Health Education Research Advance Access originally published online on January 4, 2005
Health Education Research 2005 20(3):294-297; doi:10.1093/her/cyg125
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Health Education Research Vol.20 no.3, © Oxford University Press 2004; All rights reserved
Commentary: Revitalizing research on health behavior theories
1 Department of Human Ecology, Cook College, Rutgers University, New Brunswick, NJ 08901-8520 and 2 Department of Psychology, University of Minnesota, Minneapolis, MN 55455, USA
3 Correspondence to: N. D. Weinstein; E-mail: neilw{at}aesop.rutgers.edu
 |
Introduction |
|---|
 Top Introduction What have we been...How do we move...References |
|---|
Noar and Zimmerman's article ‘Health behavior theory and cumulative knowledge regarding health behavior: are we moving in the right direction?’ (Noar and Zimmerman, 2005
) provides an informative, but disturbing, analysis of research in this area. Despite thousands of studies that use or test specific theories of health behavior [see (Noar and Zimmerman, 2005), Figure 1], innovations and advances have been quite modest. The extent to which theories have become more accurate is uncertain. In fact, comparing initial statements of the dominant theories of health behavior with current descriptions of those same theories reveals remarkably few substantive changes. Why has there been so little progress? We believe that there are two broad classes of explanations: weaknesses in the research designs and analyses used in much of this research, and problems with how investigators approach the overall research enterprise.
|
What have we been doing? Problems in research design and analysis |
|---|
|
TopIntroductionWhat have we been...How do we move...References |
|---|
Tests of health behavior theories too often fail to give careful attention to a theory's assumptions, to the operationalization of theory constructs and to the limitations of the research designs used. Furthermore, as Noar and Zimmerman demonstrate, even very similar theories—theories with many of the same constructs, applying to the same behaviors and amenable to the same types of evaluations—are seldom pitted against one another to determine whether one offers a superior explanation of behavior, neither has there been much effort to determine whether constructs that appear quite similar are redundant or not [see also (Weinstein, 1993
)]. Noar and Zimmerman were able to identify 19 articles that compared two or more health behavior theories, and, at least in this respect, are exemplary. Yet, a closer look reveals that even these studies have weaknesses that limit their ability to inform health behavior theory and that are common in health behavior research.
Problems with research designs
First, all of these theory tests (and probably the overwhelming majority of the studies represented in Figure 1 of Noar and Zimmerman's article) rely on correlational data. Although we are quick to remind our students and colleagues that correlation is not causation, we seem to forget this warning when designing our own studies and interpreting our own findings. It is almost as if there is a silent conspiracy to hold tests of health behavior theory to a lower standard than we expect in other areas of science.
Second, of the 21 different studies reported in the 19 articles, 13 rely solely on cross-sectional data. Their static tests determine only whether ‘explanatory’ constructs, such as risk perceptions or self-efficacy, are associated with intentions or behavior in the expected direction. They provide no evidence that these constructs are determinants of behavior, which is typically the theoretical question that motivated the research in the first place.
Not only will beliefs affect action, but our experiences with these actions will also shape how we think and feel about them. We may also develop post hoc explanations and justifications for our actions. Thus, in most cases, we should expect a reciprocal relationship between health beliefs and health behavior. These reciprocal relationships are especially likely with frequent, ongoing behaviors, such as condom use, exercise, diet and smoking. Improvements in health behavior theory require us to disentangle these processes, not merge them together in cross-sectional analyses. Prospective studies, especially when they concern the adoption of new behaviors, are likely to tell us more about causal relationships than cross-sectional studies.
Third, five of the 21 studies were satisfied with studying intentions to act rather than action itself. Although investigating the mediators proposed by a theory is an essential aspect of theory testing, intentions can never substitute for behavior. To cite examples from among the theory-comparing studies reported by Noar and Zimmerman, Conner and Norman (Conner and Norman, 1994) found a correlation of only 0.15 between intentions to attend a health screening and actual attendance, and Bish et al. (Bish et al., 2000) found a correlation of only 0.17 between women's intentions to get cervical screening and their behavior. Although these small correlations are distressing in and of themselves, they also offer compelling evidence for why we need to be sure that each step in the causal paths contained in a theory receives careful empirical scrutiny.
Problems with theory tests
When these 19 articles do compare different theories, the most common approach is to determine how much variance in behavior (or intentions) is ‘explained’ by each theory and then to compare these two quantities. If theory A explains 28% of the variance and theory B explains only 20%, the former is deemed the better theory. This approach to theory comparison is grossly inadequate.
The ‘biggest variance’ approach to theory testing is based on the implicit assumption that the theory that explains the most variance explains everything that the other theory does, plus more. This assumption is obviously unjustified. There are several possibilities. In our example, theories A and B might identify completely independent predictors, so that a combined theory would explain 48% of the variance in behavior. Theories A and B might identify partially independent predictors, so that a combined theory would explain less than 48% of the variance, but more than 28%. Finally, it might be that that the issues raised by theory B are already fully represented by theory A, so that a combined theory would explain only 28% of the variance. Assuming that a combined theory does make a significant improvement in the total variance explained, the researcher must determine, by further, more detailed analyses, what element or elements of theory B improve theory A. These elements might consist of different constructs or interactions among constructs, or they might consist of different relationships (e.g. non-linear) between the constructs and the behavioral outcome.
Most, if not all, researchers know how to conduct these tests. For example, In testing the Theory of Reasoned Action against the Theory of Planned Behavior [e.g. (Armitrage and Conner, 2001)], investigators assess the change in total variance in behavior accounted for by adding the construct ‘perceived behavioral control’ (PBC) to behavioral intentions; they look for possible interactions between intentions and PBC, and they look for contributions of PBC to the prediction of intentions. Similarly detailed analyses are needed when testing other theories.
Beyond these essential aspects of theory comparison, it is important to recognize that there are many ways of being ‘better’. A theory might predict well in one domain (e.g. a particular population, setting or health issue), but do poorly in another. A theory might explain more variance than another, but use variables that cannot be modified and therefore be less useful for interventions. A theory might explain little variance between individuals, but be very useful because it does a good job of predicting the overall level of behavior. We need to be sure that our research activities allow us to capture this more inclusive sense of what is ‘better.’
|
How do we move forward? Re-examining our approach to health behavior research |
|---|
|
TopIntroductionWhat have we been...How do we move...References |
|---|
Although improvements in research designs are always desirable, we believe that progress in theory development requires more fundamental changes in how investigators approach their research goals. First and foremost, investigators need to act in accord with what they know to be true: theories are not static entities to be used as initially proposed, but rather are dynamic entities that should evolve over time. Theory improvement is a cyclical process that involves the specification of relations between factors, the testing of those relations, the re-specification or rejection of initially hypothesized principles and the testing of the new relations. With each iteration of this cycle, the theory should become better articulated—clarifying the underlying processes and the conditions under which they are most likely to appear—and more accurate. Although this view of the research process is well-known and consistently advocated (Rothman, 2004
; Suls and Rothman, 2004), the research activities of the field, as illustrated by Noar and Zimmerman's review, show few signs of such a process.
If health behavior research is to inform theory development, investigators need a more programmatic approach to their research endeavors. A single study, no matter how well done, can provide only a very limited test of a theory. Both in planning their projects and in describing their findings, investigators need to better articulate the manner in which studies build on each other. As a requirement for publication, research reports should state explicitly how a given study advances what is currently known about a particular theory. Keeping track of the broader research endeavor should also help investigators recognize when their own and their colleagues' efforts have focused on only a particular aspect of a given theory, operated in a particular context, or relied on a particular methodology. It would also help prevent situations from arising in which an enormous quantity of research nevertheless provides a very narrow or incomplete test of a theory [as in the excessive reliance on cross-sectional data to test the hypothesis that risk perceptions motivate subsequent precautionary behavior (Gerrard et al., 1996)].
When investigators operate in parallel with or independently of one another, the result is likely to be a duplication of efforts and the development of measures and specification of constructs that may not be meaningfully discrete from each other. Although there are institutional and professional pressures to emphasize one's own contributions to an area of research, advancements in health behavior theory will likely come much more quickly if the activities of individual investigators build on each other. To rephrase a popular sentiment, it takes a village to raise a theory.
Another step that would speed the refinement of health behavior theories would be to urge theorists to be as precise as possible in specifying the causal relations among the components in their models. General assertions that variables are related (e.g. a list of variables hypothesized to predict a behavioral outcome) are difficult to disconfirm. For example, the Theory of Reasoned Action (Fishbein, 1980) and the Theory of Planned Behavior (Ajzen, 1985) both state that people's behavioral intentions are a function of their attitudes toward the behavior and of their subjective norms about the behavior, but they allow the relative contribution of these two factors to be fit empirically in each situation. Although this level of specification is adequate as an initial step, one would hope that the theory would be refined over time, and make more precise predictions about the factors that determine how and when each of these two factors affect behavioral intentions.
Similarly, theorists need to be more willing to address the limitations of their models. For example, is the Health Belief Model (Becker, 1974) as relevant to general health behaviors, such as exercise, as it is to behaviors directed toward specific health threats. Are stage theories needed [e.g. (Weinstein and Sandman, 2002)] in situations where potential actions arise and are decided very quickly, as with treatment choices in a doctor's office? Precise predictions are not only more testable, they make it easier to use study results to improve a theory. Laying out specific predictions increases the chance that theorists will find that their assertions are wrong, but it also increases the chance that they will know when a theoretical premise is right.
The volume of research revealed by Noar and Zimmerman demonstrates there is enormous interest in health behavior and widespread use of health behavior theories in this research. Investigators have committed and continue to commit time and effort to understanding the determinants of health behavior. The challenge for the future is to optimize the cumulative knowledge that will arise from the continued work in this area. We do not believe that what we regard as limited progress to date is due to theories that are untestable or to behavioral phenomena that are too complex to understand. Rather, greater progress will appear when we commit to making rigorous use of the methods we already have, recognize the dynamic nature of theory development, and embrace the need for both collaboration and continual reassessment of our research endeavors.
|
References |
|---|
|
TopIntroductionWhat have we been...How do we move...References |
|---|
Ajzen, I. (1985) From intentions to actions: a theory of planned behavior. In Kuhl, J. and Beckmann, J. (eds), Action-Control: From Cognition to Behavior. Springer, Heidelberg, pp. 11–39.
Armitrage, C.J. and Conner, M. (2001) Efficacy of the theory of planned behaviour: a meta-analytic review. British Journal of Social Psychology, 40, 471–499.[CrossRef][Web of Science][Medline]
Becker, M.H. (1974) The health belief model and personal health behavior [Special Issue]. Health Education Monograph, 2(4).
Bish, A., Sutton, S. and Golombek, S. (2000) Predicting uptake of a routine cervical smear test: a comparison of the health belief model and the theory of planned behaviour. Psychology and Health, 15, 35–50.
Conner, M. and Norman, P. (1994) Comparing the health belief model and the theory of planned behavior in health screening. In Rutter, D.R. and Quiine, L. (eds), Social Psychology and Health: European Perspectives. Avebury/Ashgate, Brookfield, VT, pp. 1–24.
Fishbein, M. (1980) A theory of reasoned action: some applications and implications. In Howe, H.E., Jr and Page, M.M. (eds), Nebraska Symposium on Motivation, 1979. University of Nebraska Press, Lincoln, NB, vol. 27, pp. 65–116.
Gerrard, M., Gibbons, F.X. and Bushman, B. (1996) The relation between perceived vulnerability to HIV and precautionary sexual behavior. Psychological Bulletin, 119, 390–409.[CrossRef][Web of Science][Medline]
Noar, S.M. and Zimmerman, R.S. (2005) Health behavior theory and cumulative knowledge regarding health behaviors: are we moving in the right direction? Health Education Research, 10.1093/her/cyg113.
Rothman, A.J. (2004) Is there nothing more practical than a good theory? Why innovations and advances in health behavior change will arise if interventions are used to test and refine theory. International Journal of Behavioral Nutrition and Physical Activity, 1, 11.
Suls, J. and Rothman, A.J. (2004) Evolution of the psychosocial model: implications for the future of health psychology. Health Psychology, 23, 119–125.[Medline]
Weinstein, N.D. (1993) Testing four competing theories of health-protective behavior. Health Psychology, 12, 324–333.[CrossRef][Web of Science][Medline]
Weinstein, N.D. and Sandman, P.M. (2002) The precaution adoption process model and its application. In Glanz, K., Rimer, B.K. and Lewis, F.M. (eds), Health Behavior and Health Education: Theory, Research and Practice, 3rd edn. Jossey Bass, San Francisco, CA, pp. 121–143.
Received on August 31, 2004; accepted on September 8, 2004
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Kobrin, K. L. Hall, and R. T. Croyle Advancing Theory in Cancer Screening and Beyond: A Conversation Across Fields Health Educ Behav, October 1, 2009; 36(5_suppl): 7S - 10S. [PDF]
|
|
|
|
|
|
A. J. Rothman Capitalizing on Opportunities to Refine Health Behavior Theories Health Educ Behav, October 1, 2009; 36(5_suppl): 150S - 155S. [PDF]
|
|
|
|
 |
|
 C. Keller, J. Fleury, S. Sidani, and B. Ainsworth Fidelity to Theory in PA Intervention Research West J Nurs Res, April 1, 2009; 31(3): 289 - 311. [Abstract] [PDF]
|
|
|
|
 |
|
 B. Schuz, F. F. Sniehotta, N. Mallach, A. U. Wiedemann, and R. Schwarzer Predicting transitions from preintentional, intentional and actional stages of change Health Educ. Res., February 1, 2009; 24(1): 64 - 75. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. S. Amin and R. L. Harrison Understanding Parents' Oral Health Behaviors for Their Young Children Qual Health Res, January 1, 2009; 19(1): 116 - 127. [Abstract] [PDF]
|
|
|
|
|
|
E. de Vet, J. de Nooijer, N. K. de Vries, and J. Brug Comparing stage of change and behavioral intention to understand fruit intake Health Educ. Res., August 1, 2007; 22(4): 599 - 608. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. R. Sorenson Theory and health behavior: are we making progress? Health Educ. Res., June 1, 2005; 20(3): 274 - 274. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||