Elsevier

Preventive Medicine

Volume 37, Issue 6, December 2003, Pages 635-645
Preventive Medicine

Regular article
Developing an empirical typology for regular exercise☆

https://doi.org/10.1016/j.ypmed.2003.09.011Get rights and content

Abstract

Background

Tailored interventions require the identification of distinct homogenous subgroups that will benefit from different intervention materials. One way to identify such subgroups is to use cluster analysis to identify an empirical typology.

Methods

A sample of 346 adults completed surveys through a telephone interview that included questions related to participating in regular exercise. The three variables used in the cluster analysis were the Pros of Exercise, the Cons of Exercise, and Exercise Self-Efficacy.

Results

Six resulting clusters were labeled Disengaged, Immotive, Relapse Risk, Early Action, Maintainers, and Habituated. A series of analyses tested the internal and external validity of the typology. The internal validity test revealed that four of the clusters demonstrated high stability and replicability, while the Relapse Risk and Early Action clusters were less stable. Differences among clusters on self-reported exercise behavior and a strong association with stage of change for regular exercise provided external validity evidence of the typology.

Conclusions

The resulting typology reflects a range of motivational patterns that are likely to be responsive to different types of messages and strategies regarding adoption and maintenance of regular exercise. The typology also generates a number of hypotheses about the identified clusters that can be empirically tested in further studies.

Introduction

Well-established benefits of regular exercise include improved physical and mental well-being, reduction of risk of chronic disease, and the enhancement of physical rehabilitation [1]. However, approximately 12% of all deaths in the United States are related to lack of regular exercise [2]. It has been estimated that 60% of all adults in modern Western societies are not exercising at recommended levels for achieving health benefits [3]. A recent report from CDC indicated that leisure-time physical activity levels remained stable in 1990s with only about one-fourth of U.S. adults meeting recommended levels of physical activity [4]. In addition, roughly 45% of adults reported insufficient levels of activity and 30% reported no physical activity. Health promotion interventions that increase physical activity levels could have a large impact on public health.

Much of the psychosocial research on exercise behavior has focused on the contribution of various determinants of adopting and maintaining a lifestyle of regular exercise. Several comprehensive reviews of the exercise determinant literature have been published [5], [6], [7], [8]. Three broad categories of determinants have been identified: personal attributes, environmental factors, and physical activity itself. Isolating specific determinants, such as gender, education, climate, and family makeup, has limited value because these are static variables that cannot be impacted by behavior change interventions.

Multivariate models such as the Theory of Reasoned Action [9], [10]. Social Cognitive Theory [11], [12], and Transtheoretical Model (TTM) [13], [14], [15] conceptualize the process of changing exercise behavior. Common constructs from these behavior change theories include the benefits and barriers of change and self-efficacy (SE) for making a behavior change [16]. These determinants of physical activity are important mediators in the behavior change process because they have the advantage of being dynamic variables that can be modified through interventions [17], [18], [19], [20], [21]. One of the more promising interventions for health promotion, tailored or computer-based interventions, has been developed on the basis of these theoretical models. Such interventions have been successful for smoking cessation [22], [23], [24], [25], [26], [27] but have not been extensively tested with other behaviors, including exercise.

Tailored interventions require the identification of distinct subgroups that will benefit from substantially different interventions. In this paper we take a novel approach to the analysis of psychosocial determinants of exercise by using cluster analysis to develop a multidimensional typology of motivation for regular exercise. Using measures of three theoretical constructs (Pros and Cons of Exercise and Self-Efficacy), the resulting typology classifies individuals into groups based on the similarity of variable patterns. These patterns may represent important motivational profiles useful for matching exercise interventions to individuals' needs [28].

Construct selection for the cluster analysis was based on the Criterion Measurement Model (CMM), which represents one dimension of the TTM [29]. The multivariate model was proposed as a means of assessing the full spectrum of change as individuals move from being unmotivated to change to maintaining a behavior change. The CMM includes three constructs: Habit Strength, Positive Evaluation Strength, and Negative Evaluation Strength. Positive Evaluation Strength reflects favorable beliefs about a behavior, while Negative Evaluation Strength represents the importance of the negative aspects of engaging in a behavior. The Pros and Cons scales from the Decisional Balance Inventory [30], [31] are appropriate measures of these constructs. The Pros and the Cons are similar to other expectancy-value concepts such as benefits, barriers, and costs [32]. Habit Strength reflects the psychological or learned aspects of a behavior. Self-efficacy, or the level of confidence that regular exercise can be maintained in a variety of situations, can serve as an appropriate conceptualization of habit strength for exercise behavior. The SE construct [11] indicates the perceived learned experiences of a person in specific situations. Specifically, SE is conceptualized as one's confidence to exercise under less than desirable conditions [33], and has been demonstrated to be a predictor of future exercise behavior [34], [35].

The three constructs constituting the CMM are distinct but related and together define a multivariate space in which to characterize motivation for behavior change. The CMM posits two threshold levels for each of the three constructs. The levels of the construct above the upper threshold and below the lower threshold indicate when an individual is less likely to be open to change, while the area between the thresholds defines when an individual may be amenable to change. Looking at the levels of these three constructs together as a multivariate profile may lead to a better understanding of when an individual is ready to move forward in the change process and when an individual may be at risk for a lapse or relapse.

Cluster analysis is an empirical method of defining homogenous subgroups of individuals. It can be viewed as a parsimony procedure that operates in the subject domain in contrast to a parsimony procedure that operates in the variable domain like factor analysis. As such, it is particularly appropriate for identifying subgroups that could benefit from tailored interventions. Cluster analysis creates an empirical typology where the data determine the patterns that form the typology rather than a theoretical typology. Cluster analysis is useful for studying profiles of people rather than simply examining sample averages and variable main effects [36], [37]. Rapkin and Luke [38] discuss how cluster analysis can be used in place of linear model analyses that may misrepresent complex multifaceted relationships as random error variance. When a sample contains a mixture of cases with different combinations of relationships among key variables, linear model analyses may obscure these relationships and treat them as noise. Different types may exist in the sample representing different multivariate profiles. There may be a group that conforms to the aggregate's central tendency on the key variables with the addition of several other groups that represent other patterns. For these groups their values represent “error” with respect to the central tendency of the sample but their pattern of variable relationships is also of importance to understanding the whole sample.

Cluster analysis based on the CMM has been employed in the area of smoking cessation and has produced robust replicable clusters. Two studies have examined subtypes within the stages of change for smoking cessation [39], [40]. These studies found distinct subtypes within each stage of change that shared similar characteristics across the stages of change. In each stage, a profile was found that exemplified that stage, a profile similar to the next stage, and a profile that was similar to the previous stage. Within each stage an unexpected profile was also found that reflected detachment or disinterest in the motivational aspects of smoking. The cluster subtypes were interpreted as strong support for conceptualizing the stages of change as discrete stages. Subtypes that looked more like the expected pattern of the next or previous stage may represent individuals misclassified by the stage of change algorithm. These two studies demonstrated that within stage of change subtypes are robust motivational patterns that could be used for tailoring intervention materials.

This study addressed two questions: (1) Can a meaningful empirical typology of motivation for regular exercise be created? (2) What evidence supports the validity of the resulting cluster groups? First, we conducted two cluster analyses on data from random split halves of a sample of adults. Interpreting the final cluster solution and labeling the clusters defined the resulting empirical typology. We then carried out a series of analyses on the exercise typology to determine its validity. Demonstrating that an empirical typology contains clusters that are theoretically interpretable, are internally consistent, and have expected relationships with other relevant variables provides important evidence for a typologies verisimilitude and acceptance for practical applications [41].

Section snippets

Participants

A sample of 346 adults living in southern New England completed a survey through a random telephone interview. Table 1 provides a summary of the sample characteristics. The sample was 62% women, 95% white, with a median household income between $30,000 and $40,000. Participants were between the ages of 18 and 75 (mean = 43, SD = 15) with a median education level of 14 years. U.S. Census data from 2000 for Rhode Island estimated the state to be approximately 52% female, 87% white, with a median

Results

A cluster analysis includes a series of steps where critical decisions are made at each step [41]. We present the study results as a series of analytic steps. The first section describes preliminary analysis that included data processing and decisions made prior to using cluster analysis. The second section presents the cluster analysis where the number of clusters is determined and interpreted. The third section presents an internal validation analysis of the clusters. The last two sections

Discussion

In this study we created an empirical typology of motivation for a regular exercise from a random sample of adults. Six cluster groups were found through cluster analysis using the pros and cons of exercise and exercise SE. Internal validation analyses revealed that four of the six clusters replicated well while two clusters were unstable. External validation analyses found group differences for self-reported strenuous and moderate exercise behavior and a significant association with exercise

References (67)

  • W.F. Velicer et al.

    An empirical typology of subjects within stage of change

    Addict Behav

    (1995)
  • G.J. Norman et al.

    Cluster subtypes within stage of change in a representative sample of smokers

    Addict Behav

    (2000)
  • G.J. Norman et al.

    Dynamic typology clustering within the stages of change for smoking cessation

    Addict Behav

    (1998)
  • Physical activity and healtha report of the Surgeon General

    (1996)
  • R.R. Pate et al.

    Physical activity and public healtha recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine

    JAMA

    (1995)
  • K. Buxton et al.

    How applicable is the stages of change model to exercise behaviour?

    A review. Health Educ J

    (1996)
  • Physical activity trends—United States, 1990–1998

    MMWR Morb Mortal Wkly Rep

    (2001)
  • J.F. Sallis et al.

    Determinants of exercise behavior

    Exerc Sports Sci Rev

    (1990)
  • R.K. Dishman

    Determinants of participation in physical activity

  • A.C. King et al.

    Determinants of physical activity and interventions in adults

    Med Sci Sports Exerc

    (1992)
  • R.K. Dishman et al.

    Determinants and interventions for physical activity and exercise

  • I. Ajzen et al.

    Understanding attitudes and predicting social Behavior

    (1980)
  • G. Godin

    Theories of reasoned action and planned behaviorusefulness for exercise promotion

    Med Sci Sports Exerc

    (1994)
  • A. Bandura

    Human agency in social cognitive theory

    Am Psychol

    (1989)
  • D.A. Dzewaltowski

    Physical activity determinantsa social cognitive approach

    Med Sci Sports Exerc

    (1994)
  • B.H. Marcus et al.

    The transtheoretical modelapplications to exercise behavior

    Med Sci Sports Exerc

    (1994)
  • J.O. Prochaska et al.

    The transtheoretical modelapplications to exercise

  • S.J. Marshall et al.

    The transtheoretical model of behavior changea meta-analysis of applications to physical activity and exercise

    Ann Behav Med

    (2001)
  • D.J. Plonczynski

    Measurement of motivation for exercise

    Health Educ Res

    (2000)
  • B.M. Pinto et al.

    Physician-based activity counselingintervention effects on mediators of motivational readiness for physical activity

    Ann Behav Med

    (2001)
  • B.C. Bock et al.

    Maintenance of physical activity following an individualized motivationally tailored intervention

    Ann Behav Med

    (2002)
  • W.F. Velicer et al.

    Interactive versus non-interactive interventions and dose–response relationships for stage matched smoking cessation programs in a managed care setting

    Health Psychol

    (1999)
  • V.J. Strecher et al.

    The effects of computer-tailored smoking cessation messages in family practice settings

    J Fam Pract

    (1994)
  • Cited by (0)

    This work was partially supported by Grants CA27821 and CA50087 from the National Cancer Institute. Portions of this paper were presented at the Society of Behavioral Medicine's 20th Annual Sessions, San Diego, March 1999, the American Psychological Association Meeting, Boston, August, 1999.

    View full text