0.61)	18.53 (0.73)	12.92 (0.73)	5.34 (0.83)	2.46 (1.38)

Minutes of MVPA B (SE)
(Brisk) Walking	Running, jogging, wogging	Bike riding	Swimming	Other cardio	Yoga or Pilates	Weight lifting/ strength straining
11.64 (1.87)	5.90 (4.36)	6.56 (5.36)	7.61 (5.80)	11.81 (2.55)	17.64 (4.34)	4.72 (3.10)

Potential Moderators	B (SE)		B (SE)			B (SE)
Body Satisfaction	2.43 (0.09)
Positive Affect (Composite)	3.98 (1.00)	Happiness	1.83 (0.05)	Contentment	2.21 (0.05)
Negative Affect (Composite)	4.11 (0.11)	Anger	1.41 (0.04)	Sadness	1.24 (0.03)	Stress	1.58 (0.05)

Note: n (%) yes indicates report of intention to exercise, out of 2367 valid surveys (any time window). Wogging = a combination of walking and jogging.

Relations between Exercise Intentions and Behavior

Based on observed minutes of MVPA in the following 3 hours, participants fully achieved (or exceeded) their intended exercise minutes on only 13% of occasions when intentions were set, and achieved 80% or more of their goals on 17% of occasions when intentions were set. Minutes of MVPA did not differ after times when participants set (vs. did not set) an intention to exercise (F[1,235] = 0.32, p = 0.58). At times when women did set an intention, the number of intended minutes was only weakly positively associated with minutes of MVPA in the following 3 hours (F[1,249] = 2.51, p = 0.11, sr = 0.20). Day in study did not moderate this relation (F[1,235] = 0.75, p = 0.39). Further, neither negative affect composite scores nor individual momentary ratings of anger, stress, or sadness moderated the relations between intended exercise and subsequent MVPA (ps > 0.27; see Table 3). Both positive affect composite scores and body satisfaction did moderate the relation between intended and actual MVPA minutes, however. The moderating effect of positive affect appeared to be driven by contentment ratings (F[1,227] = 7.96, p = 0.005, sr = 0.35), rather than happiness ratings (F[1,230] = 0.32, p = 0.57): the relation between intended and actual minutes of MVPA was stronger at times when women experienced less (vs. more) contentment than usual. This relation also was stronger at times when women experienced less (vs. more) body satisfaction than usual (F[1,236] = 4.94, p = 0.03, sr = 0.28).

Table 3.

Multilevel model estimates for tests of relations between exercise intentions (number of minutes) and minutes of MVPA in the 3 hours after each survey.

	Minutes of MVPA
Exercise Intentions	B (SE)
Intercept	19.87 (2.09)**
Number of intended minutes (WP)	0.04 (0.03)
Moderation by Positive Affect (PA Composite)	B (SE)
Intercept	29.76 (4.00)**
Number of intended minutes (WP)	0.05 (0.02)*
PA Composite (BP)	−2.47 (0.86)**
PA Composite (WP)	0.97 (1.16)
Intended minutes (WP)*PA composite (WP)	−0.04 (0.02)*
Moderation by Happiness	B (SE)
Intercept	25.07 (3.73)**
Number of intended minutes (WP)	0.05 (0.03)
Happiness (BP)	−2.75 (1.69)
Happiness (WP)	0.47 (2.09)
Intended minutes (WP)* Happiness (WP)	−0.01 (0.05)
Moderation by Contentment	B (SE)
Intercept	32.26 (3.81)**
Number of intended minutes (WP)	0.07 (0.23)**
Contentment (BP)	−5.81 (1.48)**
Contentment (WP)	3.66 (2.14)
Intended minutes (WP)* Contentment (WP)	−0.14 (0.05)**
Moderation by Body Satisfaction	B (SE)
Intercept	22.54 (2.97)**
Number of intended minutes (WP)	0.05 (0.03)*
Body Satisfaction (BP)	−1.14 (0.95)
Body Satisfaction (WP)	5.71 (2.43)*
Intended minutes (WP)* Body Satisfaction (WP)	−0.09 (0.04)*
Moderation by Negative Affect (NA Composite)	B (SE)
Intercept	8.17 (4.55)
Number of intended minutes (WP)	0.06 (0.03)*
NA Composite (BP)	2.88 (1.00)**
NA Composite (WP)	−0.41 (1.22)
Intended minutes (WP)*NA composite (WP)	0.03 (0.03)
Moderation by Anger	B (SE)
Intercept	4.84 (4.77)
Number of intended minutes (WP)	0.06 (0.03)*
Anger (BP)	11.16 (3.18)**
Anger (WP)	−0.38 (2.56)
Intended minutes (WP)*Anger (WP)	0.08 (0.07)
Moderation by Sadness	B (SE)
Intercept	7.29 (4.61)
Number of intended minutes (WP)	0.06 (0.03)*
Sadness (BP)	10.11 (3.38)**
Sadness (WP)	−6.13 (3.32)
Intended minutes (WP)* Sadness (WP)	0.16 (0.09)
Moderation by Stress	B (SE)
Intercept	12.26 (3.85)**
Number of intended minutes (WP)	0.07 (0.03)*
Stress (BP)	5.04 (2.12)*
Stress (WP)	2.40 (2.20)
Intended minutes (WP)* Stress (WP)	−0.006 (0.06)

Note:

^*p < 0.05;

^**p < 0.01; BP = between-person; WP = within-person; all models controlled for baseline BMI, age, weekday vs. weekend, survey of the day (1-4), and minutes of MVPA in the previous 3 hours.

Discussion

Consistent evidence has documented the intention-behavior gap with respect to PA and exercise (Sheeran & Webb, 2016). To date, however, the present study is one of the few to examine this gap in an at-risk population of women in midlife with elevated risk for CVD, as well as one of the few to do so at the within-person level. In particular, this study identified within-person moderators of the intention-behavior gap; as discussed below, this could be useful for informing improvements to intervention efforts. Consequently, the present study extends previous work in multiple ways.

Specifically, information about exercise intentions and behavior among women in midlife is scant, particularly with respect to the self-selected frequency, duration, or type of exercise intentions in this population. The current findings indicate that women in midlife with elevated CVD risk set exercise intentions approximately once per day; they are more likely to set intentions earlier in the day than later, and less likely to set intentions if they recently engaged in MVPA. Although (brisk) walking is the most popular intended exercise, women may engage in more minutes of MVPA with intentions to do yoga or Pilates. It is possible that these activities are guided with a class or video, and thus, offer support and accountability for behavior (e.g., via modeling, instruction, set duration). Walking often is recommended for its flexibility and convenience; yet, these supports may not be available for walking, making it easier to change or break intentions to walk (Im et al., 2011). As women in midlife cite lack of support as a barrier to engaging in PA and exercise (Im et al., 2013), encouragement to increase PA using guided (vs. unguided) activities may be more effective for this population (cf. Grossman et al., 2018).

Available data describing the PA intention-behavior relation among women in midlife do so between-person over considerable periods of time, with equivocal findings. For example, ratings of stronger (vs. weaker) intentions to “increase (my) PA over the next 6 months” were associated with more self-reported PA behavior 6 months later (Fortier et al., 2009), though the strength of intentions to “participate in the recommended amount of PA during (my) leisure time over the next 2 weeks” was not associated with reported PA 12 weeks later (Barg et al., 2012). To our knowledge, the present study is the first to examine the PA intention-behavior relation among women in midlife within-person over much shorter time periods and using device-based assessment of PA behavior. Findings show that these women successfully follow through on their exercise intentions on only 13%-17% of occasions when intentions are set, and that the relation between the number of intended and actual minutes of exercise is modest. Findings of minimal agreement between momentary intentions and subsequent PA behavior are consistent with investigations of PA intention-behavior coupling within days in other populations (e.g., Maher et al., 2017). Thus, the present findings add to accumulating evidence that concordance between PA intentions and behavior is much lower within days than across weeks or months (which is approximately 54%; Rhodes & de Bruijn, 2013).

Importantly, however, women’s momentary positive affect and body satisfaction moderated associations between their exercise intentions and subsequent behavior. Previous work has documented that greater positive affect on a given occasion increases the likelihood of successfully enacting PA intentions (Maher et al., 2017). To our knowledge, this is the first study to identify specific affective domains which may moderate PA intention-behavior relations: lower-than-normal levels of contentment, but not happiness, appeared to strengthen intention-behavior relations. Walking (the most common intended exercise in the present sample) at both self-selected and prescribed intensities is known to improve pleasant, deactivated affective states such as contentedness following exercise (Ekkekakis et al. 2000). Thus, moments of lower-than-normal contentment may enhance motivation for exercise to attain more positive affective states, thereby leading to greater intention-behavior coupling.

In contrast, composite negative affect and negative affective states (anger, stress, sadness) did not moderate intention-behavior coupling in this sample, whereas immediate body satisfaction did: the relation between exercise intentions and behavior was stronger at times when women experienced less (vs. more) body satisfaction than usual. These findings diverge from prior work showing that exercising for the purpose of stress reduction is associated with long-term PA engagement among women in midlife (Segar, Eccles, & Richardson, 2008) and that women in midlife with appearance-related motives for engaging in PA are less active than women with other types of motives (Segar, Spruijt-Metz, & Nolen-Hoeksema, 2006). These prior studies describe relations between-person, rather than within-person, highlighting the importance of differentiating levels of relations in daily life. In addition, it is possible that allowing participants to distinguish between negative affect more generally and the specific experience of low body satisfaction in the present study enabled the detection of the latter moderating effect, which may have been combined with other sources of negative affect in prior work. Further research is needed to better understand how and why specific momentary affective and body image states influence PA intention-behavior relations, in this and other populations.

Implications for Intentions in Daily Life and for Interventions

Women in midlife with elevated CVD risk appear to set exercise intentions on a daily basis, but rarely have success enacting these intentions. For some women, lack of success may not be salient or play a meaningful role in their daily lives; women in this age range report that PA plans are easily changed or discarded in favor of other tasks, particularly caregiving responsibilities (Im et al., 2011). For other women, however, repeated lack of success may lead to discouragement or low self-efficacy for an active lifestyle (Im et al., 2010), which could reduce the likelihood of setting future PA intentions and impede PA efforts (Barg et al. 2012). Consistent with this idea, women in the present study were less likely to set exercise intentions on later (vs. earlier) days of participation. It is possible that lack of success with enacting intentions early on led to adjustments on later days, either due to problematic discouragement or to appropriate adjustment of expectations (if initial expectations were too ambitious).

Interventions to help women in midlife set and/or strengthen intentions to be active have shown promise for increasing PA in this population, though much of this evidence relies on self-reported PA behavior. For example, relative to women who received information only, women who were guided to set intentions for exercise and to problem-solve common barriers engaged in more self-reported minutes of PA 16 weeks later (Stadler et al., 2009). Further, assigning women in midlife to set exercise intentions and detailed plans (vs. not engaging in these activities) for 2 exercise sessions per week strengthened the relation between intentions and self-reported PA behavior 8 weeks later (Arbour & Martin Ginis, 2004). Importantly, however, the same manipulation for 3 exercise sessions per week had no effect on the relation between exercise intentions and behavior. Thus, although prompting exercise intentions for this group could be a useful, low-intensity intervention to improve behavioral follow-through, the time frame and frequency of setting intentions may be critical for its effectiveness.

In the present study, women were able to set intentions for the following 3 hours at four time points during the day This approach was designed to assess the frequency with which women set intentions and their follow-through, rather than to introduce intention-setting as an intervention technique, though it is noteworthy that exercise minutes after setting versus not setting intentions did not differ. The strength of the relation between exercise intentions and behavior also did not increase over time, suggesting that the repeated act of setting and disclosing intentions did little to close the intention-behavior gap under naturalistic conditions.

Yet, the present findings may still inform improvements to interventions that promote setting PA intentions. Our observations that the extent of women’s body satisfaction and contentment moderate the relation between intentions and subsequent PA behavior may help to refine just-in-time interventions, such as those delivered via mobile devices (Schembre et al., 2018). Specifically, encouragement to set PA intentions may be most useful at times when women in midlife experience less body satisfaction than is typical for them, as this circumstance is associated with greater follow-through (relative to times when body satisfaction is higher than usual). In the same vein, encouragement to set intentions for slightly longer exercise sessions may be optimal at times when women experience less contentment than is typical for them.

This approach would require assessment of immediate affect and body image (and different prompts in response to low vs. high ratings), which is common for just-in-time interventions (Everitt et al., 2021; Hebert et al., 2018). Although such assessments can interrupt daily life, they are designed to be brief and to facilitate immediate, as-needed intervention based on contextual cues, and show high feasibility and acceptability (cf. Smyth & Heron, 2016). Consequently, they may have advantages over traditional interventions for which content is not available in moments of need (Nahum-Shani et al., 2015). Further, prompts to set PA intentions (or types of intentions) and subsequent engagement in PA during these times could have positive effects on affective state and body image (Chan et al., 2019; Hausenblas & Fallon, 2006), as well as on PA behavior, and it will be critical to identify an approach that has benefits for PA follow-through without meaningful detriment to affect or body satisfaction. An additional next step will be to experimentally test the timing and tailoring of intention-setting interventions to existing affective and body image states (with device-based assessment of PA behavior outcomes), to confirm the utility of this approach in women’s daily lives. Also important will be to learn more about women’s perceived reasons for failure to follow through on their exercise intentions. EMA could be particularly useful for this purpose, as perceptions could be captured close in time to intention-behavior lapses, limiting retrospective recall biases.

Strengths and Limitations of the Present Study

This study had several strengths, including recruitment of a sample at risk for negative outcomes that could be buffered by increases in PA (i.e., cardiovascular disease), prespecification of the study protocol (Arigo, Brown et al., 2020)), and emphasis on short-term relations between PA intentions and behavior at the within-person level. Unlike many studies of the PA intention-behavior gap, however (cf. Maher et al., 2017), intentions were assessed dichotomously, rather than using a rating scale to capture the strength of the intention. This limits our ability to make direct comparisons with existing findings, with respect to the strength of intentions. We also used an inclusive measure of MVPA (i.e., total minutes of MVPA in the following 3 hours), rather than minutes of MVPA in 10-minute bouts. The latter may be more likely to capture single episodes of exercise, and thus, could be a more accurate measure of follow-through on exercise intentions based on item wording in this study. However, instructions for the present study did not specify that MVPA intentions should be specific to bouts of 10 minutes or longer, and recent estimates of MVPA among women in midlife with elevated CVD risk indicate that their MVPA is likely to be more sporadic than 10-minute bouts would capture (Arigo et al., 2020).

It is possible that accelerometers did not accurately capture minutes spent in “strength and flexibility exercises.” These were included as an example in the exercise intention item to indicate exercise led by videos that combine cardio and muscle-strengthening activities (e.g., squats, lunges). As these activities may not be detected by accelerometers, it is possible that the observed patterns represent underestimates of behavioral follow-through. In addition, a 3-hour time frame for assessing MVPA may not fully capture women’s follow-through on their stated exercise intentions. Although this time frame roughly aligned with that specified by the intention survey item, unexpected schedule changes or delays might have led to MVPA later in the day than originally intended. In future work, it may be fruitful to examine the intended time frame and the remainder of the day following an intention report, to determine whether women are able to enact their intentions at any point following their report (vs. in the specified time frame).

Our assessments of affect and body satisfaction also offered a restricted response range (i.e., 3- and 4-point scales, respectively). Although these showed meaningful variability within-person, the restricted scales may have limited our ability to detect the true extent of within-person variability, and thereby, relevant moderating effects. As affect items also referred to emotions over the past 3 hours, the precise temporal proximity of experiencing an affective state relative to declaring exercise intentions and engaging PA behavior is not clear. (This is not the case with body satisfaction, as the assessment asked for immediate level of satisfaction.) Finally, although the present sample reflected some diversity in education level and household income, a larger proportion of women in this sample identified as White (73%) than in the population of interest (~60%; U.S. Census Bureau, 2019). Additional attention to the potential gap between PA intentions and behavior among non-White women is warranted.

In sum, employing an EMA design in this study of women in midlife with elevated CVD risk elucidates the frequency and types of exercise intentions in this population and the effectiveness of this popular motivational construct to predict behavior in the natural environment. Results suggest a strikingly large exercise intention-behavior gap in this already at-risk population, and help to identify the specific momentary states which can be targeted in exercise interventions to help close the intention-behavior gap in this group.

Highlights.

• Women in midlife with CVD risk regularly set exercise intentions in daily life

• These women achieve their exercise intentions on only 13% of occasions

• Intended exercise is weakly associated with subsequent exercise behavior

• This gap is smaller when women experience low contentment and body satisfaction