Annals of Behavioral Medicine Volume 30 issue 1 2005 [doi 10.1207 s15324796abm3001 4] Barbara A. Stetson; Abbie O. Beacham; Stephen J. Frommelt; Kerri Exercise slips in high risk situations and a

Exercise Slips in High-Risk Situations and Activity Patterns in Long-Term Exercisers: An Application of the Relapse Prevention Model

Barbara A. Stetson, Ph.D. and Abbie O. Beacham, Ph.D.

Department of Psychological and Brain Sciences University of Louisville

Stephen J. Frommelt, Ph.D.

Department of Physical Medicine and Rehabilitation St. Luke’s Medical Center

Kerri N. Boutelle, Ph.D.

Department of Pediatrics University of Minnesota

Jonathan D. Cole, Ph.D.

The Pain Treatment Center Lexington, Kentucky

Craig H. Ziegler, M.S.

Department of Bioinformatics and Biostatistics University of Louisville

Stephen W. Looney, Ph.D.

Department of Biostatistics Louisiana State University Health Sciences Center

ABSTRACT INTRODUCTION

Background: Key factors in successful long-term exercise Physical and psychological health benefits of regular exer- maintenance are not well understood. The Relapse Prevention

cise are numerous (1,2). On average, physically active people Model (RPM) may provide a framework for this process. Pur-

outlive those who are inactive (3,4). Despite these compelling pose: The purpose of this study was to examine the relationships

data, rates of regular exercise among healthy adults remain as- among characteristics of exercise high-risk situations, compo-

tonishingly low, with 25% of adults engaging in no leisure-time nents of the RPM relevant to exercise slips, and follow-up exercise

physical activity (2). It is estimated that only about 11% of outcomes in long-term community exercisers. Methods: We ob-

healthy adults engage in moderate-to-vigorous, purposeful ac- tained long-term exercisers’ (N = 65) open-ended responses to

tivity 3 or more days per week. Rates of participation in “life- high-risk situations and ratings of obstacle self-efficacy, guilt,

style-based” activity, accumulating 30 min of activity 5 or more and perceived control. High-risk situation characteristics, cogni-

days per week are higher but still remain below one fourth of tive and behavioral coping strategies, and exercise outcomes

healthy adults (5).

were examined. Results: High-risk situation characteristics in- cluded bad weather, inconvenient time of day, being alone, nega-

EXERCISE LAPSE AND RELAPSE

tive emotions, and fatigue. Being alone was associated with lower

A great deal of emphasis has been placed on encourag- incidence of exercise slip. Positive cognitive coping strategies

ing initiation of physical exercise. However, successful main- were most commonly employed and were associated with positive

tenance of exercise behavior once it has been initiated has not exercise outcome for both women and men. Guilt and perceived

been studied extensively. What constitutes successful “mainte- control regarding the high-risk situation were associated with ex-

nance” remains undefined. It has become generally accepted ercise outcomes at follow-up, but only among the men (n = 28).

that one enters maintenance after engaging in regular physi- Conclusions: Findings confirm and extend previous work in the

cal activity for a minimum of 6 months after an intervention application of the RPM in examining exercise slips and relapse.

has been completed or the behavior change has been initiated Measurement issues and integration approaches from the study of

independent of a formal intervention (6). The vast majority of relapse in addiction research are discussed.

research in this area has focused on the study of successful ex- ercise initiation and adherence as the cornerstone for main-

(Ann Behav Med

tenance. Predictors of continued maintenance among long-term exercisers have not been definitively identified (7).

Reprint Address: B. Stetson, Ph.D., Department of Psychological and Similar to the management of weight over time, the mainte- Brain Sciences, 317 Life Sciences Building, University of Louisville,

nance of exercise becomes a lifelong process. Lapses or “drop Louisville, KY 40292. E-mail: [email protected]

out” in exercise routines are quite common and are considered © 2005 by The Society of Behavioral Medicine.

to be more the rule than the exception, reaching as high as 50%

26 Stetson et al. Annals of Behavioral Medicine

dropout in some populations (8). Sallis et al. (9) suggested that a In the exercise literature to date, coping in the face of obsta- large proportion of the population may have previously initiated

cles to exercise has not been examined with regard to individu- exercise but subsequently relapsed (no exercise for ≥ 3 months)

als’ exercise history or gender. In the general coping literature, at some later time. In this community-based study of patterns of

gender differences have been examined. A meta-analysis of em- lifetime history of relapse from exercise, 20% of regular exercis-

pirical studies evaluating coping behaviors in the context of spe- ers experienced a previous relapse three or more times, 20% re-

cific situations (18) suggested that women had higher rates of ported 1 to 2 relapses, and 60% reported no history of relapse.

engaging in most types of coping strategies. Findings indicated Although injury was reported as the most frequent cause of re-

that women were more likely to use active coping, seek instru- lapse, those who cited lack of interest, low self-efficacy, and /or

mental social support, and use more problem-focused coping. other unspecified psychological factors were less able to rein-

Women were also more likely to seek emotional support, use state their exercise programs (9).

avoidance, engage in positive reappraisal, ruminate, engage in Although previous research supports the notion that exer-

wishful thinking, and employ positive self-talk. Some gender cise lapse and relapse occurs with some frequency, there is a

differences were dependent on the nature of the stressor—for lack of consensus regarding how exercise lapse or relapse are

example, personal health, others’ health, achievement, and rela- defined. For example, whereas Sallis et al. (9) defined relapse as

tionship stressors.

a period of no exercise for 3 months or more, Simkin and Gross An additional study of men and women citing high levels (10) defined a lapse as a 1-week period and relapse as a 3-week

of work or marital stress compared self-reported coping on period without exercise. Marlatt and Gordon (11,12) conceptu-

questionnaires and prospective electronic diary ratings (19). alized a slip in smoking cessation as smoking one or more ciga-

Cross-sectional assessment found that women reported higher rettes in a specific situation. Similarly, an exercise slip may be

levels of coping via social support and catharsis compared to viewed as an acute isolated missed exercise session. In this in-

men. These studies suggest that there are gender differences in stance, the slip and how a slip is perceived may be regarded as

the use of coping strategies; however, findings may be influ- the gateway to lapses and/or relapses or more prolonged seden-

enced by stressor appraisal, perceived controllability, or assess- tary periods during which time persons do not engage in planned

ment methodologies.

exercise. Another important aspect of the RPM is the self-efficacy construct. Decreased self-efficacy for exercise combined with positive outcome expectancy for a slip may increase its proba-

bility. For example, if one rationalizes “I’ll have more time and The Relapse Prevention Model (RPM) (11,13,14) has been

PREDICTION OF EXERCISE RELAPSE

energy if I skip exercise tonight,” there is an increased probabil- used extensively in the study of addiction and substance abuse

ity of missing a planned exercise session. An Abstinence Viola- (11,12). This conceptual model addresses the cyclical nature of

tion Effect then ensues in which increased feelings of guilt and long-term behavior change. It provides a valuable framework

perceived loss of situational control further contribute to the for understanding factors related to slips, lapses, and relapse in

probability of a full-blown relapse or return to the unhealthy exercise behavior among persons who are considered long-term

(i.e., sedentary) behavior for an extended period of time. The exercisers. According to the RPM, higher probabilities for re-

Abstinence Violation Effect has two component parts: one’s lapse occur when an individual with inadequate coping skills is

causal attributions for a slip and the affective (i.e., guilt) reaction faced with a “high-risk” situation. Coping behavior plays an im-

to the attribution (20). The Abstinence Violation Effect is con- portant role in understanding the relationship between exposure

sidered to be a particularly powerful determinant of the course to challenging situations and outcome (i.e., slip, lapse, or re-

of a possible lapse or relapse.

lapse). Within the addictive behaviors literature utilizing the One difficulty in understanding the impact of relapse pre- RPM, coping “refers to what an individual does or thinks in

vention interventions aimed at producing and maintaining be-

a relapse crisis situation so as to handle the risk for renewed havior change is the general lack of studies attempting to empir- substance use” (15, p. 1101) and coping is conceptualized as

ically evaluate the utility of the model. For a comprehensive

a mediator between the stress of the high-risk situation and overview of the RPM, we refer the reader elsewhere (13). Appli- relapse (15).

cation of the RPM in exercise behavior has produced inconsis- According to the RPM, it is the person’s coping response(s)

tent findings, creating some uncertainty about its pertinence to that will determine whether a slip will occur subsequent to en-

exercise behavior (21). Previous studies have examined RPM as countering a high-risk situation. Coping efforts are predomi-

it relates to exercise adherence by including relapse prevention nately linked to a desired goal (e.g., attending a planned exercise

components in interventions conducted with previously seden- session) and may be cognitive, behavioral, or a combination

tary samples with mixed results (22,23). Furthermore, the lim- thereof (16,17). Studies examining coping in relation to lapse in

ited number of studies applying the RPM have typically been exercise or dietary adherence have defined coping strategies in

plagued with a variety of methodological limitations (21). accordance with cognitive and behavioral parameters (10,17).

In our search for studies that more formally examine the Within the RPM, an inadequate repertoire of adaptive coping re-

role of the RPM in maintenance of exercise over time, we identi- sponses results in lower self-efficacy for overcoming obstacles

fied only three with more than a single intervention session. in high-risk situations.

Simkin and Gross (10) examined the RPM in exercise relapse Simkin and Gross (10) examined the RPM in exercise relapse

Belisle, Roskies, and Levesque (24) conducted a series of two studies utilizing the RPM as part of an exercise adherence intervention. Participants consisted of both men and women. The intervention included components focusing on improving recognition and awareness of barriers to exercise and adaptive coping responses, including information about recognizing and overcoming the Abstinence Violation Effect. Effects of the in- tervention were examined for differences in short-term (i.e., ad- herence throughout the intervention phase) and longer term adherence (i.e., 12 weeks postintervention). Exercisers in the condition including the RPM components exercised more days in both the short- and long-term phases of each of these studies.

In both of these studies, participants were university stu- dents (as opposed to community-based samples), and exercise was followed for a comparatively short period of time. In the first study reviewed, Simkin and Gross (10) classified par- ticipants as “exercisers” based on a self-reported self-schema descriptor as exerciser or nonexerciser. Frequency, intensity, and duration of participants’ exercise were not reported. Therefore, the degree to which the sample met behavioral exercise guide- lines was not clear. Belisle et al. (24) were able to report more objective measures of exercise participation during participants’ enrollment in exercise courses. Study participants’ exercise pat- terns were followed for periods of 10 (24) and 14 (10) weeks af- ter initiation of the exercise intervention.

Marcus and Stanton (22) conducted a three-arm random- ized intervention study with 120 previously sedentary women. The majority of participants were overweight. A structured re- lapse prevention program was compared to a reinforcement pro- gram and an exercise-only control group. The primary aim of the study was to evaluate the interventions’ impact on session at- tendance. The relapse prevention condition focused on identifi- cation of personal high-risk situations on and group discussions of effective and ineffective coping strategies for managing high-risk situations. A 10-day planned break from exercise was included as a planned relapse to demonstrate that participants could lapse and then resume exercise. A 2-month follow-up as- sessment of exercise frequency, intensity, and duration via par- ticipant and collateral self-report was also conducted. Exercise

leaders for the study were advanced undergraduate students. During the first half of the study, attendance was higher for relapse prevention program participants. However, low atten- dance was a substantial problem encountered across conditions, and attrition was high (72%) by the end of the study. No group differences were observed for posttreatment or follow up. The authors offered methodological improvements of randomized design, longer duration of intervention, follow-up assessment, and conservative definition of program adherence. However, the nature of the sample (overweight university employees led by university undergraduates) and characteristics of the interven- tion (set days and times; aerobic dance) may have influenced the observed attrition rates. With intermittent attendance, partici- pants may not have been exposed to the full complement of Re- lapse Prevention Program content.

A greater understanding of mechanisms of exercise behav- ior change and maintenance of change is needed (25). Support for theoretically driven mediators of exercise behavior change has been mixed (25–27). Previous studies examining relapse in exercise have tended to utilize samples from clinical pop- ulations, university classes, fitness programs, or previously sed- entary persons (7,9). In addition, components of the RPM typi- cally have been examined independently or following an exercise intervention. To our knowledge, the RPM has not been examined in a community sample of long-term exercisers.

The primary aim of this study was to utilize the RPM to ex- amine processes of maintenance of exercise in long-term com- munity-based exercisers. This prospective study examined long-term exercisers’ coping responses and exercise slips in re- sponse to self-identified/individualized high-risk situations. Personal high-risk situations, use of specific cognitive and be- havioral coping strategies in response to these situations, and ex- ercise self-efficacy were of particular focus within the RPM. To this end, we examined (a) types of acute high-risk situations cited by long-term exercisers, (b) cognitive and behavioral cop- ing strategies employed in high-risk situations, (c) obstacle self-efficacy, (d) outcomes (exercise vs. slip) relative to coping strategies employed, (e) Abstinence Violation Effect (guilt and perceived control regarding the personal situation), and (f) pre- dictors of exercise characteristics at 3-month follow-up. Given previous findings that women and men may have different rates of using particular coping strategies, we planned to examine gender differences and conducted analyses separately by gender where appropriate. It was hypothesized that positive coping strategies (task/problem solving, positive reappraisal, pre-exer- cise rituals, and social support) would be associated with re- duced likelihood of a slip in the reported high-risk situation and higher levels of self-reported exercise at 3-month follow-up. It was hypothesized that negative coping strategies (rationaliza- tion, procrastination/avoidance) would be associated with high- er likelihood of a slip in the reported high-risk situation and lower levels of exercise at 3-month follow-up. It was further hy- pothesized that women would report greater use of positive re- appraisal, rationalization, and seeking social support compared to men. We also hypothesized that higher Abstinence Violation Effect (higher guilt and lower perceived control scores) would

Volume 30, Number 1, 2005 Exercise Slips

27

be associated with lower physical activity level at the 3-month follow-up.

METHOD

Participants

Study participants were recruited from two YMCA exercise facilities in a large midwestern city. Fliers describing the study were posted throughout facilities. Interested individuals con- tacted the investigators by telephone and subsequently com- pleted phone screenings. Adults who were already engaged in exercise and who were free from any serious health complica- tions were eligible to participate. All participation, including in- formed consent, was conducted via U.S. mail.

Procedure

Participants (N = 65) were mailed a baseline survey con- taining questions about demographic and exercise history as well as current physical activity patterns. Questions designed to assess exercise-specific cognitions and perceptions regarding exercise maintenance were also included. Upon return of the baseline survey, participants were mailed an exercise-related high-risk situation questionnaire that assessed personal experi- ence with a self-identified situation that previously posed high risk for an exercise slip. Following completion of the high-risk situation questionnaire, participants were contacted 3 months later to assess ongoing exercise patterns. This study includes re- sults obtained from the baseline demographic/exercise ques- tionnaires and cognitions, assessment of the high-risk situation, and 3 month postbaseline follow-up. Participants were compen- sated as follows: (a) $10 baseline survey completion, (b) $10 for exercise high-risk situation questionnaire completion, and (c) entry into a $100 lottery pool upon completion of follow-up.

Measures

Baseline exercise history. The Exercise and Health His- tory Questionnaire (EHHQ) developed by Dubbert, Stetson, and colleagues (28–30) was used to obtain information on self-re- ported history of exercise and current exercise pattern. The items included (a) exercise history (total months engaged in reg- ular exercise), (b) frequency (average number exercise ses- sions/week), and (c) duration (minutes/session). Validity of this measure was examined in a pilot study of a sample of 196 com- munity exercisers recruited from an urban YMCA, community activity centers, and university undergraduates (ethnicity = 77.6% White; M age = 35.95, SD = 18.90 [unpublished data]). A composite of these three items was compared to the score on the Godin Leisure Time Questionnaire (31). Pearson correlations were .573 (p < .0001), indicating a moderately high degree of association between these EHHQ items and the previously vali- dated activity measure. Two-week test–retest of this exercise measure with a subsample (n = 83) of participants yielded a cor- relation of .255 (p = .020). One-week test–retest reliability was also evaluated in another pilot study with a sample of 29 college students (r ≥ .90).

Participants were also asked, “Rate how hard you typically work when you exercise.” A subjective rating of “typical” per- ceived exertion (RPE) during exercise sessions was obtained us- ing the 6 (very, very light) to 20 (very, very hard) Borg RPE Cat- egory scale (32,33). Reliability coefficients for this scale in submaximal exercise have ranged from .70 to .90 (34). The ret- rospective use of RPE in this study may be likened to a com- monly used practice of perceptually regulated exercise intensity (35), and in the absence of available direct physical fitness mea- sures, the derivation of exercise indexes may be regarded as an acceptable surrogate measure of physical activity (36).

Exercise cognitions and perceptions. Self-efficacy for over- coming obstacles and engaging in exercise was also examined (35,37). Participants listed their top personal obstacle to regular exercise and rated confidence for overcoming the corresponding obstacle on a 0 (I will not be able to overcome this and exercise) to 100% (I will be able to overcome this and exercise) confi- dence scale. Single item self-efficacy measures have been suc- cessfully used in the study of smoking and the RPM to identify smoking lapses and relapse following lapse (38). Zero to 100% ratings of obstacle self-efficacy have been found to be related to exercise adherence and drop out in beginning and consistent ex- ercisers (39–41) and are consistent with Bandura’s recommen- dations for measuring self-efficacy (39,42,43). The top exercise obstacles were self-generated by participants. The self-efficacy ratings for self-reported top obstacles were used in analyses, in keeping with recommendations to include personal, salient bar- riers that are anticipated to occur frequently in assessing obsta- cle efficacy (39). Self-determination of status as a regular or in- termittent exerciser was assessed dichotomously.

Self-identified high-risk situation. Each participant was asked to describe a high-risk situation of recent personal experi- ence in which he or she had planned on exercising and had felt most tempted to not engage in the exercise. Participants were in- structed to “imagine one specific situation where you felt you should exercise but felt tempted not to.” Additional items posed specific questions to elicit description of the characteristics of the situation. Participants described these characteristics using an open-ended response format. A sample high-risk situation described by a participant is presented to provide an example of

a specific situation: “I was in my car heading for the gym with all good intentions.” “It was after a very stressful day at work, I hadn’t had enough sleep, traffic was bad and the weather was overcast and cold.”

Coding of high-risk situation characteristics. Participant responses were conceptually coded by two raters into the fol- lowing categories: (a) location (i.e., home, work, or other), (b) weather (i.e., good, bad, or neutral) (c) time of day (i.e., morn- ing, afternoon, evening), (d) social characteristics of the sit- uation (i.e., alone or with others), and (e) physical state of the participant (fatigued, injured or ill, healthy/no difficulties re- ported). Participants were also asked to record their mood dur- ing the high-risk situation; this was subsequently coded as posi-

28 Stetson et al. Annals of Behavioral Medicine

Volume 30, Number 1, 2005 Exercise Slips

tive, negative, or neutral mood. The exercise outcome of the sistent with the two reviewed studies utilizing the RPM in exer- situation was assessed dichotomously—that is, “How did this

cise (10,24). Previous work in exercise adoption suggests that situation turn out—did you exercise or not?”

rates of exercise drop out have substantial magnitude in a 3-month Three independent raters were trained in the coding of the

window (8), and therefore this time frame may provide an infor- high-risk situation characteristics. Coding instructions specifi-

mative snapshot of the maintenance process . Participants rated cally guided raters to make their coding decisions independent

their perception of their current exercising status as maintaining of exercise outcome in the high-risk situation. Following calcu-

exercise or not regularly exercising. Exercise items were consis- lation of coefficient kappa, any disagreements on items were re-

tent with those administered at baseline, including frequency (av- solved via mutual agreement of the raters.

erage number of exercise sessions/week), duration (min/session) and perceived exertion (RPE) during exercise sessions during the

Abstinence Violation Effect. Cognitive and behavioral as- previous month (32,33). Participants also listed current personal pects of the response to the high-risk situation and outcome (i.e.,

obstacles to regular exercise and rated confidence for overcoming Abstinence Violation Effect) included ratings of guilt about the

the corresponding obstacle on a 0 to 100 confidence scale (48). outcome and perceived control in managing the temptation to

not exercise.

Statistical Analyses

All analyses were conducted using SPSS for Windows ® , Coding of coping responses. Coping responses were con-

Version 11.0. Odds ratios were used to measure association be- ceptually organized into two dimensions: cognitive coping re-

tween dichotomous outcome and predictor variables. Logistic sponses and behavioral coping responses. Coding criteria were

regression was used to examine the effect of confounding vari- developed based on Marlatt’s studies of coping with high-risk

ables for dichotomous outcomes. Pearson correlations were drinking situations (13) and a review of the stress and coping lit-

used to measure association between continuous outcomes and erature (16,44–46). As in previous studies assessing coping

predictors. Point biserial correlation was used to measure asso- strategies, cognitive coping was defined as nonobservable

ciation between dichotomous outcomes and continuous predic- thought processes. Behavioral coping strategies were those

tors. Partial correlation (pr) and multiple regression were used which could be observed (17,47).

to examine the effect of confounding variables for continuous Coding criteria reflected three cognitive coping ap-

outcomes. Stepwise variable selection was used to determine proaches: task-oriented problem solving, positive reappraisal,

the best set of predictors for both the logistic and multiple re- and rationalization. Task-oriented problem-solving ratings re-

gression models. T tests were used to test for significant differ- flected thought processes addressing strategies for overcoming

ences in means between two groups. Fisher’s exact test was used obstacles. Positive reappraisal ratings reflected use of positive

to test for differences in proportions between groups. Because self-statements and cost–benefit analysis. Both coping strate-

the majority of studies utilizing the RPM across the addictive gies were regarded as positive approaches. Regarded as a nega-

and health behavior literatures have been with single gender tive approach, rationalization ratings reflected personal bargain-

samples, and because gender differences have been observed in ing and justifying reasons not to exercise.

physical activity levels and use of coping strategies, it was hy- Coding criteria reflected three behavioral coping ap-

pothesized that the gender of participants might impact study proaches: engaging in pre-exercise rituals, eliciting social sup-

outcomes. Therefore, analyses were stratified by gender wher- port, and engaging in procrastination or avoidance. Engaging in

ever appropriate. The Breslow–Day method was used to test for pre-exercise rituals ratings reflected engaging in a routine to

significant differences across strata. Two-tailed tests were used pave the way for exercise (e.g., laying out exercise clothes in

for all analyses. An alpha level of .05 was used as the criterion advance). Eliciting social support ratings reflected seeking the

for statistical significance.

assistance of others to facilitate exercise. Procrastination and avoidance ratings reflected creating self-imposed barriers to ac-

RESULTS

tivity and engaging in other activities until the window of oppor-

Participant Demographic Characteristics

tunity expired (e.g., no time left, dark outside). Of 77 adults participating in a cross-sectional physical ac- Three independent raters were trained in the coping coding

tivity survey, 65 responded to an invitation to participate in this scheme. Responses to each of the coping questionnaire items

study and completed baseline measures (response rate = 84%, were subsequently coded. Coding instructions specifically guid-

with no gender differences). Participants ranging in age from 22

ed raters to make their coding decisions independent of exercise to 71 years old consisted of 28 men (M age = 38.8, SD = 12.6) outcome in the high-risk situation. Following calculation of co-

and 37 women (M age = 35.4 years, SD = 11.5). Participants efficient kappa, any disagreements on items were resolved via

were predominantly White (men = 78.6%, women = 82.9%). mutual agreement of the raters.

African Americans comprised 14.3% of men and 8.6% of women in the sample. Overall, participants were educated be-

Exercise characteristics at 3-month follow-up. Three months yond high school (M years, men = 16.4, SD = 3.1; M years, after the completion of the baseline survey, participants were

women = 15.6, SD = 2.3). The majority of participants were mailed a brief follow-up exercise questionnaire regarding their

nonsmokers (men = 85.2%, women = 85.7%) and at the upper activity over the previous month. The 3-month follow-up is con-

end of normal body weight (M BMI men = 26.6, SD = 7.8; M

30 Stetson et al. Annals of Behavioral Medicine

BMI women = 24.3, SD = 8.7). No differences between genders ing alone (75% and 59.5%, respectively), a negative mood were observed in these background characteristics.

(54.5% and 70.6%, respectively), and being physically tired (62.5% and 60.0%, respectively; all ps >.05). Of the 64 partici-

Interrater Reliability for Coding

pants who characterized their high-risk situation, 4 (6.1%) listed

of High-Risk Situation Characteristics

one characteristic, 13 (19.7%) listed two, 14 (21.2%) listed Kappa coefficients for the characteristics of the high-risk

three, 18 (27.3%) listed four, 14 (21.2%) listed five, and 3 situation (weather, time of day, social context, physical state,

(4.5%) listed six. The mean number of characteristics listed for mood) were all significantly different from 0, with moderate

the high-risk situation was 3.5 (SD = 1.3). No gender differences high to high magnitude (κ = .86 for weather, .75 for time of day,

were found in the high-risk situation characteristics. .92 for social context, .75 for mood, .85 for physical state).

Coping strategies employed in high-risk situations. Positive

Interrater Reliability for Coding

cognitive coping strategies were used by 42.9% of participants.

of Coping Responses

This included use of positive reappraisal by 35.4% of partici- pants (e.g. “I reminded myself I’ve had some of my best runs

Kappa coefficients were high across each of the coping cat- when I didn’t feel like it”) and task/problem solving by 12.3% egories, indicating strong interrater agreement. Kappas for each (e.g., “I decided to drive towards the gym. … I told myself I did- of the coping categories were as follows: task/problem solving, n’t have to exercise that long”). The negative cognitive coping .84; positive reappraisal, .83; rationalization, .88; procrastina- strategy of rationalization was used by 47.6% of participants tion/avoidance, .86; pre-exercise rituals, .95; social support, .98. (e.g., “I thought, I’ll exercise tomorrow, my figure won’t be that

much worse if I miss today”). Of those employing positive cog-

Baseline Analyses

nitive strategies, only 3.1% used more than one type of positive Baseline exercise patterns and perceptions. Baseline ex-

cognitive strategy in the high-risk situation. Given this small ercise characteristics by gender are presented in Table 1. On av-

number, the positive cognitive strategies variable was recoded to erage, participants reported maintenance of regular exercise for

reflect a dichotomous rating (used vs. did not use) in all subse- approximately 5 years, with some participants reporting over 20

quent analyses. Percentages of participants reporting use of in- years of regular exercise. Participant ratings of perceived exer-

dividual cognitive coping strategies are presented in Table 2. ciser status indicated that 61.8% of the sample rated themselves

Women were significantly more likely to use rationalization in as “regular” exercisers, whereas 38.2% perceived themselves to

the high-risk situation than men. Positive behavioral strategies

be “intermittent.” Participants averaged 3 to 4 days per week of were represented entirely by use of pre-exercise rituals (e.g., “I activity. Typical exercise sessions averaged nearly 1 hr. Ratings

laid out my exercise clothes and just started dressing for it … of perceived exertion during exercise sessions tended to be

washed my face, took two Ibuprofen … ”). None of the partici- “moderate” or “somewhat hard.” Women reported exercising

pants reported using solicitation of social support in response to for significantly fewer minutes per session than men.

the high-risk situations. The negative behavioral strategy of pro- crastination was exemplified by putting off the planned exercise

RPM Variables

(e.g., “I thought, I’m so tired, I’ll just make a few phone calls High-risk situations. The most commonly cited character-

first”). No participant used more than one type of behavioral istics of the high-risk situations were similar for both men and

coping strategy. Men were significantly more likely to use women, including bad weather (75% and 66.7%, respectively),

pre-exercise rituals in the high-risk situation than women. Only occurrence in the morning (53.8% and 43.8%, respectively), be-

two participants used a combination of both cognitive and be-

TABLE 1 Self-Reported Exercise Characteristics at Baseline and 3-Month Follow-Up Compared by Gender

3-Month Follow-Up Variable

Baseline

Men a Women b p c Men d Women e p f

Exercise characteristics Regular exercisers (%)

— — Months exercising (M, SD)

— — Frequency per week (M, SD)

3.8 (2.0) .356 Minutes per session (M, SD)

56.0 (53.7) .779 RPE g (M, SD)

66.0 (19.0) .777 Note. RPE = rating of perceived exertion.

Exercise self-efficacy for obstacles h (M, SD)

a n = 28. b n = 37. c p values reflect gender comparisons at baseline. d n = 25. e n = 34. f p values reflect gender comparisons at 3-month follow-up. Three-month follow-up activity variables did not significantly differ from their respective baseline values. g Rating of perceived exertion 6–20 category scale. h 0–100-point

confidence rating scale.

Volume 30, Number 1, 2005 Exercise Slips

TABLE 2 Rated Use of Exercise Coping Strategies in High-Risk Situations

Results of OR a Analyses Coping Strategy

% Rated as Using Strategy

CI p Cognitive

Valence

Men b Women c OR

Task/Problem solve

11.1 14.7 1.38 0.30–6.37 .680 Positive reappraisal

25.9 61.8 5.06 1.69–15.14 .004 Behavioral Procrastination/Avoid

Negative

11.1 8.8 0.77 0.14–4.18 .766 Pre-exercise rituals

Negative

22.2 2.9 0.10 0.01–0.88 .039 Social support

Positive

— — Note . OR = odds ratio; CI = confidence interval.

Positive

a Odds ratio for women, using men as the reference category. b n = 28. c n = 34.

havioral strategies (one man and one woman). Given this small test for homogeneity of odds ratios across strata indicated no number, combined use of these approaches was not included in

difference in odds ratio between men and women (p = .896); any subsequent analyses.

therefore, the men and women were pooled to calculate an over- all estimate of the odds ratio (OR = .29, 95% CI = 0.09–0.93, p =

Self-efficacy for overcoming obstacles in a high-risk situa- .038). There was no significant difference in the number of char- tion. Self-efficacy for overcoming top personal exercise obsta-

acteristics of the high-risk situation between those who slipped cles was well above the midpoint of the 100-point confidence

and those who did not.

scale (M = 71.5, SD = 18.9). There were no gender differences in self-efficacy for overcoming exercise obstacles.

Abstinence Violation Effect. High-risk situations and out- comes were significantly associated with guilt and perceived Exercise slip in a high-risk situation. Participants were near-

control. Patterns of association differed by gender. Women re- ly evenly split in the outcome (i.e., slip) of the high-risk situation

ported higher levels of guilt in response to the high-risk situation (46.2% exercised, 53.8% slipped). However, women were al-

compared to men (M = 4.30, SD = 1.85 vs. M = 3.07, SD = 1.98 , most twice as likely to report a slip (63.9% did not exercise) than

respectively), t(63) = 2.564, p =.013. However, women and men were men (37.0% did not exercise); this is equivalent to an odds

did not differ in perceived loss of control regarding the exercise ratio of 3.01 (95% CI = 1.07–8.47, p = .037).

outcome. Women who reported a slip had higher guilt ratings than women who reported exercising in the high-risk situation

Self-efficacy versus high-risk situation outcome. Baseline (M = 4.78, SD = 1.48 vs. M = 3.46, SD = 2.26), t(34) = 2.125, p = self-efficacy for overcoming obstacles was also examined in re-

.041. For women, perceived control ratings did not differ be- lation to exercise slip in the self-identified high-risk situations.

tween those who slipped and those who exercised. Men who re- No significant associations were observed for either gender.

ported a slip in the high-risk situation had far higher guilt ratings than men who reported exercising (M = 5.10, SD = .57 vs. M =

1.94, SD = 1.56), t(25) = 6.13, p < .001. Men who reported slip- both men and women, more intermittent exercisers experienced

Exercise pattern versus high-risk situation outcome. For

ping in the high-risk situation also had lower ratings of per-

a slip than did regular exercisers, and the test for homogeneity of ceived control compared to men who reported exercising (M = odds ratios across strata indicated no difference in odds ratio be-

4.00, SD = 1.73 vs. M = 5.65, SD = 1.27), t(24) = –2.77, p = .011. tween men and women (p = .170). Therefore, the men and

women were pooled to calculate an overall estimate of the odds Coping strategies employed and high-risk exercise situa- ratio (OR = 2.95, 95% CI = 0.99–8.77, p = .052). Because this

tion outcome. The relationship between use of coping strate- association is so close to being statistically significant, a dummy

gies and exercise outcome (exercise vs. slip) in the high-risk sit- variable indicating whether the participant was a regular or in-

uation was examined using odds ratios. The two positive cogni- termittent exerciser was considered as a potential confounding

tive coping strategy categories (task/problem solving and variable in the multivariate analyses.

positive reappraisal) were pooled to create a single positive cog- nitive coping variable. Similarly, the two positive behavioral

Characteristics of high-risk situation versus high-risk situ- coping strategies (pre-exercise rituals and social support) were ation outcome. The only characteristic of the high-risk situa-

pooled to create a single positive behavioral coping variable. For tion that was significantly associated with an exercise slip was

men, 38.5% of those not reporting use of positive cognitive cop- being alone; for both men and women, participants who were

ing strategies slipped, whereas 100% of those reporting use of alone during the high-risk situation were less likely to slip. The

these strategies exercised (OR = 31.18, 95% CI = 3.95–∞, p <

32 Stetson et al. Annals of Behavioral Medicine

.001). For women, 90.9% of those not reporting use of positive rated as using the cognitive reappraisal coping strategy com- cognitive coping strategies slipped, whereas 84.6% of those re-

pleted follow-up, and all participants rated as not using this porting use of these strategies exercised (OR = 44.29, 95% CI =

coping strategy dropped out. No other coping strategy ratings 5.27–721.49, p < .001). Use of positive behavioral coping strate-

were associated with drop out. Exercise self-efficacy, slip out- gies was not associated with outcome in the high-risk situation

come in the high-risk situation, and Abstinence Violation Effect for either gender.

variables were not associated with drop out at the 3-month fol- The independent effect of use of positive cognitive coping

low-up.

was examined using multiple logistic regression. The positive Of the 59 follow-up respondents, three (two men and one cognitive coping variable, along with the positive behavioral

woman) reported that they were not regularly exercising. Partic- coping variable and a dummy variable indicating whether the

ipant exercise patterns and self-efficacy at the 3-month fol- participant was a regular or intermittent exerciser, were entered

low-up are included in Table 1. These prospectively assessed into a stepwise logistic regression model with the dichotomous

activity variables did not differ significantly from their corre- exercise outcome (slipped or exercised) as the dependent vari-

sponding baseline indexes and did not differ by gender. able. This analysis indicated that there were no additional signif- icant predictors of the high-risk situation outcome other than use

Baseline self-efficacy and exercise at follow-up. Baseline of positive cognitive coping strategies. In particular, there was self-efficacy was not associated with any of the exercise mea- no significant confounding effect of exercise pattern (regular vs. sures at the 3-month follow-up for either gender (Table 3). intermittent exercisers) on this association for either gender.

High-risk situation responses and activity at follow-up. Exercise patterns at 3-month follow-up. Three men and

Participant responses to the self-reported high-risk situations three women (9% of the total sample) did not return activity

were associated with some of the exercise measures at the questionnaires at follow-up. Participants who dropped out did

3-month follow-up, but only among the men (Table 3). Exercise not differ from those who returned the 3-month follow-up sur-

slip in the high-risk situation was negatively associated with fol- vey in terms of duration of maintenance of regular exercise,

low-up duration of exercise sessions after controlling for base- baseline frequency, and duration of activity, or self-character-

line duration of exercise. Use of positive cognitive coping strate- ization as an intermittent or regular exerciser. Coping strategy

gies was positively associated with duration of exercise sessions ratings were associated with drop-out status. All participants

at follow-up after controlling for baseline exercise duration and

TABLE 3 Associations of Relapse Prevention Model Variables and Exercise Outcomes at 3-Month Follow-Up

3-Month Follow-Up RPM Variable by Gender

Exercise Self-Efficacy Men

Exercise Frequency

Exercise Duration

RPE

Coping strategy

.64 c Rationalization

Positive cognitive a .36

.53 b .13

–.57 Pre-exercise ritual

–.47 d –.05

–.09 Procrastination/Avoidance

–.13 Baseline exercise self-efficacy

.01 High-risk situation slip

–.57 Guilt rating

–.54 e –.35

–.50 Perceived control rating

–.55 f –.49 g –.39

.59 j Women Coping strategy

Positive cognitive a –.24

–.11 Pre-exercise ritual

.13 Procrastination/Avoidance

–.13 Baseline exercise self-efficacy

.44 High-risk situation slip

–.49 Guilt rating

–.40 Perceived control rating

–.12 Note . Partial correlations, controlling for baseline activity pattern or baseline exercise self-efficacy, where appropriate. RPM = Relapse Prevention Model;

RPE = rating of perceived exertion. a Task/problem solving and positive reappraisal pooled. b p = .013. c p = .024. d p =.034. e p = .010. f p = .006. g p = .021. h p = .046. i p = .001. j p = .041.

Volume 30, Number 1, 2005 Exercise Slips

with self-efficacy at follow-up after controlling for baseline lower perceived control and slip was associated with shorter ex- self-efficacy. Use of rationalization (a negative cognitive coping

ercise duration at follow-up. Abstinence Violation Effect strategy) was negatively associated with exercise duration after

variables were associated with exercise at the 3-month fol- controlling for baseline duration. Use of behavioral coping strat-

low-up, but only among the men.

egies was not associated with follow-up self-efficacy or exercise Although our findings generally support the utility of the behavior for either gender.

RPM in the study of exercise behavior in men and women who are ongoing exercisers, a series of interesting gender differences

Abstinence Violation Effect and activity at follow-up. At emerged. Women in our sample were almost twice as likely as 3-month follow-up, Abstinence Violation Effect variables—guilt

men to miss an exercise session or slip in the face of a high-risk and perceived control—were associated with some exercise pat-

situation. This finding could reflect greater prevalence of men terns, but only among the men (Table 3). Guilt regarding the

maintaining regular exercise in the adult population (2). On the high-risk situation was negatively associated with exercise dura-

other hand, this gender difference may have been partially af- tion at follow-up after adjusting for baseline exercise duration,

fected by a self-report artifact in that women may simply be and with typical frequency of exercise at follow-up after adjusting

more likely to report a slip than their male counterparts. Women for baseline frequency of exercise. Perceived control over the

also endorsed higher levels of guilt than men, which may reflect high-risk situation was positively associated with follow-up exer-

greater affective response to a slip.

cise duration after adjusting for baseline exercise duration, with Consistent with findings from studies of gender differences follow-up RPE after adjusting for baseline RPE, and with fol-

in coping, suggesting differences in use of verbal expression to low-up self-efficacy after adjusting for baseline self-efficacy.

self or others, women in this study had higher use of rationaliza- The relationship between RPM variables and each of the

tion in high-risk exercise situations than did men. Conversely, follow-up exercise measures (typical weekly frequency of exer-

higher use of the behavioral coping (pre-exercise rituals) among cise, duration per exercise session, rating of perceived exertion,

the men was an unexpected finding. In addition, some similari- and self-efficacy for overcoming exercise obstacles) was exam-

ties among men and women emerged. Neither gender reported ined using multiple regression. A separate stepwise regression

using social support as a coping strategy. Among both genders, analysis was performed with each exercise measure as the de-

the use of positive cognitive coping strategies (positive reap- pendent variable, and use of a positive cognitive coping strategy,

praisal, task/problem-focused coping) was strongly associated use of rationalization, use of a behavioral coping strategy, guilt

with exercising in high-risk situations. rating, perceived control rating, the baseline value of the DV,

Self-efficacy to overcome exercise obstacles was not pre- and dummy variables indicating whether the participant slipped

dictive of exercise outcome in the high-risk situation for either in the high-risk situation and whether the participant was a regu-

gender, which may reflect that barrier efficacy in the high-risk lar or intermittent exerciser as independent variables. For men,

situation may be context specific (49). Our single-item self-effi- guilt rating was associated with exercise frequency at follow-up

cacy measure may not have tapped barrier efficacy or variations (n = 23, p = .004), high-risk situation slip status was associated

relative to the high-risk situation. Although not assessed in this with exercise duration at follow-up (n = 22, p = .022), perceived

study, self-efficacy may be moderated by outcome expectancies control rating (n = 22, p < .001), and baseline RPE (n = 22, p =

for that situation and subsequently moderated by Abstinence Vi- .046) were associated with RPE at follow-up and use of a posi-

olation Effect variables (i.e., guilt and perceived control). Shiff- tive coping strategy was associated with exercise self-efficacy at

man and colleagues (38) suggested that the self-efficacy/relapse follow-up (n = 13, p = .022). For women, baseline exercise fre-

relationship is a dynamic and complex process and does not sup- quency was associated with exercise frequency at follow-up (n =

port the hypothesis that lowered self-efficacy precedes and helps

31, p < .001), and baseline RPE was associated with RPE at fol- to cause lapses. Modest decreases in self-efficacy following an low-up (n = 31, p = .004).

initial lapse may predict subsequent relapse. This hypothesis implies that diminished self-efficacy after an exercise slip may

be associated with subsequent lapse and relapse. This is sup- Consistent with Marlatt and Gordon’s (11) model, use of

DISCUSSION

ported by the finding that for the men in our sample, slips in the positive coping responses was associated with reduced likeli-

high-risk situation approached significant association with low- hood of a slip. Positive behavioral strategies were not associated

er levels of exercise obstacle self-efficacy at follow-up (pr = with slips or any exercise maintenance outcomes. Our results

–.57, p = .052). Future studies examining exercise obstacle effi- provide support for the RPM’s contention that exposure to spe-