In this study we examined whether the theory of planned be- havior (Ajzen, 1985, 1991) can explain the adoption and mainte- nance of a new health behavior.
Copyright 2001 by the American Psychological Association, Inc. 0278-6133/01/S5.00 DOI: 10.1037//0278-6133.20.1.12
Health Psychology 2001, Vol. 20, No. 1, 12-19
Can the Theory of Planned Behavior Explain Patterns of Health Behavior Change? Mark Conner University of Leeds
Paschal Sheeran University of Sheffield Paul Norman University of Sheffield
This article tested the ability of the theory of planned behavior (TPB) to predict patterns of behavior change associated with health screening. Attitudes, subjective norms, perceived behavioral control, and intentions were used to predict objective measures of attendance 1 month and 13 months later among participants who had never previously been screened (N = 389). Findings showed that the TPB predicted attendance on each occasion and also predicted frequency of attendance. However, the model was unable to reliably distinguish among consistent attendees, participants who delayed attending, and participants who initially attended but relapsed. Thus, the TPB needs to be extended to understand behaviors that must be performed promptly and repeatedly for health benefits to accrue. Key words: planned behavior, health screening, relapse
ior are their behavioral intentions concerning its performance (e.g., "I intend to use a condom the next time that I have sex"). Intentions, in turn, are predicted by three variables: Attitudes, subjective norms, and perceived behavioral control. Attitude refers to people's overall evaluation of their performing the behavior (e.g., "For me, using a condom the next time that I have sex would be good/bad"), and subjective norm refers to perceptions of social pressure from significant others to perform the behavior (e.g., "Most people who are important to me think that I should use a condom the next time that I have sex"). Perceived behavioral control is broadly equivalent to Bandura's (1977) concept of self-efficacy (Ajzen, 1998) and refers to people's appraisals of their ability to perform a behavior (e.g., "For me to use a condom the next time that I have sex would be easy/difficult"). The more positive people's attitudes and subjective norms are regarding a behavior, and the greater their perceived behavioral control, the stronger people's intentions to perform the behavior will be. Similarly, the stronger people's intentions, and the greater their perceived behavioral control, the more likely it is that people will perform the behavior. The theory of planned behavior assumes that intentions and perceived behavioral control mediate the effects of attitudes, subjective norms, and external variables (e.g., age, gender, socioeconomic status) on behavior. Accumulated evidence indicates that attitudes, subjective norms, and perceived behavioral control are reliable predictors of intentions to perform health behaviors and generally account for 40%-50% of the variance in meta-analytic reviews (e.g., Armitage & Conner, in press; Conner & Sparks, 1996; Farley, Lehmann, & Ryan, 1981; Godin & Kok, 1996; Hausenblas, Carron, & Mack, 1997; Sheeran & Taylor, 1999). Similarly, meta-analyses indicate that intentions and perceived behavioral control typically explain between 20% and 40% of the variance in health behaviors in
In this study we examined whether the theory of planned behavior (Ajzen, 1985, 1991) can explain the adoption and maintenance of a new health behavior. Previous research indicates that the theory of planned behavior should be capable of explaining both performance versus nonperformance, and the frequency of performance, of a health behavior (e.g., Conner & Sparks, 1996; Godin & Kok, 1996). However, performing a novel precautionary behavior can involve discrete patterns of behavior change that embrace problems with getting started on the new behavior and consistently performing that behavior that are over and above the problems of performance versus nonperformance and frequency of performance. Because the theory of planned behavior is primarily an account of goal setting rather than goal pursuit (cf. Mischel, Cantor, & Feldman, 1996), this model may be ill equipped to explain these patterns of behavior change. In the present research we tested this analysis in a longitudinal study of attendance for health screening.
The Theory of Planned Behavior The theory of planned behavior proposes that the most immediate and important predictor of whether people perform a behavPaschal Sheeran and Paul Norman, Department of Psychology, University of Sheffield, Sheffield, United Kingdom; Mark Conner, School of Psychology, University of Leeds, Leeds, United Kingdom. We thank Graham Clarke and Seraphim Alvandes of the Department of Geography, University of Leeds, United Kingdom, for coding the data relating to socioeconomic status. Correspondence concerning this article should be addressed to Paschal Sheeran, Department of Psychology, University of Sheffield, Sheffield S10 2TN, United Kingdom. Electronic mail may be sent to p.sheeran@ sheffield.ac.uk.
12
PATTERNS OF BEHAVIOR CHANGE prospective studies (e.g., Armitage & Conner, in press; Conner & Sparks, 1996; Godin & Kok, 1996; Hausenblas et al., 1997; Sheeran & Orbell, 1998). Evidence also suggests that perceived behavioral control contributes a significant increment in the variance explained in behavior after controlling for intentions (Armitage & Conner, in press; Godin & Kok, 1996; Hausenblas et al., 1997). In sum, the predictive validity of the theory of planned behavior is well established in health psychology research.
The Theory of Planned Behavior and Patterns of Health Behavior Change The present research was designed to examine predictors of attendance for health screening at two time points over a 13-month period. Health screening involves a once-yearly examination of patients by their general practitioner coupled with person-specific advice on appropriate changes in health-protective behaviors (e.g., exercise, diet, smoking). This program was part of a new health initiative by the U.K. government (Department of Health and Welsh Office, 1989), and participants had never previously attended a health screening (thus, the study involves a novel health behavior). The main aim of the study was to determine whether intentions, attitudes, subjective norms, and perceived behavioral control predict whether people attend for health screening at each time point. This was the question of attendance versus nonattendance. We also examined whether theory-of-planned-behavior variables predict how often participants attend, that is, whether participants never attended, attended once, or attended twice. This was the question of frequency of attendance. However, another question is also important: Can the theory of planned behavior predict patterns of attendance! Table 1 helps to clarify this issue. Participants can attend or not attend at two time points. If participants attend at both time points, then their attendance can be characterized as consistent. If participants fail to attend at both time points, then they can be said to have refused attendance. Two patterns of inconsistent attendance can also be discerned in Table 1. Participants may attend at Time 1 but not at Time 2 and can be characterized as exhibiting initial attendance, or participants can attend at Time 2 but not Time 1 and be characterized by delayed attendance. In sum, four qualitatively different patterns of attendance behavior are possible—and these patterns are not reducible to performance-versus-nonperformance or frequency-of-performance dimensions.1 The issues of initial and delayed attendance are important in the context of health behaviors because, as Sutton (1994) pointed out, most health behaviors must be repeated for health benefits to accrue (e.g., exercise, quitting smoking). Similarly, delaying the initiation of recommended behavioral changes could have poten-
Table 1 Patterns of Attendance Obtained by Decomposing Attendance Behavior at Two Time Points Attendance at Time 1 Attendance at Time 2
Yes
No
Yes No
Consistent Initial
Delayed Refused
13
tially serious consequences for individuals' health. Thus, knowing what variables distinguish among these patterns of attendance has both theoretical and practical significance. The proposal we tested is that the theory of planned behavior will not be able to discriminate among these patterns of behavior change. This proposal derives from the fact that the theory of planned behavior is primarily a model of intention formation (goal setting) that specifies the role of attitudes, subjective norms, perceived behavioral control, and external variables. The theory of planned behavior is not an account of the processes responsible for ensuring that intentions are turned into action quickly or effectively or of how behavior change is maintained over time (goal pursuit; cf. Mischel et al., 1996). No previous research appears to have examined this potential limitation of the theory of planned behavior. The hypotheses tested in this study, then, are that the theory of planned behavior will be able to predict attendance versus nonattendance, and frequency of attendance, for health screening; however, the theory will not be able to distinguish among the patterns of attendance described in Table 1.
Method Participants and Procedure A random sample of never-screened patients at a single general medical practice in rural England participated in the study. Data were collected at three time points over a 13-month period. At Time 1, theory-of-plannedbehavior variables were measured in a postal questionnaire that respondents had to return within 1 month. One month after the questionnaires had been distributed, participants received a postal invitation to attend a health screening from their medical practitioner, and whether they attended was recorded in their medical records (Time 2 behavior). Twelve months after the original invitation, participants received a second postal invitation, and their subsequent attendance was again recorded (Time 3 behavior). Eight hundred eighteen patients received questionnaires and invitations to attend, of whom 407 returned the questionnaire—a response rate of 49.8%. Respondents and nonrespondents were compared on demographic variables and attendance rates to determine representativeness. The data in Table 2 indicate that, compared to nonrespondents, respondents were more likely to be women (58% vs. 46%) and more likely to attend the health screening at both Time 2 (56% vs. 38%) and Time 3 (64% vs. 41%). Implications of these findings are outlined in the Discussion section. Eighteen participants (4.42%) had missing values on key variables, which left a final sample of 389. Ages ranged from 30 to 41 years (M = 36.2, SD — 3.2), because the timing and delivery of health screening was stratified by age, and patients aged 30-41 years were the first group that was offered screening.
1
The term patterns of behavior change is useful for distinguishing the present analysis from the concept of "stages" of change in the transtheoretical model (e.g., Prochaska et al., 1992). Stages of change are assumed to characterize a range of behaviors and embrace issues concerned with thinking about behavior change (precontemplation, contemplation), preparation for change, action, and maintenance of change. The patterns of change examined here are more specific to the issue of attendance behavior, although this analysis is likely to be applicable to several important health concerns (e.g., cancer screening, exercise or weight reduction programs). The issue of maintenance/relapse is examined here in terms of initial versus consistent attendance, and the present analysis also addresses the problem of delay behavior (which is not formally specified by the transtheoretical model).
14
SHEERAN, CONNER, AND NORMAN Table 2 Comparisons of Demographic Characteristics and Attendance Rates for Respondents Versus Nonrespondents Nonrespondents
Respondents
Characteristic
M
SD
M
SD
Difference
Age (years) SESa Sex (% women) Time 1 attendance (%) Time 2 attendance (%)
36.53 0.98
3.45 0.94
36.39 0.89
3.39 0.94
r(817) = 0.61
