How buildings shape occupant movement: a systematic review and framework

2.1 SEARCH STRATEGY

The information sources consulted to identify studies were the following scientific databases, to ensure a search across disciplines spanning architectural/civil engineering, transportation, psychology and public health. The databases were chosen based on those used in the reviewed systematic literature reviews (see Appendix A in the supplemental data online) and those suggested by ChatGPT. The use of ChatGPT has been recently suggested to facilitate literature reviews in health sciences research (Huang & Tan 2023). For the prompt used and full response, see Appendix B online. All artificial intelligence (AI)-generated suggestions were manually reviewed by the authors, and only relevant databases were included. The suggestions expanded the databases typically used in building science literature, enriching the pool of papers considered. The literature search was conducted on 21 July 2024 since the databases’ inception. Search term categories include physical activity, the indoor built environment (design and operation) and human-centred investigation; for the full list of search terms, see Appendix B online. Retrieved articles were imported into EndNote 21, and duplicates were removed following the automatic steps in the de-duplication method presented in Bramer et al. (2016). Steps that required manual assessment were skipped in EndNote due to the large dataset. Instead, further de-duplication was performed by checking for duplicate abstracts in RStudio.

Articles were initially screened by title and abstract using a machine-learning (ML) approach described in Section 2.2. Studies were included if they investigated the effects of indoor environmental design or operation on physical activity or sedentary behaviour in human populations without serious health conditions, reported objective or subjective outcomes, and were English-language, peer-reviewed original articles with full text available. Full eligibility criteria are provided in Appendix B in the supplemental data online. From the selected articles, additional eligible articles were obtained from their reference lists (backward citation search) and by citation search (forward citation search) using the Shiny app Citationchaser (Haddaway et al. 2022). Articles from literature reviews on the topic (see Appendix A online) were also included in the final list, if not already present.

2.2 STUDY SELECTION

An ML approach was adopted to automate the title and abstract screening process based on the eligibility criteria. Recent systematic reviews in the field focusing on the built environment (Nigg et al. 2024) adopted this technique, proving its efficacy. The open-source ML tool ASReview (van de Schoot et al. 2021) was adopted for the screening. ASReview uses active learning to mitigate 10% error rate related to false inclusion and false exclusion based on human decisions (Wang et al. 2020). The SAFE procedure proposed by Boetje & van de Schoot (2024) was referenced to conduct the ASReview. The procedure, as adapted for this review, included three rounds of model training, as described in Appendix B in the supplemental data online.

2.3 DATA EXTRACTION AND QUALITY ASSESSMENT

Comprehensive data extraction was conducted independently by both authors for studies that satisfied the inclusion criteria. The information extracted is summarised in Appendix B in the supplemental data online. Article quality of the included studies was assessed from two perspectives: overall article quality, which evaluates both the reliability of the methodology and the clarity of the article, and methodology quality. Quality assessment is a widely adopted practice in the healthcare and public health fields (Downs & Black 1998). A checklist was developed by both authors specifically for this review with criteria used in other similar reviews (Ahrentzen & Tural 2015; Engelen 2020; Ige et al. 2019; Kärmeniemi et al. 2018; Nigg et al. 2024) and from validated tools such as the Effective Public Health Practice Project (EPHPP) Quality Assessment Tool for Quantitative Studies (Thomas et al. 2004). The criteria are summarised in Table 1. A score of 0, 1 or 2 was assigned to all the items of the checklist, depending on the degree to which the specific criteria were met (‘fully met’ = 2, ‘partially met’ = 1, ‘not met’ = 0). The overall score of article quality was calculated as the percentage of the earned score for all 11 criteria out of a maximum score of 22. The methodology score was calculated as the percentage of the earned score for criteria 4, 6 and 8–10.

2.4 DATA SYNTHESIS

A meta-analysis of the selected studies was not possible due to heterogeneity in study designs, outcome measures and outcome measure assessments. A narrative approach was used to synthesise the extracted data (Petticrew & Roberts 2006), focusing on the characteristics of the studies and their outcomes.

3.1 STUDY SELECTION

The literature search yielded a total of 241,253 articles (Figure 1). After removing duplicates and screening titles and abstracts using ASReview, 63 articles remained for full-text screening. Following full-text screening, 33 articles were excluded for not meeting the inclusion criteria (n = 31) or being unavailable in full text (n = 2). Additional articles from published reviews (see Appendix A in the supplemental data online) yielded 20 further full-text records, resulting in 12 included studies. A total of 39 papers excluded at the full-text stage were incorporated into the framework development for the extraction of building-related factors to consider, as detailed in Section 4.1.

Figure 1

3.2 QUALITY EVALUATION

A total of 42 articles were included in the quality evaluation, yielding a mean quality score of 0.70 (range = 0.45–0.91, out of a total score of 1) for overall quality and 0.61 (range = 0.40–0.90) for methodology quality. The score distributions are shown in Figure 2. Both overall quality and methodology reliability fluctuated over time, showing an overall increase. However, compared with overall quality, the reliability of the methodology requires further improvement, including justification of the sample size, reliability of the study design, validity of the measurement tools for the independent and dependent variables, and robust statistical analysis.

Figure 2

3.3 STUDY CHARACTERISTICS

This review examined 42 studies investigating how building design and operational factors influence sedentary behaviour and physical activity, with most conducted in the US, Australia and UK. Research was split between the fields of architecture/environmental studies and public health, reflecting growing interdisciplinary interest. The majority were field-based studies (n = 40), primarily observational and quasi-experimental, with a wide range of sample sizes and data collection methods.

Independent variables were categorised as either design (e.g. layout, staircases) or operational (e.g. lighting, aesthetics), depending on whether they are determined at the construction or post-construction phase. Most studies focused on design features, though a few explored operational or both. Dependent variables included measures of sedentary behaviour and physical activity, assessed through people-counting sensors or cameras (e.g. Nicoll 2007), wearable devices (e.g. Sheldrick et al. 2019), self-reported surveys (e.g. Lu et al. 2015) and experimenter counting (e.g. Boutelle et al. 2001). For a detailed description of the study characteristics, including disciplinary background, study design, environmental factors investigated and data collection methods, see Appendix C in the supplemental data online.

3.4 SYNTHESIS OF FINDINGS

The systematic review reveals two key findings. First, only a limited number of studies have investigated the relationship between building-scale factors and occupant physical activity, highlighting an important gap in the healthy buildings literature. Second, the existing studies vary considerably in the building-scale factors examined, the levels of physical activity analysed, and the building programmes, populations and contexts considered. As a result, the current body of literature does not yet support definitive conclusions regarding the causal mechanisms through which building-scale factors influence physical activity. Rather, the findings from this literature review provide a structured synthesis of the factors examined to date and highlight important gaps that warrant further investigation.

The following subsections report this synthesis, organised by factors associated with building design and operation, and further categorised into building attributes. For a detailed analysis of findings for each building attribute, including consistencies and inconsistencies, see Appendix D in the supplemental data online. Based on this analysis, Figure 3 (building design) and Figure 4 (building operation) summarise the positive or negative influence of the considered factors on various activity levels. For each activity level and factor, colours indicate reported effects: red for hindering, green for promoting physical activity, grey for non-significant and blue for mixed results. Beyond summarising the findings, Figures 3 and 4 also help identify attributes that have not yet been examined for specific activity levels. For detailed information, including analysed factors, references and authors’ explanations (where available), see Appendix E online.

Figure 3

Figure 4

4.1 PROPOSED CONCEPTUAL FRAMEWORK

Building upon existing frameworks and socio-ecological models developed to describe physical movement at both the urban and building levels (Pikora et al. 2003; Zimring et al. 2005), a conceptual framework is proposed to contextualise and synthesise current knowledge on the impact of building factors on occupants’ physical activity (Figure 5). Rather than focusing solely on building-related influences, this framework takes a holistic approach, incorporating a variety of drivers and influencing factors to better understand physical movement within buildings. By doing so, it situates the role of building factors within a broader context, highlighting their interaction with individual, social and environmental determinants of movement.

Figure 5

The framework traces the progression from sedentary behaviours to active movement, ultimately linking these behaviours to secondary outcomes such as building energy savings (by reducing reliance on electrical systems, such as elevators) and occupant health, wellbeing and productivity. The transition from sedentary behaviours, including elevator use or prolonged sitting, to physical activity is driven by the decision to move. Physical activity is the consequence of the decision to move and is categorised into three intensity levels: light (e.g. standing, minimal movement of body parts), moderate (e.g. walking, stair use) and vigorous (e.g. running, dancing).

In the framework, the decision to move is influenced by ‘mediating’ and ‘moderating’ factors, similar to what was proposed in McMillan’s (2005) conceptual framework for children’s travel behaviour. Mediating factors initiate movement, while moderating factors influence how and to what extent this movement occurs within the built environment. Specifically, mediating factors serve as primary drivers to physical activity and include activity-based, leisure-based and self-motivation drivers:

• Activity-based drivers involve necessary, task-oriented movements such as walking to a meeting or moving between classrooms.

• Leisure-based drivers refer to voluntary, recreational movement undertaken for enjoyment or health, such as strolling during a break or using indoor fitness equipment.

• Self-motivation drivers represent hybrid behaviours in which individuals intentionally choose more active options—such as taking the stairs instead of the elevator—to integrate movement into daily routines, guided by personal goals or habits.

Moderating factors shape the likelihood of movement by either amplifying or reducing it, based on the initial motivation from mediating factors. For example, while someone may need to move for a task (activity-based driver), building design features—such as inviting staircases with natural light or biophilic elements—can encourage additional physical activity by making certain routes more attractive. Thus, moderating factors, such as building layout and design, can influence how people move within a space by enhancing or discouraging their movement choices.

To better understand these influences, moderating factors are categorised into three main groups, similar to those reported in previous frameworks on urban-level physical activity (Nicoll 2007; Zimring et al. 2005): personal, socio/organisational and contextual. Personal factors include demographic characteristics such as age, sex, body dimensions, education and stress levels, which have been identified as influencing physical activity (Annear et al. 2014; Arslan & Erkan 2020; Bae & Asojo 2016; Ball et al. 2006; Lindberg et al. 2018; Zimring et al. 2005). Socio/organisational factors encompass organisational goals, workplace culture, social support, management attitudes and behavioural prompts, all of which can facilitate or impede physical activity. Contextual factors involve geographical context, climate, site design, building-specific elements such as furniture (e.g. sit–stand desks), and motivational signs and technologies (e.g. nudging apps).

Building factors, encompassing several design and operational attributes (the focus of this review), are a subset of contextual factors. Within the framework, all building-related factors explored in both the included and excluded papers are listed. This choice was made to enrich the list of factors summarised in the previous sections and to provide a holistic view of potential factors to consider in future research on building design and operational attributes. Drawing from the findings summarised in the Results section, factors are categorised into building design and building operation, and further divided into specific attributes. To enhance clarity, each factor is colour-coded based on its influence on physical activity: ‘Most likely’ factors, in green, are those for which consistently significant influence was reported in the analysed literature; ‘Uncertain’ factors, in blue, refer to inconsistent or non-significant findings in the analysed literature; ‘Underexplored’ factors, in pink, represent variables studied in excluded studies (e.g. those focused on active design buildings) or factors proposed by the authors based on expert knowledge (e.g. perceived IEQ attributes). Notably, no factors were identified for exclusion from further investigation, even those with results leaning toward non-significance (e.g. space aesthetics), as summarised in Figures 3 and 4.

Finally, this framework highlights the potential interactions among moderating factors (indicated with an ‘X’ in Figure 5). These interactions illustrate the complex and interconnected ways in which personal, social and contextual factors—including building design and operational elements—shape movement patterns in indoor environments. For instance, the influence of spatial layout on an occupant’s choice of movement may depend on their age or physical ability (personal factor) as well as the presence of behavioural prompts or workplace policies (organisational factors) that encourage active choices.

The temporal dimension (namely, when a movement-supportive strategy can be implemented within the building life-cycle) is not explicitly represented in the proposed framework but should be carefully considered. Different strategies may become relevant at different stages of the design and building process. For example, decisions related to spatial configuration, circulation systems and the visibility or accessibility of stairs are typically made during the early design stages, when fundamental aspects of building form and layout are defined. In contrast, choices related to materiality, signage or electrical lighting features may occur at later design stages or even during the building’s operational phase. Recognising this temporal dimension is important, as the effectiveness and feasibility of movement-supportive strategies may depend strongly on when they are introduced in the design and decision-making process.

4.2 RESEARCH AGENDA FOR FUTURE STUDIES

Building on the literature review, the proposed framework, and existing research in medicine, public health, transportation and urban design, the following priorities are outlined.

Study design and analysis:

• Future research should explore less understood relationships between building factors and physical activity, as outlined in the proposed conceptual framework.

• It is necessary to determine whether specific relationships between indoor environments and physical activity are causal. McMillan (2005) highlights the need for deeper examination of direct and indirect relationships, interactions, and hypothesised causal paths through hypothesis building and research design. The need for controlled studies has been highlighted as field studies often lack sufficient variation (Pollard et al. 2022), while observational studies may present several confounders (Bauman et al. 2002). For example, McGann et al. (2015) attributed increased physical activity in one building to staircase features, but the results could be attributable to a population more aware of sedentary risks. Controlled studies could disentangle such confounding factors to provide robust evidence of cause and effect. If controlled studies are not feasible, intervention studies should be conducted to identify causal pathways. While cross-sectional observational approaches can identify environmental correlates that guide interventions (Saelens & Handy 2008), longitudinal and quasi-experimental studies can provide stronger evidence of behavioural changes. In addition, case studies can be used to identify promising design strategies, generate hypotheses about movement-supportive strategies, and document real-world applications in exemplary buildings.

• Understanding the interactions among personal, social and contextual factors is essential, as demonstrated at the built-environment level (Annear et al. 2014). Models such as the social–ecological model and social–cognitive theory acknowledge the interplay among intrapersonal, social, and environmental factors that influence physical activity and sedentary behaviours (King et al. 2000; Owen et al. 2011). Multilevel statistical techniques, as suggested by Ball et al. (2006), can capture these complex interactions.

Study methodology:

• Collaboration and cross-fertilisation between the fields of urban planning and public health, and between architecture and engineering, are needed (Azzopardi-Muscat et al. 2020). To enable cross-study comparisons and syntheses, there is a need to standardise data collection and analysis methodologies.

• Advancing methodological efforts is essential, particularly in assessing physical activity through both quantitative and qualitative methods. While quantitative methods capture physical activity outcomes, qualitative approaches (e.g. go-along interviews) may help explain the behavioural mechanisms underlying movement patterns in buildings, which are largely lacking in current studies. Emerging technologies, such as eye-tracking glasses, biometric wearables, laser mapping from mobile devices and geospatial AI (VoPham et al. 2018), can revolutionise movement tracking. Virtual reality has shown promise in built environment research (Ghanbari et al. 2024) and crowd behaviour studies (Haghani & Sarvi 2018). Integrating these tools with traditional methods such as behavioural mapping and wearable devices (e.g. accelerometers, pedometers, heart rate monitors) can provide more comprehensive insights. Overall, mixed-method approaches combining quantitative data with surveys or qualitative observations may further help uncover the mechanisms underlying movement patterns in buildings.

Study goals:

• While new building designs receive significant attention, renovation projects also offer substantial opportunities to promote physical activity.

• Urban studies have shown disparities in access to recreational facilities in low-income neighbourhoods (Heath et al. 2006). Similarly, research on buildings must examine whether these disparities extend to indoor environments and account for variations by race, ethnicity, socio-economic status, age and ability.

• Research should integrate physical activity with other health, wellbeing and productivity outcomes, referred to as ‘Secondary Outcomes’ in the proposed framework. For instance, the impact of physical activity on mental, social and physical health warrants further investigation. Additionally, less-explored areas, such as the relationship between physical activity and thermal comfort, or energy savings from operational changes, should be prioritised.

4.3 LIMITATIONS AND STRENGTHS

This study has some limitations, including a lack of rigorous quantitative analysis (e.g. a meta-analysis) due to limited and inconsistent evidence, potential gaps in the literature search (e.g. non-English and grey literature) and the inclusion of observational studies, which limit causal inference. It also simplifies complex interrelationships between variables and does not account for geographical or demographic differences. However, the study’s strengths include its timely focus on physical activity in buildings, its broad and interdisciplinary search strategy, the use of ML to reduce screening bias, and adherence to high standards for systematic reviews, following PRISMA guidelines and the JBI Critical Appraisal Checklist for Systematic Reviews and Research Syntheses (Aromataris et al. 2015).