The effect of transport infrastructure, congestion and reliability on mental wellbeing: a systematic review of empirical studies

ABSTRACT

There has been a growing interest in understanding the interdependencies between urbanisation and mental health. Although transportation in cities is complex and of foremost importance to support the mobility of goods and passengers, little is known about how it relates to individual psychological distress. This review aims to provide an up-to-date synthesis of research evidence about the influence of transport infrastructure and operational performance (congestion, delays and reliability) on mental health/wellbeing. It is structured around three main interacting concepts that determined the search/selection of articles: identification of the above-mentioned transport-related exposures; use of psychological and physiological validated instruments; and the outcome on mental health/wellbeing. Ultimately, 69 studies were identified involving an empirical quantitative focus that met the inclusion criteria. We summarise the instruments most reported in these studies and the findings linking transport indicators and psychological and physiological outcomes. Across the review, we identified evidence of the contribution of key transport infrastructure, congestion and delay indicators on negative affective states and psychophysiological distress. Regarding transport reliability, the scarce number of studies identified did not allow for drawing similar firm conclusions. We conclude by discussing some limitations and providing recommendations for future research and policy-making agendas.

KEYWORDS:

• Congestion
• transport delays
• transport reliability
• transport infrastructure
• mental health
• wellbeing

Introduction

Recently, we have observed an increased role of mobility in the prevention and complementary treatment of mental disorders, primarily by promoting physical activity through active mobility (WHO, 2019). Such recent policies relied on evidence of the positive effect of physical activity on mental health – “a state of well-being in which the individual realises his or her own abilities, can cope with the normal stresses of life, can work productively and fruitfully, and is able to make a contribution to his or her community” (WHO, 2004, p. 60). However, individual mobility is associated with numerous other dimensions that may impact mental wellbeing: social inclusion, stress, cognitive workload, driving anxiety, or travel satisfaction. Activity-space representation of mobility has shown to have an explanatory power on specific mental disorders such as depression (Perchoux, Chaix, Cummins, & Kestens, 2013; Saeb et al., 2015). Ultimately, mobility-related attributes experienced by individuals in each of these space–time travel will impact their mental states. Nevertheless, similarly to active mobility, robust evidence of such a relationship is required for constructive and sustainable policy recommendations.

There has been a growing interest in building evidence on the relationship between mobility attributes, urban planning, and mental wellbeing and travel satisfaction. Traditionally, studies have focused on assessments of subjective wellbeing (De Vos, Schwanen, Van Acker, & Witlox, 2013). Yet, the increased availability of relatively affordable biosensing technologies has opened the door to more objective quantification of physiological states. Although some factors are expected to impact mental wellbeing (e.g. congestion in traffic, commuting delays, uncertainty and inadequate infrastructure), evidence has rarely been systematically evaluated.

This paper provides a systematic review of studies looking at the relationship between transport infrastructure and operations with mental health/wellbeing, with an emphasis on the physiological and psychological instruments used.

Our work is based on a conceptual framework (see Figure 1) composed of three key components: (1) the exposure, or nature of the mobility attribute from the perspective of the mobility provider (i.e. infrastructure and operational performance); (2) the instrument, which can rely on physiological and/or psychological measurements; and (3) the outcome, as an impact on mental health and wellbeing.

Figure 1. Conceptual dimensions and relationships guiding the review.

With this exercise, we aim to:

• Provide an up-to-date synthesis of the empirical evidence on how transport infrastructure and its operational performance – namely congestion, delays and reliability – affect individuals’ mental health/wellbeing.

• Describe the psychological and physiological measures used in the identified evidence.

• Determine key indicators related to the transport infrastructure and its operational performance that have an impact on mental health/wellbeing, to inform future research and policy-making agendas.

Methodology

We considered empirical studies written in English from 1 January 2000 to 30 June 2021, published in full-length peer-reviewed journals. To inform design and eligibility criteria of this systematic review, we adopted a “PECO-approach” – by defining Population, Exposure, Comparator and Outcomes (Morgan, Whaley, Thayer, & Schünemann, 2018). We included studies with adult populations (healthy or unhealthy) examining the impacts of congestion, delays, reliability, and/or transport infrastructure on mental health/wellbeing. Comparisons were not specified as criteria. Studies had to provide at least one quantitative outcome measure for mental health/wellbeing – either a validated psychological instrument or an emotional- or stress-related physiological response. As “satisfaction with travel” is considered a domain-specific component of subjective wellbeing, we included studies where satisfaction components were measures of mental state (using validated scales).

Five electronic databases were searched: PubMed, ProQuest, Web of Science, Scopus and TRID. The search strategy followed our conceptual framework (Figure 1) and used search terms related to the 3-pillars structure: exposure (including general mobility and transportation terms, and more specific ones for both infrastructure and operational performance); instruments (including well-known psychological scales and specific search strings for physiological measures); and outcome (including terms covering general mental health/wellbeing, emotional and mood states). Full details of the strategy and search terms can be found in Appendix 1.

Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021), Figure 2 depicts the flow of information through the different phases of the review process. Initially, two authors independently screened the identified papers according to titles and abstracts for eligibility. Discrepancies were resolved by team discussion. Studies selected for full-text review were distributed among all the authors, who independently extracted data into a purpose-built, predesigned, and structured template. For the synthesis, studies were first grouped according to each transport-related aspect and then as a function of whether it used psychological or physiological measures.

Figure 2. PRISMA flow diagram.

Each included study was also independently assessed by two authors using a quality checklist tool (Kmet, Lee, & Cook, 2004) that involves a rating (a score between 0 and 2, wherever possible) on 14 criteria (e.g. study design and sample size). If disagreements occurred, they were resolved through discussion. All studies were included regardless of their quality scores. The vast majority of included studies were classified as high quality (80%), and only 1 (1%) had a low-quality score (see Appendix 2).

Results

This section presents the main findings obtained from our systematic review. We firstly describe the instruments – psychological scales and physiological measures – used by the included studies. We then focus our attention on the empirical data relating each transport-related attributes – infrastructure, congestion and delays, and reliability – with changes in both types of instruments. The information provided sets the ground for identifying key transport-related indicators and on how evidence should be translated to practice.

Instruments used for mental health/wellbeing assessment

Psychological scales

The mental health-related psychological constructs considered in the reviewed papers are affective states (emotions and mood); wellbeing and satisfaction with life or travel; stress and mental workload; and mental health disorders. Appendix 3 provides an overview of the psychological instruments used.

The most frequently considered emotion was anger. The State-Trait Anger Expression Inventory – STAXI (Spielberger, Sydeman, Owen, & Marsh, 1999) is the most established psychographic instrument to measure state and trait anger (Fairclough & Spiridon, 2012). The Driving Anger Scale – DAS (Deffenbacher, Oetting, & Lynch, 1994) measures anger in the context of different traffic situations and can reveal how angry people generally get in traffic as well as when provoked by different kinds of traffic situations. The Driving Anger Expression Inventory – DAX (Deffenbacher, Lynch, Oetting, & Swaim, 2002) measures to what degree and in what form drivers express anger to other drivers and differentiates between three forms of aggressive expression: verbal, physical, and using the vehicle to express anger. Anger and driving anger have most frequently been assessed in relation to an outcome of congestion, but also in relation to specific traffic infrastructure. Like the STAXI for anger, the State-Trait Anxiety Inventory – STAI (Spielberger et al., 1999) measures a general state and a trait component of anxiety. The Driving Situations Questionnaire (DSQ) – Anxiety measures anxiety specifically in driving situations (Taylor & Paki, 2008). Several other instruments have additionally been used to measure driving anxiety and avoidance.

Several studies measured a whole spectrum of different moods, which are mostly understood as more diffuse and longer-lasting affective states – as compared to more specific and shorter termed emotions (Niven, 2013). Nevertheless, there is an overlap, as state anger, for example, is also measured as part of the University of Wales Institute of Science and Technology (UWIST) Mood Adjective Checklist (Matthews, Jones, & Chamberlain, 1990). While most reviewed studies used instruments based on adjective lists to assess the current mood (e.g. Global Mood Scale – GMS, Brief Mood Introspection Scale – BMIS), one study used a non-verbal pictorial assessment, Self-Assessment Manikin – SAM (Ihme, Unni, Zhang, Rieger, & Jipp, 2018; Zhang, Ihme, & Drewitz, 2019) and another used a method related to Kahneman’s Day Reconstruction Method – DRM (Kahneman & Krueger, 2006; Morris & Hirsch, 2016). Also, in terms of mood, we find more specific traffic situation-related instruments, such as the short Travel Mood scale (TMS) and the more comprehensive Driver Behaviour Inventory (DBI), which assesses different emotions while driving.

In contrast to emotions and mood, subjective wellbeing and general satisfaction with life relate to people’s experiences of affect over a long period of time (Niven, 2013). Satisfaction with travel is considered a domain-specific component of subjective wellbeing. The most established scale to measure travel-related wellbeing is the Satisfaction with Travel Scale – STS (Ettema et al., 2011). Yet, the instrument has sometimes been adapted (He, Thøgersen, Cheung, & Yu, 2020) or only used with a focus on the affective dimension of wellbeing while travelling (Lunke, 2020).

Several scales assessed the self-reported level of stress and mental workload in general as well as specifically in driving situations. We differentiate between instruments that measure non-specific longer-lasting stress understood as a part of mental health/wellbeing (Cohen, Kamarck, & Mermelstein, 1983; Kessler et al., 2002), and those focusing on task-stress in a performance situation (e.g. during or just after a driving simulator task). Task-stress can be measured in detail by the Dundee Stress State Questionnaire – DSSQ (Matthews, Szalma, Panganiban, Neubauer, & Warm, 2013), which focuses on different facets of task engagement, distress, and worry.

Some instruments focus on stress as typically perceived in driving situations, while others focus on stress when using a public transport. Additionally, Coping Inventory for Task Stress – CITS (Matthews & Campbell, 1998) and Driver Coping Questionnaire – DCQ (Matthews, Desmond, Joyner, Carcary, & Gilliland, 1996) focus on different strategies to cope with (driver) stress, and both differentiate between task-, emotion-focused coping, and avoidance.

Mental workload is related to a specific task and is understood as the difference between demanded cognitive resources and the cognitive resources available to solve the task. The NASA-Task Load Index – NASA-TLX (Hart & Staveland, 1988) and the Rating Scale Mental Effort – RMSE (Zijlstra & van Doorn, 1985) are among the most frequently applied (in simulator studies) to measure self-reported mental workload (Widyanti, Johnson, & de Waard, 2013).

Regarding mental disorders, short forms of screening instruments like the General Health Questionnaire GHQ-12 (Goldberg et al., 1997) and Short Form Health Survey SF-12 (Knies, 2014) have been used for the probability and/or severity of a mental disorder. Two screening instruments have been used: for depression, 10-item Center for Epidemiologic Studies Depression – CESD-10 (Andresen, Malmgren, Carter, & Patrick, 1994) and for burnout, Maslach Burnout Inventory – MBI (Maslach, Jackson, & Leiter, 1997), each one only in a single study.

Physiological measures

The stress response triggers a variety of cognitive, emotional, and biological processes that alter the body’s physical state either directly (and automatically) through the autonomic nervous system (ANS) or neuroendocrine pathways; or as a consequence of the enacting behaviour (Ulrich-Lai & Herman, 2009). Most studies used the physiological measures to evaluate the impact of congestion/delay (largely on driving simulators), and fewer for transport infrastructure or reliability (Appendix 4 provides detailed information on used measures).

The ANS is a peripheral physiological system responsible for modulating a wide variety of body systems, including cardiovascular, respiratory, and temperature (Hamill, Shapiro, & Vizzard, 2004). It is made up of a complex network of neurons for interoception, as well as activation and relaxation response (sympathetic and parasympathetic pathways, respectively). Despite its general-purpose nature, the ANS assessment is often used to measure stress and emotional response (Critchley & Harrison, 2013; Mauss & Robinson, 2009; Öhman, Hamm, & Hugdahl, 2000). Commonly used measures of ANS activation include electrodermal activity (EDA) and cardiovascular response – heart rate variability (HRV) being more appropriate than heart rate (HR) alone (Camm et al., 1996). Hence, most studies included some ANS evaluation (EDA being the most used, followed by HRV and HR). Some studies used measures that are less specific of psychological/emotional distress and more influenced by other factors – such as thermal imaging and respiratory/ventilatory parameters (Stemmler, 2004).

The human brain is at the heart of our homeostatic control and the way we react to environmental stress. Part of this reactivity is advantageous in the short-term but it could become maladaptive in the long run (McEwen & Gianaros, 2010). Such a central role of the brain in coping with stress reinforces the idea that physiological correlates of psychological distress are more likely found in the brain (Buck, 1999; Izard, 2007; Panksepp, 2007). Electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) are popular, non-invasive, and well-established methods for analysing the human brain’s signals (generated by the firing of neurons or by cerebral oxygenation, respectively). Of the studies that used EEG and fNIRS, all except one used a simulated environment (or visual stimuli) in a controlled laboratory-based setting. Di Flumeri et al. (2018) took advantage of the portability of the EEG system (and of the mobile eye-tracking) to run a real driving experiment, performed under different traffic conditions (rush hour and off-peak) and along different road types (main and secondary streets).

Psychological stress also primes behaviours that could be assessed to unravel one’s emotional or cognitive state. The facial expression evaluation has shown utility in the assessment of discrete emotional reactions (Ekman, 1993; Russell, 1994), explaining its use in some studies. Changes in gaze patterns of drivers were also investigated to explore how congestion affects selective attentional processing. Acute stress can also disrupt cognitive performance or elicit errors, particularly when environmental demands are perceived as taxing or potentially exceeding one's own capacity (Calvo & Gutiérrez-García, 2016). Some papers considered cognitive tasks either to evaluate driver’s mental workload, frustration levels, or to assess the impact of unpredictability on motivation.

Finally, the neuroendocrine stress response is vital to control the energy metabolism required for health maintenance and, despite several limitations, the use of salivary cortisol as an index of psychological stress is widespread (Hellhammer, Wüst, & Kudielka, 2009). Two studies from the same group asked participants to collect their own saliva samples and, by comparing levels elicited after a daily commute routine against a non-workday at home, they investigated the impact of transport reliability and infrastructure on mental wellbeing.

Relationship between the different transport-related exposures and mental health/wellbeing

Transport infrastructure

Psychological scales

Good public transportation accessibility – as living up to 500 m from a bus stop (Lades, Kelly, & Kelleher, 2020) or within a short walking distance from home/work to the station (especially among bus travellers with long commutes) (Lunke, 2020), contributed to greater satisfaction. Residents around light rail stations (with easy access and good service) also showed greater satisfaction with travel (Cao, 2013). Within public transport stations, walking distance and level of service of the queuing and walking areas (degrees of per capita occupancy of space) can influence passengers’ perception of time, and thus their anxiety, fatigue, and satisfaction (Teng, Pan, & Zhang, 2020). Additionally, waiting time for public transport transfers negatively affected travel satisfaction (Gao, Rasouli, Timmermans, & Wang, 2017; Lunke, 2020; Wener, Evans, Phillips, & Nadler, 2003). An experimental study investigating stress of train commuters before and after a major public transport improvement, reported that reduced trip-time (due to a new direct service) was the main contributor to stress reduction (Wener et al., 2003). Further transport infrastructure-related stressors in (Mexico) public transport were the vehicles’ comfort – and maintenance-related characteristics, as well as the road networks’ characteristics. Here, stress increased with small, old, dirty buses in a bad state, with little legroom, few/uncomfortable seats and the bad state of roads (Dorantes-Argandar, Rivera-Vázquez, & Cárdenas-Espinoza, 2019).

As for driving, the quality of road infrastructure – for example roads with potholes, irregular or unpaved (Dorantes Argandar, Tortosa Gil, & Ferrero Berlanga, 2016), narrow roads and tight corners (Rodrigues, Kaiseler, Aguiar, Cunha, & Barros, 2015) – influenced driving stress. Moreover, people with the highest anger scores were living in neighbourhoods with narrower and more mountainous roads (Hernández-Hernández, Siqueiros-García, Robles-Belmont, & Gershenson, 2019). Intersection design (Samuel et al., 2019) and the presence of (working) lights at intersections (Dorantes Argandar et al., 2016) were also influencers of the emotional states of motorcycle riders and car drivers, respectively. Other identified road-related variables were the presence of road signs (Høye, Fyhri, & Bjørnskau, 2016); adaptive signal control technology, which in turn corresponded with higher speeds and fewer stops (Khattak, Fontaine, & Boateng, 2018); and challenges like roadworks, difficult junctions, and lack of car parking (Lades et al., 2020; Pykett et al., 2020). Interestingly, a road sign with the text “share the road” (with a picture of a smiling cyclist and a car) improved the interaction between car drivers and cyclists (Høye et al., 2016).

It is noteworthy that the effect size of transport infrastructure depends on the scale of the analysis and used measures. For instance, a study in California using built and natural environment data at the block-group level and a large health interview survey did not find a significant relationship between mental wellbeing and job accessibility by public transport, percentage of jobs near public transport stations, street network density, or street intersection density (Schoner et al., 2018).

Regarding active travel and travel satisfaction, relationships with transport infrastructure were also found. An auto-oriented environment was found to decrease commute wellbeing for cyclists (Smith, 2017). A study in Delft and Atlanta found that the top stressors for cyclists were the presence of motor vehicles; poor pavement condition; and poor infrastructure (Gadsby, Hagenzieker, & Watkins, 2021). The most stressful factor was the absence of bicycle facilities; components like sight distance or low number of travel lanes (Gadsby et al., 2021); but difficult junctions (Pykett et al., 2020) were also relevant.

Satisfaction with travel is not only affected by objective trip characteristics, but also by moods (feeling happy or relaxed) – and, for example, walking safety at night was related to travel satisfaction (Gao et al., 2017). Glasgow, Le, Geller, Fan, and Hankey (2019), studying mood during travel with an experience sampling method app, reported lower mood when individuals travelled through places with a higher “Walk Score®” (versus higher when individuals travelled through natural environments). The negative effect of walkability is probably related to the fact that walkability is higher in dense urban environments (with environmental stressors, such as noise and air pollution).

Several intercorrelated neighbourhood satisfaction factors were associated with mental health/wellbeing (Leslie & Cerin, 2008). These include safety and walkability (related positively with infrastructure for walking/traffic safety, but negatively to traffic load); social network (related positively with street connectivity, but negatively with traffic load); and traffic and noise. Older people who are more satisfied with walking and public transport infrastructure and services, experienced higher social inclusion, physical health, and psychological wellness (He et al., 2020). Interestingly, street connectivity was positively correlated with a greater social network, which in turn has a positive association with mental health (Leslie & Cerin, 2008). Chng, White, Abraham, and Skippon (2016) found that individuals with good public transport connectivity reported fewer symptoms of psychological distress.

Transport mobility has been found to be related to subjective wellbeing through the mediating variables of environmental mastery, positive relations with others, and self-acceptance (Vella-Brodrick & Stanley, 2013). This would justify expenditure in improving transport mobility through infrastructure and increased service provision. Feng, Feng, and Astell-Burt (2017) also suggest that investments in walkability and public transport infrastructure are required to promote (mental) health – as people who more frequently use public transport tend to report better mental health/wellbeing. However, frequent public transport users who perceived the public transport infrastructure to be lower in quality reported poorer psychological outcomes.

Physiological measures

The impact on cyclists of bicycle facilities and related road infrastructure was the main target of studies using ANS measures (Caviedes & Figliozzi, 2018; Cobb, Jashami, & Hurwitz, 2021; Doorley et al., 2015; Fitch, Sharpnack, & Handy, 2020; Teixeira et al., 2020; Werner, Resch, & Loidl, 2019).

Increased EDA stress-response in cyclists was related to factors involving motorised vehicles (being too close to the bicycle lane; and interference at intersections), pedestrians (invading the bicycle lane or with unexpected movement that conflicts with cyclists’ trajectories) and bicycle facilities (when lanes were narrow, rough, unpaved, or noisy) (Caviedes & Figliozzi, 2018; Teixeira et al., 2020). Cycling on physically segregated facilities (off-street) was associated with less stress than on shared roadways – except if cyclists were confined to a narrow lane or “sandwiched” between parked and motorised vehicles (Caviedes & Figliozzi, 2018). Another study (Werner et al., 2019) compared EDA response of cyclists as a function of different infrastructures (e.g. dedicated/mixed ways; few/many road crossings); but found no significant associations. Finally, in a more recent study, greater stress levels were observed while riding in the absence of a bicycle lane and with high vehicular flows – inexperienced cyclists being the ones with lower levels of comfort and yielding higher EDA values (Cobb et al., 2021). Notably, the riders showing more stress were also the ones more prone to deviate from the centre and with greater lateral position variations.

Stress-related cardiovascular response of cyclists was also investigated (Doorley et al., 2015; Fitch et al., 2020). In one study, among five road environments tested, only the local road with no centreline and low car speed/volumes consistently provided greater HRV – suggesting that “bicycle boulevard” designs (where traffic is calmed and bicyclists are given travel priority) may reduce stress bicycling environments (Fitch et al., 2020). For roads that provide important connections for vehicles (collectors/arterials), reducing outside traffic lane widths and increasing bike lane widths are likely to reduce bicyclist stress. Furthermore, another fully controlled experiment (Doorley et al., 2015) found a linear trend between cyclists’ HR and busy roads without a bicycle lane or roundabouts where the cyclist continued straight on.

An on-road car driving experiment took place where participants had to drive through two real traffic corridors – with and without adaptive signal control technology (ASCT), while their HR was measured (Khattak et al., 2018). The results showed higher speeds and fewer stops on the ASCT corridor relative to the non-ASCT corridor, accompanied by a significant reduction in driver HR over the ASCT corridor.

In a study with HRV assessment on bus drivers, urban planning issues – narrow roads and tight corners, were also considered stressors (Rodrigues et al., 2015). The only non-cycling experiment with EDA assessed public transport users travelling with several biosensors – and reported that waiting for the bus (likely due to fewer buses) increased stress levels, although examining the EDA readings of only one person (Castro, Guevara, & Jimenez-Molina, 2020).

Finally, one study (Wener et al., 2003) measured train commuters’ salivary cortisol and applied a proofreading task in response to an improvement in public transport infrastructure (a connection to existing tracks and no need to transfer to another train to reach Manhattan), over a 3-years longitudinal study (a caveat being the high drop-out rate). They observed that both salivary cortisol and motivation task scores of users before and after this new infrastructure was built showed strong differences, this being a rare case where physiological data was compared pre- and post- a major public transport intervention.

Congestion and delays

Psychological scales

Associations between congestion/delays and psychological variables related to mental health/wellbeing have been addressed in several studies. They either examined affective and cognitive states that may arise directly in or after the congestion – such as higher workload/stress (Teh, Jamson, Carsten, & Jamson, 2014), anger and frustration (Ihme et al., 2018; Parker, Lajunen, & Summala, 2002); or the relationship between congestion and long-term states, including mental health/wellbeing (Smyth et al., 2011; Wang, Rodríguez, Sarmiento, & Guaje, 2019).

Most studies focused on car drivers. Short-term effects were often measured in driving simulator experiments, where congestion was, for example, used as part of specific scenarios to induce stress, anger, anxiety, or aggression; combined with self-reports to assess if those scenarios successfully induced such desired effects (Zhang et al., 2019). Other studies (Brewer, 2000; Rodrigues et al., 2015) induced stress or frustration to examine its effect on driver performance – either generally or for specific groups of drivers (e.g. bus drivers).

In a simulator study (Fairclough & Spiridon, 2012) frustration was induced by preventing individuals from reaching their destination on time due to congestion at the beginning and end of the route. The results showed a significant increase in self-reported anger and a decrease in perceived control comparing pre- and post-task answers (Fairclough & Spiridon, 2012). Another simulator experiment included constraints in scenarios to generate time pressure during a ride in a highly automated vehicle (Techer et al., 2019) – participants’ mood/emotional state deteriorated when facing congested conditions, regardless of whether they were under time pressure. Similarly, more results support that after simulated driving in slow traffic, participants express increased anger and stress (Emo, Matthews, & Funke, 2016). Although they found that anger induced by congestion did not predict risky behaviours, other simulation experiments reported that road congestion around railway level crossings increased drivers’ frustration, making them more prone to risky behaviour (Larue, Blackman, & Freeman, 2020). Moreover, satisfaction with travel was also affected by traffic conditions as heavy traffic was associated with a less positive evaluation (Ettema, Gärling, Olsson, Friman, & Moerdijk, 2013; Lades et al., 2020).

Teh et al. (2014) examined the effects of traffic density on self-reported workload and found that driver workload increased with traffic flow up to moderate levels and levelled off beyond that. Furthermore, participants scored high for frustration when the slow lane had vehicles moving faster and less congested than the fast lane. Another study found that drivers’ stress increased (and satisfaction with the chosen route decreased) as traffic complexity was greater (Ringhand & Vollrath, 2019).

Another study focusing on the interaction between traffic situations (including congestion) and emotions, presented respondents with “realistic and not frequent” descriptions of hypothetical traffic scenarios – and participants reported higher levels of anger and anxiety for higher traffic density and time pressure (Roidl, Frehse, Oehl, & Höger, 2013). Experienced drivers reported both less anger and anxiety, while those who reported more intense anger and anxiety did not present it more frequently throughout the experiment. Furthermore, it was found that high traffic density was likely to lead to more impatient driving, triggering anger in other drivers (Parker et al., 2002). Also, while older drivers were less reactive to other drivers, males reacted significantly more to most anger-provoking situations than women.

Deffenbacher (2008) did not find a significant relation between anger and its expression in urban (ordinary and rush-hour traffic) versus rural (open country roads) areas, which was expected due to higher levels of congestion in urban areas. Men driving in those reported less driving anger than other groups; yet they reported more aggressive and risky behaviour than women. In a rural community sample, driving in heavy traffic was among the situations generating anxiety the most; consequently, it was highly avoided by drivers (Taylor & Paki, 2008). Stress and higher mental workload (as an outcome of congestion) have more extensively been examined for professional drivers, especially bus drivers. Brewer (2000) investigated road anger motivations and activation contexts by comparing bus drivers with non-bus drivers. It was found that bus drivers felt less impatient and irritated and overtook less frequently than non-bus drivers during peak-hour, but the latter felt overall less tense than bus drivers. Also, although both groups experienced similar levels of driving anger, bus drivers were less likely to engage in aggressive behaviours; indeed, they tended to feel more irritated about passengers than about other road drivers. Among non-bus drivers, women tended to report more impatience during peak-hour compared to men (Brewer, 2000). One study examining how metro drivers perceived workload (Huang, Li, Liu, Zhu, & Zhu, 2019), found that drivers’ workload was mainly affected by the complexity of the driving environment (associated with peak passenger flow) and long-term accumulation of operations.

Another study (Rodrigues et al., 2015), with both self-reported and physiological assessment of stress in public transport bus drivers, indicated that 41% reported stress due to events impacting time schedule (e.g. congestion) – with the city centre (dense urban area) being marked as more stressful. While congestion was found to impact the quality of life and mental health in general (resulting in increased stress, tension, and anxiety for those living in congested areas), this effect was particularly stronger for public transport drivers and their families (Nadrian et al., 2020). Moreover, the frequency of job-related problems – including high-density road traffic, was also associated with public transport operators’ burnout (Cunradi, Chen, & Lipton, 2009).

Considering drivers in general, Morris and Hirsch (2016) measured the intensity of happiness, sadness, fatigue, and stress of 14,000 U.S. drivers. Instead of observing congestion during trips, they relied on indicators (time of the travel, population of the city and city centre vs. suburban residential location). They found that drivers had a slightly less positive mood when driving in the largest cities between 5:00pm-6:00pm (shorter peak hour), but no significant effect was found when using any other definition of peak period (Morris & Hirsch, 2016). While their findings reveal limited effects of congestion on emotions, they show the difficulty in assessing emotions and level of congestion at the same time based only on classic self-report measures.

Another set of studies focused on the persistent effects that individuals experience due to congestion. Examining the influence of urban environment on wellbeing, it was found that living in cities with high congestion was associated with lower mental wellbeing (Smyth et al., 2011). Living in a more congested city was associated with higher levels of anger (Li, Yao, Jiang, & Li, 2014) and related to lower life satisfaction among bus/coach commuters, despite lower mental distress for bus/coach commuters (Chng et al., 2016). Road traffic congestion also significantly decreased commute wellbeing for cars and public transport (Smith, 2017). Moreover, travelling in congested public transport vehicles (crowded) also impacted negatively on commuting wellbeing (Smith, 2017). Survey data from eleven Latin American cities (Wang et al., 2019) found commuting time during congested conditions positively related to the probability of being diagnosed with depression – with public transport users having a lower likelihood than drivers.

With the aim of automatically detecting frustration, Ihme et al. (2018) conducted a simulation experiment to trigger frustration (goal-blocking and time pressure) with red lights, construction sites, slow lead cars and pedestrians slowly crossing the street. Frustrating drives were rated as more negative and more arousing than the non-frustrating drives. The work by van Hooff (2015) found that the occurrence of a stressful delay during the work-home commute had a negative effect on employee recovery status after returning home (but not at the end of the evening) seeming to be relatively short-lived.

Friman (2010) examined affective reactions to waiting for public transport (namely bus) and their impact on overall service satisfaction. Based on text descriptions of waiting scenarios, positive and negative waiting experiences produced consistently more affective reactions and changes in the level of satisfaction than neutral experiences, while in-process waiting (during the service) seemed to be more acceptable than pre-process waiting (before service started).

Lunke (2020) investigated how different public transport journey characteristics affected commute satisfaction, including distance to stations, need to transfer, and associated waiting time. Waiting time and reliability were more important than short distances and direct routes, although their effects were stronger on long commutes. This suggests that providing more direct public transport routes, efficient transfers and improving public transport can improve satisfaction and make public transport the preferred transport mode for commuters. However, potential confounding variables (e.g. congestion, seat availability, and commuters’ attitudes toward public transport modes) were not considered in the study.

He et al. (2020) considered walking, rail, bus, and minibus travels to study the correlation between satisfaction with transport systems (impacted by congestion/delays) and social inclusion and wellbeing in the elderly. Although causality was not implied, it proved to be (positively) statistically significant.

Teng et al. (2020) focused on mobility within metro stations, under the hypothesis that passengers’ psychological state could be affected by space congestion, slow walking speed, and increasing queue length. They observed that an enhanced time-lapse effect reflected an increase in anxiety and decrease in service satisfaction (also observed when the queuing time or walking distance of a station exceeded a certain value).

All in all, most reviewed studies provide evidence that congestion and delays are related to negative affective states and mental health problems, the latter especially for professional drivers. They are also a source of road anger and frustration, so that they are not only likely to impact mental health/wellbeing, but also traffic safety (Wickens, Mann, Ialomiteanu, & Stoduto, 2016).

Physiological measures

The impact of congestion and delays on mental health/wellbeing have been explored mostly by ANS (cardiac or skin conductance responses) and EEG measures; and predominantly using car simulated driving – with only a few performing real on-road experiments.

One experiment combined three simulated driving scenarios and a similar on-road experiment to study the effect of delay (due to frequent red lights), traffic congestion and surrounding vehicle interference (Yan, Wan, Qin, & Zhu, 2018). Drivers’ beta-band on EEG showed a significant increase when they felt angry in all conditions (likely due to arousal); while delta-band activity (likely reflecting relaxation) reduced when the driver was angry on-road (but not in the simulation). Drivers’ cardiac response showed significant volatility in the three induced delay scenarios, being more sensitive (together with EDA) after red light encountering. Overall, drivers’ stress-related cardiac response increased in reaction to adverse traffic events, being greater on-road than in the simulation. Another EEG study exploring post-congestion effects also observed changes in beta-band in brain areas associated with the perception of driving speed (Li et al., 2020) – likely related to a more aggressive attitude in the post-congestion situation and decreased perception of speed.

Di Flumeri et al. (2018), using another type of EEG measure (usually associated with mental workload) reported a significant brain activation under real (on-road) settings in rush hours when compared to normal traffic (Di Flumeri et al., 2018). These findings support the idea that congestion leads to higher mental workload (eye tracking and self-reported measures were not as sensitive as EEG). Left-hemispheric EEG activation – often associated with anger and worry, was found during a simulated traffic jam (including a delay) scenario (Fairclough & Spiridon, 2012). In the same condition, both HR and facial muscle activity were significantly elevated, consistent with a negative emotional and stress-related response. Halim and Rehan (2020) used the spectral power of differential hemispheric asymmetry across five frequency bands, to distinguish brain dynamics as a response to stress stimuli. A drop in alpha rhythm power and initial spike in beta rhythm with driving speed was observed, indicating that subjects had to be more attentive to complete the driving task under time pressure (but with an added level of stress they lost attention).

In an experiment evaluating fNIRS response to frustrating (road blocking) and non-frustrating driving simulator scenarios, brain areas with higher activity in the former scenario were in line with the other neuroimaging literature on frustration; and related with cognitive appraisal, impulse control, and emotion regulation processing (Ihme et al., 2018).

Studies more focused on the cardiovascular responses of drivers facing urban traffic congestion contexts or frequent lane changing (Cardone et al., 2020; Qi, Wen, & Wu, 2015; Qi, Wen, Wu, & Qin, 2017) revealed changes in HR, HRV or electrocardiography (ECG) stress levels, reflecting a rapid sympathetic activation as observed in other stressful situations. The results on questionnaires about drivers’ feelings, recorded thermal features and respiration patterns as a function of congestion, further corroborated the cardiac findings, robustly indicating high-stress levels. Similar evidence (with EDA, respiration/ventilation, HR and HRV) was gathered by another study (Singh, Conjeti, & Banerjee, 2013) investigating busy driving on-road situations (including traffic congestion).

In contrast to other environmental variables such as weather, Rigas, Goletsis, Bougia, and Fotiadis (2011) indicate that heavy traffic did not have a significant contribution to stress classification based on ECG, EDA, or respiration patterns. However, the limited recordings and small sample size of this work compromise their conclusion. In fact, another study with limited sample size and confounding variables found the opposite – traffic jams significantly affected the stress level of drivers as measured by EDA (Bitkina, Kim, Park, Park, & Kim, 2019). Moreover, using an on-road driving experiment where EDA was considered in addition to cardiac activity, Wang and Guo (2021) concluded that foot EDA was able to predict stress in busy roads (congestion). Similar conclusions could be drafted from Dogan, Bogosyan, and Acarman (2019), where EDA stress when driving an electric car was consistent with the ones identified in a post-driving survey.

Studies have also focused on the relationship between congestion/delays and mental wellbeing in public transport (Castro et al., 2020; Huang et al., 2019; Rodrigues et al., 2015). Contrary to previous results, a thorough analysis of HRV in bus drivers (Rodrigues et al., 2015) was not conclusive in showing a statistically different distribution for tagged stressful events, such as traffic congestion (presumably due to higher signal noise in real-world). In a study evaluating metro drivers (Huang et al., 2019), both HRV and reaction time were collected for two driving modes (manual/automatic) and three daily shifts. Although the physiological indicators and performance had a positive correlation with fatigue, the same was not obvious in reaction times. Whilst the arousal of biological rhythms in the morning relieved fatigue, the large amount of information processing and accurate judgement required during peak passenger flow and associated time pressure led to increased workload and fatigue. The effect of delays was also studied for bus and metro commuters (Castro et al., 2020) with EDA and HRV. The authors observed that waiting for the bus increased the probability of feeling sad, with EDA showing a significant association with self-reported emotions.

Finally, two studies have focused on cycling using EDA (Caviedes & Figliozzi, 2018; Werner et al., 2019). In one study, EDA captured increased stress responses during peak hours and when the bike route was blocked (by cars or people) (Caviedes & Figliozzi, 2018). However, the latter experiment found no significant correlation between stress and traffic load (Werner et al., 2019) – possible reasons for this include cognitive biases, memory errors and/or small sample size.

Reliability

Psychological scales

We now focus on reliability – i.e. the probability of service failure from the user perspective (as opposed to the operator’s). Depending on the transport mode, different variables that concern travellers can be considered: travel/queueing time in personal transportation or waiting time, boarding and alighting time, seat availability and public transport transfer time. Looking at the distribution of such variables helps in estimating operational reliability, which is often a factor of interest to satisfaction/wellbeing/stress scales since its perception can lead to behavioural change or impact mental health/wellbeing. Associations between reliability and psychological scales have been considered in 10 studies. We can classify these into three reliability measures: direct, combined and proxy measurements.

Few studies directly measured reliability, but surprisingly show strong evidence on the link between predictability and stress – both in commuting settings (Evans, Wener, & Phillips, 2002; Sposato, Röderer, & Cervinka, 2012; Wener et al., 2003) and for professional drivers (Rodrigues et al., 2015). Evans et al. (2002) assessed predictability of travel time, perceived stress and salivary cortisol to investigate the role of predictability in commuter rail. All stress-related assessments were related to more unpredictable commutes. Their small sample, limited collection of control variables and absence of an objective reliability measure led the authors to classify its findings as preliminary. Wener et al. (2003) studied new rail commuting options with an ecological experiment. They found that riders on a selected new rail line had lower stress levels, mediated by reduced travel time. Yet, the predictability of the trip was also inversely correlated with stress regardless of changing to the new or staying with the old commuting option. Later, Sposato et al. (2012) extended this to other modes, and through an online survey investigated the relative impact of predictability on commuting stress. They collected self-reported measures and proposed a conceptual model of commuting stress. Similarly to Wener et al. (2003), the authors sustained that predictability is a measure reflecting temporary incidents, which forces the commuter to adapt; rather than being a constant feature in the commuting journey (such as control). Yet, their findings also point to an interactive (negative) effect of predictability and control on commuting stress. Rodrigues et al. (2015) focused on stress responses of a small sample of bus drivers to self-reported and self-labelled events, including deviations to time schedules. Despite not including direct reliability measures, nor relying on a causal model (rather on statistical tests), the work pointed to the negative impact of unreliability sources on stress, along with a unique showcasing of longitudinal (detailed spatial and temporal) data collection.

Another thread of research (De Vos, Schwanen, Van Acker, & Witlox, 2019; He et al., 2020; Vella-Brodrick & Stanley, 2013) combined reliability measures with other travel attributes and focused on its relationship with subjective wellbeing (rather than trip-specific stress). Vella-Brodrick and Stanley (2013) relied on an Australian survey to validate the hypothesis that psychological measurements mediate transport mobility and subjective wellbeing. Perceived reliability was an exposure considered for estimating transport mobility. Yet, the combination of reliability with other exposures into a single variable in the analysis prevented the identification of isolated effects of potential triggers. He et al. (2020) specifically considered a satisfaction score for the service reliability of different modes of transport (rail, bus, and minibus) and for the reliability of the bus information provision system. Such measures were combined into a single (average) measurement of transport satisfaction which was then integrated into a Structural Equation model to estimate the effect of satisfaction with the most commonly used transport systems on the wellbeing of the ageing population. They found that an intermediate effect of social aspects should be considered when modelling the relationship between transport and mental wellbeing. De Vos et al. (2019) brought the dimension of individual decision-making by testing (and confirming) the hypothesis that a positive evaluation of walking and cycling trips (including reliability aspects) positively affects the attitude towards the respective mode (which then positively impact on choosing that mode in leisure trips).

The third thread of reviewed research (Fan, Chen, & Zhang, 2021; Larue et al., 2020; Lunke, 2020) relied on proxy measures of reliability. Larue et al. (2020) focused on congestion at railway level crossings and its impacts on drivers’ mood and potentially (risky) behaviour. Insights about how reliability decreases frustration levels were obtained (despite a small sample size) – once drivers got used to extended waiting times (low variability) at a crossing, they were more likely to violate the crossing rules; and reducing waiting times did not result in an immediate return to baseline behaviour. Lunke (2020) measured (Oslo) public transport commuters’ satisfaction through a cross-sectional Stated Preference survey. Direct reliability measures were not considered in the survey since the user's last trips were the object of the survey. Nevertheless, through ANOVA, perceived service frequency seemed to have no significant effect on commuter satisfaction. Fan et al. (2021) investigated travel happiness for different modes through a web-based survey. Perceived trip duration together with arrival time flexibility was collected considering a recent trip performed by the respondents, but the latter was not significant in the quantitative (regression) analysis.

Physiological measures

We only found two studies that directly measured reliability; and another study (Wener et al., 2003) that considered aspects of reliability together with some infrastructure changes. In the first study, greater stress levels measured by HRV monitoring were reported by bus drivers facing unpredictable situations related to schedule changes (Rodrigues et al., 2015). In the other one, both salivary cortisol and cognitive effort (with a proofreading task) of train commuters were measured. It was found that rail commuters who perceive their daily commutes as more unpredictable experienced greater stress (elevations in salivary cortisol), while the motivational index (provided by the proofreading task) was not significantly related to commute unpredictability (Evans et al., 2002).

Discussion

Summary of evidence

How individuals use transportation systems is critically dependent on its attributes and, while their influence in shaping travel behaviour has been a central focus for research, their consequent impact on mental health/wellbeing has received less attention. Our review aimed to fill this gap and to provide insights for the future role of transport attributes in mental health/wellbeing. To help translating the evidence into their use for policy-making and research, we compiled in Figure 3 a list of key transport-related indicators. Although not exhaustive, it captures the most relevant findings on how transport infrastructure and operational performance relate positively or negatively (or if no association/correlation was found) with mental health/wellbeing.

Figure 3. Key indicators with an impact on mental wellbeing/health.

Reviewed studies on transport infrastructures that used psychological scales focused either on stress/anxiety or levels of satisfaction with travel; whereas those reporting physiological measures mainly targeted cyclists. Our summary addressed distinctively the different travel modes. Regarding public transport, shorter walking distances to stops, greater connectivity, shorter waiting time in transfers, increased level of service of the queuing and vehicles’ comfort- or maintenance-related characteristics contribute to higher levels of self-reported mental wellbeing. Both psychological and physiological stress-related effects were also observed for bus drivers in narrow roads and tight corners; and for passengers waiting for the bus. Regarding car drivers, relevant self-reported factors include the quality/maintenance of road infrastructure, width of the roads and availability of parking; whereas, higher speeds and fewer stops induced stress-related biosignals. Notably, the presence of road signs or ASCT showed a positive impact on car drivers in both psychological and physiological assessments. Regarding active travel, several transport infrastructures variables were found to be perceived as stressful – including the presence of motor vehicles, poor pavement condition, absence of bicycle facilities and difficult junctions. Both psychological and physiological cyclists’ stress were related to other traffic being too close or on the bicycle lane, shared (rather than segregated) roadways and narrow, rough, unpaved or noisy bicycle facilities.

The evidence on congestion and delays robustly showed a negative impact across different types of mental health/wellbeing assessments. Such effects were most consistently observed for indicators such as high traffic density and slow traffic speed, which are also expected to happen in urban areas, increased population density and in peak hours. Interestingly, studies also reported stress-induced biological signals corroborating the negative effects of perceived travel duration, waiting time and time pressure.

Fewer studies addressed reliability and mental health/wellbeing, but some indicators related to stress and travel satisfaction in public transport were encountered. Predictability results in lower self-reported stress in commuters and in professional drivers. Stronger stress-related biosignaling was also observed in bus drivers facing unpredictable situations related to schedule changes; and for rail commuters who perceive their daily commutes as more unpredictable. Finally, a positive evaluation of reliability aspects on walking/cycling trips also positively affects the attitude towards these modes.

Limitations

Despite the structured search strategy on multiple databases and empirical evidence targeting, our study has limitations. When discussing transport-related indicators, no weighted contribution as a function of study design or effect sizes was considered – although a quality assessment and individual limitations are described. Furthermore, by applying strict criteria for self-reported scales – focusing on outcomes closely related to psychological distress, we potentially excluded relevant topics (e.g. travel satisfaction without a valid psychological measure; or safety-focused road anger).

Additionally, the reviewed studies also have limitations. While analysing the data we have identified gender (more males) and age (preference for the young) imbalance; and few studies focused on clinical populations or had a longitudinal design (limiting causality inference). Adapted versions or sub-scales of psychological instruments were occasionally used with questionable reliability. Regarding the use of physiological measurements, its associated data collection effort often led to studies with small sample sizes. Despite the advantages in manipulating variables and eliciting scenarios with driving simulators (some scenarios not possible in real-life), such controlled environments are oversimplified and also have power/sample size issues.

Finally, the impact of potential confounding variables (e.g. noise exposure, social interactions, participants’ attitude towards transport services, personality traits, and other context variables) was frequently disregarded.

Policy and research recommendations

Recent policies have started to link mobility-related solutions and mental health/wellbeing by relying on actions for increasing social inclusion, the share of active mobility and green areas, and decreasing congestion (Cooper et al., 2004; Dora, Hosking, Mudu, & Fletcher, 2011; WHO, 2021; Wild et al., 2021). Our review shows that the mobility policy agenda can now include further dimensions on action for improved mental wellbeing (Figure 3). These indicators are intrinsically linked with individual cognitive, physical, time and uncertainty challenges and have proven to impact mental wellbeing. Despite such evidence, little is known on their quantified benefits in terms of complementary treatment/prevention. This is mainly due to (mental) public health impact assessments requirements during mobility planning and transport operations; pre–post studies that can effectively quantify the benefits of interventions; and linked quantitative measures throughout the conceptual chain presented in Figure 1. Thus, the introduction of key performance indicators as drivers for policy-making is still not possible. Yet, the inclusion of valid psychological measures with robust evidence (Appendix 3) can potentially be introduced in practice for service assessment (e.g. at the level of contract or regulation for public transportation operators).

After reviewing the empirical evidence, we have also identified some needs for future research. New studies should explore built environment characteristics of travel routes and how these are related to mental health/wellbeing. Further work is also needed for better combining the different methodologies to study the relationship between congestion, emotions, and wellbeing – as inconsistencies between reported outcomes still exist due to heterogeneous selection of exposure- and instrument-related measurements. Transport reliability is surprisingly neglected, despite old evidence found in studies related to the impacts of public transport reliability. Additionally, few studies combined both psychological and physiological measures, despite the increased availability of mobility sensing and biosensing technologies. Also, we focused our review on infrastructure and operational attributes which are critical in travel behaviour and decision-making. Such decisions ultimately will influence mental states through subsequent changes in an individual's activity and travel patterns. Understanding such processes is a promising and yet-to-be-explored research path.

Finally, the critical link in our proposed framework (Figure 1) relies on the evaluation of mental health/wellbeing outcomes. The most challenging and needed research stream will be on the systematic quantification of potential gains and losses with interventions within transport attributes and mobility decision-making.

Disclosure statement

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

Additional information

Funding

MAC, MM, SH, CLA and BM are supported by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No 945307 (eMOTIONAL Cities).

Appendices

Appendix 1. Search terms

Table

Exposure				Instrument							Outcome
Transport Infrastructure Operational Performance (Congestion, delays, unreliability)				Physiological measures				Psychological scales			Mental health / Wellbeing
Title / Abstract / Keywords	AND	Full-text	AND	Title / Abstract / Keywords	AND	Full-text	OR	Title / Abstract / Keywords	Full-text	AND	Full-text
travel* OR		infrastructure OR		cortisol OR		cortisol OR “galvanic skin response” OR			“Beck Depression Inventory-II” OR “BDI-II” OR		irritat* OR
transport* OR		access* OR		skin OR		“skin conduct*” OR			“Center for Epidemiologic Studies Depression Scale” OR “CES-D” OR		irritab* OR
mobility		transit OR		“electrodermal activity” OR		temperature OR			“Geriatric Depression Scale” OR “GDS” OR “Hospital Anxiety and Depression Scale” OR “HADS” OR “Patient Health Questionnaire-9” OR “Patient Health Questionnaire” OR “PHQ-9” OR “PHQ-2” OR “PHQ” OR “EQ-5D” OR “Hamilton Depression Rating Scale” OR “HAM-D” OR “HDRS” OR “HRSD” OR “Montgomery-Åsberg Depression Rating Scale” OR “MADRS” OR “Social Problem-Solving Inventory-Revised” OR		anger OR
		car OR		sweat* OR		perspir* OR sweat* OR			“SPSI-RTM” OR “Beck Hopelessness Scale” OR “BHS” OR		angry OR
		bus OR		heart OR		ppg OR “heart rate” OR			“Quick Inventory of Depressive Symptomatology-Self-Report” OR “QIDS-SR” OR “Reminiscence Functions Scale” OR “RFS” OR		psych* distress OR
		cycl*		pulse OR		oxym* OR oxygen OR			“Short Form Health Survey” OR “SF-36” OR “Social Adjustment Scale-Self Report” OR “SAS-SR” OR “Social Functioning Questionnaire” OR “SFQ” OR “Life Satisfaction Index” OR “Life Satisfaction Ratings” OR “LSR” OR		psych* OR
				“blood pressure” OR		satur* OR heart OR			“General Health Questionnaire” OR “GHQ” OR “Mini-International Neuropsychiatric Interview” OR “MINI” OR “Structured Clinical Interview for DSM” OR “SCID” OR		depress* OR
				EEG OR		“blood pressure” OR			“Generalised Anxiety Disorder measure” OR “GAD-7” OR		stress* OR
				autonomic OR		pulse OR “heart-rate” OR			“21-item Depression, Anxiety, Stress Scale” OR “DASS-21” OR “State-Trait Anxiety Inventory” OR “STAI” OR “Life Satisfaction questions” OR “LS Questions” OR “Life Satisfaction Questionnaire” OR “LiSat-9” OR “Satisfaction With Life Scale” OR “SWLS” OR “Pearlin Mastery Scale” OR “PM Scale” OR “Personal Well-Being Index” OR “PWI” OR “Scales of Psychological Well-Being” OR “RPWB” OR “PWB”		distress* OR
				autonomous OR		“HRV” OR ECG OR					anxi* OR
				sympathetic OR		electrocardio*OR catecholamine*OR					tense OR
				respirat*		electroencephalo* OR EEG OR neuroendocrine OR hormon* OR autonomic OR autonomous OR					tension OR
						adren* OR noradren* OR					pressure OR
						Hypothal* OR					mood OR
						sympathetic OR parasympathetic OR					mental ill OR
						epinephrine OR					mental illness OR
						norepinephrine OR					mental sick* OR
						fight-or-flight OR breath*					mental disab* OR
											lonel* OR
											sad OR
											sadness
OR				OR				OR
travel* OR	AND	congestion*OR		vital OR			self-reported OR
transport* OR		delay OR		sensor* OR			questionnaire OR
commut* OR		uncertaint* OR		sensing OR			survey OR
mobility OR		reliab* OR		monitor* OR			scale OR
traffic OR		disrup* OR		wearable* OR			index OR
driv*		real-time information		smartphone OR			inventory OR
				device* OR			psychometric
				biomarker OR
				“biological marker” OR
				Measurement OR
				Biometric OR physiologic*

Appendix 2. Quality assessment

Table

Study	Score	Quality
Bitkina et al. (2019)	1	High
Brewer (2000)	0.91	High
Cao (2013)	0.91	High
Cardone et al. (2020)	0.77	Medium
Castro et al. (2020)	0.93	High
Caviedes and Figliozzi (2018)	0.88	High
Chng et al. (2016)	1.0	High
Cobb et al. (2021)	0.86	High
Cunradi et al. (2009)	0.95	High
Deffenbacher (2008)	0.91	High
De Vos et al. (2019)	1	High
Di Flumeri et al. (2018)	0.82	Medium
Dogan et al. (2019)	0.86	High
Doorley et al. (2015)	0.91	High
Dorantes-Argandar et al. (2019)	0.82	Medium
Dorantes Argandar et al. (2016)	0.95	High
Emo et al. (2016)	0.79	Medium
Ettema et al. (2013)	0.95	High
Evans et al. (2002)	0.86	High
Fairclough and Spiridon (2012)	0.86	High
Fan et al. (2021)	0.86	High
Feng et al. (2017)	1.0	High
Fitch et al. (2020)	0.86	High
Friman (2010)	0.95	High
Gadsby et al. (2021)	0.82	Medium
Gao et al. (2017)	0.95	High
Glasgow et al. (2019)	0.95	High
Halim and Rehan (2020)	0.91	High
He et al. (2020)	0.95	High
Hernández-Hernández et al. (2019)	0.95	High
Høye et al. (2016)	1.0	High
Huang et al. (2019)	0.77	Medium
Ihme et al. (2018)	0.86	High
Khattak et al. (2018)	0.86	High
Lades et al. (2020)	0.91	High
Larue et al. (2020)	0.75	Medium
Leslie and Cerin (2008)	1.0	High
Li et al. (2014)	1.0	High
Li et al. (2020)	0.86	High
Lunke (2020)	0.86	High
Morris and Hirsch (2016)	0.91	High
Nadrian et al. (2020)	0.55	Low
Parker et al. (2002)	0.95	High
Pykett et al. (2020)	0.95	High
Qi et al. (2015)	0.73	Medium
Qi et al. (2017)	0.73	Medium
Rigas et al. (2011)	1	High
Ringhand and Vollrath (2019)	0.91	High
Rodrigues et al. (2015)	0.95	High
Roidl et al. (2013)	0.95	High
Samuel et al. (2019)	0.91	High
Schoner et al. (2018)	0.95	High
Singh et al. (2013)	0.73	Medium
Smith (2017)	1.0	High
Smyth et al. (2011)	1.0	High
Sposato et al. (2012)	1.0	High
Taylor and Paki (2008)	0.82	Medium
Techer et al. (2019)	0.71	Medium
Teh et al. (2014)	0.95	High
Teixeira et al. (2020)	0.95	High
Teng et al. (2020)	0.86	High
van Hooff (2015)	1.0	High
Vella-Brodrick and Stanley (2013)	0.95	High
Wang and Guo (2021)	0.94	High
Wang et al. (2019)	1.0	High
Wener et al. (2003)	0.91	High
Werner et al. (2019)	0.82	Medium
Yan et al. (2018)	0.95	High
Zhang et al. (2019)	0.86	High

Notes: Rating considering the Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields (Kmet et al., 2004). The outcome score for each study is reported as being either high (0.85–1), medium (0.70–0.84) or low (<0.70).

Each summary score is computed through a quality checklist tool that considers 14 criteria (each scored between 0 and 2, wherever possible), namely:

1. Question or objective sufficiently described?

2. Design evident and appropriate to answer study question?

3. Method of subject selection (and comparison group selection, if applicable) or source of information/input variables (e.g. for decision analysis) is described and appropriate.

4. Subject (and comparison group, if applicable) characteristics or input variables/information (e.g. for decision analyses) sufficiently described?

5. If random allocation to treatment group was possible, is it described?

6. If interventional and blinding of investigators to intervention was possible, is it reported?

7. If interventional and blinding of subjects to intervention was possible, is it reported?

8. Outcome and (if applicable) exposure measure(s) well defined and robust to measurement / misclassification bias? Means of assessment reported?

9. Sample size appropriate?

10. Analysis described and appropriate?

11. Some estimate of variance (e.g. confidence intervals, standard errors) is reported for the main results/outcomes (i.e. those directly addressing the study question/ objective upon which the conclusions are based)?

12. Controlled for confounding?

13. Results reported in sufficient detail?

14. Do the results support the conclusions?

Appendix 3. Summary table of psychological scales used in reviewed studies

Table

		Studies that included the psychometric scales (or some adaptations)
Instrument	Description	Congestion/ Delays	Reliability	Transport Infrastructure
Anger and driving anger
STAXI 2 – State-Trait Anger Expression Inventory (Spielberger et al., 1999)	44-item instrument to measure facets of anger as emotional states and personality traits, 2 subscales: S-Anger; T-Anger	Fairclough and Spiridon (2012)
DAS – Driving Anger Scale (Deffenbacher et al., 1994)	33-item driving anger scale (14-item short form) with 6 subscales involving anger triggers: hostile gestures, illegal driving, police presence, slow driving, discourtesy, and traffic obstructions	Brewer (2000); Deffenbacher (2008); Li et al. (2014); Ringhand and Vollrath (2019); Roidl et al. (2013)		Hernández-Hernández et al. (2019); Høye et al. (2016)
DAX – Driving anger Inventory (Deffenbacher et al., 2002)	49-item driving anger expression scale; 4 subscales: verbal aggressive expression (VAE), personal physical aggressive expression (PPAE), use of the vehicle to express anger (UoV); adaptive/constructive expression (A/C)	Deffenbacher (2008)		Hernández-Hernández et al. (2019); Høye et al. (2016)
DAQ – Driver Anger Questionnaire (Parker et al., 2002)	22-item instrument on driving anger with 5 types of anger-provoking situations: Progress impeded; Reckless driving; Direct hostility; Inconsiderate driving; Impatient driving	Parker et al. (2002)
DDDI – Dula Dangerous Driving Index (Dula & Ballard, 2003)	31-item instrument measuring 3 types of driving behaviour: aggressive driving (7 items); negative emotional driving (9 items); and risky driving (12 items)			Hernández-Hernández et al. (2019)
Anxiety and driving anxiety/avoidance
STAI – State-Trait Anxiety Inventory (Spielberger et al., 1999)	40-item instrument to measure facets of anxiety as emotional states and personality traits, 2 subscales: S-Anxiety; T – Anxiety	Roidl et al. (2013); Taylor and Paki (2008)
DSQ – Driving Situations Questionnaire (modified) (Ehlers et al., 1994; Taylor & Deane, 2000)	211-item instrument that measures the degree of anxiety and avoidance in response to a range of driving situations	Taylor and Paki (2008)
DCQ – Driving Cognitions Questionnaire (Ehlers et al., 2007)	20-item scale to measure the frequency of 3 negative cognitions in driving fear: panic-related, accident-related, and social concerns	Taylor and Paki (2008)
General affect, mood, emotional state
SAM – Self-Assessment Manikin (SAM) (Bradley & Lang, 1994)	Non-verbal pictorial assessment technique to measure pleasure, arousal, and dominance associated with a person’s affective reaction to various stimuli.	Ihme et al. (2018); Zhang et al. (2019)
PANAS – Positive and Negative Affect Schedule (Watson et al., 1988)	20-item scale measuring general tendencies to experience positive affect (PA, e.g. active, enthusiastic) and negative affect (NA, e.g. distressed, nervous)	Zhang et al. (2019)	Vella-Brodrick and Stanley (2013)	Vella-Brodrick and Stanley (2013)
PANAS-X (Watson & Clark, 1994)	60-item, expanded version of the PANAS additionally measuring 11 specific affects: fear, sadness, guilt, hostility, shyness, fatigue, surprise, joviality, self-Assurance, attentiveness, serenity	van Hooff (2015)
General Mood Scale (Quinn & Shepard, 1974)	8-item scale measuring general mood (downhearted through – optimistic) and associated feelings	Brewer (2000)
UWIST Mood Adjective Checklist (Matthews et al., 1990)	Measurement of different dimensions of mood by 28 mood-related adjectives related to 3 factors: energetic arousal (EA), tense arousal (TA), hedonic tone (HT)	Emo et al. (2016)
BMIS – Brief mood introspection scale (Mayer & Gaschke, 2001)	16-item scale with adjectives referring to anger, happiness, sadness, or serenity	Techer et al. (2019)
SCAS – Swedish Core Affect Scale (Västfjäll et al., 2002)	12-item scale with adjective pairs related to 4 factors: Unpleasant activation (UA), pleasant activation (PA), unpleasant deactivation (UD), and pleasant deactivation (PD)	Friman (2010)
TMS – Travel mood scale (Glasgow et al., 2018)	9-item instrument to measure travel mood			Glasgow et al. (2019)
Driver Behaviour Inventory (Glendon et al., 1993)	Instrument to measure the frequency of different feelings/emotions while driving, such as: (a) driving aggression; (b) driving alertness; (c) dislike of driving; (d) general driver stress; (e) irritation when overtaken; (f) frustration in overtaking; and (g) feelings about driving	Brewer (2000)
DRM – Day Reconstruction Method (Kahneman et al., 2004)	Measure to assess how people spend their time and how they experience their activities combining time-budget measurement and experience sampling	Morris and Hirsch (2016) (adaption of the method)
General wellbeing, satisfaction with life
PWI – Personal Wellbeing Index (International Wellbeing Group, 2006)	Generic measure of subjective wellbeing based on satisfaction ratings for 7 life domains: living, personal health, achievement in life, personal relationships, personal safety, community-connectedness and future security	Smyth et al. (2011)
SPWB – Scales of Psychological Well-Being (Ryff, 1989)	42 item scale of wellbeing divided in 6 sub-scales: autonomy, environmental mastery, personal growth, positive relations with others, purpose in life and self-acceptance		Vella-Brodrick and Stanley (2013)	Vella-Brodrick and Stanley (2013)
SWLS – Satisfaction with life scale (Diener et al., 1985)	5-item global satisfaction with life scale		De Vos et al. (2019); Vella-Brodrick and Stanley (2013)	Cao (2013); Vella-Brodrick and Stanley (2013)
WHOQOL BREF – WHO Quality of Life Survey (Skevington et al., 2004)	26-item instrument covering four domains (physical health, psychological, social relations, environment) with 24 facets (one item per facet)	He et al. (2020)	He et al. (2020)	He et al. (2020); Pykett et al. (2020)
WEMWBS – Warwick Edinburgh Mental Well Being Scale (Tennant et al., 2007)	14-item wellbeing scale covering hedonic and eudaimonic aspects of mental health, positive affect, satisfying interpersonal relationships and positive functioning			Feng et al. (2017)
Travel and commuter satisfaction
STS – Satisfaction with Travel Scale (Ettema et al., 2011)	9-item scale to measure travel-related subjective wellbeing covering affective and cognitive wellbeing	Ettema et al. (2013); He et al. (2020); Lades et al. (2020); Lunke (2020)	De Vos et al. (2019); Fan et al. (2021); He et al. (2020); Lunke (2020)	Cao (2013); Fan et al. (2021); Gao et al. (2017); He et al. (2020); Lades et al. (2020); Lunke (2020) (some are adapted versions)
Commute well-being (CWB) (Smith, 2017)	7-item scale adapted of STS measuring the affective and cognitive responses to work commuting	Smith (2017)		Smith (2017)
Stress and coping with stress (in general and while transport use)
Kessler-6 (Kessler et al., 2002)	6-item mental health screening instrument measuring non-specific psychological distress			Schoner et al. (2018)
PSS – Perceived Stress Scale-10 (Cohen et al., 1983)	14-item instrument to measure to what degree situations in life are perceived as stressful (e.g. feeling upset, stressed, in control)			Khattak et al. (2018); Pykett et al. (2020)
Scale to measure stress in public transportation (Dorantes-Argandar et al., 2019)	10-items to measure stress as public bus users, resulting from driver and other passenger’s behaviour, infrastructure/surrounding consisting of 2 factors: “People and Interaction” & “Infrastructure”			Dorantes-Argandar et al. (2019)
Stress on the commute-scale (Evans et al., 2002)	10-item commuter stress scale		Sposato et al. (2012), Wener et al. (2003) (previous version based on different studies)	Wener et al. (2003) (previous version based on different studies)
Inventory of Stressful Situations in Traffic (Dorantes Argandar et al., 2016)	22-item instrument to measure how stressful different situations in traffic are perceived; 3 factors: “Authority, Respect, Prudence and Social Order”; “Respect for Norms, Infrastructure and Environment”; “Inside the Car and Visibility”			Dorantes Argandar et al. (2016)
DSI – Driver Stress Inventory (Matthews et al., 1997)	48-item instrument to measure driving stress; 5 sub-scales: aggression, dislike of driving, thrill-seeking, hazard monitoring, fatigue proneness	Emo et al. (2016)
CITS – Coping Inventory for Task Stress (Matthews & Campbell, 1998)	21-item instrument measuring what strategies people apply to cope with task stress, 3 dimensions: task-focused coping, emotion-focused coping, avoidance	Emo et al. (2016)
DCQ – Driver Coping Questionnaire (Matthews et al., 1996)	35-item scale to measure coping with driving stress (7 items per dimension): Confrontive coping, Task-focused, Emotion-focused, Reappraisal, Avoidance	Emo et al. (2016)
Task stress, mental workload
DSSQ – Dundee Stress State Questionnaire (Matthews et al., 2013)	96-item for situational task stress in performance settings: 10 primary scales: Mood: Energetic (EA) and Tense (TA) arousal; Hedonic tone (HT); Motivation: Intrinsic motivation (MOT); Cognition: Self-focus of attention (S-FOC); Self-esteem (ESTEEM); Confidence and control (CONF); Concentration (CONC); Cognitive interference: task-related (CI-TR) and task-irrelevant (CI-TI); 3 higher order factors: Task Engagement, Distress, Worry	Emo et al. (2016); Fairclough and Spiridon (2012)
NASA-TLX – NASA-Task Load Index (Hart & Staveland, 1988)	Multi-dimensional measurement of overall subjective workload consisting of 6 items assessing Mental Demand, Physical Demand, Temporal Demand, Own Performance, Effort, Frustration Level	Huang et al. (2019); Ihme et al. (2018); Larue et al. (2020); Teh et al. (2014)	Larue et al. (2020)
RSME – Rating Scale Mental Effort (Zijlstra & van Doorn, 1985)	Uni-dimensional measurement of overall subjective workload on a 150-point, 15-cm long vertical line marked at 1-cm intervals, containing 9 anchor points with descriptive labels ranging from “absolutely no effort” to “extreme effort”.	Teh et al. (2014)
Mental disorders
GHQ-12 – General Health Questionnaire (Goldberg et al., 1997)	12-item mental health screening instrument (focussing on i.e. sleep loss, stress, feeling worthless) for detection of psychiatric morbidity	Chng et al. (2016) (1-item subscale)		Chng et al. (2016) (1-item subscale); Lades et al. (2020)
SF-12 – Short Form Health Survey (Knies, 2014)	12-item version of a screening instrument for general health and functioning covering 8 dimensions of mental health, i.e. physical functioning, social functioning, vitality, mental health			Feng et al. (2017) (only social functioning = 1 item); Leslie and Cerin (2008)
CESD-10 - 10-item Center for Epidemiologic Studies Depression (Andresen et al., 1994)	Screening tool for probably of clinically significant depression	Wang et al. (2019)
MBI – Maslach Burnout Inventory (Maslach et al., 1997)	Clinical instrument to measure burnout consisting of 3 scales: emotional exhaustion; depersonalization; personal accomplishment	Cunradi et al. (2009)
Other scales considered appropriate or relevant
		Nadrian et al. (2020); Rodrigues et al. (2015); Teng et al. (2020)	Evans et al. (2002); Rodrigues et al. (2015)	Gadsby et al. (2021); Leslie and Cerin (2008); Rodrigues et al. (2015); Teng et al. (2020)

Appendix 4. Summary table of physiological measures used in reviewed studies

Table

		Studies that included the physiological measures (or some adaptations)
Instrument	Description	Congestion/Delays	Reliability	Transport Infrastructure
Electrodermal activity (EDA)	An index of sympathetic nerve activity affecting skin sweat (and related to temperature control); transient increases are closely related to arousal in emotional contexts.	Bitkina et al. (2019); Castro et al. (2020); Caviedes and Figliozzi (2018); Dogan et al. (2019); Rigas et al. (2011); Singh et al. (2013); Wang and Guo (2021); Werner et al. (2019); Yan et al. (2018); Zhang et al. (2019)		Castro et al. (2020); Caviedes and Figliozzi (2018); Cobb et al. (2021); Teixeira et al. (2020); Werner et al. (2019)
Heart rate variability (HRV)	By analysing variations in the beat-to-beat interval, HRV can be calculated and is related to stress and cardiovascular risk; low (vs. high) frequency reflects sympathetic stress (vs. parasympathetic) reaction.	Cardone et al. (2020); Fairclough and Spiridon (2012); Huang et al. (2019); Qi et al. (2015); Rigas et al. (2011); Rodrigues et al. (2015); Singh et al. (2013); Zhang et al. (2019)	Rodrigues et al. (2015)	Doorley et al. (2015); Fitch et al. (2020); Rodrigues et al. (2015)
Heart rate (HR)	In general, HR increases with acute stress, but other factors are also important, such as respiration; influences HR and in some contexts, HR decreases despite sympathetic stress response occurs	Qi et al. (2017); Singh et al. (2013); Wang and Guo (2021); Yan et al. (2018)		Doorley et al. (2015); Khattak et al. (2018)
Electroencephalography (EEG)	Non-invasive electrophysiological monitoring process for capturing brain’s electrical activity; several types of analysis could be used, but often EEG signals are classified according to their frequency bands (alpha, beta and gamma frequency bands are associated with relaxation, arousal, and anxious conditions, respectively).	Di Flumeri et al. (2018); Fairclough and Spiridon (2012); Halim and Rehan (2020); Li et al. (2020); Yan et al. (2018)
Cognitive task	Cognitive tasks are either used to induce stress or to evaluate how stress (or cognitive load) influences one’s ability to behave; attention, executive functions and short-term memory are the domains most used for testing.	Huang et al. (2019); Ihme et al. (2018)	Evans et al. (2002); Wener et al. (2003)	Wener et al. (2003)
Facial expression	Facial behaviour is commonly used for emotion recognition; it assumes that discrete emotions could be elicited and recognised via facial muscle configurations (captured either by muscle electrodes or video-recordings).	Fairclough and Spiridon (2012); Ihme et al. (2018); Qi et al. (2015); Rigas et al. (2011); Wang and Guo (2021)
Cortisol levels	It is a hormone produced by the adrenal gland important for many metabolic functions, which increases in response to various stressors; salivary (rather than blood) cortisol level is convenient and often used as a biomarker of psychological stress.		Evans et al. (2002); Wener et al. (2003)	Wener et al. (2003)
Other types
Thermal imaging	Acute stress might cause a rapid drop in skin temperature (due to peripheral sympathetically mediated vasoconstriction), but an increase in the core body temperature (due to a net influx of blood to key organs).	Cardone et al. (2020); Di Flumeri et al. (2018); Li et al. (2020); Zhang et al. (2019)
Respiratory assessment	An increase in the respiratory rate (resulting in lowered end-tidal CO2) is often noted during psychological distress. On some occasions, rapid drops in peripheral capillary oxygen saturation could be associated with stress.	Rigas et al. (2011); Singh et al. (2013); Wang and Guo (2021)
Eye movements	Eye-movement patterns (i.e. saccades, fixations, dwell times, etc.) could be used to evaluate cognitive load, attention, and fatigue; they can also be influenced by visual and emotional salience.	Di Flumeri et al. (2018); Li et al. (2020)
fNIRS	It is a non-invasive optical imaging technique that uses near-infrared light to measure cerebral blood flow changes when neural activity is elicited.	Ihme et al. (2018)