Development of a dynamic framework to explain population patterns of leisure-time physical activity through agent-based modeling

Despite the increasing body of evidences on the factors influencing leisure-time physical activity, our understanding of the mechanisms and interactions that lead to the formation and evolution of population patterns is still limited. Moreover, most frameworks in this field fail to capture dynamic processes. Our aim was to create a dynamic conceptual model depicting the interaction between key psychological attributes of individuals and main aspects of the built and social environments in which they live. This conceptual model will inform and support the development of an agent-based model aimed to explore how population patterns of LTPA in adults may emerge from the dynamic interplay between psychological traits and built and social environments. We integrated existing theories and models as well as available empirical data (both from literature reviews), and expert opinions (based on a systematic expert assessment of an intermediary version of the model). The model explicitly presents intention as the proximal determinant of leisure-time physical activity, a relationship dynamically moderated by the built environment (access, quality, and available activities) – with the strength of the moderation varying as a function of the person’s intention– and influenced both by the social environment (proximal network’s and community’s behavior) and the person’s behavior. Our conceptual model is well supported by evidence and experts’ opinions and will inform the design of our agent-based model, as well as data collection and analysis of future investigations on population patterns of leisure-time physical activity among adults. Electronic supplementary material The online version of this article (doi:10.1186/s12966-017-0553-4) contains supplementary material, which is available to authorized users.

Operational definition of constructs and meaning of the relationships contained in the version of the conceptual model assessed by experts. Model's delimitation: our conceptual model tries to encompass the main psychological and environmental variables and mechanisms that might be involved in the emergence of collective patterns of LTPA practice among adults. Some aspects (such as demographic attributes and constructs related to the volitional phase of behavior adoption) were not included for the sake of simplicity, and because we feel they are not highly relevant to our research questions. 1 Self-efficacy (person's self-perception about his/her competence to execute or control the behavior. It comprises barrier self-efficacyconfidence to overcome possible barriers to performing repeated bouts of physical activityand task self-efficacyconfidence to perform the specific physical activity act itself) is a function of: a) Social network's habit (leisure-time physical activity habit of those people with whom the person has proximal relationship; similar to social support), acting on barrier self-efficacy; b) Socioeconomic status (person or group's economic and social position in relation to others), acting on barrier self-efficacy; c) Habit (person's regularity of leisure-time physical activity practice), acting on task selfefficacy; d) Perceived environment to practice (person's perceptions about available places for leisuretime physical activity practice and their featuresavailable activities, quality and condition, and accessibilitycombining spatial distance and affordability), acting on barrier self-efficacy. 2 Attitude (person's evaluation and expectation about the behavior and its outcomes. It comprises affective attitudeenjoyment and pleasure expected from physical activity -, instrumental attitudeutility of physical activity practice -, and social reactionsocial approval or disapproval the behavior produces in one's interpersonal relationships) is a function of: a) Social network's habit, acting on instrumental attitude and social reaction; b) Community's habit (leisure-time physical activity habit of those people living in the same relatively large, geographically delimited settlement, such as a city; similar to social norm), acting on social reaction; c) Habit, acting on affective and instrumental attitudes. 3 Intention (person's conscientious inclination to practice leisure-time physical activity) is a function of: a) Attitude; b) Self-efficacy. 4 Features of the places where leisure-time physical activity is practiced is a function of socioeconomic status. 5 Perceived environment to practice is a function of: a) Intention; b) Features of the places where leisure-time physical activity is practiced. 6 Habit is a function of: a) Intention; b) Features of the places where leisure-time physical activity is practiced. 7 Social environment (social network and community's habit) is a function of the habit of each person. Would change the term "habit" to "behavior" I'm not certain what "perceived environment to practice" means In general, the model covers psychological variables better than environmental/policy variables and the latter could be improved.
-3 4 I'm not quite sure but time is sometimes a constraint to LTPA. Secondly, it seems that acessibility may overlap the idea of facilities (having facilities), which, in fact are not the same.
-Well done. Thank you 4 4 I think that 'habit' is developed after initiation of a behavior. In my opinion, intention is enough to initiate a behavior, but a setting that facilitates the behavior is needed for the maintenance/habit. Given that individual-level interventions focusing on motivation are often successful in the short term, but sustaining behavioural change is a real challenge, you may want to clarify that intention and environment play a complementary role in the process toward achieving a habit.
I am not sure about dividing 'social environments' into social network and community (these two sound similar). I wonder if you could simply list this as something like 'social norm'.
Intention is considered to influence people's environmental perceptions (5.a). I assume that people with and without intention see the environment differently. But, some particular settings may provide a cue for physical activity. In such a case, environmental perception may also influence intention. Secondly, perception of the environment is affected by previous experiences in the environment, which are also related to the time one has been living in that area or neighborhood. Therefore, these two variables should be better explained in the model.
I'm not convinced by the actual model and the references included that "habit" is a good description for the outcome variable. What do you want to predict? Participation in any activity? The time or the frequency of LTPA?
The definition of habit is not clear and I'm not even sure this is applicable to leisure physical activity since this is not a very stable behavior. For instance, if for any reason you feel ill and can't practice any PA for one or two weeks how would you classify this person? If you enjoy being active once a month doing a monthly field trip for camping would this be enough for you to be classified and having a PA habit? This is critical and should be really refined or excluded to better reflect the outcome variable.
-6 3 The interplay among various domains of physical activity should be considered. For example, someone who travels by foot for four miles each day (to and from work for example) may not be physically active in leisure time.
-This is interesting work and should be continued. I do hope you also incorporate systems thinking into your model as it matures. Thank you for including me. You should not exclude any constructs My only concern in the way you have developed your conceptual framework is related to the definitions of some of your constructs. For instance, the definition of social environment provided in the model, as a function of social network and community habit, I would add that it is also a function of socio-economic status.  Also, will your model be behaviorspecific? LTPA is a very broad concept: walking for recreation, cycling for recreation, jogging, gym, sport... In my experience, the environment is not a great predictor of non-specific LTPA (see my early paper in SSM for the distance of decay parameters which shows that people will travel further to use LTPA facilities than those for walking).
No -I think you have covered the pathways well -could 5a also be bidirectional pathway (rather than unidirectional)? could there also be a direct pathway from features of the places where LTPA is practiced to self-efficacy and habit (rather than through perceived environment to practice?) -I'm not sure, but it could be direct and mediated through perceived -I guess it will depend on how you measure "perceived".
Nice survey -thank you and good luck Your introduction mentioned the interaction between env and psych variables, but the model does not reflect that (i.e. it is assuming a linear relationship between the variables). Recent work has started to show that PApromoting environmental features may be more important for those with low attitude/intention, and so assuming that associations work the same for all may be a bit simplistic.
It may be useful to distinguish more between social network and communitythey could overlap.
Good luck with the work. 15 4 I do not think new variables are needed. The problem is that the model needs to take into account accelerants, delays, competing actions, as suggested in some system approaches.
No. I like the conceptual model, but think you should try to incorporate some system thinking into it. Due to the influence of the stages of change, it seems to be a huge "leap" between intentions and environment perception and "habits". I believe that "trying out the practice" or doing so for an experimental period do not constitute a "habit" yet, but it's almost unavoidable in the process of behavioral change. 17 (continues in the next page) 5 1. Self-efficacy: I consider that figure and the textual description are consistent.

Attitude:
The figure is clear and it has the necessary link. In the textual part, I believe that the item "Social network's habit" also acts on the "affective attitude", because social networks are formed mostly for the sake of collective practice. For example, some people like and feel more comfortable in performing collective PA due to the affective connection that originated this practice. I agree that the other degrees of attitude are contemplated here.

Intention:
The arrows in the figure are well positioned. The only caveat I do is for ordination. I am not convinced that the attitude precedes the self-efficacy in the order of arrows! On the other hand, I think, attitude has a more critical input to generate an intention than just feeling able to execute or control a given behavior.

Features of the place where LTPA is practiced:
The figure and the textual description are consistent.
5. Perceived environment to practice: I agree with the two mentioned paths and their order. I think the way one perceives the environment for the practice may be influenced by the social environment (i.e., by both social networks and the community where I belong). It may be worth testing the inclusion of an arrow from the social environment for the perceived environment to practice. 6. Habit: I agree.
7. Social environment: I agree. -I was wondering about the influence of perceived environment on habit. Do you assume a direct relation or does the perceived environment act as a moderator between intention and habit? For example, when a person has a very positive intention but they perceive barriers in the perceived environment (e.g. no adequate facility nearby) they may not be able to turn this intention into action. On the other hand, if a person has a very positive intention and the facility is available, they are able to be active as planned. Is it possible to include interactions in your model?
Are the features of the places where LTPA is practiced considered to be constant (apart from the influence of SES but that is static since SES does not change in this model)? In other words, are people assumed to only be exposed to the PA places in their own neighbourhood/environment? Or are they allowed to go to other neighbourhoods/environments? For example, when people develop a PA habit, they may choose to move to a different neighbourhood with better facilities or they may travel further for better facilities. 3. Would you consider habit as influencing the perceived environment as well? When people are more exposed to their PA-relevant environment (by being active in them), they may change their perception of this environment. Furthermore, it is possible that people who are not active judge the PAenvironment less positive due to cognitive dissonance mechanisms? This last aspect may already be covered with the link between intention and perceived environment.
4. I noticed that subjective norm, one of the factors in the Theory of Planned Behaviour, is not included in the model. The concept of 'social reaction' within 'attitude' did seem to overlap with this concept. Did you merge these two concepts deliberately?
In the introduction of the model, it is stated that some constructs are not included in the model due to simplicity considerations. I understand that it is not possible to include everything but some of these factors do seem to be very important in determining physical activity and they may also interact with some of the other variables. Do you somehow adjust for the variation in these underlying determinants in your model? Especially health status (which is also socioeconomically determined) may be important to consider.
There is a small mismatch between the graphical representation and textual representation of the model: In the graphical representation, habit is a function of the perceived environment, wherein the textual format, it is said to be influences by the features of the places where ltpa is practiced. In my comments, I assumed the figure as correct.
I think it is a very well developed model and it contains most important features. I wish you good luck with your further research and I am looking forward to the results! Please do not hesitate to contact me if any of my comments is unclear or if you have additional questions.