The effects of daily spatiotemporal variables and physical activities at workplaces on physical and mental health

Abstract

Previous studies have shown the effects of health on activity-travel patterns or spatiotemporal variables. Health is seen as another aspect of capability factor in the time-space prism perspective. However, it is hardly ignored that the effects of activity-travel patterns or other spatiotemporal variables also shape people’s health. Physical activities were assumed to mediate the effects of activity-travel patterns on health. This study addresses how vigorous and moderate physical activities at workplaces (vigorous and moderate PA work) mediate the relationship between spatiotemporal variables in particular activity-travel patterns and physical and mental health. Using the modified Structural Equation Modelling, this study also includes the effects of day-to-day variability of activity-travel patterns disregarded by previous studies. Different from previous studies, the definition of PA work also incorporates the interactions of the average time per day and the average number of days per week to engage in vigorous and moderate PA work. By including day-to-day activity-travel patterns, some activity-travel pattern variables showed significant impacts on physical health contradicting previous studies. However, as hypothesized, vigorous and moderate PA work mediated the effects of activity patterns on mental health. Against results in developed countries, PA work positively correlates with positive mental and physical health.

Keywords:

• Spatiotemporal variables
• activity-travel patterns
• physical activity
• physical health
• mental health

Reviewing Editor:

• Shukla Sanjay Kumar
• Senior Editor
• Edith Cowan University
• Australia

Subjects:

• Transport Policy
• Transport & Vehicle Engineering
• Civil, Environmental and Geotechnical Engineering
• Engineering & Technology
• Built Environment
• Health and Safety
• Health and Social Care
• Health & Society

Introduction

In 2015, all the United Nations members adopted the 2030 Agenda for Sustainable Development, which provides the objectives of sustainable goals in the future. Sustainable Development Goals (SDGs) have highlighted 17 indicators, which are an urgent call for action by all countries, either developing or developed countries, to collaborate to achieve these goals (The United Nations, 2015). Good health and well-being have been highlighted under SDGs as one of the goals. Health has a key position on the agenda and is closely linked to over a dozen other priorities related to public development, equitable access to care, and non-communicable diseases. However, given the increased GDP over the years in developed and developing countries, many countries have shown obesity problems among their adults in 2016 (World Health Organization, 2016).

In time-space prism studies, health has been argued as a component of capability constraints that will affect the individual’s daily activity-travel pattern (Dharmowijoyo et al., 2015; Dharmowijoyo & Joewono, 2020; Hunt et al., 2015). The capability constraint has been described by Hägerstrand (1970) as the limitation of a person to carry out his/her daily activities. For example, people who are sick or have a walking disability will have limitations to some activities such as travelling or working instead of staying at home most of the time. Having social and mental health problems can also limit people’s participation in social activities (Dharmowijoyo et al., 2017; Shergold, 2019).

However, it is difficult to ignore that the activity-travel patterns (Dharmowijoyo et al., 2020) and the built environment (Tajalli & Hajbabaie, 2017) also play important roles in influencing physical and mental health. Dharmowijoyo et al. (2020) argued that organising more regular activity-travel patterns directly or indirectly shapes people’s mental health, whereas more varied activity-travel patterns directly correlate with better social health. It is also found that organising more regular or more varied activities is better connected to activity-travel patterns and health than time-use and activity participation.

Although it seems that travelling and activity participation can directly influence health, Zhang (2013) and Dharmowijoyo et al. (2015, 2017) found that the relationship between activity-travel patterns and health is insignificant. Meanwhile, Van Wee and Ettema (2016) stated that the relationship between travel and health is not direct. Physical activities (PA) and intensities are factors that might mediate the relationship between travel and health. The physical intensities can be defined as the rate of the activity type carried out. Susilo and Liu (2017) stated that physically active travel modes such as walking and cycling contributed to more positive health effects.

This study tries to answer the research gaps ignored by previous studies especially Dharmowijoyo et al. (2015)’ and Ali et al. (2020, 2021)’s studies. As the expansion of Ali et al. (2020, 2021), this study does not only focus on inferring the complex interaction among multiple spatiotemporal variables (e.g. activity-travel patterns, socio-demographic and built environment), physical activities at workplace (PA work) and physical health but also mental health. This study infers that PA work can mediate the relationship between multiple spatiotemporal and physical and mental health variables. As the extension of Dharmowijoyo et al. (2015) and Ali et al. (2020, 2021), in inferring the relationship between multiple spatiotemporal and health mediated by PA work, this study also includes the day-to-day variability effects of people’s activity-travel patterns on vigorous and moderate physical activities at working places (PA work), and physical and mental health. In applying day-to-day variability effects, multilevel modelling which takes into account the people’s variation of daily activity-travel patterns is utilised in the model. When accounting for the day-to-day variability effects, besides socio-demographic and built environment variables, some activity pattern variables are hypothesised to significantly correlate with physical and mental health as opposed to Dharmowijoyo et al. (2015)’s results. This study only accounts structural model in showing whether PA work in different intensities will mediate the effects of multiple spatiotemporal variables on physical and mental health by combining multilevel and structural equation modelling using the modified Structural Equation Modelling (the modified SEM). Since the modified SEM is a multilevel model modified to solve the structural effects of SEM using the Instrumental Variable (IV) method as shown by Susilo and Kitamura (2008) and Dharmowijoyo et al. (2018, 2021), it will not be able to solve the reciprocal effects between the health and PA work. This is the limitation of this study. The details of the modified SEM will be explained in the method section.

In addition, different from Ali et al. (2020, 2021), PA work in this study is defined as the interaction of the average time per day and the average number of days per week to perform PA during a working time. Confirmatory Factor Analysis (CFA) will be used to measure the PA work from those two variables as the method was also suggested by Booth (2000) and Hagströmer et al. (2007). Factor scores will be estimated to create the composite value of PA work for the subsequent analysis as suggested by Hair et al. (2014). Moreover, activity time-use is not the only dimension hypothesised to significantly correlate with physical and mental health. The similar or different sequence of activity-travel patterns on weekdays and weekdays compared to other weekdays is also hypothesised to significantly correlate with physical and mental health. That variable will be included in this study. Moreover, even though there were some series of Susilo’s, Dharmowijoyo’s, Verma’s, and Varghese’s, the use of time-space prism and activity-based analysis in developing countries are very rare compared to the developed countries (Nayak & Pandit, 2023). Therefore, this research tries to fill the research gap concerning the use of time-space prism or activity-based analysis on health in developing countries.

This study used the 2013 Bandung Metropolitan Area dataset (2013 BMA dataset). As the authors are aware, the dataset was the first multi-dimensional dataset captured in developing countries. Even though the dataset is quite old, physical activities at workplaces (PA Work) seem not to change so much. Workers from low and middle-high-income households tend to have different PA work and authors assumed that the condition is similar between the current and 2013 situation. PA work is part of mandatory activities at working places and obligations to do some physical activities at working places might not change for some types of occupations. Therefore, the dataset can still be relevant for the current context.

After this section, literature review and method sections will be presented. Two analysis, bivariate and multivariate analysis, will be applied in this study. Lastly, conclusion and recommendations will be explained in the end of paper.

Literature review

The effects of time-space prism on health

Time-space prism reveals that capability, coupling, and authority constraints correlate with each other in shaping activity-travel patterns (Dharmowijoyo et al., 2015, 2017; Hägerstrand, 1970; Miller, 2017; Neutens et al., 2011). Coupling constraints refer to individuals’ limitations due to their interactions with other people in various activities (e.g. interactions during working, social, and leisure time) and materials (e.g. taking letters to the post office or taking money from an ATM). In contrast, authority constraints are defined as limitations imposed by written and unwritten regulations (e.g. regulations, norms, cultures, etc.). According to the definition of constraints in a time-space prism, travel is a permanent constraint that can limit people’s engagement in some activities (Dharmowijoyo, 2016; Hägerstrand, 1970; Miller, 2017; Nayak & Pandit, 2023). On the other hand, travel is often connected to certain activities that are fixed in time and space. To participate in these activities, individuals are required to take their time to travel and be at this location at their available times (Miller, 2017; Nayak & Pandit, 2023; Varghese et al., 2022) as the grocery store is situated in various places and people need to travel to reach the destination. In the relationship with physical and mental health, people’s participation in some activities can reduce their time allocation to do some physical activities or sports activities that might correlate with low or high physical health conditions. Moreover, spending a certain kind of physical activity at the workplace might limit participation in physical activities at home and vice versa (Voulgaris et al., 2019; Wicaksono et al., 2023). However, not all types of physical activity contribute to positive health. PAs in the workplace and household domains have negative and insignificant effects on mental-ill health, whereas PAs in leisure have positive impacts (White et al., 2017; Wicaksono et al., 2023).

In the space-time prism, activity-travel patterns are also influenced by resources (e.g. transport networks, the availability of pedestrians and cycling paths, land use shape, income, and access to private motorized mode). For instance, a person with access to a private motorised mode can travel farther than people with non-motorised modes such as bicycles. People with more income can use faster transportation modes, which allow them to travel much farther and faster. The availability of public transport networks and pedestrian and cycling paths might also trigger high intensities to do PA or non-motorised transportation (Andersen et al., 2015; Brown et al., 2019; Ewing & Cervero, 2010; Liu et al., 2017; Sener et al., 2016; Shanahan et al., 2015; Varghese et al., 2022).

However, the effects of spatiotemporal variables (e.g. activity-travel patterns and built environment) on health are unknown. No significant correlation is found the effects in the effects of those spatiotemporal variables on health (Dharmowijoyo et al., 2015, 2017; Syahputri et al., 2022; Van Wee & Ettema, 2016; Zhang, 2013). Physical activities were suggested to mediate the effects (Frank et al., 2019; Pereira et al., 2020; Susilo & Liu, 2017; Van Wee & Ettema, 2016; Wicaksono et al., 2023). Conventional activity diary survey does not provide detailed information on people’s physical activities and/or well-being evaluation which may have significant correlations on health variables. Therefore, some detailed variables such as PA might be able to mediate the effects of spatiotemporal variables on health and suggest which activity and travel types have positive and negative correlations on health. Understanding this would answer a research question of which activity and travel types should be encouraged or discouraged for better health conditions. The research question is relevant in the time-space prism study because people are believed not to be able to perform all activities, either healthy or unhealthy, due to the existence of constraints, unavailability of resources, and time limitations (Panik et al., 2019; Voulgaris et al., 2019) and energy limitation (Fedewa et al., 2017).

Many studies have investigated the effects of PA or active travel on general health (e.g. Standage et al., 2012; Wanner et al., 2012; Wicaksono et al., 2023), physical health (Frank et al., 2019; Stubbs et al., 2017) and proxy of health such as obesity, high blood pressures, diabetes and other chronic diseases (Tajalli & Hajbabaie, 2017). Active travels were found to correlate positively with mental health or proxy of mental health in longitudinal (Kroessen & De Vos, 2020) and cross-section studies (Currier et al., 2020). However, those studies ignored how PA can mediate the effects of activity-travel patterns on health, in particular physical and mental health.

Ali et al. (2020, 2021) showed the effects of PA on physical and social health. In that study, physical activities were defined as time engagements to perform moderate- and strenuous-intensity physical activities or M-IPA and S-IPA. S-IPA was found to positively correlate with physical and social health, whereas M-IPA showed negative effects. It might be because those who had no time to perform S-IPA or had physical problems (or higher capability constraints) might reschedule their time to perform M-IPA. Therefore, M-IPA was an escape to perform PA, but it did not result in positive health effects. Performing S-IPA might be a platform to socialise with other persons, whereas M-IPA can be done without the involvement of other persons. However, Dharmowijoyo et al. (2015) and Ali et al. (2020, 2021) performed cross-sectional studies and ignored the effects of day-to-day variability on physical health. Ali et al. (2020, 2021) also disregarded mental health. This study is the extension of Dharmowijoyo et al. (2015)’s and Ali et al. (2020, 2021)’s studies. This study focuses on investigating the effects of PA in mediating day-to-day activity-travel patterns and built environment variables on physical and mental health which account for the day-to-day variability effects using multilevel modelling.

Activity-travel patterns

Activity-travel patterns are seen as a proxy for capturing people’s daily time-space constraints (Dharmowijoyo, 2016; Schwanen et al., 2008). Activity-travel patterns contain some activities such as sleeping and taking a bath which may represent people’s capability constraints, whereas working, schooling, socialising, meeting, and dropping and picking up children, other household members and friends or colleagues correspond to coupling constraints.

Regarding activity classification, time-space prism or geography studies define activities whether the activities can be scheduled and performed flexibly at different time and places or not. Cullen and Godson (1975) and Schwanen et al. (2008) defined mandatory activities as activities that must be undertaken at a certain place for a fixed time (such as working, school, and pick-up/drop activities). In general terms, this activity will be prioritized more by individuals, although it is found that there is a different definition of fixed and flexible activities in developed and developing countries (Dharmowijoyo et al., 2018; Schwanen et al., 2008). Out-of-home mandatory activities can be described as activities to meet other individuals or materials at a certain location outside of private accommodation such as workplace locations, studying, meeting, and dropping off/picking up children to/from the school location. On the other hand, activities such as sleeping, eating, and personal care were defined as in-home mandatory as these activities are undertaken at home.

Meanwhile, discretionary activities tend to be re-scheduled easily (Cullen & Godson, 1975) and to be less prioritized by individuals (Dharmowijoyo, 2016; Nayak & Pandit, 2023). Discretionary activities were separated into maintenance and leisure. Maintenance activities were defined as discretionary activities for meeting the requirement of personal physiological and biological needs and household (Akar et al., 2011). Examples of in-home maintenance activities were housekeeping and nursing, whereas grocery shopping, health treatment, and other service activities were out-of-home maintenance. Leisure activities were conducted to fulfil cultural and physiological needs (Akar et al., 2011; Sobhana & Verma, 2023). These activities included entertainment (such as watching TV, accessing the Internet, and listening to music), social and family activities (such as visiting other family members/friends, sports activities, voluntary activity, and going on a vacation).

Methods

The 2013 BMA dataset contained various information such as socio-demographics, activity-travel patterns, household information, physical activity, and health-related quality of life. The study includes 513 people from all over BMA aged 17 and above which were from 191 households. The time-use and activity-travel participation were collected consecutively for a total of 21 days.

The household data section consisted of information on socio-demographics, physical activities (PA) at the workplace and home, and health-related information. PA questions were derived from International Physical Activities Questionnaires (IPAQ) (Booth, 2000; Hagströmer et al., 2007). PA in workplaces (PA work) is used in this study. PA at home (PA home) is not used in this study because, after the COVID19 pandemic, there might be a change in in-home activities due to the high intensities of digital activities. On the other hand, PA work in some work types in particular performed by workers from middle-low-income household tend to be similar to the 2013 situations or before the outbreak of COVID19. PA questions contained two types of questions: "How many days within a week do you do PA in a particular dimension in the last four weeks?" and "How many hours per day do you spend PA in a particular dimension in the last four weeks?". The profile of the dataset is shown in Table 1.

Table 1. Profile of the respondent.

Variables	Percentage	Mean
Socio-demographic characteristics at an individual level
Male	52.43%
Worker	56.33%
Non-worker (student)	43.67 (10.14%)
Young ages (<22 years old)	14.42%
Productive ages (23-55 years old)	73.29%
Senior citizens (>55 years old)	12.28%
Part of low-income households (< IDR 3 million/month)	75.05%
Part of medium-income households (IDR 3-6 million/month)	15.82%
Household characteristics
Number of household members		4.52
Single and couple (1 or 2 household members)	10.13%
A household with more than 2 members	89.87%
Trips engagements and travel time spent on weekdays (weekends)
Percentage of using private motorised transports	66.43% (64.85%)
Percentage of using public transports	13.86% (10.61%)
Percentage of using a non-motorised mode	19.71% (24.54%)
Total travel time spent from Monday-Friday (minutes)		782.58
Time spent for different activities on weekdays (weekends)
Time spent for in-home mandatory activities (minutes)		658.81 (708.35)
Time spent for in-home leisure & maintenance activities (minutes)		358.5 (402.04)
Time spent for out-of-home mandatory (minutes)		277.07 (154.87)
Time spent for out-of-home grocery shopping (minutes)		12.97 (22.42)
Time spent for out-of-home leisure & maintenance activities (minutes)		50.07 (75.95)
Built environment variables
Population density		11430.6
The density of industrial area per square km within the respondents’ residential location^1		0.0237
The density of government office^1		0.0118
The density of settlement area per square km within the respondents’ residential location^1		0.4872
How many days, on average, within a week, do individuals spend time for vigorous PAs like heavy lifting, digging, heavy constructions, or climbing stairs at working places (Out-of-home/O.H. strenuous/vigorous)? (days)		3.871
How long, on average, do individuals spend out-of-home vigorous PAs at workplaces (OH strenuous/vigorous) during a day? (minutes)		35.88
How many days on average within a week do individuals spend time for moderate PAs like carrying light loads at workplaces (OH moderate)? (days)		4.28
How long, on average, do individuals spend moderate PAs at workplaces (OH moderate) during a day? (minutes)		30.15

Note: ¹The percentage of a type of land use was calculated based on the size of built area in km² of the corresponding land use within a particular individuals’ residential location zone divided by the total area in km² of the individuals’ residential location zone multiplied by 100%. This measure is computed only in the horizontal plane; it excludes the area computed in a vertical plane.

The dataset recorded people’s activity-travel participation for 21 consecutive days in terms of activity-travel participation. Twenty-three types of activities were used to classify people’s daily activities and travels; thus, it was downsized into two types of activities: mandatory and discretionary, and whether the activities were conducted in-home or out-of-home. Travel mode is also included in the dataset. These travel modes are categorized into three classifications: private transport, public transport, and non-motorized transport. The private motorised mode can be seen from those using a car, motorcycle, or other privately owned transportation modes. Meanwhile, the non-motorized mode utilizes active travel such as walking and cycling. Public transport is a transport system available for use by the general public such as buses, trains, and others as defined in the survey. Table 1 shows the sample profile used in this study.

Built environment data used digital land-use data captured by Bappeda Kota Bandung (2011). Some built environment indicators are used in this study, namely the density of industrial, government, and settlement areas.

Health-related quality of life (QoL) questions are incorporated in a section in the questionnaire. SF-36 (Short-Form 36) was used to capture health-related QoL. SF-36 measured eight subscales (physical functioning (PF), limitations on role functioning because of physical health (RP), bodily pain (BP), general health (GH), mental health (MH), limitations on role functioning because of an emotional problem (RE), social functioning (SF), and vitality (VT) and totalled to 36 items included. According to Suzukamo et al. (2011), physical health is defined by PF, RP, BP, and GH, whereas mental health is composed of GH, BP, SF, VT and MH.

Following Suzukamo et al. (2011), physical and mental health were defined using Confirmatory Factor Analysis (CFA). Vigorous and moderate physical activities are derived from the International Physical Activities Questionnaire (Booth, 2000; Hagströmer et al., 2007), and CFA is also used to define vigorous and moderate physical activities as suggested by Booth (2000) and Hagströmer et al. (2007).

As suggested by Hair et al. (2014), factor scores were estimated to create a composite value for subsequent analysis that reflects the relative contributions of each of the observed variables as a result of CFA. Factor scores have no unit, and it is a standardized value as a Z score metric with a mean of zero and a value ranging from −3 to 3 across the sample Hair et al. (2014). Equation (1)(1) $$\widehat{F_i}=\mathrm{\Lambda }\prime {\sum }_{}^{-1}{y}_i$$(1) shows how to find the factor score value $\widehat{F_i}$ as a product of the factor loading matrix (Λ') as a result of CFA, the inverse of the covariance matrix (Σ⁻¹) and observed variables (y_i). (1) $$\widehat{F_i}=\mathrm{\Lambda }\prime {\sum }_{}^{-1}{y}_i$$(1)

The results of factor score estimation, including mean, maximum and minimum values, are shown in Table 2. SPSS version 20 was used to examine factor scores. As shown in Table 2, the loading factors show different weights; thus, factor scores represent health variables in the regression analysis than summated scales (‘average’ or ‘mean’ values).

Table 2. Loading factors of each observed variable and factor scores of health and physical intensity variables.

Observed variables	Factor loadings	Latent variables (mean, minimum and maximum value of factor scores)
Physical functioning (PF)	0.5	Physical health/PH (−0.09, −4.04, 2.13)
Limitations of role functioning because of physical health (RP)	0.493
General health (GH)	0.418
Bodily pain (BP)	0.401
General health (GH)	0.296	Mental health/MH (0.89, −3.46, 1.97)
Bodily pain (BP)	0.254
Social functioning (SF)	0.263
Vitality (VT)	0.413
Mental health (MH)	0.432
How many days in a week do you perform heavy lifting, digging, heavy constructions, or climbing stairs at working places	0.210	Vigorous physical activities at working places (0.00, −1.30, 5.94)
How many hours per day do you perform heavy lifting, digging, heavy constructions, or climbing stairs at working places	0.498
How many days in a week do you carry light loads at working places	0.391	Moderate physical activities at working places (0.00, −1.08, 6.00)
How many hours per day do you carry light loads at working places	0.520

As shown in Table 2, the loading factors show different weights; thus, factor scores represent health variables in the regression analysis than summated scales ('average’ or 'mean’ values).

The bivariate analysis

Bivariate analysis is used in the analysis to reveal the indication of correlation between activity-travel patterns and PA work, activity-travel patterns and physical and mental health, and PA work and physical and mental health. In the analysis, since factor scores will value vigorous and moderate PA work, and physical and mental health range from −3 to 3, high vigorous and moderate PA work, and high physical and mental health are defined as people who have factor scores above zero (>0) of the respective variables. However, on the opposite, low vigorous and moderate PA work, and low physical and mental health refer to people with the same or below zero (≤0) vigorous and moderate PA work, and physical and mental health, respectively.

Figures 1–10 exhibit the indication of the correlation between different types of activities and travels (represented by the use of private motorised mode) and vigorous and moderate PA work, and health variables. The use of motorised mode is a representation of travel activity here because motorised mode is the most dominant mode in the studied area as shown in Table 1 (the percentage of using private motorised mode is 66.43% on weekdays and 64.85% on weekends). Therefore, the effect of private motorised mode is expected to have higher magnitudes in explaining the PA work and health variables.

Figure 1. Daily time use of mandatory and in-home activities on strenuous PA work activities.

Figure 2. Daily time use of out-of-home discretionary activities and private motorised mode on strenuous PA work activities.

Figure 3. Daily time use of mandatory and in-home activities on moderate PA work activities.

Figure 4. Daily time use out-of-home discretionary activities and private motorised mode on moderate PA work activities.

Figure 5. Daily time use of mandatory and in-home activities and in-home activities based on physical health (PH).

Figure 6. Daily time use out-of-home discretionary activities based on physical health.

Figure 7. Daily time use of mandatory and in-home activities and in-home activities based on mental health.

Figure 8. Daily time use out-of-home discretionary activities, travel time and percentage of using motorised mode based on mental health.

Figure 9. Effects of the socio-demographic and built environment on physical health and PA work. *A low household size is defined as households with ≤ 4 household numbers, whereas a high household size is defined as households with > 4 household numbers. ** Low and high population densities are defined as areas with ≤ 12,000 people/km² and >12,000 people/km². *** Low and high settlement densities are defined as areas with ≤ 0.35 and > 0.35.

Figure 10. Effect of different intensities of PA work on physical and mental health.

The effects of daily activity-travel patterns on strenuous and moderate PA work

Figures 1–4 show the indication of the correlation between different activity-travel patterns, and vigorous and moderate PA work. Those who had high vigorous PA work tended to be those who undertook shorter in-home maintenance, but longer out-of-home leisure, out-of-home maintenance, and travel time using private transport than those who had low vigorous PA work as shown in Figures 1 and 2. On the other hand, in Figures 3 and 4, the ones who engaged in high moderate PA work were the ones who performed longer in-home mandatory and working/schooling time but shorter time to in-home discretionary activities than the ones who had low moderate PA work. Similar to people who performed high intensities of vigorous PA, people who performed more out-of-home leisure, maintenance, and travel time with private transport tended to perform high intensities of moderate PAs.

The effects of daily activity-travel patterns on physical and mental health

Figures 5 and 6 show the relationship between activity-travel patterns and physical health. As illustrated in Figure 5, people who tended to have better health were the ones who worked and did in-home maintenance more than leisure activities. However, there is a tendency for healthier persons concerning physical health were those who spend longer time for out-of-home leisure on weekdays but spent it shorter on weekends and performed longer shopping time as shown in Figure 6. Moreover, healthier people were those who tended to travel more time using private motorised transport (Figure 6).

Figures 7 and 8 illustrates the effects of activity-travel patterns on mental health. The figures show that those who had worse mental health were those who tended to work more, sleep less and performed in-home discretionary activities longer (as displayed in Figure 7), but undertook out-of-home discretionary and travelled less (as displayed in Figure 8). Those who spent less travel time using motorised mode tended to also have low mental health.

The effects of socio-demographic and built environment and different physical intensities on physical and mental health

In terms of socio-demographic variables shown in Figure 9, males, younger adults, and people with small household sizes tended to have better physical and mental health. However, those who were of productive ages tended to perform their days and time during the day with more vigorous and moderate PA work than their counterparts. Still, from Figure 9, people who resided in a high population density tended to have better PA work, and physical and mental health. On the other hand, people who stayed in areas with low settlements tended to have better work PA but low physical and mental health. High population density does not necessarily have high settlement density. This is because the density of a land use in this study only accounts for measurements in the horizontal plane and not in the vertical plane as noted in Table 1. It means that high population density areas might be located in the area with low settlement areas with slump areas and apartments as in suburban areas of BMA near the city centre and in the western part of BMA, whereas high settlement areas are located in suburban areas in the east of BMA and greater areas (Dharmowijoyo et al., 2023).

Figure 10 shows how people with high vigorous and moderate PA work tended to have worse physical and mental health than those who performed with low PA work. The results show an opposite indication as stated in the earlier hypothesis of this study.

The bivariate analysis shows some evidence that is in line with the earlier hypothesis, but some others are not. Health might not be influenced by only one variable but by multiple variables in a complex system. Therefore, multivariate analysis is worth conducting.

The multivariate analysis

In multivariate analysis, this study infers whether vigorous and moderate PA mediate the effects of multiple spatiotemporal variables on physical and mental health. This study also includes day-to-day variability effects that were ignored by Dharmowijoyo et al. (2015) and Ali et al. (2020, 2021). The model structure is shown in Figure 11. Since the objective is to infer relationship between spatiotemporal and health variables mediated by different intensities of PA work in considering complex interactions among those variables, having better goodness fit of model is not the objective of this study. Therefore, improving the goodness of fit of this study is not necessary the aim and be discussed. This study focuses on whether the model framework shown in Figure 11 works well or not, and the correlation tendencies of empirical evidence among the relationships of spatiotemporal, PA work and health variables.

Figure 11. Model structure.

To solve the model structure as in Figure 11 and to include the day-to-day variability effects of activity-travel patterns, the modified Structural Equation Modelling (modified SEM) is applied in this research. The modified SEM is actually a multilevel model modified to tackle some endogeneity problems as shown in Figure 11 as used as well by Susilo and Kitamura (2008) and Dharmowijoyo et al. (2018, 2021). It is assumed that vigorous and moderate PA work has a high correlation with spatiotemporal variables such as socio-demographic and daily activity-travel patterns. Therefore, a structural equation is proposed. The Instrumental Variable (IV) method is used to tackle the endogeneity between vigorous and moderate PA work and spatiotemporal variables. The IV method was used by 2SLS and 3SLS. However, the IV method used in this study is more similar to 2SLS because it is omitting the simultaneous effects of the SEM by disregarding the correlation among various error terms of the lower- and upper-level equations. Therefore, similar to 2SLS, this method can solve the structural form of SEM, but cannot solve the reciprocal effects of SEM because of omitting the third step or disregarding the correlation among various error terms.

Since it is a multilevel modelling modified to solve the structural effects of SEM, the goodness-of-fit used in the modified SEM is the good-ness-of-fit of the multilevel modelling different from traditional SEM which uses AMOS, MPLUS or LISREL software. Since the modified SEM uses the NLME package in R, the goodness-of-fit is the NLME’s goodness-of-fit such as AIC, BIC, and log-likelihood, and not traditional SEM’s goodness-of-fit. Moreover, as a common multilevel model modified to solve the structural effects of SEM, another SEM analysis such as direct and indirect effects is not provided as well.

Using the modified SEM, the coefficients (βn) (Equations (2)–(5)) include activity-travel patterns made by individual i on day t. The u_i, and ε_h,i,t are the individual specific error terms, and the uncorrelated combined household, individual, and time error components, respectively. Those error terms have a mean value of zero and variance σ_u (for u_h, u_i) and σ_ε (for ε_h,i,t). The mathematical models are shown below: (2) $${\mathit{\text{VPA}}}_{i,t}=({\alpha }_{1,i}+u_{1,i})+{\mathbf{\beta }}_1{W}_i+{\mathbf{\beta }}_{\mathbf{2}}{R}_i+{\beta }_3\mathit{\text{Time}}-{\mathit{\text{Act}}}_{i,t}+{\mathbf{\beta }}_{\mathbf{4}}{\mathit{\text{Travel}}}_{i,t}++{\epsilon }_{1i,t}$$(2) (3) $${\mathit{\text{MPA}}}_{i,t}=({\alpha }_{2,i}+u_{2,i})+{\mathbf{\beta }}_5{W}_i+{\mathbf{\beta }}_{\mathbf{6}}{R}_i+{\beta }_7\mathit{\text{Time}}-{\mathit{\text{Act}}}_{i,t}+{\mathbf{\beta }}_{\mathbf{8}}{\mathit{\text{Travel}}}_{i,t}++{\epsilon }_{2,i,t}$$(3) (4) $$\begin{array}{l}{\text{PH}}_{\mathrm{i}}=({\alpha }_{3,\mathrm{i}}+{\mathrm{u}}_{3,\mathrm{i}})+{\mathbf{\beta }}_{\mathbf{9}}{\mathrm{W}}_{\mathrm{i}}+{\mathbf{\beta }}_{\mathbf{10}}{\mathrm{R}}_{\mathrm{i}}+{\mathbf{\beta }}_{\mathbf{11}}\text{Time}-{\text{Act}}_{\mathrm{i},\mathrm{t}}+{\mathbf{\beta }}_{\mathbf{12}}{\text{Travel}}_{\mathrm{i},\mathrm{t}}+\widehat{{\gamma }_1{\mathit{\text{VPA}}}_{i.t}}\\ +\widehat{{\gamma }_2\mathit{\text{MPA}}{}_{i.t}}+{\epsilon }_{\text{3i},\mathrm{t}}\end{array}$$(4) (5) $$\begin{array}{l}{\text{MH}}_{\mathrm{i}}=({\alpha }_{4,\text{i}}+{\mathrm{u}}_{4,\mathrm{i}})+{\mathbf{\beta }}_{\mathbf{13}}{\mathrm{W}}_{\mathrm{i}}+{\mathbf{\beta }}_{\mathbf{14}}{\mathrm{R}}_{\mathrm{i}}+{\mathbf{\beta }}_{\mathbf{15}}\text{Time}-{\text{Act}}_{\mathrm{i},\text{t}}+{\mathbf{\beta }}_{\mathbf{16}}{\text{Travel}}_{\mathrm{i},\mathrm{t}}+\widehat{{\gamma }_3{\mathit{\text{VPA}}}_{i.t}}\\ +\widehat{{\gamma }_4\mathit{\text{MPA}}{}_{i,t}}+{\epsilon }_{4,\text{i},\mathrm{t}}\end{array}$$(5)

Linear and non-linear mixed effect model through NLME package (Pinheiro et al., 2017) is used to apply the modified SEM. Pinheiro et al. (2017) explained the maximum likelihood estimation in NLME as the Equation (6)(6) $$L^{\mathrm{*}}\left(\sum _b,\beta ;y\right)=\int \left(P(y\right|\beta ,b,i)\left(P(b\right|\sum _b\mathrm{ })\mathit{\text{db}}$$(6) . (6) $$L^{\mathrm{*}}\left(\sum _b,\beta ;y\right)=\int \left(P(y\right|\beta ,b,i)\left(P(b\right|\sum _b\mathrm{ })\mathit{\text{db}}$$(6)

The salient results of the multivariate analysis are shown in Table 3. As hypothesised, socio-demographic, activity-travel patterns and built environment variables showed significant effects on vigorous and moderate PA work and physical helath. The trip characteristics showed the highest magnitudes on all models. It is indicated that those who performed regular vigorous and moderate PA at working places are those who undertook more trips within fewer trip chains. However, those persons were categorized as unhealthy concerning physical and mental health. The healthier persons were those who performed more trips within more than one trip chain.

Table 3. The estimated results (using standardized coefficients and only significant variables with p-value < 0.1 are shown).

Variables	Vigorous PA at the workplace		Moderate PA at the workplace		Physical health		Mental health
	Coeff	T-stat	Coeff	T-stat	Coeff	T-stat	Coeff	T-stat
Intercept	1.131	4.821	−0.730	−3.116	−0.0481	−0.239	0.007	0.030
Individuals’ characteristics
Males	0.043	3.937	0.084	7.652			−0.086	−4.472
Females	Ref	Ref	Ref	Ref	Ref	Ref
Below 23 years old	0.043	2.951	0.074	5.056	0.223	11.259	−0.063	−3.030
Between 23 and 45 years old	0.104	6.933	0.185	12.342	0.221	6.381	−0.253	−6.942
Between 46 and 55 years old	0.046	3.589	0.102	7.915	0.198	9.209	−0.089	−3.924
Above 55 years old	Ref	Ref	Ref	Ref	Ref	Ref
Workers	−0.027	−2.358			0.147	12.657	0.100	0.179
Students	−0.064	−5.126	0.041	3.294	0.153	11.511	0.040	0.113
Non-workers	Ref	Ref	Ref	Ref	Ref	Ref
Households’ characteristics
From high-income households	−0.022	−5.990	−0.096	−9.746			0.138	7.405
From middle-income households	−0.057	−2.262	−0.084	−8.754	0.068	3.811	0.133	6.665
From low-income households	Ref	Ref	Ref	Ref	Ref	Ref
Number of household members							−0.051	−4.676
Number of dependent children	0.054	5.166	0.037	3.523	−0.037	2.715	−0.095	−7.094
Having access to motorcycles and cars	−0.066	−6.930	0.109	11.387	−0.134	−7.470	−0.163	−8.576
No access to motorcycles and cars	Ref	Ref	Ref	Ref	Ref	Ref
Having access to cars	0.073	7.064	0.038	3.645	0.067	5.118
No access to cars	Ref	Ref	Ref	Ref	Ref	Ref
Travel patterns
Percentage of using the motorised mode	−0.019	17.891	−0.084	−4.806			0.172	7.339
Percentage of using public transport	−0.053	−0.361	−0.071	−5.690	0.067	3.843	0.138	7.496
Percentage of using non-motorised transport	0.031	2.193					0.039	2.630
Number of trips	0.404	17.891	0.358		−0.512	−7.690	−0.777	−11.048
Number of trip chains	−0.286	−13.244	−0.167		0.342	8.704	0.463	11.172
Time-use and activity patterns
In-home mandatory activities	−0.083	−3.273			0.149	5.802
In-home maintenance activities	−0.101	−4.019			0.084	3.299
In-home leisure activities	−0.103	−3.893			0.097	3.586
Out-of-home mandatory activities	−0.091	−2.574	0.076	2.140	0.132	3.638
Out-of-home socialising activities	−0.033	−1.921			0.090	5.245
Out-of-home sports activities			0.019	1.737
Out-of-home grocery shopping	−0.066	−6.046			0.089	7.841
Out-of-home other maintenance
The average variability in individuals’ activity-travel patterns on particular weekdays compared to other weekdays	0.033	2.884	0.067	5.947	−0.059	−3.828	−0.223	−13.770
The average variability in individuals’ activity-travel patterns on particular weekends compared to other weekdays	−0.033	−3.265	−0.021	−2.081	0.037	3.413	0.050	4.368
Built environment patterns
Average distances to various public amenities	−0.045	−4.753			−0.107	−11.370	0.047	4.744
Number of public transport lines	−0.070	−7.497	−0.050	−5.345
Diversity index within individuals’ residential locations	−0.117	−7.237	0.150	9.297	−0.177	−7.623	−0.308	−12.584
Population density/km^2 within individuals’ residential locations	0.193	14.094	0.059	4.197			0.077	3.495
The density of settlement areas/km^2 within individuals’ residential locations	−0.223	−11.624	−0.085	−4.417	0.116	4.219	0.058	2.001
The density of industry areas/km^2 within individuals’ residential locations	0.200	17.224	−0.054	−4.670
Endogenous variables
The endogenous variable of vigorous PA at the workplace					0.217	10.208	0.213	9.473
The endogenous variable of moderate PA at the workplace					0.156	2.524	0.528	8.116
Mean	0.000		0.000		−0.089		−0.032
SD	1.000		1.00		0.940		0.895
Individual household error terms	0.00003		0.00003		0.00003		0.00003
White noise	0.912		0.911		0.851		0.855
AIC	30774.56		30739.70		29184.93		29285.14
BIC	31032.03		30997.17		29442.40		29542.61
Log-likelihood	−15352.28		−15334.85		−14557.47		−14607.57

Vigorous and moderate PA at working places

In the vigorous PA work (VPA_i) model, the objective-built environment characteristics significantly correlated with higher magnitudes than socio-demographic and activity pattern variables, but lower than trip characteristics. Those who resided within areas with denser industrial areas and denser populations positively correlated with vigorous PA work. Those areas might indicate the residential locations of the blue collars or people who work as labourers in many industrial areas. People who lived near the city centre or those who resided in areas with high mixed-land use negatively correlated with vigorous PA work. Younger males particularly at age 23–45 years old who are part of low-income households and have dependent children were those who undertake vigorous PA work more often. Non-workers engaged in vigorous PA work more often than workers and students. In the 2013 BMA dataset, the part-time workers were defined as non-workers. Therefore, non-workers, here, might spend time at working places. Having commitments to longer working and studying and in-home activity time are found to have higher magnitudes in reducing vigorous PA work than commitments to perform out-of-home discretionary activities.

In the moderate PA work (MPA_i) model, besides having commitments to do vigorous PA at workplaces, younger males particularly at age 23–45 years old with more dependent children and from low-income households had commitments to do moderate PA work. Different from the vigorous PA model, students were those who had more commitments to do moderate PA at their school. No in-home activity variables significantly correlated with moderate PA work. Moderate PA work was shaped by out-of-home activities particularly working/studying and sporting time.

As hypothesised, having a different or similar sequence of activity-travel patterns on weekdays and weekends compared to other weekdays showed significant impacts on vigorous and moderate PA work. Those who have more varied activity-travel patterns on weekdays compared to other weekdays positively correlated with vigorous and moderate PA. However, those who have more regular activity-travel patterns on weekdays but more varied on weekends showed the opposite effects.

Physical and mental health

As shown in Table 3, endogenous variables of vigorous and moderate PA work positively correlated with physical and mental health as expected. When a health model includes various time-space prism variables and PA, built environment variables show a low magnitude compared to trip characteristics, endogenous variables of vigorous and moderate PA work and age. No activity patterns are found to significantly correlate with mental health as hypothesised, but activity patterns are found to significantly correlate with physical health as expected.

In the physical health model, age, endogenous variables of vigorous and moderate PA work and employment status significantly correlated with physical health with higher magnitudes than activity patterns and travel mode. It is quite plausible that younger persons particularly people aged below and at 45 years old had better physical health than people at age above 45 years old. Workers and students had better physical health than non-workers, whereas people from middle-income households showed more positive effects on physical health than people from low- and high-income households. Performing longer time to mandatory activities either out-of-home or in-home showed more positive effects on physical health than discretionary activities.

In the mental health model, endogenous variables of vigorous and moderate PA work are seen to mediate the effects of activity patterns on mental health. However, having more varied or more regular activity-travel patterns was found to significantly correlate with mental health. More travel with any travel mode showed positive effects on mental health. Taking travel using private motorised vehicles showed higher magnitudes than other travel modes. Older workers and students from middle-high-income households had better mental health than younger non-workers from low-income households. Access to jobs and having more financial security might be the reasons why older workers and students from middle-high-income households had the highest mental health as also indicated by Irish Aid (2016) and Dharmowijoyo et al. (2020).

Those who arranged activity-travel patterns more regularly during weekdays but had more variability during weekends showed better physical and mental health as indicated by Hunt et al. (2015) and Dharmowijoyo et al. (2017, 2020). The diversity index is the strongest built environment variable. However, those who resided in areas with high mixed land use negatively correlated with physical and mental health.

Conclusions

This study tries to expand previous studies by including the day-to-day variability effects of people’s activity-travel patterns. Moreover, this study also tries to explain whether vigorous and moderate physical activities at workplaces (vigorous and moderate PA work) mediate the effects of time-space prism variables on physical and mental health. As hypothesized, this study shows that PA work with different intensities positively correlates with physical and mental health as opposed to Ali et al. (2021) results. Against results in developed countries (Voulgaris et al., 2019) and another study in developing countries with different methods and datasets (Wicaksono et al., 2023), PA work with different intensities showed positive effects on physical and mental health. As explained above, the Ali et al. (2021) study disregarded day-to-day variability and had a different physical intensity definition. Different from Dharmowijoyo et al. (2015), when including day-to-day activity-travel patterns, some activity-travel pattern variables showed significant impacts on physical health. However, as hypothesised, vigorous and moderate PA work mediated the effects of activity patterns on mental health. In detail, this study confirms that:

1. Having more regular activity on weekdays showed higher magnitudes on mental health than socio-demographic, travel modes, activity patterns and built environment variables.

2. Balancing life with having jobs, enough rest and more breaks to perform leisure activities both at home and out-of-home showed more positive effects on physical health than performing a more regular activity-travel pattern on weekdays.

3. Stage of life and employment status also showed higher magnitudes on physical and mental health than activity patterns and built environment variables.

4. Built environment settings had fewer impacts on physical and mental health than the stage of life and commitments to perform vigorous and moderate PA work.

5. Having more mixed land use without providing walking or cycling paths negatively correlated with physical health.

6. Younger males from low-income households and having more dependent children are seen to be those who are exposed to more vigorous and moderate PA work which to some extent, they potentially had better physical health. However, having more dependent children and part of low-income households tended to reduce their physical health.

Recommendations and limitations

This study highlights that the natural behaviour of people with high vigorous and moderate PA work were usually those who performed more trips within one trip chain and more varied activity-travel patterns during weekdays, but more regular activity-travel patterns on weekends in comparison to other weekdays. Those persons likely had negative physical and mental health. People with high vigorous and moderate PA work were recommended to change their natural behaviour in order to improve their physical and mental health. To improve their physical and mental health, those people should arrange trips within more trip chains and perform more regular activity-travel patterns on weekdays with more variability during the weekends. This study also finds that having more dependent children and part of low-income households should be approached to change their activity-travel behaviour arrangements to improve their physical health.

Regarding mental health, this study recommends providing financial security to senior citizens or people at aged > 55 years old to shape people’s positive mental health as also found in other studies in developing countries (e.g. Dharmowijoyo et al., 2020; Irish Aid, 2016). More travel with private vehicles also shapes people’s mental health which might ensure travellers reach some leisure destinations more often with shorter travel time in combination with non-discretionary activities.

This study advocates that balancing life with having jobs, enough sleep and more breaks both in-home and out-of-home by performing leisure activities was found to shape better physical and mental health compared to setting the built environment. Moreover, having a diverse land use, but with fewer incentives to do more travel to those from low income and less employment opportunities to senior citizens likely correlated with negative mental health.

Even though younger males from the low-income group might still have to do more vigorous and moderate PA work now in comparison to 2013 as reported in this study, there are so many changes in situations. More online activities were found according to 2016 data collection in comparison to 2012–2014 (Rizki et al., 2021). Moreover, the COVID-19 pandemic is believed to give more exposure to ICT penetrations and more online activity engagements. More flexible working hours and changing situations might make the relationships among activity-travel patterns, PA work and PA non-work, and physical and mental health different from this study. More data collection is required.

Disclosure statement

No potential conflict of interest was reported by the author(s).