Effectiveness of physical activity promotion based on inter-sectoral collaboration - a meta-analysis of complex interventions in high-income western countries

Although the prevalence of insufficient physical activity (PA) in adults in high-income western countries has fallen from 31.6% in 2000 to 27.7% in 2020, insufficient PA. Among adolescents, the prevalence of insufficient PA in high-income countries (HICs) is still one of the most relevant causes of numerous diseases and premature death in these countries (Strain et al. 2024 ) has also fallen slightly from 79.8% in 2001 to 78.2% in 2016, although it has risen slightly in some of these countries (e.g. Germany) (Guthold et al. 2020 ). More recent data for Germany show that the prevalence of the recommended level of PA fell from around 14% to around 11% among girls between 2009 and 2022, while remaining stable at around 20% among boys (Bucksch et al. 2024 ). Around 7.2% of all deaths worldwide can be attributed to insufficient PA (Katzmarzyk et al. 2022 ). In addition, a substantial proportion of the prevalence of most major common diseases (e.g. CVD) can be attributed to insufficient PA (Katzmarzyk et al. 2022 ). The proportion of deaths and the prevalence of major common diseases that can be attributed to insufficient PA is more than twice as high in HICs as in low-income countries (Katzmarzyk et al. 2022 ). Although most countries have succeeded in reducing death and disease attributable to insufficient PA, insufficient PA is still one of the major determinants of population health (Xu et al. 2022 ). Therefore, Guthold and colleagues (Guthold et al. 2020 ) as well Strain and colleagues (Strain et al. 2024 ) call for inter-sectoral efforts to further reduce the prevalence of insufficient PA. A call to promote PA through inter-sectoral collaboration (ISC) can also be found in the Toronto Charter for Physical Activity from 2010 (Global Advocacy Council for Physical Activity (GAPA) of the International Society for Physical Activity and Health 2010 ) and in many papers of the World Health Organization (World Health Organization 2022 , 2024 ). The concept of ISC has been a central concept in health promotion since the Ottawa Charter for Health Promotion 1986 (World Health Organization. Regional Office for Europe 1986 ). It is assumed that certain health promotion goals such as promoting PA can be achieved more effectively, efficiently and sustainably through ISC than if individual actors and organizations pursue these goals on their own (Nutbeam and Muscat 2021 ). ISCs can be established at the municipal, regional, national and/or international level. In this paper, we focus on ISCs that are established at the municipal level in HICs. With regard to recent reporting guidelines for systematic reviews (Guise et al. 2017 ) these ISCs can be understood as interventions with intervention complexity (multiple components like information campaigns and enhance the walkability in neighbourhoods), pathway complexity (complicated/multiple causal pathways, feedback loops, synergies, mediators, and/or moderators), and population complexity (target multiple participants, groups, and/or organizational levels). For the majority of these ISCs, there is also implementation complexity (require multi-facetted adoption, uptake, or integration strategies) and contextual complexity (work in a dynamic multidimensional environment). Although ISCs to promote PA and health have been increasingly established in recent years, their effectiveness, efficiency and sustainability have rarely been studied (Corbin et al. 2018 ). Literature reviews on ISCs show that the approaches labeled with this term are very heterogeneous (Corbin et al. 2018 ; Quilling et al. 2020 ). In addition, sometimes they are labeled with other terms (Corbin et al. 2018 ; Quilling et al. 2020 ). Despite these difficulties, there are meanwhile some meta-analyses on the effectiveness of ISCs or similar approaches to promote health and health-relevant living conditions and behaviours (Hayes et al. 2012 ; O’Mara-Eves et al. 2013 ; Anderson et al. 2015 ; O'Mara-Eves et al. 2015 ; Brunton et al. 2015 ; Bagnall et al. 2019 ), but none of these meta-analyses focused on the effectiveness of ISCs to promote PA. All such meta-analyses we are aware, expect one, include only few studies evaluating complex community-based interventions (with or without ISCs) to promote PA (Hayes et al. 2012 ; Anderson et al. 2015 ; O'Mara-Eves et al. 2015 ; Brunton et al. 2015 ; Bagnall et al. 2019 ). These meta-analyses therefore do not provide an evidence synthesis on whether ISCs are effective in promoting PA. The other meta-analysis included 32 studies with PA as an outcome (O’Mara-Eves et al. 2013 ). But this meta-analysis is 12 years old, includes some other forms of community-based interventions than ISCs and studies from middle- and low-income countries. Therefore, this meta-analysis does not provide an evidence synthesis on the question of whether PA can be effectively promoted in western HICs through ISCs. The ISCs we examine in this meta-analysis are partnerships of two or more groups or organizations from different sectors that share the goal of promoting PA. In simple cases it is just a science-community partnership (Schulz et al. 2015 ; Higgerson et al. 2018 ), in the more complex cases the ISC consists of a community network of actors and organizations from different sectors (Dzewaltowski et al. 2010 ; Hoelscher et al. 2010 ; Wells et al. 2013 ; Wright et al. 2013 ; Phillips et al. 2014 ; Nettlefold et al. 2021 ). In the majority of the ISCs we investigated, a key feature was the existence of an inter-sectoral coalition (network or board). Another key feature of the ISCs we studied is that they do not simply change the attitudes, skills and behaviours of their target group, but also change their environment to facilitate the increase in PA. These more complex cases can be described using the Community Coalition Action Theory (Fig. 1 ). This model illustrates the formation of inter-sectoral coalitions, the synergies within inter-sectoral coalitions, and how these change local practices and policies as well as the local environment (community change outcomes) and ultimately promote PA among the target group (health and social outcomes). Logic model of the Community Coalition Action Theory (Butterfoss and Kegler 2022 ) The primary question addressed in our study is the extent to which strategies for promoting PA based on ISCs. We were not primarily interested in any specific source of complexity. The target group of the ISCs we examined were residents of municipalities. The meta-analysis presented here is based on a systematic review of the effectiveness of ISC in primary prevention and health promotion. The literature search was conducted for several health-related outcomes (see inclusion criteria). For the meta-analysis, the inclusion criteria, selection processes, and data extraction were adapted to the selected outcome of PA. Thus, only data on PA were included in the meta-analysis. This meta-analysis adheres to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al. 2024 ) and aims to evaluate the effectiveness of ISC-based strategies for promoting PA. The studies for the systematic review were selected using the PICO-S schema: Population (P) : General population, including both children and adults, without any specific health conditions. Population (P) : General population, including both children and adults, without any specific health conditions. Intervention (I) : ISC for health promotion and/or primary prevention. Intervention (I) : ISC for health promotion and/or primary prevention. Comparison (C) : No specific comparison required. Comparison (C) : No specific comparison required. Outcomes (O) : Morbidity, mortality, quality of life, social inequality, attitudes (e.g., health-related self-efficacy), competencies (e.g., social-emotional competence), behaviours, living conditions (incl. physical environment, social environment, policies/laws). Outcomes (O) : Morbidity, mortality, quality of life, social inequality, attitudes (e.g., health-related self-efficacy), competencies (e.g., social-emotional competence), behaviours, living conditions (incl. physical environment, social environment, policies/laws). For the meta-analyses, the outcome PA was selected from these outcomes. Study type (S) : Randomised controlled trial (RCT), cluster randomised controlled trail (CRCT), controlled before-after study (CBA), interrupted time series (ITS), cross-sectional study (CS). Study type (S) : Randomised controlled trial (RCT), cluster randomised controlled trail (CRCT), controlled before-after study (CBA), interrupted time series (ITS), cross-sectional study (CS). For the meta-analyses only RCT, CRCT, CBA, and ITS were eligible. Publications in German and English were included, the results of which were transferable to the German context. Excluded were studies (i) from countries whose results are not meaningfully transferable to the German context, (ii) with a purely qualitative/explorative approach, (iii) with a limitation to conditions of success or outputs, (iv) on exclusively health care-related ISC (such as integrated care and disease management programs), (v) with a focus on curative, therapeutic or rehabilitative measures, e.g., studies only aiming at secondary or tertiary prevention and (vi) non-controlled before-after studies and non-comparative studies were excluded. The systematic literature search for the systematic review and meta-analysis was conducted in the electronic databases PubMed/MEDLINE, Web of Science and LIVIO on December 17, 2021. The search strategy includes the selection of keywords for the effectiveness of PA promotion based on ISC, including: ‘promotion’, ‘prevention’, ‘intersectoral’, ‘interorganizational’, ‘cooperation’, ‘collaboration’ etc. Appendix 1 shows the complete search strategy. It was filtered according to German and English publications from 2010 onwards. In addition, a hand search was carried out in the reference lists of the relevant articles. The literature selection for the systematic review was evaluated in three steps. In the first screening, only the titles of the references were analysed, in the second the summaries and in the third the complete publications. In a fourth step, the full texts included in the systematic review were screened for the outcome of PA for the meta-analysis. Two independent reviewers (DR, RB) were involved in the selection of the relevant publications. In the event of differences between these two persons, a consolidation was carried out. In the first two reviews, references were only excluded from further analysis if there were clear indications that a publication was not a study or clearly did not fit the topic under review. Duplicates of publications, conference papers, case reports and abstracts were excluded. Data were extracted independently by two reviewers (RB, HH) using the Cochrane data collection form intervention review - Randomised studies and non-randomised studies . No automation tools were used in the data collection process. Where necessary, study authors were contacted to clarify or obtain missing information. Discrepancies between reviewers were resolved through discussion. In addition to outcome measures, the following study-level data were collected: title, authors, year of publication, language, study design, number of participants, demographic characteristics (e.g., age, sex), inclusion and exclusion criteria, intervention details (type, duration, delivery), control or comparator conditions, timing and type of outcome measures (primary and secondary), reported results per outcome, funding sources, and declarations of conflicts of interest. If data were missing or unclear, assumptions were made based on context or the authors were contacted. The quality and risk of bias of included studies was assessed using the Cochrane Risk of Bias Tool (RoB 2) (Sterne et al. 2019 ) for randomised controlled trials and the ROBINS-I tool (Sterne et al. 2016 ) for non-randomised studies. Two reviewers (RB, HH) independently performed the risk of bias assessment. Disagreements were resolved through discussion. No automation tools were used during this process. The meta-analyses are limited to the PA outcome domain. PA may have been measured using objective methods (e.g., accelerometers or pedometers) or subjective methods (e.g., self- report questionnaires). If several results for the PA outcome domain were reported in a study (e.g., several measures, time points, analyses), we only considered the result for the primary outcome measure from the main analysis at the last measurement time point. If a publication did not explicitly state which of several analysed outcomes was the primary outcome measure, we researched whether this information was contained in a previously published study protocol or in a study registry for this study and then aligned ourselves with it. If this information could not be found for a publication at all, we took the result from the analysis that referred to the entire study population and controlled for the most confounders for the outcome measure reported at the top. Due to variation in the reported effect measures, standardization was applied to ensure comparability. Continuous outcomes (n = 6) were converted to risk-based measures in line with the recommendations of Anzures-Cabrera et al. (Anzures-Cabrera et al. 2011 ). Dichotomous outcomes (n = 2) were used as reported. Odds ratios (OR) and their logarithmic forms (logOR) were calculated using the Campbell Collaboration’s web-based effect size calculator (Wilson 2023 ), based on formulas from Practical Meta-Analysis (Lipsey and Wilson 2001 ). Missing data were requested from study authors. In some cases, such as missing standard deviations, data could not be obtained. As a result, certain effect size estimates may be subject to imprecision. Meta-analyses were conducted using RevMan 5.4 software with a random-effects model due to anticipated clinical and methodological heterogeneity across studies. Forest plots were used to visualise the results. Heterogeneity was assessed via visual inspection, Chi tests, and the I statistic. According to Higgens et al. ( 2024 ) I values above 75 % indicate considerable heterogeneity. Sensitivity analyses were performed by excluding outlier studies to reduce heterogeneity and assess the robustness of the findings. Subgroup analyses were conducted based on participant age. Potential reporting bias was assessed visually using funnel plots. No formal tests for asymmetry (e.g., Egger’s test) were conducted. The certainty of evidence was not formally assessed using GRADE or similar frameworks. The systematic literature search in the databases identified 5093 publications. After excluding 27 duplicates, 5066 titles were screened and 268 abstracts were selected. Of these, 102 publications were screened in full text (Fig. 2 ). During the review, 71 of the 268 full texts were excluded. The intervention was inappropriate for 35 publications. 19 publications were excluded due to their publication type. Eight publications each had an inappropriate outcome or an inappropriate survey method. One publication included a different population. An additional 59 publications were identified by hand-searching relevant articles. Of these, eight full texts were excluded due to an incorrect publication type (n = 6) and an inappropriate intervention (n = 2). Thus, a total of 82 publications on 60 studies were included in the analyses. As the following is a partial analysis of the effectiveness of PA promotion based on ISC, only eight studies with the outcome “physical activity” were included in the meta-analysis (Dzewaltowski et al. 2010 ; Hoelscher et al. 2010 ; Wells et al. 2013 ; Wright et al. 2013 ; Phillips et al. 2014 ; Schulz et al. 2015 ; Higgerson et al. 2018 ; Nettlefold et al. 2021 ). PRISMA 2020 flow diagram of study selection The meta-analysis included eight studies (Tab. 1 ) published between 2010 and 2021, utilizing various designs such as cluster-randomised controlled trial (six studies), controlled before-after (one study), and interrupted time series (one study). Five studies were conducted in the United States (US) (Dzewaltowski et al. 2010 ; Hoelscher et al. 2010 ; Wells et al. 2013 ; Wright et al. 2013 ; Schulz et al. 2015 ), and three were based in the UK (Phillips et al. 2014 ; Higgerson et al. 2018 ; Nettlefold et al. 2021 ). Sample sizes ranged from 251 to 147,489 participants. The populations studied were diverse, encompassing various age groups, genders, and ethnic backgrounds. Four studies specifically addressed school-aged children. The interventions focused on ISCs in health promotion strategies, compared to standard programs, alternative interventions, delayed interventions, or different community approaches. PA was assessed using objective methods (e.g., pedometers, attendance records) and subjective, self-reported measures. Outcomes included daily step counts, minutes of moderate-to-vigorous PA (MVPA/VPA), adherence to activity guidelines, self-rated activity levels, and participation in sports or physical education. The cluster-randomised studies by Dzewaltowski et al. ( 2010 ), Nettlefold et al. ( 2021 ), Phillips et al. ( 2014 ), Schulz et al. ( 2015 ), Wells et al. ( 2013 ), and Wright et al. ( 2013 ) were assessed for potential bias risks using the ROB II tool (Fig. 3 ). The analysis revealed that all six studies presented a low to moderate risk (some concerns) across various domains. In outcome measurement (Domain 4), five of the six studies raised some concerns regarding bias risk, as the individuals collecting the data were not blinded and were aware of the assigned interventions. Domain 5 also showed some concerns regarding the selection of reported outcomes in all six studies. Overall, all six studies assessed with the ROB II tool exhibited some concerns regarding the overall risk of bias. Risk of bias summary for cluster-randomised trials: review authors' judgements about each risk of bias item for each included study The studies by Higgerson et al. ( 2018 ) and Hoelscher et al. ( 2010 ) were evaluated using the ROBINS-I tool. The results indicated that both studies (Fig. 4 ) had a low to moderate bias risk across all domains. Higgerson et al. ( 2018 ) showed a moderate bias risk related to missing data, while Hoelscher et al. ( 2010 ) presented a substantial bias risk in this domain. Both studies were deemed appropriate regarding participant allocation, intervention classification, and outcome measurement. The overall risk of bias for Higgerson and Hoelscher's studies was classified as moderate. Risk of bias summary for non-randomised studies: review authors' judgements about each risk of bias item for each included study A random-effects model was applied to synthesise findings from the included studies (Fig. 5 ). The studies by Dzewaltowski et al. ( 2010 ), Nettlefold et al. ( 2021 ), and Wright et al. ( 2013 ) reported their results only separately for different groups (e.g., overweight and normal-weight participants, or male and female participants) rather than as combined samples. Therefore, these groups are displayed separately in the figure. The pooled analysis yielded an odds ratio (OR) of 1.44 (95% CI: 1.14–1.81), indicating a significant positive effect of inter-sectoral health promotion strategies on PA. considerable heterogeneity was = = = that the was not attributable to In such cases, the pooled effect not be reported at all, or it be out that the pooled effect be with heterogeneity is it is with in the of the effect of the individual studies. As 5 the of the effect is with one In our the high heterogeneity is primarily by two In such cases, the meta-analysis be excluding the In addition, analyses be performed in such cases to whether the heterogeneity can be attributed to in effect sizes between We both of ISCs on PA = = = = A analysis excluding outlier studies (Fig. ) a pooled of (95% CI: Although this a a positive it did not = heterogeneity was = = = excluding outlier studies of ISCs on PA After the outlier the funnel (Fig. ) a of study results around the central A funnel a low risk of publication bias. excluding outlier studies Subgroup analyses examined the of interventions across different age groups (Fig. ). For adults, the pooled was (95% CI: a positive that did not = Heterogeneity was high = and the Chi was significant = In the analysis for children revealed a pooled of (95% CI: indicating a significant positive effect = with high heterogeneity = = and a significant Chi = These findings that inter-sectoral health strategies may a more effect on PA among children than if the were excluded here as the pooled effect be by of ISCs on PA The funnel (Fig. ) shows a of study results around the central line for both groups and two are among the studies children. Potential risks of publication bias be meta-analysis by

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.