Effects of interventions on physical activity behavior change in children and adolescents based on a trans-theoretical model: a systematic review

Search strategy

This study was searched using both Chinese and English search terms, and the search year was from database creation to March 2023. Chinese literature was searched in CNKI, Wan-Fang and VIP core databases; English literature was searched in Web of Science (WOS), PubMed and EBSCO databases. Retrieval strategies: ((“trans-theoretical model (TTM)” OR “stage of change” OR “process of change” OR “self-efficacy” OR “decisional balance”) AND (“child” OR “children” OR “adolesent” OR “young” OR “teenager” OR “juvenile” OR “student”) AND (“physical activity” OR “exercise” OR “sedentary” OR “weight” OR “overweight” OR “obesity” OR “obese” OR “health” OR “health behavior”)). This study imported the retrieved literature into EndNote X9 software for management. Two researchers (Z.Z. and X.Z.) independently searched for Chinese and English literature based on the above search strategy, with X.Z. responsible for searching Chinese literature and Z.Z. responsible for searching English literature.

Selection criteria

This study designed literature selection criteria based on Population, Intervention, Comparison, Outcomes and Study (PICOS) principles [30]. Inclusion criteria: (1) participants were typical children and adolescents aged 3 to 18 years old; (2) interventions based on TTM; (3) control measures mainly include: health education measures such as lectures, continuation of the original lifestyle, and blank controls with no intervention, etc.; (4) outcome variables include SOC, POC, SEI, DBL, behavior, and health indicators; (5) the study design included randomized controlled trial (RCT) and quasi experiment (QE). Exclusion criteria: (1) qualitative research; (2) children and adolescents with atypical development such as deafness and mental retardation; (3) for duplicate publications of the same subjects, only the higher quality literature was included. Selection was made independently by two researchers (Z.Z. and Y. L.) each according to the literature selection criteria, in the order of title, abstract, table, figure, and full text. The selection results were evaluated secondarily by two other researchers (X.Z. and S.W.), and if there was a dispute, the group discussed and decided together.

Quality assessment

In this study, the Risk of Bias 1.0 tool, which is a relatively mature and widely used tool recommended by the Cochrane Collaboration Network, was adopted to assess RCTs [31, 32]. The tool assesses the included literature from the following aspects: randomization methods, blinded implementation, allocation concealment, completeness of outcome data, selective reporting of study results and other biases. Quality assessment of QEs was performed using Risk Of Bias In Non-randomised Studies-of Interventions (ROBINS-I) [33]. The tool assesses the following areas: possible pre-intervention confounders, subject selection bias; possible intervention classification bias during the intervention; and possible post-intervention intention-to-intervention bias, missing data, outcome measures, and selective reporting bias. Two researchers (Z.Z. and Y.L.) independently made judgments based on the above assessment tool. If there were serious differences of opinion on any item, a group discussion would be held among all researchers (Z.Z., X.Z., Y.L., and S.W.).

Data extraction

Based on reading the full text of the included literature, this study extracted the following data: (1) bibliographic information, including the first author and publication year; (2) study design, including RCTs and QEs; (3) characteristics of participants, including sample size, age, proportion of females, and country; (4) whether a stage-matched intervention program was adopted; (5) elements of TTM involved in the intervention, including one, some, or all of SOC, POC, DBL, and SEI; (6) behavioral change techniques used in the intervention, mainly including four categories: health information intervention, psychological skills training, PA program, and health behavior feedback; (7) follow-up time points; (8) control measures; (9) outcome variables and main findings. Furthermore, for the coding of research results, this study uses “+” to indicate a positive and significant change after the intervention, and “-” to indicate either no significant change or a negative and significant change after the intervention. Extraction was made independently by two researchers (X.Z. and Z.Z.). The extraction results were evaluated secondarily by two other researchers (P.S. and C.X.), and if there was a dispute, the group discussed and decided together.

Statistical methods

In this study, frequency analysis is primarily used to synthesize similar data. Firstly, factors such as whether the intervention stages match, whether the complete TTM structure is employed, and the type of behavior change techniques used are considered as grouping variables. Secondly, based on each outcome variable, the studies under each group are summarized, and the studies that achieve significant results are counted. Finally, the proportion of studies achieving significant results in each group is calculated relative to the total number of studies in that group. Based on these steps, an initial comparison of the effectiveness of the interventions is conducted.

Selection results

A total of 1469 articles were retrieved, and 587 articles were obtained after de-duplication using Endnote X9 software, including 89 Chinese articles and 498 English articles. After reading the titles and abstracts, 48 articles were obtained. Among these 48 articles, 6 qualitative studies and 13 meeting abstracts were excluded, resulting in 29 articles. Furthermore, 6 intervention studies without support from the TTM model and 1 duplicate publication were excluded, ultimately incorporating 22 articles. The PRISMA flow diagram is detailed in Fig. 2.

Basic characteristics of included studies

Quality assessment results

Out of the 14 RCTs, 10 studies [34, 41, 42, 44, 45, 47, 49, 51, 52, 54] reported specific randomization methods, only 3 studies [44, 52, 54] reported strategies for implementing blinding, and only 2 studies [47, 52] reported strategies for allocation concealment. None of the studies exhibited selective reporting of study results. Additionally, the outcome data in 10 studies [34–36, 44, 46, 47, 49, 51–53] were relatively complete, and the missing samples did not interfere with the accuracy of the research results (Table 2). Four studies [34, 41, 42, 49] adopted cluster random sampling methods, while two studies [44, 47] used block random sampling methods. One study each adopted stratified random sampling [45], simple random sampling [51], random number Table [52], and parallel grouping [54] methods. Three studies [44, 52, 54] all employed a single-blind strategy, and two studies [47, 52] both adopted a third-party allocation strategy. Additionally, three studies [42, 45, 54] had excessive sample losses, with respective losses of 28.3%, 34.3%, and 151 cases.

For QEs, the likelihood of bias due to subject selection before the intervention was low, but consideration of confounding factors such as age and proportion overweight/ obese may have been missing. During the intervention classification period, the participants were relatively evenly distributed in the two controlled quasi- experimental studies [37, 55]. However, the study by Frenn et al. [43] has not yet demonstrated whether the experimental and control groups were homogeneous before the experiment. The likelihood of bias for intentional deviation, outcome measures and selective reporting of findings after the intervention was low, but a few studies [39, 43] may have had bias for missing data due to participant’s dropout. The dropout rates of participants in two studies [39, 43] were 9.9% and 28.3%, respectively. Detailed results are shown in Table 3.

Effects of TTM-based interventions on PA behavior change in children and adolescents

The promotional benefits of interventions based on TTM

Based on a systematic review of relevant research findings, this study has yielded the following main findings.

1. Stage-matched interventions are more likely to increase PA in children and adolescents, and promote their progress through the stages. However, Romain et al. [27] found inconsistencies in their meta-analysis on adult populations. Their study revealed that while stage-matched interventions had a greater effect size, stage- mismatched interventions also contributed to promoting PA. Children and adolescents are in a critical stage of growth and development, with high behavioral plasticity. Using targeted, stage-matched interventions can promote their understanding of health knowledge and development of PA plans. However, adult groups generally have stronger autonomy and self-management skills, enabling them to choose and maintain PA based on their own needs and preferences. Therefore, they are more likely to benefit from broad health information and general guidance. In addition, Stage- matched measures are more easily understood, recalled, read through, considered relevant, help to trigger central processing pathways, that is, subjects are more likely to use their central vision to scrutinize the intervention information, are more likely to get the subjects’ attention, and are perceived as credible and more appropriate [56]. For example, the inclusion of tailored photos, pictures and other vivid and effective information in health education print materials can increase the interest of subjects. Also, individuals who “know but don’t do” are aware of the benefits of PA but do not act on them, so training to provide the knowledge and value of exercise is not effective and the transition from “knowing” to “doing” must be facilitated by setting goals and activity plans. However, stage-mismatched interventions use a uniform standard for all individuals, ignoring inter-individual differences, and the intervention does not reach all individuals.

2. In general, measures designed with more focus on the TTM structure are more likely to increase PA in children and adolescents, and promote their progress through the stages. This is supported by the Meta-analysis by Romain et al. [27] on PA in adults. The complexity of PA requires not only educational interventions to enhance behavioral change intentions and attitudes, but more importantly, a variety of intervention strategies are used in conjunction with the psychological state of the subject at the stage they are in to reinforce and sustain the effects of the intervention. It is noteworthy that changes in PA stages do not necessarily imply an increase in actual PA levels [57]. The process of human behavior change is extremely flexible and individuals do not spend a fixed amount of time at any stage, but depending on the conditions required for individual behavior change, instantaneous change may occur if the conditions are present at the same time [58–60]. Overall, interventions involving more structure are more likely to promote behavioral change in children and adolescents.

3. There are differences in the roles played by different behavior change techniques in promoting PA and its health benefits. Firstly, interventions based on health information or health behavior feedback are more conducive to promoting the increase in PA and changes in behavioral stages. Interventions rooted in health information dissemination not only promote the enhancement of health knowledge and awareness, but also boost exercise intention. Furthermore, they can provide necessary skills and resources for PA plans, assisting individuals in successfully implementing behavioral changes [34–36]. Health behavior feedback is the main measure for providing targeted interventions, and tailoring to a particular individual increases the likelihood of being perceived as closely related to the individual and being perceived as credible by the participants [56]. Secondly, interventions that incorporate PA programs more likely to enhance health benefits, but the shorter cumulative duration of the intervention may limit its effectiveness. This is because improvements in health-related indicators require regular physical exercise over a long period of time to become evident, and short-term interventions are unable to produce effective stimulation on the body. However, given that the effective stimulation time is influenced by other behavior change techniques and the PA programs employed, it is difficult to determine the effective PA program duration based on the TTM. Further research is urgently needed to explore this matter.

4. Measures based on health information interventions, health behavior feedback, and psychological skills training have relatively limited effects on improving health benefits. The main reasons are as follows. First, promoting the health of children and adolescents is a long-term and complex process that is not only influenced by PA but also by factors such as dietary habits and sleep quality [38, 39], which are not strictly controlled in relevant studies. Second, progression through stages does not necessarily mean an increase in PA, and the health benefits of PA require accumulation to a certain dose and duration to become apparent [61]. Third, cross-seasonal investigations may confound the results, as cold winter weather reduces opportunities for PA and may lead to changes in dietary patterns, which can result in seasonal increases in blood lipid concentrations and fasting blood glucose [62].

The deficiencies of interventions based on the TTM

In addition, researchers have the following deficiencies in the formulation of intervention measures.

1. Stage progression is not the same as an increase in PA. While some studies [38, 41, 49] have interpreted stage increases as indicative of intervention effects, it’s important to note that such increases serve only as proxies for behavioral change. They do not invariably correspond with enhancements in key exercise parameters, including actual time spent, intensity, and frequency of exercise [59]. Consequently, these stage increases should be regarded as secondary outcomes, not primary ones. For example, in the early stages of behavior change, the increase in stages simply reflects an increase in the individual’s intention to change behavior, while more effort is required to turn intention into action. Therefore, some studies [43, 46, 52, 54] have also shown that the actual increase in PA after their intervention was not significant.

2. The lack of normality in the model may be the reason for the inappropriate design of the intervention. Firstly, the advantage of TTM is that it provides stage-matched interventions that are tailored to the characteristics of individuals at different stages, but there is still a large body of studies [38, 40, 43, 44, 46, 47, 50, 52, 53] that ignore the issue of stage specificity. Secondly, although each component of TTM plays an important role in individual behavior change, a large number of studies [37, 40, 43, 44, 46, 47, 50, 52, 53] in practice have used only one or a few of these components to design interventions. Thirdly, some studies [41, 50] have designed interventions only for the SOC, ignoring other important components such as the POC, SEI and DBL. Although the SOC is conceptualized as a central component of the model, in the context of intervention it is usually considered as a variable rather than a theory.

3. The lack of a framework for targeted behavior change techniques. While TTM provides a theoretical framework for achieving stage progression through the POC, it does not provide specific operational techniques. For example, consciousness raising is a cognitive change process, and while we should clearly inform intervention design, there is no normative technical framework for consciousness raising, so researchers design interventions that are inconsistent with each other, creating not only heterogeneity in outcomes, but also difficulties in subsequently exploring the effectiveness of interventions through combined effects.

4. The lack of dynamic feedback constrains the effectiveness of the intervention. TTM asserts that individual behavioral change is a slow, dynamic process of stage change, and that along with the stage change in behavior, individuals also need multiple points of intervention exposure to get feedback that matches the stage they are currently in [21]. Therefore, the researcher is required to reassess the individual stage and relevant theoretical variables several times during the intervention until the individual achieves and maintains behavioral change. However, most interventions [34, 37, 40, 43, 44, 48–55] are static and not tailored in the full sense of the word.

Future research prospects

Based on the quality assessment results of the included studies, this research recommends strengthening the randomized design to minimize potential selection bias and the interference of confounding factors. It also advocates for the introduction of blinding methods and allocation concealment strategies to eliminate subjective bias of researchers and participants towards the intervention measures, as well as to prevent potential operational bias during data collection and analysis. In summary, researchers need to continuously improve the quality and rigor of their studies in order to more accurately evaluate the effects of intervention measures and promote knowledge advancement in this field. In addition, future research can enhance the external validity and generalizability of studies by increasing sample sizes and including a more diverse group of children and adolescents.

Based on the issues existing in the development of intervention strategies in the original research, this study proposes the following prospects for future work. Firstly, it is recommended that subsequent practice examine the psychological pathways of TTM to promote PA, using actual behavioral changes in PA as the main outcome variable, to reveal the intrinsic associations and interactions between structural factors within TTM and to improve the predictive effect of the model on PA in children and adolescents. In addition, in the interventions based on the TTM, it is recommended that future research should focus on actual PA energy expenditure and duration, rather than the stages of PA. Secondly, it is recommended that the practical application of subsequent models should adopt the complete TTM framework to design intervention strategies in a standardized manner, with detailed reporting of staged interventions targeting the POC, SEI, and DBL. On this basis, it is recommended that subsequent researchers report in detail on behavior change techniques that promote changes in the POC, SEI and DBL, condense the effects of interventions, and provide a standardized framework of behavior change techniques for targeted interventions. In practice, it is often recommended to continuously assess SOC in order to provide dynamic, multi-point feedback for truly tailored interventions. Finally, given the complexity of PA, it’s challenging to envision a single model effectively driving behavioral change. Therefore, a more effective approach lies in employing a combination of intervention strategies. This approach should also consider the impact of external environmental factors—such as the accessibility of exercise facilities, the ease of road connectivity, and macro policy support—which extend beyond the scope of the TTM framework. Based on this, this study argues that a more integrated (informational, behavioral, social, policy, etc.) intervention can be developed to promote PA behavior change in children and adolescents, based on the SOC and incorporating the perspectives and methods of the social-ecological model.

The limitations of this study

Firstly, due to language limitations, this study only included English and Chinese articles, and has not further searched for potentially relevant articles in other languages. Secondly, the study included both RCT and QE, and some RCTs did not describe specific methods for random allocation, blinding, and allocation concealment, which may introduce various biases. Finally, due to the limitations of the original data, especially the serious heterogeneity in the assessment of PA, this study finds it difficult to estimate the effect size of interventions based on the TTM and to identify potential moderating factors. Therefore, more high-quality studies are needed to confirm the reliability of this research.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.