Correlates of the Stages of Change for Physical Activity in a Population Survey

Abstract

Objectives. We sought to identify variables associated with being in a particular stage of change for physical activity—a measure of behavioral intention to engage in regular physical activity. Understanding behavioral intentions can be useful in explaining why individuals are physically inactive or active.

Methods. Data from the Rhode Island 2000 Behavioral Risk Factor Surveillance System were used to evaluate predictors of stage of change for physical activity. There were 3454 observations in the data set, representing a weighted population of 742636 people. Estimates were obtained from polytomous multiple logistic models.

Results. Being a woman, Hispanic, non-Hispanic Black, and older than 55 years of age were associated with being in precontemplation and contemplation stages of change rather than maintenance. Self-perceived health status and rarely feeling healthy or full of energy were strongly predictive of stage of change. Having a health limitation was a dichotomous predictor, predicting being in action and precontemplation stages.

Conclusions. Several sociodemographic and health variables were associated with varying patterns of stages of change for physical activity. The complexity of individual intentions for physical activity provides evidence for the potential existence of mediating, effect-modifying, and confounding variables that differ depending on individual characteristics.

The Transtheoretical Model combines key social–cognitive models of health behavior change with the core construct, stages of change, reflecting behavioral intention.1^,2 The Transtheoretical Model purports that people progress through 5 stages as they change health behaviors.1 First applied to physical activity behavior by Marcus et al.,3^–6 the model has been validated against measures of physical activity behavior, physical fitness, and health,7^–16 and has been successfully applied in an array of populations and settings.17^–24

Despite public health efforts, the proportion of the population engaging in healthful levels of physical activity is low.25 Between 1990 and 1998, the prevalence of adults engaging in optimal levels of physical activity in the United States increased from 24.3% to 25.4%.26 At the same time, prevalence of walking remained stable, although select subgroups showed improvement.27 These data intimate that there is a need for improved understanding of this complex behavior.28^,29

Epidemiological studies have described correlates of physical activity behavior and inactivity,30^–36 but few37^,38 have studied the intentions for physical activity. To gain a better understanding of physical activity, it is important to understand not only the behavior, but also the intentions for the behavior.15 Studying the stages of change for physical activity should improve the understanding of physical activity behavior, because of its focus on intentions for behavior.

There are few studies describing the stages of change for physical activity in large populations, and only limited data on factors associated with stages of change.38^–43 We sought to identify variables associated with the probability of being in a stage of change for physical activity in a population sample.

METHODS

Self-reported data were prospectively collected via a random-digit-dial phone survey as part of the Behavioral Risk Factor Surveillance System survey (BRFSS) conducted in Rhode Island during 2000.44 The survey included the standard BRFSS module, supplemental BRFSS modules, and state-added modules.45^,46

Stages of Change Questions

Five questions assessing intention to participate in regular physical activity consistent with the recommendations of the US Surgeon General were used to measure stage of change.46 Respondents were queried about their engagement in leisure and nonleisure physical activities maintained for at least 10 minutes performed over the course of the day and which totaled 30 or more minutes per day on 5 or more days per week. The algorithm of Marcus et al. was used to classify stages of change.3^,4 Participation in physical activity meeting the criterion was classified as maintenance if performed for 6 months or more, and action if performed for less than 6 months. Physical activity not meeting the criteria for action or maintenance was categorized as preparation. For example, an individual was categorized as being in preparation if he or she engaged in some physical activity on an irregular basis or if they were regularly active, but were not active for at least 30 minutes per day on 5 or more days per week. Contemplation was the intention to engage in activity within the next 6 months, and no intention to engage in physical activity was precontemplation.

Other Correlates

Several potential correlates were evaluated, including the following sociodemographic characteristics: gender, age (18–34, 35–54, or 55 years and older), race/ethnicity (non-Hispanic White, non-Hispanic Black, His-panic, or other), marital status (married or unmarried couple, divorced or separated, widowed, or never married), and highest year of school completed. Some self-reported health history variables were evaluated, including current smoking status (every day, some days, former smoker, or never smoked), body mass index (BMI; calculated as weight in kilograms divided by height in meters squared), whether they had been told by a doctor that they had diabetes, self-perceived health (excellent, very good, good, fair, or poor), whether they were limited by an impairment or health problem, and the number of days in the past 30 they felt very healthy and full of energy.

Statistical Analyses

We fit a polytomous multiple logistic model using the SVYMLOGIT procedure in Stata version 7.0 (Stata Corp, College Station, Tex). This model, described elsewhere,47 permits the ascertainment of the independent effect of a predictor variable on all levels of the dependent variable simultaneously, with 1 level of the dependent variable as the reference category. This approach yields greater precision in estimates and increased power than if a dichotomized outcome were used.48

The measure of effect derived from polytomous logistic regression models is the ratio of the relative risk of being classified in 1 stage-of-change category instead of a reference stage if the respondent is negative for a predictor variable. For this analysis, we chose maintenance as the reference category. We used Stata to estimate the ratio of relative risks (RRRs) and associated 95% confidence intervals (CIs).

An example of the RRR for a predictor variable is demonstrated using the dichotomous response (yes, no) to the variable, “[are you] limited in any way by impairment or health problem?”(health limitation):

(1)

Thus, the probability of being in precontemplation, compared with being in maintenance, is 1.8 times greater in those with a health limitation compared with those without a health limitation.

RESULTS

There were 3454 observations in the data set, representing a weighted population of 742 636. The stages of change in the population (weighted estimate) are shown in Figure 1 ▶. Sixty-four percent were in action or maintenance for physical activity, whereas 36% were in preaction stages (preparation, contemplation, or precontemplation). In response to the BRFSS exercise question, 72% of adults surveyed participated in “any physical activities during the past month,” and 27.5% were sedentary.

FIGURE 1—

Percentage of adult sample (n = 3454) in stages of change in physical activity: Rhode Island, Behavioral Risk Factor Surveillance System Survey, 2000.

Note. Presented are weighted estimates.

Characteristics of the respondents by stage of change are presented in Table 1 ▶. Of the total sample, 68% of the respondents were aged 18 to 54 years, and 17% were members of a racial/ethnic minority group. Fourteen percent of respondents had less than a high school education, 29% had an annual income of less than $25 000 per year, and 64% were employed. Fifty-eight percent of respondents were married, and 58% reported that their health was very good to excellent. The RRRs and corresponding 95% CIs for each predictor variable are presented in Table 2 ▶.

TABLE 1—

Sample Demographic and Clinical Characteristics Among Adults (n=3454), by Stage of Change in Physical Activity: Rhode Island, Behavioral Risk Factor Surveillance System, 2000

	Precontemplation, %	Contemplation, %	Preparation, %	Action, %	Maintenance, %
Gender
Men	34	43	43	43	51
Women	64	57	57	57	49
Age category, y
18–34	18	27	29	44	32
35–54	31	36	44	31	38
≥55	51	38	28	25	30
Race/ethnicity
Non-Hispanic White	74	72	88	78	85
Non-Hispanic Black	4	5	6	5	3
Hispanic	18	20	4	15	9
Other	3	4	2	2	3
Marital status
Married or unmarried couple	56	55	61	54	58
Divorced or separated	12	17	13	9	12
Widowed	17	9	6	4	7
Never been married	15	18	20	32	23
Highest year of school completed
Grade 8 or less	10	11	1	6	4
Grades 9–11	21	11	8	12	9
Grade 12 or GED	32	35	24	27	30
1–3 years of college	21	21	27	32	26
4 or more years of college	16	21	41	23	31
Smoking status
Current smoker—every day	16	15	19	25	18
Current smoker—some days	4	5	5	3	6
Former smoker	25	26	26	23	28
Never smoked	54	54	50	49	49
Body mass index,a kg/m2
< 20 (underweight)	10	3	5	9	7
20–24 (normal weight)	34	39	38	32	42
25–29 (overweight)	34	32	38	42	37
≤ 30 (obese)	22	25	19	17	15
Given diabetes diagnosis	11	9	7	7	5
Self-perceived health
Excellent	10	16	22	18	28
Very good	23	25	36	31	36
Good	27	34	28	29	27
Fair	30	18	10	15	8
Poor	10	7	4	7	2
Limited in any way by impairment or health problemb	34	25	19	24	12
No. of days in past 30 felt very healthy and full of energy
0–10	50	41	36	34	24
11–20	12	19	25	26	22
21–30	37	40	38	41	54

Notes. GED = general equivalency diploma. Percentages may not add up to 100 because of rounding errors.

^aBody mass index is weight in kilograms divided by height in meters squared.

^bMissing data.

TABLE 2—

Ratio of Relative Risk (RRR; With 95% Confidence Intervals [CIs]) for Being in a Stage of Change in Physical Activity Rather Than Maintenance Among an Adult Sample (n= 3454): Rhode Island, Behavioral Risk Factor Surveillance System, 2000

	Precontemplation, RRR (95% CI)	Contemplation, RRR (95% CI)	Preparation, RRR (95% CI)	Action, RRR (95% CI)
Gender
Men	0.61 (0.43, 0.85)	0.74 (0.55, 1.01)	0.71 (0.57, 0.88)	0.62 (0.44, 0.90)
Women (Ref)	1.00	1.00	1.00	1.00
Age category, y
18–34	0.45 (0.26, 0.78)	0.73 (0.45, 1.20)	1.09 (0.78, 1.53)	1.66 (0.95, 2.91)
35–54	0.70 (0.48-1.03)	0.88 (0.60, 1.27)	1.28 (0.97, 1.67)	1.09 (0.66, 1.78)
≥55 (Ref)	1.00	1.00	1.00	1.00
Race/ethnicity
Non-Hispanic White (Ref)	1.00	1.00	1.00	1.00
Non-Hispanic Black	1.81 (0.88, 3.72)	1.92 (0.95, 3.88)	1.88 (1.09, 3.24)	1.42 (0.60, 3.39)
Hispanic	2.02 (1.17, 3.48)	2.14 (1.33, 3.44)	0.58 (0.35, 0.96)	1.35 (0.64, 2.87)
Other	1.89 (0.60, 5.96)	1.99 (0.67, 5.88)	0.67 (0.33, 1.35)	1.05 (0.35, 3.14)
Marital status
Married or unmarried couple	1.45 (0.85, 2.46)	1.13 (0.48, 1.80)	1.19 (0.88, 1.60)	0.86 (0.54, 1.38)
Divorced or separated	0.93 (0.51, 1.72)	1.29 (0.75, 2.19)	1.13 (0.76, 1.67)	0.54 (0.29, 1.00)
Widowed	1.43 (0.73, 2.80)	0.93 (0.48, 1.80)	1.00 (0.61, 1.63)	0.44 (0.19, 1.02)
Never been married (Ref)	1.00	1.00	1.00	1.00
Highest year of school completed
Grade 8 or less	1.87 (0.85, 4.15)	2.16 (1.05, 4.47)	0.18 (0.08, 0.42)	1.48 (0.61, 3.59)
Grades 9–11	2.27 (1.28, 4.00)	1.08 (0.61, 1.93)	0.60 (0.38, 0.94)	1.14 (0.56, 2.35)
Grade 12 or GED	1.37 (0.86, 2.16)	1.35 (0.91, 2.01)	0.56 (0.42, 0.74)	0.96 (0.59, 1.57)
1–3 years of college	1.25 (0.77, 2.02)	1.05 (0.67, 1.62)	0.72 (0.55, 0.93)	1.33 (0.85, 2.07)
4 or more years of college (Ref)	1.00	1.00	1.00	1.00
Smoking status
Current smoker—every day	0.74 (0.46, 1.18)	0.70 (0.47, 1.06)	1.02 (0.75, 1.38)	1.29 (0.81, 2.05)
Current smoker—some days	0.67 (0.34, 1.32)	0.73 (0.35, 1.53)	0.89 (0.56, 1.42)	0.48 (0.22, 1.04)
Former smoker	0.66 (0.46, 0.96)	0.89 (0.63, 1.25)	0.87 (0.66, 1.13)	1.00 (0.63, 1.59)
Never smoked (Ref)	1.00	1.00	1.00	1.00
Body mass index,a kg/m2
< 20 (underweight)	1.31 (0.72, 2.39)	0.33 (0.16, 0.69)	0.58 (0.36, 0.95)	1.04 (0.53, 2.04)
20–24 (normal weight)	0.81 (0.55, 1.20)	0.72 (0.49, 1.06)	0.83 (0.62, 1.11)	0.80 (0.49, 1.31)
25–29 (overweight)	0.80 (0.52, 1.21)	0.62 (0.42, 0.89)	1.03 (0.76, 1.38)	1.32 (0.83, 2.10)
≤30 (obese) (Ref)	1.00	1.00	1.00	1.00
Given diabetes diagnosis	0.93 (0.51, 1.71)	0.95 (0.54, 1.69)	1.18 (0.75, 1.85)	1.27 (0.62, 2.61)
Self-perceived health
Excellent	0.15 (0.06, 0.35)	0.26 (0.10, 0.66)	0.32 (0.16, 0.61)	0.18 (0.66, 0.50)
Very good	0.25 (0.12, 0.51)	0.30 (0.12, 0.71)	0.41 (0.22, 0.77)	0.23 (0.09, 0.55)
Good	0.28 (0.14, 0.56)	0.43 (0.19, 0.98)	0.44 (0.24, 0.80)	0.26 (0.11, 0.63)
Fair	0.62 (0.32, 1.23)	0.48 (0.21, 1.10)	0.54 (0.28, 1.05)	0.37 (0.15, 0.94)
Poor (Ref)	1.00	1.00	1.00	1.00
Limited in any way by impairment or health problemb	1.76 (1.22, 2.59)	1.42 (0.96, 2.10)	1.23 (0.92, 1.65)	1.79 (1.03, 3.13)
No. of days in past 30 felt very healthy and full of energy
0–10	1.78 (1.22, 2.61)	1.81 (1.28, 2.56)	1.78 (1.37, 2.32)	1.22 (0.79, 1.86)
11–20	0.82 (0.49, 1.39)	1.16 (0.78, 1.73)	1.45 (1.12, 1.88)	1.36 (0.89, 2.09)
21–30 (Ref)	1.00	1.00	1.00	1.00

Notes. The reference group was the maintenance stage of change.

^aBody mass index calculated as weight in kilograms divided by height in meters squared.

^bMissing data.

Stage of Change Results

Precontemplation.

Positive predictors of being in precontemplation, compared with maintenance, included being Hispanic, completion of some high school, having a health limitation, and rarely feeling healthy and full of energy (less than 10 days during the past month). Men were less likely than were women to be in precontemplation compared with maintenance. Other negative predictors of being in precontemplation included being aged 18 to 34 years, being a former smoker, and having good-to-excellent perceived health.

Contemplation.

Being Hispanic, completing less than grade 8, and rarely feeling healthy and full of energy were positive predictors of being in contemplation rather than maintenance. Conversely, negative predictors of contemplation, indicating those less likely to be in contemplation compared with maintenance, were BMI that was less than 20 kg/m² (underweight) or BMI between 25 kg/m² and 29 kg/m² (overweight), and good-to-excellent perceived health.

Preparation.

Men, Hispanics, and those with less than an undergraduate degree, a low BMI, and good-to-excellent health were less likely to be in preparation compared with maintenance. Blacks and those reporting feeling healthy and full of energy on 0 to 20 days in the past month were more likely to be in the preparation stage than the maintenance stage.

Action.

Having a health limitation or feeling healthy and full of energy between 0 and 20 days in the past month made it more likely that a person would be in action compared with maintenance. Men and those in fair-to-excellent health were less likely to be in action.

DISCUSSION

The correlates of stages of change analyses, with maintenance as the referent category, showed that both common and unique factors predicted each stage of change for physical activity. These data are consistent with and extend the findings of previous studies of the correlates of physical activity25^,34^–36^,49^–51 and the distribution of stages of change in the population.39^,41^,52

Our study provides a comprehensive depiction of stages of change in a large population sample through the use of an analytic model that allowed for prediction of the likelihood of being in a stage other than maintenance. Our data demonstrated that people with certain characteristics were not only less likely to be in maintenance for physical activity, but were often unlikely to be thinking about or preparing to become more active. This information about behavioral intention for physical activity provides insight into the small degree of success achieved by public health efforts promoting physical activity and has important implications for the design of future community-based physical activity interventions.

Specific Correlates of Being in a Stage of Change Other Than Maintenance

Being a woman predicted being in stages of change other than maintenance in our study. Race and ethnicity were strong predictors of some stages of change. Compared with non-Hispanic Whites, Hispanics were twice as likely to be in precontemplation or contemplation, and half as likely to be in preparation, rather than maintenance. About 50% of Rhode Island residents originating from Latin America arrived in the United States over the previous decade.53 The degree of acculturation is an important determinant of physical activity behavior.54^,55 Berrigan et al.54 reported that 47% of the most acculturated Hispanics engaged in regular exercise, compared with 22% of the less acculturated. In addition, the types and patterns of physical activity varied with degree of acculturation. The high probability that the Hispanic respondents were in preaction stages rather than maintenance probably reflected the degree of acculturation of our sample.

Non-Hispanic Blacks were nearly 2 times more likely than were non-Hispanic Whites to be in preparation rather than maintenance but were no more or less likely to be in any of the other stages. There were no significant associations between stages of change in other racial/ethnic groups compared with non-Hispanic Whites. The small proportion of persons identifying as non-Hispanic Black (4.5%) or as a member of other racial/ethnic groups (7.9%), combined with the high degree of heterogeneity within these groups, probably made the detection of associations between stages of change in these groups difficult. A recent study56 found that racial and ethnic differences in physical activity behavior nearly disappeared when the statistical analysis was adjusted for social class: within each racial/ethnic group, physical activity was gradated by social class, with lower social class associated with higher levels of sedentary behavior.

Educational attainment was also a determinant of stage of change. The lowest levels of educational attainment were associated with a higher likelihood of being in precontemplation or contemplation, whereas more-educated persons had a greater probability of being in preparation. Educational status was not associated with being in action rather than maintenance. Recent analyses of BRFSS data57 showed that the health behavior with the greatest inequality related to educational status was physical inactivity. Our data extended these findings by demonstrating that not only were more-poorly educated persons more likely to be sedentary, but that they were also more likely to have no or little intention for increasing their physical activity.

Perceived health, health limitations, and feelings of energy and wellness, each representing a unique dimension of health, were associated with the probability of being in a stage of change other than maintenance. Persons in good or better health, compared with those in poor health, were increasingly likely to be in maintenance compared with other stages of change. Previous studies34^,58^–60 have reported a strong association between health status and physical activity behavior in disabled and nondisabled adults; our data provided the insight that health status was also associated with behavioral intention.

Health limitations were a dichotomous predictor of being in action and precontemplation. This implies that having health impairment is both a barrier and a promoter of intentions to engage in physical activity. Physical activity is recommended as an essential component of disease management for some chronic diseases,61^,62 which may explain some of this dichotomy. Conversely, patients rarely receive counseling about physical activity from their physician,63^–65 but it is not known if physicians are more or less likely to recommend physical activity depending on the disease or disability.

The finding that rarely feeling healthy or full of energy was a strong predictor of being in preaction stages (precontemplation, contemplation, and preparation), but not in action, augments understanding of the relationship between health and physical activity intentions. These results suggested that a common symptom, fatigue,66 and generally feeling unwell affect intentions for physical activity. Previous studies have shown that fatigue is associated with sedentary behavior and reduced adherence to exercise.67^,68 The results of this and other studies suggested that further evaluation of how health conditions, symptoms, and functional impairment affect intentions for physical activity may provide important insights to guide development of successful interventions.

Weight status was also a dichotomous negative predictor of being in some stages of change rather than maintenance. Persons who reported being underweight were more likely to be in maintenance for physical activity and less likely to be in contemplation or preparation stages compared with obese persons. Overweight persons were about 33% less likely to be in contemplation rather than maintenance compared with obese persons. Normal weight was not associated with any stage of change. Underweight may be a surrogate for an eating disorder, often characterized by excessive exercise,69^,70 as well as poor health, which is associated with sedentary behavior. Overweight may provide a motivation for a program of weight loss involving increased physical activity,71 whereas obesity can present a physical challenge to movement. Extremes of weight can affect the desire and motivation to engage in physical activity and physical activity behavior.72 The results of our study demonstrated that overweight and underweight can affect the intentions for physical activity.

Some variables that we measured showed no significant relationship or inconsistent associations with stage of change. Age was not a significant predictor of stage of change, with the exception that younger persons were less likely to be in the precontemplation stage rather than maintenance compared with older persons. It is well known that increasing age is associated with higher levels of physical inactivity,25 although the aging effect on physical activity may be explained in part by increased levels of chronic disease and disability in older adults.73 Our results that demonstrated the strong associations between variables representing self-perceived health lend support for the assertion that it is health status, rather than age, that affects intentions for physical activity.

Marital status was not associated with stage of change, which is in contrast to other studies that have found that divorced or widowed persons were more often sedentary than were their married peers.74^,75

Current smoking was not predictive of being in any stage of change. Former smoking was associated with a smaller likelihood of being in preparation, but it did not predict being in other stages of change. Previous studies have shown that smokers were more likely to be sedentary than were nonsmokers.75^–78 A study conducted in the Netherlands found that former smokers were of higher socioeconomic position, which is associated with higher levels of physical activity.79

Diabetes was unrelated to any stage of change. Plotnikoff et al.80 showed that persons recently diagnosed with diabetes were more active than were persons who have had a diagnosis of diabetes for a longer time. Differences in physical activity levels have also been reported in type 1 and type 2 diabetes, with persons with type 1 diabetes more likely to be active, probably because of younger age and less perceived disability.80^,81

Nearly 6 of every 10 survey respondents were physically active at a level meeting the recommendations of the US Surgeon General,74 whereas the remainder were in preaction stages. The prevalence of persons in action and maintenance stages in our sample was somewhat higher than that reported in other community samples.38^,39 Studies from Australia, Canada, and the United States have found that the distribution of the stages of change for physical activity varied by country and sociodemographic status.37^,39^,42^,43^,52^,74^,82

Population Distribution of Stage of Change for Physical Activity

A study of health behaviors in 1387 randomly selected Rhode Island adults38 reported that 43.5% were in maintenance, 5.9% in action, 18% in precontemplation, 11.8% in contemplation, and 20.7% in preparation stages. The differences in stage distribution may be a function of the differences in the types of data collection (i.e., surveillance vs research), the stage of change questionnaire, and sample size between the 2 studies. Alternatively, the discrepancies may reflect a trend of increased intention for physical activity occurring in Rhode Island through the efforts of the Rhode Island Prevention Coalition, a public–private partnership that has been working to increase physical activity.

More recently, a nonconcurrent cohort study by LaForge et al.52 compared stages of change for 5 health behaviors, including regular exercise, in datasets from the United States (2 samples) and Australia (3 samples), each using identical survey questions. There were similar distributions of stages for physical activity and other health behaviors across the 5 samples and between the 2 countries. About 50% were in preaction stages, 39.5% to 48.8% were in maintenance, and 7% to 10% were in action. The discrepancies in the proportion of persons in the action stages compared with that in our study was likely because of the differences in the criterion used: 30 minutes of exercise per session performed 3 or more times per week was used in the study of LaForge et al., whereas we assessed stages of change for leisure and nonleisure physical activity (accumulation of 30 minutes per day, 5 or more days per week).

Limitations

We cannot exclude the possibility that the stage of change questions resulted in an inflated prevalence of physically active persons, as Ronda et al.39 have suggested. How questions are asked and the staging algorithm used can make a difference in the stage in which an individual is classified.83 Nevertheless, we believe it is more likely that the discrepancies in prevalence had more to do with differences in the constructs that were being measured.15^,84 Questions about exercise or physical activity address the duration and frequency of respondents’ participation in specific physical activities or categories of physical activities performed in the past. Conversely, “stage” questions query individuals about their current physical activity behavior and intentions for future behavior compared with a particular standard, which requires a person to compare his or her perceived physical activity with a described behavior.

We used a self-report instrument to measure stages of change; thus, the results were limited by the inherent biases of self-report data.83 These biases were minimized by the use of a validated algorithm of stages of change for physical activity,3 although the criterion we used for physical activity was somewhat modified from previous studies of exercise behavior.

There was also the possibility of several sources of sampling bias: (1) adults without telephones were necessarily excluded from the phone survey (no contact), (2) there was no attempt to recontact persons who did not answer their telephones (nonrespondents), and (3) some persons refused to participate in the survey (refusals). We cannot determine whether there were any differences between these groups of residents (no contact, nonrespondents, refusals) and those who responded to the survey; therefore, we cannot ensure that our sample was representative of the population of the state.

Conclusions

We used multivariable statistical techniques that permitted the simultaneous evaluation of the correlates of stages of change on all levels of the outcome. We identified numerous differences in the correlates of stages of change for physical activity and demonstrated the complexity of the correlates of stages of change, representing behavioral motivation and intention for physical activity. The limitations of our study and the difficulties in comparing data between studies provide some challenges to the application of the results of our study directly to practice. However, our study and others have illustrated that many people believe they are getting enough physical activity, which has great implications for the probability of successful physical activity promotions. The complexity of individual intentions for physical activity shown in our study provides evidence for the potential existence of mediating, effect modifying, and confounding variables that differ depending on individual characteristics and helps to explain in some measure why physical activity interventions have been modestly successful.28^,85 Better understanding of past, current, and future intentions for physical activity behavior is important in improving our ability to develop effective interventions to increase physical activity. The results of our study lend support to the need to develop interventions differently for persons with different characteristics.