IMPROVING PLANNING INTERVENTIONS: A META-ANALYTIC
REVIEW IN HEALTHY EATING
The present study represents the largest meta-analysis performed to test the efficacy of implementation intentions to achieve healthy eating goals. Our research confirms previous findings and provides additional evidence with implications for future researchers and policy makers. Overall, our results indicate that implementation intentions in the healthy eating domain have a low to medium effect. In the situations where goal attainment is more complicated, i.e. when the intervention targets complex and avoidance behavior goals, implementation intentions show a low effect. However, the intervention boosts behavior change in the case of simple and approach eating goals, i.e. when goal attainment is easier. Our study suggests that to avoid methodological biases when studying the effect of implementation intentions, researchers should
analyze separately experimental and correlational studies, and student and non-student samples.
Keywords: Implementation intentions, Healthy-eating behavior, Meta-analysis Track: Consumer Psychology
1. Introduction
From the Transformative Consumer Research perspective, conducting practical research that can be used by consumers, activists, policy makers and industries to increase the wellbeing of consumer citizens is important (Mick, 2006; Ozanne et al., 2011). Healthy eating is becoming an increasingly serious behavior due to problems resulting from poor dietary habits. Being overweight or obese has a negative effect on quality of life and has significant psychological, sociological and economic costs (Bublitz, Peracchio, & Block, 2010). Despite consumer seem to be motivated to follow a healthy diet, strong intentions do not always lead to corresponding actions (Rhodes & Bruijn, 2013; Sheeran, 2002). In this particular case, our role as consumer researchers is to develop, test and perfect the instruments that are used to design social policies aimed at helping consumers to achieve their healthy eating goals. The current study sought to evaluate the efficacy of implementation intentions (Gollwitzer, 1990) in the healthy eating domain.
Implementation intentions are volitional planning interventions that delegate the control of goal-directed responses to anticipated situational cues that (when actually encountered) elicit these responses automatically. Planning-based interventions mainly involve two different formats: action plans or coping plans. On one hand, action plans specify “when, where, and how” the individual is going to perform a certain behavior leading to goal attainment. On the other hand, coping plans link anticipated critical scenarios with goal-directed responses following the structure of "If situation x arises, then I will perform response y!. For example, “If I go to a restaurant, then I will order a salad as a side dish instead of chips”. Because implementation intentions imply the selection of a suitable response applicable to a future situation (i.e., a good opportunity), it is assumed that the mental representation of this situation becomes highly activated and thus more easily accessible (Gollwitzer, 1990). Previous meta-analyses have shown the efficacy of implementation intentions on health behavior, as shown in Bélanger-Gravel, Godin, & Amireault (2013) and Gollwitzer & Sheeran (2006). In particular, in healthy eating, the average effect of forming implementation intentions is small to medium (Adriaanse, Vinkers, De Ridder, Hox, & De Wit, 2010). Moreover, past evidence shows that the effect size is lower when the intervention aims at reducing unhealthy behaviors, such as non-healthy snacking or fat consumption, since it is easier to initiate a goal behavior than to break an habit (Adriaanse, Vinkers, et al., 2010; Karimi-Shahanjarini, Rashidian, Omidvar, & Majdzadeh, 2013; Verplanken & Faes, 1999). In the last decade there has been a proliferation of research on the effects of planning
interventions in the healthy eating domain (Hagger et al., 2016). The analysis of these studies shows a significant amount of heterogeneity in the effect sizes that could be due to the
substantial differences found in the design of the interventions (Hagger et al., 2016; Sullivan & Rothman, 2008). These differences can be categorized into three groups; first, the type of behavior targeted (approach versus avoidance eating goals and complex versus simple goals); second, methodological issues (experimental versus correlational studies and the type of sample used); third, substantive characteristics refer to the nature of the intervention (implementation intention format: action plan or coping plan, the use of monitoring during the intervention, the use of personalized plans versus prompted plans, the presence of initial training, among others).
Consequently, there is a need to shed light on the efficacy of planning interventions in healthy eating. Moreover, as the method used is a meta-analysis, the authors aim to identify possible moderator variables that may influence the process with the final goal of improving this intervention and propose a research agenda for future studies in the implementation intentions domain. This study makes a contribution to the theory of implementation intention suggesting the possible moderating effect of a new variable unused in previous meta-analysis, goal complexity, although some authors have pointed out their possible influence on the
intervention (Luszczynska, Scholz, & Sutton, 2007; Verplanken & Faes, 1999). Drawing from the literature on task complexity (Liu & Li, 2012), a goal behavior is complex when the individual requires some degree of knowledge on the subject, especially because the course of action to achieve the goal is unclear due to the existence of more than one alternative, and each course of action has a different degree of effectiveness (De Vet, 2007). For instance, people can reduce their fat intake in ways ranging from exchanging a portion of chips for a salad, to eating half of their plate instead of all of it. This complexity contributes to make these goals more abstract, increasing their difficulty in terms of goal attainment. If the person thinks about a task in a concrete way instead of in abstract terms, that would affect the likelihood to completing the task because forming a concrete representation of the task will reduce the procrastination (Liberman & Wakslak, 2007). Adopting a healthier diet is a complex goal per se due to the different tasks that it includes, such as planning, shopping and cooking among others (Benyamini et al., 2013). However, there are eating behaviors more complex than others: it is more complex to reduce fat intake than to increase fruit
consumption (Luszczynska et al., 2007).
As the goal of our research is to verify the efficacy of the use of implementation intentions and in which conditions they work best, our results will be useful for designing efficient health promotion planning interventions to help citizens to improve their healthy eating behavior.
2. Method
2.1. Search and selection of studies
The literature search and data extraction were performed in January 2016. We searched for studies published in Web of Science (Core Collection) and MEDLINE (1990 - January 2016). The search was conducted using the keyword search terms “implementation
intention”, “if-then”, and “action plan*”, in all combinations with the following words: “eat*”, “diet”, and “nutrition”, excluding combinations with the word “plant*”. To increase the scope of our search, cross citations from previous narrative reviews were explored as well (i. e. the review conducted by Adriaanse et al., 2010). The results produced 524 articles for this search (some of them were repeated as Web of Science and Medline have articles in common). From these articles which included different types of healthy eating consumer behavior, we found 69 articles that were deemed relevant to the meta-analysis because they had the following characteristics. First, the studies promoted a healthy diet. Second, the studies assessed all formats of implementation intentions. Third, the papers were published in peer-reviewed journals. Fourth, the use of planning could be either measured or manipulated, and a variety of reliable (food diaries) or less reliable (single item assessments of food intake) outcome measures could be employed as long as the unique effect of planning on eating behavior could be extracted from the results. Finally, the studies were quantitative.
Application of these inclusion criteria resulted in 69 relevant articles describing 91 empirical studies involving 15.661 participants. From those 91 studies, 9 of them measured more than two different food goals (e. g. healthy snacks and unhealthy snacks). Whenever possible, we have used both measures (e. g. to calculate the d for unhealthy snacks and healthy snacks separately) resulting in 103 different measures of d.
2.2. Data collection
This meta-analysis uses Hedges’ standardized mean difference d to measure the effect-size which is the difference between the means for two groups divided by a pooled standard deviation. According to Cohen's (1992) power primer, d = .20 is a “small” effect, d = .50 is a “medium”-sized effect, whereas d = .80 is a “large” effect. In those articles that did not report the effect-size, it was calculated.
Data were abstracted independently by two reviewers (IC and IV) and disagreements were resolved by consensus with a third reviewer (RR). Before analyzing the data set, some decisions were made. First, in some studies, the format of delivery of implementation intentions varied across experimental groups (e.g. including coping plans or barrier identification). Only groups with equivalent interventions were included in order to favor homogeneity with the purpose of isolating the effect of implementation intentions. Second, since only one measure per study could be included in the meta-analysis, only the final follow-up measure was used when multiple follow-ups were conducted.
2.3. Meta-analysis procedure
We determined the effect-size for all the studies selected according to the mentioned criteria using Hedges’ standardized mean difference (d), linear correlation coefficient or standardized regression coefficient depending on data. All these measures were transformed into d following the conversion formulae of Borenstein, Hedges, Higgins, & Rothstein (2011) and Laroche & Soulez (2012). The initial meta-analysis was performed with pooled effect-sizes using the inverse variance statistical method with random effects models (REM). The pooled effect-sizes were reported as d with its 95% CI, and heterogeneity is reported by I2, Q and p-value. In order to search for homogeneity, a second set of analyses was performed to test the possible presence of moderator variables by means of subgroup analyses.
3. Results and discussion
Overall, implementation intention yielded a small-to-medium significant pooled effect size on healthy eating according to Cohen’s (1992) classification (k=91; d=.347; 95% IC= (.226; . 418)), although homogeneity was not found (I2=77.1%, Q=393.7, p= .000).
This effect size is smaller than those reported in previous meta-analysis of health-related behaviors such as Gollwitzer & Sheeran (2006) where the effect size was .65. As said before, eating is a complex goal per se. Previous research has already shown that formulating good quality specific implementation intentions is difficult for people when targeting complex behaviors (Verhoeven, Adriaanse, Vet, & Fennis, 2014). Our results are also lower than those reported by Adriaanse et al (2010) in the healthy eating domain (d =.43). The reason for these differences could be the sample size of their meta-analysis. They used a limited number of studies, most of them related to promote eating behaviors in which implementation intentions have a greater effectiveness (healthy snacking or fruit and vegetable consumption). Since we are using a sample with a wider variety of studies, a lower effect size can be expected. In order to seek for homogeneity, certain category variables were analyzed by means of group analysis.
The first moderator tested was whether the corresponding study was experimental or correlational. Results are shown in table 1. The effect size in the correlational group (.63) more than doubled the effect size in the experimental one (.27), and differences were statistically significant at the 1% level (p=.000). These results seem to be consistent with Gollwitzer and Sheeran meta-analysis which found that correlational studies presented a higher effect size (.70) than experimental studies (.65). As this moderator variable is not related to intervention (but it is methodological). As the homogeneity degree in the
experimental group was larger, we decided to continue the analysis just with those studies in the experimental group, in order to avoid any methodological biases.
Table 1. Moderator Variables Group Analysis
The second moderator tested was the type of sample, establishing 3 groups: group 0,
integrated by healthy people, group 1, composed by unhealthy people (with physical illnesses or overweight problems), and group 3, formed by students. Results are pictured in table 1. Effect size for students is significantly lower than those for group 0 and 1. These results are consistent with data obtained from Gollwitzer & Sheeran (2006). Additionally, the degree of heterogeneity was smaller for students. Previous literature on social science research, states that, in general, student’s responses are more homogeneous than non-students’ and the effect sizes are significantly different but without a systematic pattern (Peterson, 2001). In planning interventions in particular, a possible explanation for a lower effect size in students might be that they could be more motivated for obtaining the credits than for changing their eating behaviors. Consequently, that motivation might be artificial and not enough to promote goal striving.
In order to obtain a clear knowledge of moderator variables, we decided to duplicate the rest of our study, analyzing separately groups 0 and 1 (non-students) and 2 (students). For these two groups, the following moderator variables were explored independently: (1) number of interventions (one intervention/more than one), (2) initial training (yes/no), (3) format (if-then/ when-where-how/ both), (4) monitoring (yes/no), (5) personalization (yes/no), (6) implementation intention check (yes/no), (7) complex goal (simple/complex) and (8) type of behavior (approach/avoidance). Table 1 shows the results for those that produce significant differences across groups.
For students, only the variable format establishes significant differences across groups: when the format is if-then, coping planning, the effect size is lower (.15) than when format is when-where-how, action planning, (.23), that in turn it is lower than when format is complex, including both formats, if-then plus when-where-how (.45). These differences can be explained in part by the fact that writing down a coping plan, or an action plan, is easier and faster than both formats. When participants make the effort to write those complex plans that include both formats, it may be easier for them to remember that plan and consequently act to perform the plan, achieving better results (Sniehotta et al, 2005).
For non-students, the results show the presence of two moderator variables, both of them related to the type of goal behavior. First, the type of behavior gives bigger effect size in the approach group (.42) than in the avoidance one (.22). This result is coherent with previous findings (Adriaanse, Vinkers, et al., 2010; Karimi-Shahanjarini et al., 2013; Verplanken &
Faes, 1999). Second, the complexity of the goal yields larger effect size when the goal is simple (.49) than when it is complex (.24). The results obtained for complex goals are similar to those reported by Bélanger-Gravel et al. (2013) on exercise.
4. Conclusions and Implications
This study represents the largest meta-analysis performed in the healthy diet domain pertaining to the efficacy of forming implementation intentions. Due to the large sample of studies included it assists in our understanding of the role of planning to promote healthy eating and it helps to improve the intervention design. The present study confirms previous findings and provides additional evidence with implications for future researchers and policy makers.
Overall, our results indicate that implementation intentions have a lower effect than expected. The study of moderator variables shows that it is even lower if we exclude correlational studies. One reason for these low results might be the sample used in the interventions, mainly students. The intervention has a significant lower effect in this population especially when they formulate coping plans instead of more elaborate plans.
The intervention increases its efficacy when it is used in non-student samples although the eating behavior targeted moderates the effect of the intervention. Volitional interventions aim at reducing the behavior gap, especially in difficult situations (Gollwitzer & Brandstätter, 1997). Nevertheless, in the light of our results, this is not the case. In the situations where goal attainment is more complicated, i.e. when they target complex and avoidance behavior goals, implementation intentions show a low effect. However, the intervention boosts
behavior change in the case of simple and approach goals, i.e. when goal attainment is easier. More research is required to understand why the cognitive link between the critical situation and the response is stronger in the case of certain eating goals, in the vein of the work of Adriaanse, Oettingen, et al. (2010) including the role of goal complexity.
Our results have several implications for future studies. To avoid methodological biases when studying the effect of implementation intentions, future researchers should analyze separately experimental and correlational studies, and student and non-student samples.
The ultimate purpose of our research was to help policy makers design interventions intended to help people achieve their goals of healthy eating. The intervention will be reasonably helpful when used to increase simple healthy behaviors, such as fruit and vegetable intake, but not much to for diminish complex unhealthy behaviors, e.g. fat consumption.
References
The references do not present the articles included in the meta-analysis, due to the lack of space. Any person interested on them can contact the first author.
Adriaanse, M. A., Oettingen, G., Gollwitzer, P. M., Hennes, E. P., De Ridder, D. T. D., & Wit, J. B. F. (2010). When planning is not enough : Fighting unhealthy snacking habits by mental contrasting with implementation intentions ( MCII ),1293(March), 1277–1293 Adriaanse, M. A., Vinkers, C. D. W., De Ridder, D. T. D., Hox, J. J., & De Wit, J. B. F. (2010). Do implementation intentions help to eat a healthy diet? A systematic review and meta-analysis of the empirical evidence. Appetite, 56(1), 183–193.
Bélanger-Gravel, A., Godin, G., & Amireault, S. (2013). A meta-analytic review of the effect of implementation intentions on physical activity. Health Psychology Review, 7(1), 1–32 Benyamini, Y., Geron, R., Steinberg, D. M., Medini, N., Valinsky, L., & Endevelt, R. (2013). A Structured Intentions and Action-Planning. American Journal of Health Promotion, 28(2), 119–128
Borenstein, M., Hedges, L. V., Higgins, J. P., & Rothstein, H. R. (2011). Introduction to meta-analysis. John Wiley & Sons.
decision making and dietary restraint. Journal of Consumer Psychology, 20(3), 239–258
Cohen, J. (1992). Quantitative Methods in Psychology. Psychological Bulletin,112(1),155–159
De Vet, E. (2007). Implementation intentions and diet. Journal of Psychosomatic Research, 63(5), 499–500.
Gollwitzer, P. M. (1990). Action phases and mind-sets. In E. Tory & M. R. Sorrentino (Eds.), Handbook of motivation and cognition: Foundations of social behavior, Vol. 2 (pp.53–92). New York: Guilford Press.
Gollwitzer, P. M., & Brandstätter, V. (1997). Implementation intentions and effective goal pursuit. Journal of Personality and Social Psychology, 73(1), 186–199.
Gollwitzer, P., & Sheeran, P. (2006). Implementation Intentions and Gol Achievement: a Meta-Analysis of Effects and Processes. Advances in Experimental Social Psychology. Hagger, M. S., Luszczynska, A., Wit, J. De, Benyamini, Y., Burkert, S., Chamberland, P.,Gollwitzer, P. M. (2016). Implementation intention and planning interventions in Health Psychology : Recommendations from the Synergy Expert Group for research and practice. Psychology & Health, 31(7), 814–839.
Karimi-Shahanjarini, A., Rashidian, A., Omidvar, N., & Majdzadeh, R. (2013). Assessing and comparing the short-term effects of TPB only and TPB plus implementation intentions interventions on snacking behavior in Iranian adolescent girls: a cluster randomized trial. American Journal of Health Promotion, 27(3), 152–162.
Laroche, P., & Soulez, S. (2012). La méthodologie de la méta-analyse en marketing. Recherche at Applications En Marketing, 27(1/2012), 79–105.
Liberman, N., & Wakslak, C. (2007). Construal Level Theory and Consumer Behavior. Journal of Consumer Psychology, 17(2), 113–117.
Liu, P., & Li, Z. (2012). Task complexity : A review and conceptualization framework. International Journal of Industrial Ergonomics, 42(November), 553–568.
Luszczynska, A., Scholz, U., & Sutton, S. (2007). Planning to change diet : A controlled trial of an implementation intentions training intervention to reduce saturated fat intake among patients after
myocardial infarction. Journal of Psychosomatic Research, 63, 491–497
Mick, D. G. (2006). Meaning and Mattering Through Transformative Consumer Research. Advances in Consumer Research, 33.
Ozanne, A. S. J., Pettigrew, S., Crockett, D., Firat, A. F., Downey, H., & Pescud, M. (2011). The Practice of Transformative Consumer Research - Some Issues and Suggestions. Journal of Research for Consumers, (19), 1–7.
Peterson, R. A. (2001). On the use of college students in social science research: Insights from a second-order meta-analysis. Journal Of Consumer Research, 28(3), 450-461. Rhodes, R. E., & Bruijn, G. De. (2013). How big is the physical activity intention – behaviour
gap ? A meta-analysis using the action control framework. British Journal of Health Psychology,
18, 296–309.
Sniehotta, F. F., Schwarzer, R., Scholz, U., & Schüz, B. (2005). Action planning and coping planning for long‐term lifestyle change: theory and assessment. European Journal of Social Psychology, 35(4), 565-576.
Sheeran, P. (2002). Intention — Behavior Relations : A Conceptual and Empirical Review. European Review of Social Psychology, (September 2013), 37–41.
Sullivan, H. W., & Rothman, A. J. (2008). When planning is needed: Implementation intentions
and attainment of approach versus avoidance health goals. Health Psychology : Official Journal of the
Division of Health Psychology, American Psychological Association, 27(4), 438–44.
Verhoeven, A. A. C., Adriaanse, M. A., Vet, E. De, & Fennis, B. M. (2014). Identifying the “if ” for “if-then” lans: Combining implementation intentions with cue-monitoring targeting unhealthy
snacking behaviour, 29(12), 1476–1492.
Verplanken, B., & Faes, S. (1999). Good intentions , bad habits , and effects of forming