Background

Physical activity (PA) is defined as any movement produced by the skeletal muscles that require energy consumption or modification of the metabolic rate (Vidarte et al. 2011). It is used worldwide as a strategy to promote health, reduce the burden of disease and improve the quality of life in populations (World Health Organization 2018) throughout the life cycle (Bidzan-Bluma and Lipowska 2018; Lees and Hopkins 2013; Prakash et al. 2015), and aid the prevention and treatment of chronic diseases such as hypertension, diabetes, heart disease, osteoporosis, chronic obstructive pulmonary disease (COPD) and Parkinson’s disease (Billinger et al. 2014; Mantoani et al. 2017).

Interventions or strategies that promote PA can be classified at either the individual or community level. Where such strategies are part of public policies, the evidence suggests they have a significant positive effect on PA levels (Pratt et al. 2014; Shi et al. 2017). Some interventions that are environmental at the physical level, such as modifications to parks, bicycle paths, cycle lanes, areas and spaces at work or in schools, industrial areas or transportation systems, are understood as strategies to promote healthy lifestyle habits in the individual but not necessarily the community (Chow et al. 2009; Stankov et al. 2020). The theoretical concept of purposely designed and built environments to facilitate PA makes it clear that the promotion of PA depends not only on the individual but also on the surrounding environment (“Overview | Physical activity and the environment | Guidance | NICE,” n.d.). Thus, having schools with activity spaces for the children implies greater PA and having fewer obesogenic environments impacts adults and older adults´ sedentarism (Seefeldt et al. 2002; Corsi et al. 2013; Sisson et al. 2019). In this vein, the development of policies and programs in the nations that stimulate PA have been proposed as strategies at the population level (Overview | Physical activity and the environment | Guidance | NICE n.d.).

Because of the above, global environmental structural modifications have been implemented as a public policy measure in several countries with a key method to measure their impact being through natural experiments (Craig et al. 2012). The natural experiment as a research study has two characteristics including the examination of the natural variation of an exposure an outcome (Craig et al. 2017) and the intervention takes place in the natural environment without the intervention of the researcher (Leatherdale 2018).

The evidence from natural experiments developed in various contexts (Risica et al. 2019; Mancipe Navarrete et al. 2015; Hernández-Álvarezet al. 2015; Sandu et al. 2018) examining the effect of structural-level environmental interventions on promoting PA (Becerra et al. 2013; Sugiyama et al. 2014; Alexandre 2013; Alexandre 2013) is widespread making it necessary to aggregate and review these data. This systematic review aims to determine the effectiveness of environmental interventions with structural or physical modifications on the changes in the population’s PA level. Our hypothesis is that structural or physical modifications at the environmental level lead to an increase in the PA level in the population.

Methods

This study is a systematic review conducted according to Cochrane Collaboration (Deeks et al. 2011) and PRISMA (Moher et al. 2009) guidelines. The protocol is registered in the PROSPERO international database of systematic reviews under the number CRD42021229718.

Selection criteria

Type of studies

Natural experiments that involve structural- or physical-level environmental intervention programs to promote PA. “Understanding a natural experiment as the research study where the exposure to the event or intervention of interest has not been manipulated by the researcher.” (Craig et al. 2012).

Type of participants or population of interest

General or specific population studies examining PA in educational settings (university level and below), the workplace, in cities and neighborhoods, and older institutionalized adults or adults in neighborhoods or communities. Studies aimed at specific populations with diseases or conditions, such as neuromuscular diseases (multiple sclerosis, cerebrovascular disease, muscular dystrophies), musculoskeletal diseases (lupus, arthritis, osteoarthritis) or cardiovascular diseases (myocardial infarction, cardiac arrhythmias, valvular heart diseases, etc.) were excluded. Studies on athletes were also excluded.

Type of interventions

Studies that evaluated the promotion of PA from the perspective of structural- or physical-level environmental modification such as the construction of parks, bicycle paths, cycle lanes or modifications in cities were included.

Type of outcomes

The main outcome was PA in terms of variation or levels measured objectively or subjectively. At the objective level, measurements included pedometers, accelerometers, heart rate monitors and direct and indirect calorimetry. Subjective measurements included self-reports or questionnaires like the International Physical Activity Questionnaire (IPAQ) System outlines expected behavior for students in each activity throughout the daily (CHAMPS) or the Physical Activity Recall, among others. Both the objective and subjective measurements could be expressed continuously [such as total energy expenditure (Kcal/Kg/week, kcal/week), metabolic consumption in METS, oxygen consumption or differences in respiratory exchange ratio (CO2/VO2), heart rate, variability in heart rate, total minutes of PA or number of steps, among others] or categorically (such as of light, moderate or vigorous PA). Participation in the programs, percentage of physical activities performed, amount of PA performed, measures of fitness level if they were reported in metabolic expenditure or oxygen consumption, and measurement scales of an individual or group PA were also included.

Search strategy

To identify the studies, a search was carried out up to January 2022 in electronic databases: Medline/Pubmed, Web of Science, Science Direct and CINAHL, using Mesh, Decs and Emtree terms. Supplementary material 1 provides the search strategies. Additionally, a search of crossed references was done manually as well as a search of gray literature in specialist journals, university repositories or general websites related to the topic.

Study selection

For the selection, two reviewers (EH, EC) screened the records by title and abstract according to the inclusion criteria. The selected studies were then blinded, read in full-text format by both reviewers and their results and conclusions compared. In the case of disagreements, a third reviewer acted as a peer evaluator to enable a definitive conclusion and to settle such disagreements. To optimize the work at this stage, the Rayyan© software (Elmagarmid et al. 2014) was used.

Data extraction and risk of bias evaluation

For the data extraction, a spreadsheet was created, in which the characteristics of the studies were recorded and included the title, authors, year and place of publication, program undertaken and its characteristics, start and end date of the program, scope of intervention, outcomes and measures considered, and reported results.

The risk of bias was assessed using the recommendations and evaluation criteria of the ROBINS I tool for non-randomized studies of interventions (Bero et al. 2018). This made it possible to evaluate specific risks of bias at three points in the study: point 1 or pre-intervention, where the biases of confounding and participant selection were considered; point 2 or during the intervention, where measurement bias was assessed; and point 3 or post-intervention, where the biases in the interventions performed, outcome measurement and attrition bias were considered. Each risk of bias domain was graded by considering the criteria as described by the ROBINS I tool and evaluated qualitatively. The risk of bias level was not evaluated quantitatively, since the ROBINS 1 tool is designed for observational studies, but it provided an assessment of the risk of bias that is often associated with natural experiments. When the risk of bias was not apparent in a given study, a third evaluator assisted with the assessment of the risk of bias.

Data analysis

The data analysis was undertaken qualitatively through figures and tables that show the information revealed in the data extraction process. This way, the scope of the intervention programs, methodological findings and results of each study could be visualized in an organized manner conducted according to Cochrane Collaboration (Deeks et al. 2011).

Results

In total, 3865 articles were found in the 4 interrogated databases. After eliminating the duplicates (200), two evaluators screened 3665 titles and abstracts and excluded 3566 for not fulfilling the inclusion criteria. Thus, 99 full texts were reviewed in depth to determine that 24 studies fulfilled the eligibility criteria and 2 new studies were included for the expert’s recommendation. The reasons for exclusion were: 20 articles did not report on natural experiments, 26 did not have outcomes of interest, and 29 reported on natural experiments but focused on social programs with no structural or physical environmental modification. It is important to note that, of the 26 studies included, 9 worked jointly on physical environmental modifications and social programs. The selection process is presented in Fig. 1.

Fig. 1
figure 1

Selection process Prisma flow chart

Full size image

Characteristics of the studies included

The 26 studies included (Schultz et al. 2017; Dill et al. 2014; Auchincloss et al. 2019; McCormack et al. 2016; Ogilvie et al. 2016; Simões et al. 2017; Lee et al. 2017; Cohen et al. 2012, 2019; Jindo et al. 2020; Cranney et al. 2016; Brittin et al. 2017; Andersen et al. 2017; Zhu et al. 2018; Sun et al. 2020; Mölenberg et al. 2019a; Fitzhugh et al. 2010; Veitch et al. 2012; Aittasalo et al. 2019; Hooper et al. 2018; Quigg et al. 2012; Giles-Corti et al. 2013; McGavock et al. 2019; Demant Klinker et al. 2015; Ward Thompson et al. 2013; Jancey et al. 2016) were natural experiments published between 2010 and 2020. These studies were developed in different parts of the world (Fig. 2). The studies examined interventions from different perspectives including structural modifications for access and the use of parks and green areas (8 studies), installation of fitness areas or playground equipment in parks (2 studies), construction of new active transportation alternatives such as bicycle lanes or other transportation systems (3 studies), modification of cities, urban spaces or neighborhoods (7 studies), modification of school environments (2 studies) and work spaces (4 studies). The general characteristics of each study are shown in Table 1.

Fig. 2
figure 2

Study distribution around the world

Full size image
Table 1 Characteristics of the studies included and areas of emphasis
Full size table

Measurements of the levels of physical activity in the studies

All outcome measures have previously published evidence of validity and reliability globally and while not all measures are used in every study, some of the same measurement tools were used in more than one study. Instrumental measures were identified as the accelerometer in 8 studies (Dill et al. 2014; Ogilvie et al. 2016; Jindo et al. 2020; Brittin et al. 2017; Andersen et al. 2017; Quigg et al. 2012; Demant Klinker et al. 2015; Jancey et al. 2016), pedometer in 1(McGavock et al. 2019), and satellite location and data processing systems like Global Positioning System (GPS), Active Physical Activity Location Measurement System (ACTIVE PAL 3C), Physical activity location measurement system (PALMS) in 4 studies (Dill et al. 2014; Andersen et al. 2017; Zhu et al. 2018; Demant Klinker et al. 2015). Non-instrumental measures recognized worldwide such as the System to Assess the Practice of Physical Activity and Recreation in Parks and Natural Surroundings (SOPARC) were identified in 6 studies (Schultz et al. 2017; Auchincloss et al. 2019; Cohen et al. 2012, 2019; Cranney et al. 2016; Veitch et al. 2012), the IPAQ, a PA questionnaire, in 2 studies (Dill et al. 2014; Simões et al. 2017), and the Recreation Physical Activity Questionnaire (RPARQ), in 1 study (Ogilvie et al. 2016).

It should be mentioned that 20 of the studies reported measuring PA levels with scales of light, moderate or vigorous. In some, the application of general surveys of perception of activity and its level or walking was reported, up to counting the activities performed in each of the protocols. (See Table 2).

Table 2 Main methodological characteristics of the studies
Full size table

The reports of PA were also expressed by counting the number of subjects participating in the programs, the levels of PA in percentage, minutes or total time, and the number of steps or metabolic expenditure. In addition, there were reports on changes in PA patterns, either in the use of transportation, management of the activity or performance. Some studies reported on the time spent sitting as a significant measurement to define PA.

Effectiveness of the structural modifications

The outcome of PA was presented in the 26 studies (Schultz et al. 2017; Dill et al. 2014; Auchincloss et al. 2019; McCormack et al. 2016; Ogilvie et al. 2016; Simões et al. 2017; Lee et al. 2017; Cohen et al. 2012, 2019; Jindo et al. 2020; Cranney et al. 2016; Brittin et al. 2017; Andersen et al. 2017; Zhu et al. 2018; Sun et al. 2020; Mölenberg et al. 2019a; Fitzhugh et al. 2010; Veitch et al. 2012; Aittasalo et al. 2019; Hooper et al. 2018; Quigg et al. 2012; Giles-Corti et al. 2013; McGavock et al. 2019; Demant Klinker et al. 2015; Ward Thompson et al. 2013; Jancey et al. 2016) shown in Table 3 in which the effectiveness of the programs was variable. The level of PA was greater in some structural-level environmental modifications than in others while no difference between intervention and control groups were observed. Table 3 shows the participation in the activity and global measures time of the activity in minutes, percentage increase or decrease in PA.

Table 3 Effectiveness of the structural modifications
Full size table

Activity level

Seventeen studies (Schultz et al. 2017; Dill et al. 2014; McCormack et al. 2016; Ogilvie et al. 2016; Simões et al. 2017; Cohen et al. 2012; Jindo et al. 2020; Brittin et al. 2017; Andersen et al. 2017; Mölenberg et al. 2019a; Veitch et al. 2012; Aittasalo et al. 2019; Hooper et al. 2018; McGavock et al. 2019; Demant Klinker et al. 2015; Ward Thompson et al. 2013; Jancey et al. 2016) reported an increase in the PA level post-structural modification. For example, an increase in total activity time, varying between 10 and 30 min, a 6–28% increase and an increase of between 38 and 369 users participating in the activity were found between the pre- and post-structural modifications. The measurement of PA in the area of parks, especially in those with outdoor gyms included, were more effective at improving PA. It is important to emphasize that these changes were dependent on factors that were both internal and external to the practice of activity, such as the physical environment, sex, socioeconomic level of the participant, and personal characteristics.

The intensity, or level, of PA was often described as light, moderate, or vigorous levels were used as a measure of PA in these studies. However, although the use of light PA was consistent throughout all studies, some studies further divided PA and reported the intensity of PA as moderate, moderate-vigorous, and vigorous levels while other studies combined them into one level. At the specific intensity of PA, the duration of light PA increased between 22 and 67 min or between 6 and 8% while the duration of moderate PA increased from 1 to 43%. The moderate-vigorous level of PA increased from 6 to 36%, but in 3 studies no change in the level of PA was observed (Dill et al. 2014; McCormack et al. 2016; Lee et al. 2017). At the vigorous level, no increase in PA was reported in 80% of the studies while the remaining 20% reported only a slight increase of 5–7%. A major difference between the studies is that those in the area of work and transportation report greater increases in light and moderate activity. Finally, in one of the largest studies conducted on urban modification, RESIDE in Australia, a 54% post-renovation increase in minutes/week of recreational walking was reported along with a reduction of minutes walking in active transport, suggesting participants preferred recreational walking over active transport walking when both options are available. Similarly, an increase in moderate-vigorous activity was observed in this population, with an Odds Ratio 3.17, (I.C 1.02-9.81) demonstrating the effect of activity in the area of the new neighborhood which is summarized in Table 3.

Counts

Only 9 studies reported participation counts which were defined as the number of individuals using a new structural-level environmental modification, of which 8 described modifications of parks (Schultz et al. 2017; Auchincloss et al. 2019; Cohen et al. 2019; McCormack et al. 2016; Veitch et al. 2012; Quigg et al. 2012; McGavock et al. 2019; Ward Thompson et al. 2013) and 1 in cities (Fitzhugh et al. 2010) with all studies showing an increase in the number of subjects using the area. In contrast, the 1 study of active transport in bicycle lanes (Dill et al. 2014), showed a decrease in post-participation measurement as shown in Table 3. Regarding the increased number of participants in the studies examining the modification of parks, there was a greater increase in participant use when the structural modification included outdoor gyms while there was less of an increase observed when the parks were modified only for recreation and exercise purposes (Veitch et al. 2012).

One study described the construction of a green area in the city (Auchincloss et al. 2019) and found a 45% increase in walking, cycling, and running in the park post-remodeling with an odds ratio (OR) of 1.46 (CI 1.06–1.98). In contrast, a study examining the remodeling of areas in parks for users with pets found an increase in the number of subjects in the remodeled parks increased, but without a significant increase in the PA of those subjects (McCormack et al. 2016).

Metabolic consumption and number of steps

The third and fourth outcomes are metabolic consumption in METS and the time spent sitting versus active time. METS were reported by 5 studies (Schultz et al. 2017; Ogilvie et al. 2016; Cohen et al. 2012, 2019; Brittin et al. 2017), of which 3 were in parks, 1 was in schools, and 1 in the area of transportation. Fourth of the 5 studies found an increase in metabolic work (Ogilvie et al. 2016; Cohen et al. 2012, 2019; Brittin et al. 2017) and 1 found a decrease (Schultz et al. 2017) as shown in Table 3. The time spent sitting versus active time was reported in 4 studies, with 3 in the workplace (Jindo et al. 2020; Zhu et al. 2018; Jancey et al. 2016) and 1 in a school (Brittin et al. 2017) and all of the studies found not only an increase in PA but less time spent sitting in both the work and school populations. Furthermore, in the workplace studies, as shown in Table 4, there was a reduction in sitting time and of seated work, with an increase in both standing work and light PA. In fact, light PA increased from 130 to 150 min while no increase in moderate-vigorous PA was observed in the work areas. Similarly, in schools there was a reduction in sitting time during the students’ day, although no increase in light PA was reported (Table 3). It is worth noting that these 4 studies are more controlled given that they are in closed areas like a company or a school, which enables a direct and accurate follow-up of the above outcomes.

The outcome, number of steps, was reported in 2 studies using pedometers or accelerometers with (McGavock et al. 2019; Jancey et al. 2016), 1 study examining modifications in parks with frozen waterways and the other study examining the effect of modifications to the workplace on PA. Both studies found an increase in moderate-vigorous activity with a greater number of measured steps related to the use of frozen waterways and to the use of the space for physical activity in the workplace as shown in Table 4.

City-wide changes

Seventh studies examined the impact of structural modifications in cities (Simões et al. 2017; Andersen et al. 2017; Hooper et al. 2018; Fitzhugh et al. 2010; Mölenberg et al. 2019a; Giles-Corti et al. 2013; Demant Klinker et al. 2015), with several favorable results on the increase in PA. For example, in the modification of 184 cities in Brazil it was found that the population of each city reached the light activity goals previously set forth in the project. In particular, increased light-PA was observed in men with a higher level of education yielding an OR of 1.46 (CI from 1.11 to 1.92). Furthermore, the study found that programs lasting longer than 6 months increased the OR for PA as shown in Table 3 (Simões et al. 2017). Fourth related studies on urban renovation described the modification of spaces in cities to increase PA particularly aimed at the populations of young people and children (Mölenberg et al. 2019a; Andersen et al. 2017; Fitzhugh et al. 2010; Demant Klinker et al. 2015) and found an increase in PA post-renovation, with greater activity in minutes/week for both active transport and use of green areas, especially in sporting scenarios and playgrounds.

In these studies, an increase in moderate-vigorous PA was reported, although the increase was not significant, post-renovation. However, greater attendance time with an average of 24 min was observed in the modified spaces. Greater light PA was found globally reflected by an increase in the minutes in the populations attending the activity.

Furthermore, urban modification studies in Denmark led to an average increase in PA per week of 7.8 min for light activity, 4.5 min for moderate-vigorous, and 13.1 min less sitting time. Overall, an average increase in play time for students of over 40 min/week was observed, which was particularly pronounced for children with mothers having a lower level of education. Interestingly, there were no significant differences in the groups according to the distance from the new areas modified (Mölenberg et al. 2019a).

Commuting

In regard to commuting via active transport, several studies report the use of alternative transport modifications, such as cycling and walking as well as the measurement of the commute (Dill et al. 2014; Lee et al. 2017; Sun et al. 2020; Aittasalo et al. 2019). For both school and workplace cycle programs, an increase in both the number of users and the number of trips by bicycle was noted, especially where the user lives 1 mile near the travel area.

In a study on the construction of a subway, an increase in active transport was measured by an increase in cycling time, for one work group, as well as an increase in the walking activities associated with taking the subway (Sun et al. 2020). One of the largest studies in the United Kingdom, where commuting changed from car to bicycle, reported an increase in minutes of bicycle use and a reduction in car use lead to an overall rise in commuting time from 10 to 150 min per week with a rate risk ratio RRR 2.5 CI 1.2–5.0. Additionally, greater METs were found for the use of the bicycle and walking in the moderate-vigorous activity and an increase of 150 min/week of walking was observed (Ogilvie et al. 2016), as shown in Table 3.

Risk of bias in the included studies

Table 4 illustrates the results of the evaluation of the risk of bias in the studies included in this systematic review. Although the studies have different degrees of risk of bias, risk of bias is inherent in natural experiments. The observed biases are present in pre-intervention, during the intervention and post-intervention, due to access and follow-up of the population. However, where applicable, efforts were made to control for confounding variables.  

Table 4 Risk of bias in the studies included
Full size table

Discussion

The results presented in this systematic review provide a better understanding of the myriad of strategies encountered in the worldwide literature regarding the promotion of PA via structural-level environmental interventions. Twenty-six studies were found that identified 4 fundamental areas of structural modifications to promote PA including schools, work environments, streets or cities, and neighborhoods or parks. Of the 26 studies, 21 examined the effect of environmental modifications like parks, cities, pedestrian walkways, or steps, while 5 examined modifications in indoor or closed environments like in schools and the workplace. A summary of the findings is presented in Supplementary material 2.

Environmental modification strategies at both the physical and social level appear to bring about changes in a subjects’ behavior, which was observed in this study. One key message from this systematic review is that structural-level environmental interventions and modifications such as providing physical space or programs to facilitate PA does in fact increase PA leading to healthier lifestyles that may decrease the development of chronic diseases (Sisson et al. 2016; Sarmiento et al. 2020). There is also evidence of a worldwide increase in PA levels and a positive association between the amount of walking and use of the bicycle in adults and older adults with the development of paths and bicycle paths, respectively, as reported in the manuscripts included in this review (Gomez et al. 2015). In the same sense, neighborhoods that facilitate spaces for walking and green spaces demonstrate a positive association with an increase in PA (Sarmiento et al. 2020). Similarly, structural modifications to promote additional standing and walking in schools and universities showed increased PA in children and young adults (Sisson et al. 2019). In addition, having outdoor spaces and infrastructures in parks, along with paths for bicycles or walking increases the use of bicycles and the development of exercise programs (Mölenberg et al. 2019b; McCormack and Shiell 2011), producing favorable changes in the level of activity, METS or time spent in PA.

These environmental modifications have been studied from different perspectives, such as the implementation of paths and different methods of transportation, changes in schools to control obesity and strategies in an open environment to promote healthy behaviors (Hirsch et al. 2018; Sarmiento et al. 2020; Sisson et al. 2016). A systematic review of the natural experiment found evidence for the modification of transportation paths to increase PA (Hirsch et al. 2018) in which changing the type of transportation resulted in an important an environmental strategy to promote PA with findings similar to our systematic review. In contrast to our study, the authors performed a meta-analysis of the available evidence and found high levels of statistical heterogeneity with I2 values of 98%, which is to be expected due to the differences in the protocols and the populations measured in a natural experiment.

Another study was an integrative review (Gomez et al. 2015) examining the effects of urban modifications on active transportation to promote PA. Fourteen reviews were included in the study and all related to environmental modification of spaces for transportation or bicycles to increase active rather than passive transportation. Unlike the present systematic review which focuses on natural experiments, the authors combined and analyzed a variety of study designs such as cohort, cross-sectional, experimental, case, and controlled studies. Although the current systematic review examined only natural experiments making direct comparisons between the two studies challenging, results broadly similar to ours were reached including an increase in the use of transportation and the number of travels like as active transports.

Another study of note was a systematic review of the interventions used to promote bicycle use through structural modification (Mölenberg et al. 2019b). Similar to the current systematic review, data from natural experiments were reviewed yielding a total of 31 studies. Of these 31 studies, 20 focused on structural modification in the environment and 16 examined the combined effect of the structural modification and use of the infrastructure of bicycle paths. The authors found improvements in performance outcomes measured for both the use of bicycles and bicycle paths in 11 countries studied. Of these 11 countries, 7 have similar outcomes in PA to those described in our study, while the remainder focus on the measurement of the use of bicycles rather than the effect on PA.

An additional review article related to environmental modification (McCormack and Shiell 2011), but not focused on transportation, evaluated 33 studies all of which had different methodologies that included quasi-experimental and cross-sectional designs. All of the studies examined the effect of environmental modifications on PA, but as in the present review, from the perspective of modifications to neighborhoods, land use for activity, parks, tracks and walking trails or transportation structures. All studies were found to improve walking time and active movement either by walking or cycling more and promoting an active lifestyle among the populations studied. Similar to the current systematic review, instrumental and non-instrumental measures of PA were reported in 31 of the 33 studies and 2 were excluded because such measurements were not performed. Instrumental measurements such as accelerometers and pedometers were used in 4 studies, while questionnaires on PA levels, including walking, cycling or moderate-vigorous level of activity were used in the remaining 27 studies. Although an increase in PA resulting from the environmental modification was found, an increase in the walking indices following neighborhood modifications was not consistently observed.

It is important to note that there is evidence regarding the potential influence of diet and nutrition influencing environmental modifications such as the use of supermarkets, roads and neighborhoods. A systematic review (Macmillan et al. 2018) examined the influence of diet and nutrition on PA in 15 natural experiments of which 4 were on supermarkets and foods specifically and the remaining 11 studies examined the effect of diet and nutrition on walking, footpaths, parks, green areas, and the modification of roads and byways. Like in the present systematic review, most of the studies were conducted in the United States (10 of the 15 studies), and the remainder in the United Kingdom, New Zealand and Brazil. Two of the studies used accelerometers, 3 accelerometers combined with GPS, and the remainder used questionnaires and self-reports as outcome measures. It is worth noting that a control group was included in some of the studies while the remaining studies examined outcomes pre- and post-urban remodeling. The overall results of the above study found that diet and nutrition or modifications in the environment were not effective at increasing PA.

Limitations to this study include the quality of the data due to the challenge of follow-up in certain populations as well as attrition of participants and fluctuating populations that participate voluntarily in the experiment rather than participant recruitment into a study. The examination of PA in studies that are natural experiments with their implicit risk of bias due to the lack of a control group, variability in populations and environments, different structural interventions, follow-up and attrition of subjects, and population migration. Regarding the risk of bias we chose to assess the risk of bias and its impact on the results of the studies examining the effect of environmental modifications on PA using the criteria of the ROBINS I instrument for non-randomized studies (Benton et al. 2016). We assessed the risk of bias, based on the previously outlined approach, and found a moderate to severe risk in many of the criteria due to the nature of natural experimentation. However, we decided to take the criteria from the list for each of the phases but not the qualifiers, which differentiates the risk analysis in this study compared to others. Therefore, placing the qualitative emphasis, according to the design, on its importance for public health and decision-making in health policy and care.

One significant strength of the study is the measurement of several specific interventions in large groups of individuals in different parts of the world. In addition, unusual community-level interventions have been employed and examined in the included studies such as the creation of cycle lanes, bicycle paths, outdoor gyms, and others.

Another important aspect is that the current study found that the instrumental methods used and analysis of outcomes described in the reviewed studies are consistent with management guidelines for natural experiments according to the Research Council (Craig et al. 2012). The suggested analyses for such designs including logistic regression, the difference in differences, propensity scores and multiple regressions, are present in 80% of the studies that comprise this review. Thus, we provide a more robust and comprehensive systematic review examining the effect of structural-level environmental interventions on PA.

Conclusion

The results of the above systematic review highlight the important role that natural experiments have on decision-making in public health and health care, especially in regards to the quality of life and reduction of risk factors by means of PA through environmental interventions. Based on the results of this systematic review it seems prudent that a natural experimental study considers controlling for confounding variables and establishing outcome measures that can readily identify the impact of interventions on health indicators in the countries where they are conducted to incorporate actions and public policy guidelines in such countries and worldwide.

As with any study based on the results from natural experiments, there are potential limitations in the evaluation of data. However, at the same time, we believe that from a public health perspective this type of study plays an important role in ensuring the continued application of such experiments in real-life situations. The current systematic review suggests that modification of the environment is an important element in improving public health through increased PA for which reason we encourage the use of natural experiments to examine such interventions.

Based on our systematic review the environmental interventions of structural modification in parks and active transportation demonstrated greater effectiveness in increasing the level and time of PA in the population study.

In schools and work areas, there is evidence of both a decrease in sedentary time and an increase in PA related to environmental modification of space or change in office structure.

Finally, given that the economic and cultural setting is a key variable when considering the effectiveness of structural interventions, and since only 1 of the 26 reviewed articles included such data, more studies examining economic factors are needed especially in low- and middle-income countries like in South America.