A Daily Diary Study of Rumination and Health Behaviors: Modeling Moderators and Mediators

Abstract

Objective

Rumination, thinking about a negative mood repetitively, is a common cognitive process that may affect health behavior engagement or avoidance. Little research has examined relations between rumination and health behaviors.

Purpose

We aimed to test links between rumination and health behaviors as well as possible moderators and mediators of those links.

Methods

We used an 11-day online daily diary design. Health behavior outcomes included fruit intake, vegetable intake, exercise, alcohol intake, sexual risk taking behavior, and cigarette smoking.

Results

Rumination was related to alcohol intake at the within-person level. Using multivariate modeling, we found that significant within-person mediators for rumination to health behaviors included impulsivity, amotivation, self control, and using health behaviors as coping, with each of these mediating relationships for one to four out of the five health behavior outcomes. A significant between-person moderator includes perceived behavioral control for alcohol intake only, and intention was not a significant moderator of the rumination to health behavior relationships.

Conclusions

Rumination affects various maladaptive health behaviors differentially, through a number of mechanisms and under a moderating condition whereby those who feel more control are better able to buffer rumination’s deleterious effects. Future interventions can apply the results to individual and multiple behavior change interventions for chronic disease prevention, especially for those who are particularly suffering from ruminative thoughts.

Rumination causes maladaptive health behavior patterns through impulsivity, not acting at all, and decreased self-control, and this is especially true when individuals report low perceived control.

Maladaptive lifestyle factors, such as poor diet, lack of exercise, excessive alcohol intake, cigarette smoking, and sexual risk taking, are among the biggest threats to future health of the U.S. population. Poor health behaviors are epidemic and can contribute to myriad health issues, including obesity, heart disease, diabetes, and cancer [1, 2]. A number of risk factors have been identified, and intervention programs based on them have been developed and implemented [3]. However, poor health behaviors continue to be highly prevalent and problematic [4, 5].

It is therefore important to identify additional targets for intervention for public health. Rumination, thinking about a negative mood or event passively and repetitively [6, 7], is a common cognitive process with deleterious effects on physical as well as mental health [7–10]. Brooding rumination specifically refers to this “moody pondering,” or “thinking anxiously or gloomily about” [11, 12], as opposed to reflective rumination, or a more adaptive type of reflective contemplation [12]. In this study, we examine the more maladaptive type of rumination, or brooding rumination, hereafter referred to simply as “rumination.” Rumination is a variable ripe for intervention, as it is not as pervasive as stress, and it is more pervasive than clinical targets such as depression and anxiety; it is the negative cognitive style through which stress affects mood and clinical diagnosis [13]. Rumination has also been shown to be readily malleable through interventions in previous literature [14, 15].

Rumination has been thought to affect physical health by interfering with engagement in healthy behaviors [16], but few studies have examined these proposed relationships [17]. A recent review and meta-analysis that summarized this small but growing literature indicated that increases in rumination were related to health risk behaviors [17]. Even fewer studies have posited why or how rumination may affect health or health behaviors. No studies, to our knowledge, have empirically tested possible mechanisms of action between rumination and health behaviors or the conditions under which this relationship occurs. In the present study, we aim to develop a comprehensive model of rumination’s impact on health behaviors, to make recommendations for interventions. Specifically, this study examined possible mediators and moderators of the relationship between rumination and health behaviors. Hypothesized mediators include two behavioral pathways: impulsivity and amotivation, as well as two coping/regulation pathways: self-control and using health behaviors as coping. Hypothesized moderators include intention and perceived behavioral control.

Mechanisms Through Which Rumination May Influence Health Behaviors

Impulsivity

Rumination may lead to more maladaptive health behavior or less adaptive health behavior by causing an individual to act too quickly. If a person is stuck in his or her ruminative thoughts, he or she may put off making decisions about health behaviors and may need to make spontaneous decisions as a result. For example, if someone is ruminating, he or she may put off health related decisions (e.g., to go to the gym, to cook a healthy meal, to plan for a healthy snack) and may have less time as other deadlines approach to engage in healthy behaviors (e.g., run out of time to be able to go to the gym or cook a healthy meal [and need to grab junk or packaged food for a snack or meal instead], or need to make a snap decision about attending a party). In other words, if someone is ruminating, he or she may have a more full cognitive load, and be able to dedicate fewer resources for thoughtful or planful decisions about adaptive health behavior engagement. Rumination has been linked to impulsivity [12, 18], and impulsivity has been linked to maladaptive health behaviors [19, 20]. Impulsivity on its own may lead individuals to make spontaneous decisions that are at odds with thoughtful planning that is usually required for adaptive health behavior engagement. For example, impulsivity may lead one to eat junk food instead of cooking a healthy meal, or to decide to partake in another activity (e.g., a party, a nap) instead of going to the gym for exercise. For example, in studies of smokers, impulsivity mediated associations between rumination and depression [21] and moderated associations between depressive rumination and number of failures to quit [22]. However, no studies have yet tested whether impulsivity mediates relations between rumination and health behaviors.

Amotivation

Alternately, rumination can sap individuals’ motivation and initiative. Rumination maintains one’s focus on depressive or sad thoughts, which may distract individuals from their desire to engage in constructive behavior [23]. Rumination may lead to anhedonia and hopelessness in the same way depression does: Sad and repetitive thoughts focused on a negative event can lead to a more negative perspective of other events, perhaps perceiving them as out of the person’s control or hopeless to change. Several studies suggest that people who focus on negative feelings show reduced motivation. For example, one study asked students to report their biggest problems and then come up with solutions to them; trait ruminators were less likely than nonruminators to follow through with their planned solution [24]. In another study, ruminators reported believing that pleasant activities would lift their mood, yet they did not engage in them [25]. Another study found that women with breast cancer who had a tendency to ruminate reported having delayed a doctor’s appointment, and therefore diagnosis of cancer, for an average of 2 months longer than nonruminators [26]. A lack of motivation as a result of rumination is distinct from intention to engage in specific health behaviors. Intention implies a specific plan, while motivation speaks to a general desire to engage in the health behavior [27]. Motivation and amotivation have not yet been studied as mediators between rumination and behaviors.

Self-control

While motivation captures the desire to engage in a behavior, self-control captures the perceived ability to override incipient responses and regulate one’s behavior toward one’s goals. In the face of the above challenge of amotivation, self-control may provide additional willpower to overcome temptation (to either not do something or do something; [28]). Even in the face of depressive ruminative thought, self-control may provide an added buffer to shifting locus of control from internal to external during times of rumination. Low self-control has been shown to mediate the rumination–aggressive behavior relationship [10], and it is plausible that it may mediate the relationship between rumination and health behaviors.

Health Behaviors as Coping

Finally, there is some evidence that demonstrates that maladaptive health behaviors may also result from efforts to distract oneself from one’s distress or cope with the distress of the repetitive negative thoughts involved in ruminative processes [29]. Previous studies have found that rumination is positively related to avoidance coping, the tendency to avoid one’s mood through reckless behaviors such as excessive alcohol consumption [30, 31]. Maladaptive health behaviors often offer an opportunity to escape one’s ruminative thoughts, albeit briefly. In a behavioral activation framework, individuals often cope with depressive, sad, or ruminative thoughts with coping mechanisms that work in the short term, but not in the long term, like laying on the couch or eating junk food [32]. This relationship of health behaviors as coping as a mediator of rumination to maladaptive health behavior patterns has not thus far been tested empirically, so we tested the use of health behaviors to cope as a mediator in the relationship between rumination and health behavior [33].

Moderating Factors of the Rumination–Health Behavior Link

Intention

According to the theory of planned behavior, intention is the most important predictor of behavior [8]. In addition, theoretically, if we are hypothesizing that rumination causes a person to be impulsive or not act at all with health behaviors, that it interferes with what one would otherwise do, it is important to know about participants’ intentions. For example, if a person eats unhealthily, but has no intention to eat healthfully, there would not be a process for rumination to disrupt. However, if a person eats unhealthily, but had intended to eat healthily, rumination may have been involved in that person’s not following through on that intended behavior, through various mechanisms. It is therefore reasonable to posit that the relationship between rumination and health behaviors is strengthened, or only exists, when intention is present. This hypothesis was borne out in a pilot study conducted in 2013, in which rumination only had minimal bivariate correlations with health behaviors, but when intention was included as a moderator in linear regressions, more relationships between rumination and health behaviors emerged [33].

In addition, because intention is an important theoretical moderator of the rumination to health behavior relationship, such that there would likely not be a relationship without it, we included intention as a moderator for the mediation relationships as well. Intention is proposed to moderate the rumination to mediator relationship such that the rumination to mediator (impulsivity, amotivation, low self-control, using health behaviors as coping) link will be stronger should there be intention to engage in an adaptive health behavior and less strong should there be no intention to engage in maladaptive health behaviors. In addition, intention is proposed to strengthen the relationship between the mediator and the outcome, such that without intention to engage in the health behavior, there is no process for the mediator (e.g., impulsivity) to affect health behavior outcomes.

Perceived Behavioral Control

Perceived behavioral control refers to an individual’s perceived ease or difficulty in performing a particular behavior [34]. Perceived behavioral control predicts engagement in a variety of health behaviors [8]. This relationship is not as strong as the relationship between intention and behavior, as intention has been shown to mediate the relationship between perceived control and behavior with indirect effects [8] but is still important to measure as a potential moderator in the relationship between rumination and health behaviors.

Study Overview

Because health behaviors and rumination are ongoing daily processes and the measurement of each on a daily basis has been shown to be useful [35, 36], we conducted a daily diary study. A daily diary design allowed us to examine whether on days when people ruminate more, they may engage in less healthy behaviors, and further, to examine mediators and moderators of this relationship. This study hypothesizes that (H1) rumination will be negatively related to adaptive health behaviors (fruit intake, vegetable intake, and exercise) and positively related to maladaptive health behaviors (cigarette smoking, alcohol intake, and sexual risk taking) in bivariate correlations as well as in multilevel models at both the within-person and between-person levels; (H2) rumination will be associated with more maladaptive and less adaptive health behaviors through four pathways (mediators): impulsivity, amotivation, low self-control, and health behaviors as coping motivation; and (H3) both (trait) perceived behavioral control and daily intention will moderate the rumination–health behavior links.

Method

Participants

285 participants (mean age = 19.3; 76.8% female; 79.4% Caucasian, 6.3% Black/African American, 4.2% Asian, and 3.1% “Other,” with 9.3% identifying as Hispanic/Latino) were recruited via participant pool at a large Northeastern university. Participants were compensated with credit for an introductory psychology course.

Procedure

Participants completed online questionnaires using the Qualtrics online survey software. Students completed a battery of questionnaires at baseline, followed by 11 days of 5-min nightly diary assessments. This study was approved by the institutional review board. A certificate of confidentiality was obtained from National Institute on Alcohol Abuse and Alcoholism to protect underage participants’ report of alcohol intake. All participants were provided with contact information for mental health resources.

Measures

Baseline

The following measures were collected once on Day 1 of the study. Reliability estimates include inter-rater reliability (α) and an estimate of total factor saturation (Ω) to using a confirmatory factor analysis approach in R [37, 38], as alpha coefficients can often be flawed, especially for psychological scales [39].

Demographics

We asked the participants to identify their age, year in school, gender, race, and ethnicity with five questions.

Perceived behavioral control

Behavioral control of health behaviors was measured using the Health Specific Self Efficacy Scales (Nutrition Self-Efficacy Scale and Physical Exercise Self Efficacy Scale; [28]). Internal consistency reliability in the present sample was good (Ω = .69).

Daily diary

The following measures were collected on each of 11 days; alphas reported refer to averages across daily measures.

Fruit and vegetable intake

Daily diet was assessed using the Dietary Screener Questionnaire (DSQ) from National Cancer Institute. Questions were modified to inquire for daily intake, a practice used often [40]. Scoring algorithms produce a single number each for fruit and vegetable intake. The DSQ has been used in large-scale and more focused studies and has been shown to have reasonable validity.

Exercise

Exercise was assessed by the Godin Leisure-Time Exercise Questionnaire (LTEQ), a four-item measure that assesses leisure-time exercise [41]. The LTEQ assesses frequency per day of strenuous (heart beats rapidly), moderate (not exhausting), and mild (minimal effort) exercise practiced for at least 15 min. A composite score was used for our daily exercise index [41].

Sexual risk taking

Participants reported occasions on which they had unprotected sex with a monogamous partner (e.g., sex without protection against sexually transmitted diseases [STDs] or pregnancy with an exclusive dating partner) and a nonmonogamous partner (e.g., sex without protection against STDs or pregnancy with a nonexclusive dating partner) [42]. Participants completed both items on 7-point scales from 0 times to 6+ times. These items have been used and validated in college student samples [42].

Alcohol

Participants reported total alcohol consumption in the previous 24 hr. One drink was defined as one 12-oz bottle of beer, one 4-oz glass of wine, one 12-oz bottle of wine cooler, or 1-oz of liquor straight or in a mixed drink. This measure has been successfully used in college student samples [43].

Cigarettes

Participants reported how many cigarettes they had smoked in the past 24 hr, a common method of assessing cigarette smoking [44].

Rumination

The five-item brooding subscale [36] of the Ruminative Styles Questionnaire (RRS; [12, 31]) was adapted for individuals to report their daily levels of ruminative thought. Items include, “Think ‘What am I doing to deserve this?’” “Think ‘Why do I always react this way?’” “Think about a recent situation, wishing it had gone better,” “Think ‘Why do I have problems that other people don’t have?’” “Think ‘Why can’t I handle things better?’” on a 4-point Likert scale from “Not at All” to “Constantly”. Internal consistency reliability in the present sample was good (Ω = .81).

Impulsivity

Impulsivity was measured with the eight-item Barratt Impulsiveness Scale, brief version (BIS-11; [45, 46]). The BIS-11 has been shown to be reliable and valid in studies of college students [37] and has been adapted for use as a state measure, the directions being adapted from Stanford and colleagues (2009) [47]. Internal consistency reliability in this study was good (α = .82).

Amotivation

The amotivation subscale from the Global Motivation Scale (GMS-28; [48]) was used to assess daily lack of motivation about health behaviors. This scale has been shown to have good reliability and validity. Internal consistency reliability in this study was acceptable (Ω = .76).

Self-control

Self-control was assessed with the Brief Self-Control Measure, a 13-item scale that measures ability to override incipient responses and regulate behavior. It has demonstrated reliability and validity [25]. Internal consistency reliability was only fair in this study (Ω = .84).

Health behaviors as coping

These were assessed by health behaviors as coping items [49]. Namely, we took the two highest loading items from each of the four health subscales (exercise, eating, self-care, smoking). These items have been validated and were used to assess using health behaviors to cope with distress from ruminative thought. Internal consistency reliability in this sample was fair (Ω = .69).

Intention to be healthy

We asked questions for each health behavior for the next day (i.e., How much do you intend to engage in ____ tomorrow?) based on Ajzen’s [50] recommendations for asking about health behavior intention on a daily level.

Power Analysis

We used the software Optimal Design to determine the balance of days to participants based on expected effect sizes for outcome variables [51]. This study chose to sample over the course of one work week and two weekends, when health behaviors vary more [35, 52–54], balancing the number of days recommended for intensive longitudinal modeling [55], participant pool credits allotted for this study, and sufficient power to power the study at 80%, yielding 11 observations. With 11 observations, we required a minimum of 250 participants to power at 80%.

Data Analysis Plan

We examined normality of data and made appropriate adjustments in preparation for multilevel modeling, including considering whether categorization of outcomes as continuous, count, or categorical was appropriate; more detail about this procedure is provided in the Results section below. This study followed steps to ensure conforming to all best standards and practices for intensive longitudinal modeling [51, 55]. Each predictor variable was group mean centered. Between-subject components of predictor variables were created and included in all models using means of daily items per subject. We plotted multilevel moderation effects using recommendations from previous research [55–57]. All main models tested used Mplus to estimate multilevel models [58]. Mplus uses an Structural Equation Modeling (SEM) approach and takes into account the types outcomes specified in its modeling. We modeled count variables using Poisson distribution and using negative binomial (NB) regression for any variables that exhibited a zero-inflation pattern after examining variables with frequency distributions [58–60]. Item- and construct-level intraclass correlations (ICCs) were also examined prior to data analysis. Item-level ICC values and construct level ICC values (.20–.82) indicated support for multilevel analysis. We ran separate mediation models for each health behavior outcome, with four proposed mediators and five outcome variables, so 20 mediation models. We ran moderation models including all five outcome variables, with four proposed moderators, so four moderation models. We provide only one figure (conceptual) and one table of results in the main paper (Multilevel model of daily rumination to daily health behavior outcomes over time) but we include many additional tables and figures in the supplemental material (see Supplementary Tables 1–32 and Supplementary Figures 1–9 for additional information on all results)

Results

Health Behavior Frequencies

On average, per day, participants reported exercising for 23.11 min (SD = 30.38), eating 1.18 fruits (SD = 1.08) and 0.75 vegetables (SD = 0.93), drinking .81 alcoholic beverages (SD = 2.22), and engaging in 0.06 incidents of sexual risk taking (SD = 0.24). Only eight participants reported any smoking incident over the entire study, for a total of 60 observations, or 1.9% of all daily observations (11 for 285 participants = 3,135 observations). The rate of smoking was so low that this outcome variable was excluded from further analysis.

Normality and Variable Categorization

We examined normality of data and made appropriate adjustments in preparation for multilevel modeling, including categorization of outcomes as continuous (exercise), count (fruit and vegetable, alcohol), or categorical (sexual risk taking). For mediation analyses, to more easily interpret effects, we treated fruit and vegetable intake as continuous outcomes as well. In hierarchical linear modeling, it is important to examine outcome data distribution carefully to model associations appropriately, increasing the accuracy and interpretability of results [55]. As such, consideration of continuous variables as count or categorical (dichotomous) was conducted thoughtfully according to observed data frequencies, distributions, skewness, and kurtosis. Specifically, the scored exercise variable was a continuous outcome, with variation in responses; the scored fruit and vegetable outcomes were small integers, as per a count variable [55]. Alcohol was also reported in an integer, count format (e.g., 0–7 drinks per day), and exhibited zero-inflation, so were modeled using NB modeling [59, 60]. Sexual risk taking was a low frequency and not-often-endorsed variable that fell into a categorical dichotomous (yes or no) pattern.

Missing Data

No student withdrew from this study prior to its completion. Of valid responses, meaning of days on which participants did fill out a survey, missing data was fairly low: Missing data at baseline for study variables varied from 1.0% to 4.2%, and missing data on the daily variables varied from 1.7% to 3.0%. On average, participants completed 9.72 out of 11 days of daily diary responses, or 88.3% of daily assessments, a typical amount for daily diary completion [61].

Missing data were accounted for by full information maximum likelihood estimation. Although most unplanned “missingness” in psychosocial research is at some level missing not at random, we sought to reduce bias produced by this mechanism by including variables associated with missingness in our models; parameter estimates are valid under the assumption of data missing at random [55, 62]. Specifically, all missing data were marked as missing. All models were run with MPLUS, which uses maximum likelihood estimation [58].

Multilevel Modeling Results

Relationship with time

Modeling each health behavior variable with time within- and between-persons, there was no significant change. Fruit decreased at a trending significance level (p = .081).

Rumination and health behaviors

Primary hypothesis 1 (H1), that rumination will be related to health behaviors at the within- and between-person levels, was tested with multilevel modeling taking into account time. Within-person, daily rumination was significantly related to alcohol intake (B =.252, p = .042) but was not significantly related to fruit intake (B = −.085, p = .084), vegetable intake (B = −.060, p = .092), exercise (B = −.379, p = .806), or sexual risk taking (B = .181, p = .382; see Table 1). At the between-person level, daily rumination was significantly related to vegetable intake (B = −.092, p = .053) and alcohol intake (B = .252, p = .041), but not related to fruit intake (B = −.182, p = .053), sexual risk taking (B = .122, p = .092), and exercise (B = −.339, p = .112; see Table 1).

Table 1.

Multilevel model of daily rumination to daily health behavior outcomes over time

	Estimate	SE	t a	p b	CI95
					Lower	Upper
Within-person level
Fruit intake on
Rumination (within)	−0.085	0.051	1.673	.084	−0.015	0.185
Time	0.003	0.006	0.454	.650	−0.009	0.015
Vegetable intake on
Rumination (within)	−0.060	0.039	1.546	.092	−0.040	0.136
Time	−0.015	0.005	−3.010	.003	−0.026	−0.005
Exercise on
Rumination (within)	−0.379	1.540	−0.246	.806	−3.398	2.640
Time	−0.494	0.164	−3.014	.003	−0.816	−0.173
Alcohol on
Rumination (within)	0.252	0.140	1.796	.062	−0.225	0.527
Time	0.003	0.015	0.214	.830	−0.073	0.032
Sexual Risk on
Rumination (within)	0.181	0.207	0.874	.382	−0.255	0.588
Time	−0.008	0.025	−0.315	.753	−0.057	0.042
Between level
Fruit intake on
Rumination (between)	−0.182	0.062	2.342	.053	−0.015	0.185
Vegetable intake on
Rumination (between)	−0.092	0.041	2.146	.047	−0.040	0.136
Exercise on
Rumination (between)	−0.339	1.540	−0.246	.112	−0.498	1.640
Alcohol on
Rumination (between)	0.252	0.140	1.796	.041	−0.125	0.927
Sexual risk on
Rumination (between)	0.122	0.97	1.3744	.092	−0.127	0.238
Intercept (within)
Fruit	1.171	0.057	20.475	<.001	−0.015	0.121
Vegetable	0.830	0.043	19.474	<.001	−0.016	0.136
Exercise	25.629	1.603	15.992	<.001	−3.398	2.640
Alcohol	1.543	0.221	13.782	<.001	−0.023	0.527
Sexual risk	2.682	0.177	15.149	<.001	−0.225	0.588

Mediation

Mediation models were conducted with intention as a moderator for both the path from rumination to the mediator and the mediator to the outcome. If moderation terms were not significant, we reran the model without the moderator included, as simpler mediation models. These models test the second hypothesis (H2), that rumination will relate to health behaviors through impulsivity, amotivation, self-control, and health behaviors as coping.

Impulsivity

Impulsivity significantly mediated the rumination to alcohol pathway (alcohol on rumination B = .500, p = .049; alcohol on impulsivity B = .189, p = .038; impulsivity on rumination B = −.309, p = <.001) and rumination to sexual risk taking pathway (sexual risk on rumination B = .763, p = .018; sexual risk on impulsivity B = .203, p = .011; impulsivity on rumination B = .302, p = .003), with intention as a significant moderator in each path from rumination to impulsivity in these models. Impulsivity mediated relationships between rumination and alcohol, and sexual risk taking with indirect effects: 11% of the average relationship between rumination and alcohol can be explained by impulsivity, and 23% of the average relationship between rumination and sexual risk taking can be explained by impulsivity. Impulsivity did not mediate the rumination to fruit intake, vegetable intake, or exercise relationships. Intention was not a significant moderator in any model, so simple mediation models are presented.

Amotivation

We examined amotivation measured at the daily within-person level. Daily amotivation was only a significant mediator in the rumination to exercise relationship (exercise on rumination B = −.327, p = .044; exercise on amotivation B = −.919, p = .033; amotivation on rumination B = .182, p = .042). At the within-person level, for each unit of rumination, amotivation increased 18 units (SE = .031, z = −.274, p = .082). Every unit of amotivation decreased exercise by 13.72 min. Exercise did not appear to have a direct relationship with rumination, indicating mediation. c = c′ + ab + o_ajbj = −.32 + (−.18 × .92) + .03 = .456. Given these results, we can calculate that 56% of the average relationship between rumination and exercise can be explained by amotivation. Daily amotivation was not a significant mediator of the rumination to fruit, vegetable, alcohol, or sexual risk taking models. At the between-person level, average amotivation was significantly related to rumination in the fruit (B = .181, p = .044), vegetable (B = .204, p = .038), and sexual risk taking models (B = −.225, p = .027), but amotivation was not related to any health behavior outcome at the daily level. Intention did not moderate any relationship, so simple mediation models are presented.

Self-control

At the within-person level, while rumination was related to self-control, self-control was only significantly related to vegetable intake. Self-control was therefore a mediator of rumination to vegetable intake (vegetable intake on rumination B = −.127, p = .024; vegetable intake on self control B = .082, p = .036; self control on rumination B = −.582, p = .012). Each unit of rumination increase decreased vegetable intake by 0.74 vegetables. Self control was not a mediator of the rumination to fruit, exercise, alcohol, or sexual risk taking relationships. Intention did not moderate any relationship, so simple mediation models are presented.

Health behaviors as coping

Using health behaviors as coping was a significant mediator between rumination and all health behaviors besides sexual risk taking, where there was a significant path from rumination to using health behaviors as coping (B = −.019, p = .046), but not from health behaviors as coping to sexual risk taking (B = −.560, p = .160). Intention as a moderator was significant in the rumination to health behaviors as coping path for the exercise model and for health behaviors as coping to vegetable intake in that model, so these two were estimated as moderated mediation models.

Moderation

Moderation models test the third hypothesis (H3), that rumination will affect health behaviors variably based on presence or absence of intention and perceived behavioral control.

Intention

Intention was not a significant moderator of the rumination–health behavior relationship for any health behavior. For fruit intake, vegetable intake, and exercise, such that there was significance for intention and the interaction term, but not rumination, at both the within- and between-person levels for these variables.

Perceived behavioral control

Perceived behavioral control was examined as a between-person moderator. At the within-person level, perceived behavioral control significantly moderated only alcohol intake, such that perceived behavioral control buffered the relationship between rumination and alcohol intake (B = −.682, p = .011). At the between-person level, perceived behavioral control buffered the relationship between rumination and fruit intake (B = .229, p = .003).

Discussion

Although rumination is a common cognitive process, it is widely unstudied as a possible cognitive predictor of maladaptive health behavior patterns. Understanding the possible mediators and moderators of rumination via this comprehensive modeling will allow us to develop and improve interventions for health behavior change for increased public health.

Rumination and Health Behaviors

First, this study hypothesized that rumination will be related to health behaviors (H1). This hypothesis was only partially supported. Multilevel modeling showed that, on the within-person level, rumination was only significantly related to alcohol intake, but not fruit intake, vegetable intake, exercise, or sexual risk taking. At the between-person level, rumination was related to all health behaviors except exercise. However, given the marginal significance level of rumination’s impact on the other health behaviors at the within-person level (e.g., exercise: p = .089), and moderately strong negative effect observed (e.g., exercise: B = −.339), perhaps impulsivity, motivation, and using health behaviors as coping, all significant mediators, completely mediated some relationships between rumination and health behaviors. Further exploration is warranted. Furthermore, the frequency of sexual risk taking on a daily level is low, so it is possible that change is not detectable on the daily level due to a floor effect. Perhaps additional reported sexual risk taking, sampling over a longer period of time, or with an ecological momentary assessment (EMA) method whereby a person reports only when a sexual risk taking incident occurs, would yield more incidents of sexual risk taking to model statistically. Or, perhaps rumination is truly not related to sexual risk taking on the daily level, within person at all, and is more appropriately measured at the between-person level.

Mediation

The second hypothesis (H2), that rumination will be associated with more maladaptive and less adaptive health behaviors through four pathways (mediators), impulsivity, low motivation, low self-control, and using health behaviors as coping, was generally supported (see Fig. 1).

Fig. 1.

Rumination and health behaviors model.

Impulsivity

Impulsivity was a significant mediator of the rumination to alcohol and sexual risk taking relationships, but not for the rumination to fruit or vegetable intake or exercise relationships. Indeed, impulsivity may not affect eating fruits and vegetables but rather may be more likely related to ingestion of unhealthy foods. Eating unhealthy foods is not necessarily related to not eating fruits and vegetables [49]. Future research should explore the relationship between rumination, impulsivity, and other types of food intake, such as junk food or comfort food.

The negative effects of excessive alcohol intake, including death, assault, sexual assault, injury, academic problems, and physical and mental health issues, necessitate interventions on college campuses [63]. The strong relationship between alcohol use and impulsivity has been widely documented [64]. However, researchers have noted that the etiological nature of this relationship remains unclear [64, 65]. The current study, which demonstrated that a cognitive process (rumination) was antecedent to this impulsivity to alcohol relationship, provides insight into one of the processes responsible for this connection, and suggests an additional area for interventions to target for decreasing excessive alcohol intake.

A broad literature already links impulsivity to sexual risk taking [66], a construct also strongly associated with alcohol intake; in fact, a recent report was published by the Centers for Disease Control and Prevention [63]. Hoyle, Fejfar, and Miller [66] called for increased study of factors associated with the impulsivity to sexual risk taking relationship. The present results contribute an additional construct that warrants future study.

While rumination, to our knowledge, has never been studied with regard to sexual risk taking behavior, the present results extend the current literature linking “hot thoughts,” or a lack of awareness due to exciting, stimulating positive thoughts, a possible proxy of rumination, to sexual risk taking [67]. Rumination could be an important unstudied precursor to sexual risk taking, or it could be one of the primary paths between depression and sexual risk taking [68].

Amotivation

Surprisingly, daily amotivation was only a mediator for the rumination to exercise relationship. This may be because the daily measure used, four items from the GMS-28, tapped a general notion that healthy behaviors do not matter, or some negative affect, general pessimism, or even depression, rather than feeling unmotivated to act because of ruminative thoughts. The mean responses to these items were low, ranging from 1.3 (SD = .24) to 1.7 (SD = .42). This may have meant that there were floor effects (i.e., not enough variance in endorsement of these items to detect significant relationships). These items may not capture the construct we intended to measure; in the future, researchers should ask about lack of motivation regarding health behaviors specifically in response to rumination, using a different measure.

That motivation is related to exercise is not surprising; much previous literature has shown a robust relationship between motivation and exercise [69]. Exercise may require more motivation than do other health behaviors studied because there are a number of specific barriers that one is often required to overcome for exercise, as opposed to fruit intake, vegetable intake, alcohol intake, and sexual risk taking.

We theorized that not acting at all in response to rumination, because one feels too preoccupied by the rumination to act on anything else, is an important construct. The construct of motivation may not be the best proxy for that inability to act. It is possible that in the future researchers may have to develop a measure specific to this effect, which may be important for better examining relationships between cognitive constructs and health behaviors.

Self-control

Self-control was a mediator for vegetable intake only, in an indirect effects relationship at the within-person level; it was anticipated that self-control would relate to other health behaviors, especially exercise, as motivation and exercise are highly correlated and self-control and motivation are related constructs [70]. For example, in one study of college students, depletion, having used self-control in other situations recently, combined with a lack of motivation, predicted self-control failures [70]. More studies should examine this combined effect of self-control and motivation. It is surprising that self-control did not mediate the rumination to fruit relationship even though it mediated the rumination to vegetable relationship. One potential explanation for this differential relationship is the greater variance in vegetable intake and thus an increased ability for this study to detect relationships with more variability. Alternatively, eating vegetables may require more self-control than does eating fruit.

Health behaviors as coping

Daily use of health behaviors as coping mediated the rumination to health behaviors link at the within-person level, besides the rumination to sexual risk taking relationship. This measure of coping is unique because it measures the widely acknowledged use of maladaptive health behaviors to cope with negative affect or stress. However, this measure does not measure using sexual risk taking to cope with stress; it could and should be adapted to include sexual risk taking items. These findings also demonstrate that, while alcohol intake and sexual risk taking tend to be highly correlated, they do have some differential predictors. Developing interventions that focus on teaching new coping strategies to cope with rumination, as well as practice implementing these strategies in lieu of engaging in maladaptive health behaviors, will be important, due to the strong mediation effect observed here.

Moderation

Third, the hypothesis (H3) that the relationship between rumination and health behaviors would be moderated by intention (increase) and perceived behavioral control (decrease) was mostly supported. Intention strengthened the rumination to health behavior relationships for fruit intake, vegetable intake, and exercise, and perceived behavioral control did moderate significantly in the direction expected, but only for alcohol (see Fig. 1).

Intention

At the within-person level, intention was not a significant moderator of rumination to any health behavior relationship. For most models, intention independently predicted health behaviors, meaning it is likely still an important piece of predicting health behaviors. Intention appeared to moderate only a few mediation paths. Therefore, contrary to our prediction, intention may not be important to include as a moderator for mediation paths of the rumination to health behavior associations.

Perceived behavioral control

Within-person, perceived behavioral control moderated the rumination to alcohol relationship. It appears that, within person, increased perceived behavioral control may limit the negative impact of rumination on excessive drinking. This is an important finding for college campuses [63]. Glanz and Bishop [71] suggest focusing on coping strategies and behavior change that increase perceived behavioral control on college campuses. However, on average, between-person, perceived behavioral control appears to increase the maladaptive relationship between increased rumination and decreased fruit intake. This finding again speaks to the potential drawbacks of a mostly adaptive variable; perceived behavioral control may backfire if individuals believe they have so much control that they do not need to prepare for any potentially difficult situations or barriers to health behavior engagement or lack thereof, an important consideration in health behavior change [72].

Limitations

Study limitations must be acknowledged. Our sample may not be generalizable to the broader college student or young adult population. However, health behavior frequencies in this study were similar to those that have been found in previous studies of university students’ health behaviors [73]. Methodologically, though multilevel modeling of estimated effects between variables suggests causality, this study design is still primarily correlational [74]. Further insights might be gained by examining longer lagged effects between predictor and outcome variables. Finally, measuring health behaviors at the day’s end, relying on recall and accurate reporting throughout the day, can be somewhat problematic. While asking about health behaviors for the previous 24 hr at day’s end is preferable to even longer recall periods, as is common in many other types of studies, data show that people tend to not estimate their day’s activities and intake particularly accurately, due to individual bias, poor estimation, or poor memory of the day’s events [75]. Technology that measures behaviors objectively, such as accelerometers that measure movement as well as wearable cameras that measure all activities will be useful in obtaining the most valid data [76].

Future Directions and Clinical Implications

Specific recommendations regarding future research include measuring change in multiple behaviors to design the most efficacious interventions and developing a measure for the inability to act/amotivation. In addition, because impulsivity, low motivation, low self-control, and coping mediate the relationship between rumination and health behaviors, future research should examine what causes rumination to lead to each of these mechanisms.

As perceived behavioral control was implicated as a moderator in this study but intention was not, a more in-depth and comprehensive study of the theory of planned behavior and its associated variables (i.e., adding attitude and subjective norm) in the context of rumination and health behaviors would be a good addition to the literature.

Finally, an aim of this research is to inform health behavior change interventions. Interventions that target rumination could lead to multiple behavior changes. Mindfulness interventions, behavioral activation, and cognitive behavioral therapy have been shown to decrease rumination [15, 77], with wide support and evidence for strong and various adaptive outcomes [78], making them good candidates for widespread dissemination. Watkins and colleagues [77] recently developed an intervention called rumination-focused cognitive-behavioral therapy, a manualized treatment showing promise for decreasing rumination. The treatment, which consists of up to 12 individual weekly sessions, is designed to coach individuals to shift from unconstructive rumination to constructive rumination through the use of functional analyses, experiential/imagery exercises, and behavioral experiments. Developing interventions aimed at identified moderation and mediation relationships may also be useful for multiple behavior change interventions. The mediators that seem to have a significant salutary effect on the greatest number of health behaviors are impulsivity and using health behaviors as coping. Impulsivity can be decreased with cognitive behavioral therapy [79], and using health behaviors as coping could be decreased through interventions focusing on teaching coping skills, through a stress management program such as Cognitive Behavioral Stress Management [80]. Mind-body practices such as yoga and mindfulness meditation, which are growing in popularity, may have impacts on decreasing impulsivity, increasing coping skills, and decreasing rumination [81, 82]. More research about maximally efficacious interventions targeting rumination for multiple behavior change, including dissemination to those in particular need, is warranted.

Concluding Remarks

Rumination, an understudied construct, has clear relationships to health behavior engagement, and the moderators and mediators of this relationship studied here present unique new combinations of predictors of health behavior engagement and avoidance. More studies should be conducted to further explicate these relationships and to develop appropriate interventions aimed at decreasing rumination in college students, or decreasing rumination’s effect on health behaviors. This approach holds further promise for broader populations, pending future research.

Compliance With Ethical Standards

Primary Data These findings have not been previously published, and this manuscript is not being simultaneously submitted elsewhere. There has been no previous reporting of this data. The authors have full control of all primary data and agree to allow the journal to review the data if requested.

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards There are no actual or potential conflicts of interest to report. The authors hereby adhere to all ethical standards.

Authors’ Contributions All authors earned authorship as per the definitions outlined by this journal.

Ethical Approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Informed Consent All participants were provided an information sheet for informed consent. These procedures were approved by the University of Connecticut IRB.