The Effectiveness of Planning Interventions for Improving Physical Activity in the General Population: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Planning interventions such as action planning (AP) and coping planning (CP) have been recognized as influential strategies in promoting physical activity (PA), but mixed results of existing evidence have been observed. This study aims to perform a systematical meta-analysis to evaluate the efficacy of planning interventions for improving PA in the general population. Eight databases, including Medline, Embase, PsycINFO, Cochrane Library, Web of Science, ProQuest, CNKI, and Wanfang Data, were searched to locate relevant randomized controlled trials (RCTs) from their inception to 31 December 2021. In total, 41 trials with 5439 samples were included in this systematic review, and 35 trials were used in our meta-analysis. The results showed that PA was better promoted in the planned intervention group compared to the control group (SMD = 0.35, 95% CI = 0.25–0.44, I2 = 61.4%). Based on the subgroup analyses, we found that planning strategies were more effective among patients, males, when adopting AP intervention, when using the face-to-face sessions delivery mode, and when reinforcements were conducted during the follow-up. The findings of this study indicate that planning interventions significantly improved PA behavior, and, in some contexts, the effects performed better. Future research needs to be conducted to explore the underlying mechanisms of planning interventions and validate their effects more extensively.


Introduction
Insufficient physical activity has become a public health issue globally [1]. Regular physical activity (PA) reduces the risk of noncommunicable diseases and all-cause mortality and delivers important prevention and treatment benefits for many different physical and psychological conditions [2]. Nevertheless, according to a global survey of 1.9 million individuals in 168 countries, over one-quarter of people engage in minimum or no PA (150 min of moderate exercise per week or comparable) [3]. Globally, 81% of teenagers aged 11-17 years are insufficiently physically active [4], and older adults engage in the least amount of physical activity of all age groups [5,6]. In Canada, only 9% of children meet PA recommendations [7]. Thus, it is critical for public health practitioners to encourage regular PA by developing more effective interventions.
Despite persistent efforts to enhance physical activity (PA) through public health policies and behavior change techniques (BCTs) [8], interventions aimed at increasing public involvement in and adherence to PA have generated limited outcomes [9,10]. For instance, a comprehensive meta-analysis of 27 PA intervention studies found that the overall

Methods
This study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [26] and Cochrane Collaboration Handbook recommendations [27].

Search Strategy
A comprehensive literature search that has no language and publication time constraints was conducted to identify the relevant randomized controlled trials (RCTs) that probe the effectiveness of planning or implementation intention interventions on PA in the following academic databases: Medline (via PubMed), Embase, PsycINFO, Cochrane Library, Web of Science, ProQuest, Chinese National Knowledge Infrastructure (CNKI), Wanfang Data, from their inception to 31 December 2021.
An exhaustive search was performed utilizing the medical subject headings (MeSH) combined with free text terms by employing Boolean logical operators, with the terms "Physical Activity", "Exercise", "Planning", "Implementation Intentions", "Action Planning", "Coping Planning", and "Randomized Controlled Trial" taken into consideration. Furthermore, as a supplementary search, we conducted a series of recursive screenings of top journals (top international journals: Health Psychology Review, Annals of Behavioral Medicine; top China journals: Sport Science, Advances in Psychological Science), grey literature, well-known publishers, and significant international academic proceedings to reduce the damage caused by the exclusion of suitable items that match our inclusion criterion. After a manual screening of the selected articles, supplementary searches were performed for other articles by important authors. In the supplement search techniques, details from all databases of search methods were displayed.
All records originally retrieved were imported into EndNote 20 (Thomson ISI Research Soft, Philadelphia, PA, USA), confirmed, and managed by two authors (S.Y.P. and F.Y.) independently and concurrently. Disagreements in this process were resolved through discussion in a meeting with the other authors (A.T.O. and J.H.L.). These steps ensure the completeness and accuracy of this study.

Eligibility Criteria
The following PICOS criteria were used to include relevant studies.

Population
The participants included in this study were not restricted by age, gender, health status, region, or nationality. Given that they have freedom of movement and can participate in the physical activity promotion program set up by the researchers, they are all included in the scope of this study.

Interventions
Acceptable treatments included planning interventions (AP or the combination of AP and CP), i.e., the core components of planning strategies, such as when, where, and how a goal action is to be performed or added and how to deal with the barriers to goal pursuing. Prerequisites such as an introduction to recommended amounts of PA or education on the benefits of PA were permitted in the intervention. Interventions that included other psychological treatments were excluded.

Comparators
Studies were included if the control group was set up as a waiting group with no treatments or if the control group only received some PA recommendations or education about the benefits of PA. Studies with control groups other than those were excluded.

Outcomes
Interventions that reported PA measures (e.g., number of steps, frequency of PA participation, amount of light, moderate-or high-intensity PA in time units) were included. Outcomes can be measured by any measurement instrument, including subjective selfreport questionnaires, self-administered items, and objective instruments. Outcomes must include baseline and post-test measurement data.

Study Design
Only published paralleled-group RCTs were included, including pilot RCTs, with no language limitations.
Duplications were deleted first, as indicated in our included and excluded criteria. The studies were independently chosen by two authors by examining their titles and abstracts. Following that, full-text reviewing was carried out to locate possibly suitable studies. Inconsistencies that arose in this section were avoided.

Data Abstraction and Risk of Bias Assessment
According to the Cochrane Consumers and Communication Review Group's data extraction template [27], two authors performed an independent double-blind method to investigate and extract the key data of the included studies, as follows: the first author, publication year, participants, sample sizes of different groups and total of all groups, gender, age, intervention details (strategies, duration, delivery mode, and reinforcement or not), instrument and various outcomes, etc. Missing data were obtained by emailing the corresponding author of the related study or retrieving other systematic reviews that included the studies. Based on the Cochrane Risk of Bias tool [27], seven items were structured to assess the risk of bias for the included studies: (1) random sequence generation: whether the random sequence method and process are described in detail; (2) allocation concealment: whether to allocate participants strictly according to the results of random numbers; (3) blinding of participants: whether subjects and investigators were blinded; (4) blinding of outcome assessment: whether the evaluators were blinded; (5) incomplete outcome data: whether the drop-out rate was controlled within 10%; (6) selective data reporting: whether only favorable results were selectively reported; (7) other undefined biases: bias such as conflict of interest, small sample size, and baseline imbalance. Each of the items was graded as high risk of bias, low risk of bias, and unknown risk of bias, respectively. The study was assessed as low risk if more than four items were low risk. If most of the study was unknown, the study was assessed as unknown. When more than one item was high risk and less than four other items were rated as low risk, the study was considered high risk. Two reviewers independently assessed the bias risk of the included literature, and disagreements were judged by a third reviewer.

Statistical Analyses
According to Cochrane Collaboration Handbook recommendations [27], the present meta-analysis was conducted by a conventional pair of crossed trials for each comparison at post-intervention. To begin, the omnibus homogeneity test (Q) and I 2 statistics were used to assess study heterogeneity, and p < 0.1 for Q was regarded as statistically significant, while I 2 statistics had values of 25%, 50%, and 75%, indicating mild, moderate, and high heterogeneity [28]. Second, means and standard deviations of outcomes reported postintervention were uniformly extracted as numerical variables for meta-analysis to ensure more accurate analysis outcomes. Moreover, to evaluate the primary effect of planning interventions, the effect sizes were pooled using the inverse variance statistical approach with random effect models. The pooled effect sizes were provided as standardized mean differences (SMDs) with respective 95% confidence intervals (CIs). SMDs were calculated by the mean and standard deviation of each comparison group. When standard errors and confidence intervals were reported instead of standard deviations, they were transformed according to statistical methods. Third, comparison-adjusted funnel plots were drawn to visually detect various forms of potential publication bias. Egger's test was used as a supplement to the quantitative evaluation of the funnel plot to test the significance level [29]. Then, a sensitivity analysis was performed, removing the studies deemed to be at high risk of bias. Lastly, given that the role of moderator variables may be the source of heterogeneity between studies, subgroup analyses were performed to ensure the stability of the overall effect sizes. The subgroup analysis performed in this study is shown as follows: Intervention Strategy (AP vs. AP + CP); Duration (≥5 weeks vs. <5 weeks); Publication Year (year ≥ 2012 vs. year < 2012). Delivery Mode (Sessions vs. Online vs. Sessions and Online); Reinforcement or not (Yes vs. No); Participants (Patients vs. Healthy Populations); Students or not (Yes or No); Female-to-Male Ratio (≥1 vs. <1); Sample Size (≥100 vs. <100); Instrument (Objective vs. Self-report vs. Items). The above analyses were carried out in STATA 14.1 (StataCorp, College Station, TX, USA).

Literature Selection
The initial database search yielded a total of 1790 records. Before screening, 374 duplicate records and 794 records with only simple protocols were removed. After the first round of careful screening of titles and abstracts, 254 search records entered the next step for re-examination. After 131 records were excluded, 123 studies were left for full-text review. As a result, 41 articles  were included in this systematic review, and only six articles [38,40,63,65,68,70] were not suitable for quantitative meta-analysis. The selection process is shown in Figure 1.

Literature Selection
The initial database search yielded a total of 1790 records. Before screening, 37 plicate records and 794 records with only simple protocols were removed. After the round of careful screening of titles and abstracts, 254 search records entered the nex for re-examination. After 131 records were excluded, 123 studies were left for ful review. As a result, 41 articles  were included in this systematic review, and six articles [38,40,63,65,68,70] were not suitable for quantitative meta-analysis. The s tion process is shown in Figure 1.   [64,66], and 3 studies that did not report this cond   [64,66], and 3 studies that did not report this condition [38,40,63]), with publication years ranging from 2002 to 2021. All studies were published in English except for one study published in Chinese [62] and another in Spanish [59]. In total, 2936 participants were randomized to the planning interventions group, while 2948 participants were assigned to the controlled group, with the mean age ranging between 8.06 and 73.30 years old. There were 5 studies targeting women [32,40,43,45,58] and 28 studies with more than 50% female participants [30][31][32][38][39][40][41][42][43][44][45][48][49][50][51][52][53][54][58][59][60][61][65][66][67][68][69][70]. Table 1 provides a detailed introduction to these demographic characteristics. As regards the type of planning strategies chosen by the PA promotion intervention, the number of studies that chose AP (n = 21) as an intervention strategy and the number of studies that chose AP combined with CP (n = 20) as an intervention strategy were very similar.

Characteristics of Studies
The delivery of planning interventions was characterized by three typical modes. The first mode (n = 24) is to implement the intervention through face-to-face individual or group sessions; the second (n = 5) is to conduct online delivery modes, such as phone calls, emails, postal mail, pedometers, phone text messages, APP tracking, and website feedback; and the third mode (n = 12) is to combine face-to-face sessions with online delivery.
The duration of the intervention (from baseline to the last endpoint) of the included studies ranged from 1 week to 14 months. After the baseline interventions were applied, some studies reinforced the effects of the baseline interventions through telephone surveys, text message reminders, diary records, and face-to-face sessions. Across all studies, 20 studies used standardized self-report questionnaires tested for reliability and the validity of previous studies to assess physical activity, 15 studies employed measurement items (three of which were validated), 5 studies used objective instruments (two accelerometers and two pedometers), 1 study used diaries, and 1 study used checklists.
In most of the studies, the control groups received some motivational education, including the benefits of physical activity and WHO physical activity recommendations through face-to-face sessions, leaflets, emails, or text messages. This motivational education was also implemented simultaneously in the intervention group. Moreover, intervention completers were considered for statistical analysis in most studies and only eleven studies employed an intention-to-treat approach [36,42,[45][46][47]51,52,58,60,64,69].
The duration of the intervention (from baseline to the last endpoint) of the included studies ranged from 1 week to 14 months. After the baseline interventions were applied, some studies reinforced the effects of the baseline interventions through telephone surveys, text message reminders, diary records, and face-to-face sessions. Across all studies, 20 studies used standardized self-report questionnaires tested for reliability and the validity of previous studies to assess physical activity, 15 studies employed measurement items (three of which were validated), 5 studies used objective instruments (two accelerometers and two pedometers), 1 study used diaries, and 1 study used checklists.
In most of the studies, the control groups received some motivational education, including the benefits of physical activity and WHO physical activity recommendations through face-to-face sessions, leaflets, emails, or text messages. This motivational education was also implemented simultaneously in the intervention group. Moreover, intervention completers were considered for statistical analysis in most studies and only eleven studies employed an intention-to-treat approach [36,42,[45][46][47]51,52,58,60,64,69].

Primary Outcome
Overall effect sizes were combined for the 35 trials included in the meta-analysis; 6 studies [38,40,63,65,68,70] were excluded due to unavailable data. All studies used the measurement of PA as the outcome indicator. By means of the random effect model, the planning interventions group yielded a small-to-medium significant pooled effect size (SMD = 0.35, 95% CI: 0.25, 0.44, I 2 = 61.4%) compared to the controlled group (shown in Table 2). The effect sizes varied between −0.12 and 0.94 across the studies. A funnel plot indicated that there was no publication bias (shown in Figure S1), but the quantitative Egger test did not reveal the same result (p-value = 0.001) (shown in Figure S2).

Primary Outcome
Overall effect sizes were combined for the 35 trials included in the meta-analysis; 6 studies [38,40,63,65,68,70] were excluded due to unavailable data. All studies used the measurement of PA as the outcome indicator. By means of the random effect model, the planning interventions group yielded a small-to-medium significant pooled effect size (SMD = 0.35, 95% CI: 0.25, 0.44, I 2 = 61.4%) compared to the controlled group (shown in Table 2). The effect sizes varied between −0.12 and 0.94 across the studies. A funnel plot indicated that there was no publication bias (shown in Figure S1), but the quantitative Egger test did not reveal the same result (p-value = 0.001) (shown in Figure S2).

Subgroup Analyses
The results of the predefined subgroup analyses, separated into ten subgroups, are shown in Table 2

Discussion
Thirty-five high-quality RCTs were included in the present study for meta-analysis, and the results found that the planning strategies intervention significantly promoted physical activity in the general population, with the overall effect size (SMD = 0.35, 95% CI: 0.25, 0.44) being "small-to-medium" according to Cohen's classification criteria of effect size (Cohen, 1988). Subgroup analyses were conducted and revealed that the planning interventions were more effective in the patient group, the group with fewer females. Moreover, the delivery mode of individual or group face-to-face sessions during the imposition of the intervention and the group that underwent post-intervention reinforcement performed better. We also found that the effects of different measurement instruments and sample sizes on the pooled effect sizes suggested that they may be sources of heterogeneity between studies.
A positive and significant intervention effect, as revealed by this study, identified that planning strategies can improve PA successfully. Bélanger-Gravel et al. [23] conducted the first meta-analysis of an AP-induced trial, showing that the planning intervention had a significant effect on physical activity, both post-intervention and at follow-up. Almost simultaneously, Carraro and Gaudreau [24] also conducted a meta-analysis combining data from both correlational and experimental studies; it showed that both spontaneous and experimentally induced AP and CP were successful in promoting physical activity. A recent meta-analysis of BCT interventions incorporating AP, conducted by Howlett et al. [16], also showed a significant small-to-moderate effect size effect of BCTs on initiating PA behaviors. This study, the largest meta-analysis of high-quality RCTs to date, further validated the significant effect of planning interventions to promote physical activity, which results from the key role that planning strategies play in behavioral change as self-regulatory strategies [71]. According to the health action process approach (HAPA), two types of planning strategies, AP and CP, play a key role in the initiation and maintenance of intended behavior [72]. The function of AP is to enhance awareness in the individual about potential future scenarios in which the behavior may be performed by making clear when, when, and how the individual would initiate the behavior [73]. CP focuses on the anticipation of barriers that may interfere with a desired activity and how to choose alternative behaviors that may be implemented to overcome those barriers [74,75]. As mental simulations of a series of behavioral processes, planning strategies facilitate the successful translation of good intentions into action through the pre-construction of situations that initiate behavior and the management of possible anticipated obstacles [76,77]. Moreover, AP and CP have been designed in HAPA as mediating variables between intentions and behaviors, helping to bridge the gap between intention and behavior [78]. Some studies have empirically confirmed that AP and CP can also moderate the relationship between intention and behavior [79][80][81][82]. From the above analysis, it can be identified that AP and CP are crucial psychological determinants of PA initiation, and future research should explore the deeper mechanisms of action of AP and CP in promoting PA [83].
The exploratory subgroup analyses conducted in this study revealed that the effects of the planning interventions differed across conditions and contexts, which contributes to a cautious interpretation of the overall effect sizes. In the intervention strategy group, the intervention effects of AP were superior to AP combined with CP, which may stem from the fact that the AP conducted in the trials was more acceptable to the participants, while the combined strategy added an extra CP to the psychological process of behavior change using an if-then format "cue" in response to behavior obstacles [24], which may have led to a decline of intervention effects. However, the combination of AP and CP is also a promising choice of an effective strategy for increasing PA, and its efficacy needs to be verified by more RCTs. Moreover, in terms of intervention delivery modes, face-to-face sessions were the most effective, with online sessions alone (e.g., telephone calls, emails, or visiting websites) being the least effective; post-intervention groups with reinforcement achieved better results. As self-regulatory strategies, planning strategies need to control the details of the interventions to be effective in promoting complex behaviors such as physical activity, so the interventions were more effective in the cases using the delivery mode of face-to-face sessions that were adept in focusing the participants' attention and the addition of reinforcements during the follow-up period. Subgroup analysis by publication year showed a higher effect size for studies published before 2012 than those published after 2012, indicating a decreasing trend in overall effect sizes for studies in the last decade; this trend needed to be verified by more evidence. Furthermore, in the subgroup analysis of the different samples, it was found that the planned interventions were more effective with the patients than with the healthy population, which supports the idea that planning interventions were important interventions for rehabilitating patients [84]. In addition, the interventions were less effective in the population with a high proportion of females, which may be more related to the intention status of the study sample. The planning interventions had a better effect among those with PA intention [85,86], while most females are usually unintentional due to a lack of interest in PA. Although no visible difference was observed between the student and non-student groups, planning strategies remain promising interventions to promote students participating in PA because of their low cost and ease of implementation in campus settings.
Of note, the results from the instrument group in the subgroup analyses suggest that differences between instruments may have contributed to the heterogeneity of the studies. Future studies should employ validated instruments of PA, such as objective instruments (e.g., accelerometers and pedometers) or widely recognized self-report questionnaires (e.g., LTEQ [87] and IPAQ [88]). Given that objective instruments and self-reported questionnaires measure the different parts of PA and that such measurement outcomes are not equivalent [89], further investigation of more appropriate approaches to merging objective instruments and self-reported questionnaires would contribute to improving the validity of evidence based on PA measurements.
The present study is the first meta-analysis of planning interventions for PA that uses RCTs and includes a significant amount of literature covering a wide range of populations (mean age from 8.06 to 73.30 years old). Although only 11 studies used intention-totreat analysis as a method of calculating outcome indicators, the inclusion of far more than 15 trials gives credibility to the outcomes [90]. Based on the high-quality literature included and the rigorous research procedure, the findings of this study elucidate the broad effectiveness of planning interventions. As low-cost interventions that can be delivered in a variety of ways, planning interventions can be easily disseminated and promoted to a wide range of populations, providing them with promising strategies used in the public health domain to increase physical activity and prevent noncommunicable diseases caused by sedentary and physical inactivity.
Several uncontrollable limitations also affected the results of this study. Firstly, most of the included trials in this study were completed in developed countries, and they fail to reflect the actual characteristics of the broader sample. In addition, few trials completed the registration process on the relevant trial platforms, which may directly affect the stability of our evidence. Finally, although we used quantitative analyses to ensure the accuracy of the effects of planning interventions in promoting PA, the sources of moderate heterogeneity observed (e.g., different planning intervention strategies, intervention delivery models, samples, and sample size selection, etc.) need to be further explored.

Conclusions
This review identifies that planning interventions are effective in improving PA behavior among the general population. In addition, the results of this review provide sufficient evidence that the effects of planning interventions vary according to different moderators and contexts. As effective intervention strategies with low cost, planning intervention should be broadly promoted and applied by health practitioners and policymakers.