Physical workplace adjustments to support neurodivergent workers: A systematic review

Institute of Facility Management, School of Life Sciences and Facility Management, Zurich University of Applied Sciences, Wädenswil, Switzerland Environmental Psychology Department, School of Psychology, University of Surrey, Guildford, UK Institute of Occupational Therapy, School of Health Science, Zurich University of Applied Sciences, Winterthur, Switzerland Department of Organizational Psychology, Birkbeck University of London, London, UK

Regarding such sensory needs, difficulties in modulating sensory responses 1 and overresponsivity (overload) are common symptoms shared by the main ND groups; either as part of the diagnostic criteria (ASD), as comorbidity or as an explanatory (developmental) theory (ADD/ADHD, Dyslexia, Dyspraxia). To illustrate, according to the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association, 2013), people with ASD 2 'may struggle with hyper-or hyporeactivity to sensory input or unusual interest in sensory aspects of the environment. This may present as indifference to pain or temperature and adverse responses to specific sounds or textures, excessive smelling or touching of objects, visual fascination with lights or movement ' (p. 50). While the two response patterns/sensitivity states, hyper-responsiveness/hypersensitivity (exaggerated response to stimuli) and hyporesponsiveness/hyposensitivity (lack of response to stimuli) are most prevalent in autistic populations, they are also present in populations with ADD/ADHD (Panagiotidi et al., 2018;Perrachione et al., 2016) and Dyspraxia (Mikami et al., 2021), affecting different sensory domains. In ADD/ADHD populations, hyper-or hyposensitivity concerns visual (Panagiotidi et al., 2017a), auditory (Ghanizadeh, 2009), tactile (Hern & Hynd, 1992), olfactory (Romanos et al., 2008), vestibular modalities (Sergeant et al., 2006) and multisensory integration (Panagiotidi et al., 2017b). In dyslexic populations, hypersensitivity or hyposensitivity concerns visual, visual-spatial and auditory modalities (cf. Goswami, 2015). In populations with Dyspraxia, multiple sensory modalities related to movement planning and processing are affected, including vestibular, tactile, visual-spatial but also oral and auditory sensory domains (Allen & Casey, 2017;Blank et al., 2019;Gomez & Sirigu, 2015;Goyen et al., 2011;Loh et al., 2011;Mikami et al., 2021).
Sensory interventions and adjustments such as promoting optimal arousal, using sensory modalities for self-regulation or lower over-reactivity for ND individuals appear to be widely applied, accessible and well researched, especially in childhood (e.g. Case- Smith et al., 2015). However, a sound evidence base does not exist for the work domain, where a range of recommended adjustments was found to be poorly implemented and their efficacy poorly understood (Doyle, 2020;Doyle & McDowall, 2021). Despite the unclear evidence base, workplace design solutions have been proposed by the building industry and design advisor groups (e.g. Forbo Flooring System, 2020; Group GSA, 2020HOK, 2019HOK, , 2020. Although such developments suggest a positive trend in a growing preoccupation within the design industry to build inclusive workplace environments, the utility of such design solutions is unclear. Physical workplace adjustments, which may benefit neurodiverse workers in regard to work performance and health/well-being, may have wider reaching benefits, such as occupational longevity and systemic inclusion (e.g. Doyle & McDowall, 2021). As posited by Doyle and McDowall (2019, p. 3), "the overarching issue of (lack of) social inclusion and equality … creates a moral imperative to ensure any adjustments are substantiated by evidence and guiding theoretical frameworks." We undertook a focused evaluation and synthesis of the available evidence, specifically of physical workplace adjustments (e.g. design solutions), on occupational and health-related outcomes for ND workers. In line with relevant guidelines, we adopted a broad conceptualization of environmental adjustments to the physical environment to elicit all relevant evidence.

Review aim
Given the need to identify the evidence base for physical workplace adjustments addressing sensory difficulties, this systematic review examined the following research question in line with the CIMO-logic: What is the evidence (in terms of extent, nature, and quality) for physical workplace adjustment [I] to support occupational longevity, performance, and health/well-being [O] in ND workers [C] as of specific sensory needs [M]?

METHOD
In conducting our review, we adopted a systematic approach as outlined by Briner and Denyer (2012) and informed by PRISMA guidance (Liberati et al., 2009) to locate all relevant empirical, peer-reviewed evidence for the relationship between physical workplace adjustment (that directly or indirectly address sensory needs) and occupational longevity, performance and health/well-being outcomes in ND office workers.

Eligibility criteria
We used the CIMO framework to guide the specific inclusion and exclusion criteria due to its appropriateness in social science settings (Denyer & Tranfield, 2009). We included all types of empirical study designs published in English from 2000 until September 2021 as follows: Context: adult neurodivergent workers including Attention Deficit Disorder, Autism Spectrum Disorder, Dyslexia, Dyspraxia, Tourette Syndrome, Dyscalculia, Dysgraphia (Doyle, 2020) and/or their family members, colleagues, employers, experts and vocational programme staff; office environments or mixed work settings; Interventions: studies referencing any types of physical workplace adjustmentsfor example, office layout considerations, special furniture, acoustic adjustments, ergonomics considerations, sensory/fidget toys, either in their results section or in any specific recommendations; Mechanisms: all mechanisms that explain the relationship between the intervention and outcome and the conditions under which these are activated or accessed; Outcomes: all outcome measures that encompass concepts relevant to changes in occupational longevity, performance and health/well-being.
Exclusion criteria were: participants that were exclusively self-diagnosed (studies with a mix of self-and formal diagnosis were included), under the age of 18 years; solely technological workplace adjustments, such as use of computer software or programmes; workplaces that were not/did not include office environments; non-empirical studies, such as secondary data analyses-for example, systematic literature reviews and meta-analyses, reports or guidelines. Although we had initially aimed to exclude any non-peer-reviewed studies, we revised our approach given the limited search and include such sources to ensure breadth of primary evidence (Kendall,

Data sources
We searched across multiple databases to ensure coverage of health, psychology and management research: (1) PubMed; (2) ABI/Inform; (3) PsychINFO; (4) Web of Science Core Collection and (5) Scopus. We then hand-searched reference lists of all included and excluded studies (non-eligible or non-empirical) and those within guidance documents by charities, government, business, psychological and workplace/design advisor groups.

Conducting searches
We completed iterative searches from July 2020 to September 2021 to locate as many relevant papers as possible (Greenhalgh & Peacock, 2005). These identified 312 records, which included at least one term of each search string in the title or in the abstract (S1-S4) or text (only S4) (see Figure 1). We identified seven records by hand-picking. After duplicate removal, two researchers (CW and EH) assessed the eligibility of studies using a four-step process: (1) screening titles; (2) screening abstracts; (3) full text skimming and (4) full text review and application of exclusion-inclusion criteria. At each screening step, we assigned sources to a different researcher (EH and CW). We discussed uncertainties at each screening stage on a regular basis to achieve consensus. A total of 20 studies remained for inclusion and synthesis.

Data extraction
Our bespoke extraction template designed by two members of the author team (CW, EH) included first author, year, country, study design (qualitative, quantitative and mixed methods), target population by condition, participants (e.g. family members, employers and co-workers of ND workers) and sample size, participants' characteristics (e.g. age, gender, education level and occupation), study objectives, methods/measures, key findings pertinent to this review, workplace adjustments suggested or implemented in studies (e.g. adjusting light and providing private offices) and outcome domain (occupational longevity, performance and/or health/well-being). Two researchers extracted and reviewed the data (CW and EH). S3 Intervention Intervention terms Accommoda*, Adapt*, Adjust*, Change*, Intervention*, Modifi*, "Occupational intervention*", "Reasonable accommoda*", "Workplace accommoda*", "Workplace adapt*", "Workplace adjust*", "Workplace intervention", "Workplace modifi*"

Quality assessment
To address the heterogeneous nature of the primary studies, we used widely accepted tools (Noyes et al., 2018) for quality assessment: the Critical Appraisal Skills Programme (CASP Qualitative Checklist, 2013 for qualitative studies (n = 13) and the Mixed Methods Appraisal Tool (MMAT, Hong et al., 2018) for mixed methods (n = 6) and for quantitative nonrandomised studies (n = 1), 3 to evaluate (1) method appropriateness, (2) rigour in data collection and bias, (3) ethical issues, (4) analytical rigour and bias and (5) reporting. We elicited overall quality scores aligning with recommendations (Pluye et al., 2009) and prior practices (e.g. Bury et al., 2020) for further evaluation. The maximum overall score varied by tool and study category (10 CASP; 5-25 mixed methods MMAT; 5 quantitative non-randomised MMAT). In both tools, rating questions followed the format of 'Yes', 'No' and 'Cannot tell'. Aligning with Long et al. (2020, p. 36), 'we nominated to use "can't tell" when there was insufficient information reported to make a judgement (i.e. a reporting issue)'. Scoring procedure followed Pluye et al.'s (2009, p. 540) guidelines: 'Score presence/absence of criteria respectively 1/0 … F I G U R E 1 PRISMA flowchart Calculate a "quality score" [(number of "presence" responses divided by the number of "relevant criteria") Â 100]'. Quality was assessed by three researchers (CW, BK and EH); any potential divergence was discussed in the wider group (CW, BK, EH and JY) to achieve consensus, but ratings did not result in exclusion to ensure an inclusive approach to the extant evidence base.

Data analysis: Idealist narrative synthesis
Neither quantitative meta-analysis nor qualitative meta-synthesis were appropriate due to the heterogeneous nature of the retrieved studies (Popay et al., 2006). Instead, we extracted, categorised and narratively synthesised relevant information from each study aligned to the sixstep framework by Braun and Clarke (2006) with a hybrid approach to inductive and deductive thematic analysis (cf. Fereday & Muir-Cochrane, 2006). We combined a data-driven induction (e.g. Boyatzis, 1998) with a deductive a priori template of master themes (e.g. Crabtree & Miller, 1999) to conceptualise the sensory physical environment. Such hybrid approach to thematic analysis is considered useful in occupational contexts, provided a clear framework for the underlying phenomenon under investigation is available (Fereday & Muir-Cochrane, 2006;Ligurgo et al., 2018). Given the complexity of our research question, which referenced context as well as individual experience, our coding ensured consideration of all aspects of interest, such as comprehensive coverage of sensory modalities, while allowing specific codes and subthemes to be interpreted from the data. First, we familiarised ourselves (CW and EH) with the data by reading and rereading with a data-led inductive approach through identification of relevant information, such as participant quotes and relevant information in the respective results and discussions. Second, we inductively coded the data set to represent nuanced types of demands or resources in the physical environment, concrete adjustments or related outcomes; we used a deductive framework-informed approach to refine the codes, e.g. Fereday and Muir-Cochrane (2006). Third, we interpreted subthemes from the initial codes through a deductive scan, where we clustered each code using the a priori template based on the master themes of the framework. We then collated codes in each of the five dimensions/master themes into subthemes (e.g. Dimension/Master theme 1: Adjustments for Managing Environmental Stimulation; Subtheme 1: Auditory; Example code: Reducing noise levels with tools). Fourthly, we reviewed and revised the subthemes by checking whether they represent the extracted codes adequately and whether any aspect (e.g. sensory modality) had been overlooked. Finally, we defined the subthemes through iterative analysis of the themes by relating them back to the framework and dimensions. Our approach was marked through an iterative and reflexive process, with several cycles of revision within the group of researchers (CW and EH) rather than linear rigidity as our illustration might suggest.
We took an idealist approach including quotes from participants or observational findings and authors' interpretations (Barnett-Page & Thomas, 2009;Noyes et al., 2018). This allowed us to include mixed-methods studies (of which the majority had rich qualitative participant data) and one cross-sectional study as the title authors' interpretations were treated as relevant data. Where possible, we retained the original participants' or title authors' wording for coding and subtheme development. We undertook outcome domain categorisation (occupational longevity, performance and health/well-being) from the evidence provided in each primary study. If authors were unclear or did not use any terminology that related to the broader outcome domains (e.g. health/well-being), we classified based on our own interpretation of participant quotes (e.g. headache à health/well-being). We retained the title author's wording where possible to maintain transparency and minimise the risk of undue interpretations and categorisations. We undertook the narrative synthesis within-and across-study .

Framework: Conceptualising the sensory physical environment
Our framework for the narrative synthesis is rooted in Lawton's (1987) revised ecological model of Person-Environment (P-E) fit, which has been widely used in vulnerable group research (cf. Cvitkovich & Wister, 2001) and in workplace design research (e.g. privacy fit; Weber & Gatersleben, 2021). This P-E fit model explains the multi-tiered subjective dimensions of P-E fit where behaviour and well-being are a result of how the environment meets the needs of the individual. Lawton (1987) specifies that the environment consists of social and physical dimensions, the latter differentiates between the objective measurable environment and the phenomenal environment. Thus, we frame the physical environment as offering demands, resources, affordances and meanings.
Thus, we conceptualised the P-E fit perspective and understanding of the physical environment by drawing on the International Classification of Functioning, Disability and Health (ICF), which considers physical environmental factors that facilitate or hinder the performance of daily activities (World Health Organization, 2009). In line with other ND researchers (cf. Pfeiffer et al., 2017), we focused on sensory physical factors that present demands or resources: hypersensitivity and hyposensitivity needs. We refined environmental demands and stresses using environmental stress theories concerning social stimulation (e.g. crowding and privacy) and environmental control (Evans & Cohen, 1987). The evidence was therefore categorised as outlined below in Table 2.

Study and participant characteristics
The full data extraction is summarised in Table 6.
T A B L E 2 P-E fit framework

Dimension Conceptualisation
Further detail
Environmental resource stimuli positioned as desired sensory stimulation related to hyposensitivity (cf. Lawton, 1987;Pfeiffer et al., 2017) Dimension (2) Managing social stimulation in space (cf. Evans & Cohen, 1987;Lawton, 1987;Pfeiffer et al., 2017) Aspects that relate to reducing social interaction, demands, and sensory stress linked to hypersensitivity incorporating socio-environmental phenomena such as privacy, crowding, or personal.
Aspects that relate to increasing social interaction in space Environmental control which relates to a sense of control afforded by the environment (cf. Evans & Cohen, 1987;Lawton, 1987) N/A

Dimension (4)
Captures all physical environment adjustments that are non-environmentalor social-stimulation specific.
Summarised occupational longevity, performance, and health/well-being outcomes of suggested or provided physical workplace adjustments (cf. Lawton, 1987;ICF, World Health Organization, 2009) N/A

ND conditions
Overall, 916 participants in this review were indicated to be ND. Conditions included Autism (ASD; Asperger's disorder; high-functioning autism, PDD-NOS); attention deficit disorders (ADD, ADHD); dyslexia; and another learning disabilities, which were not specified. The majority of studies (18) focused on ASD, whereas one focused on attention deficit disorders (Schreuer & Dorot, 2017), and only one focused on a variety of ND conditions (autism, attention deficit disorders and dyslexia) (Morris et al., 2015). Regarding diagnosis, eight studies had participants with formal diagnoses, five with formal, informal and self-reported diagnoses, three studies gave no information and for three studies diagnosis type was not relevant as a non-ND sample. Results are not differentiated by type of diagnosis (self or formal). Furthermore, six studies had samples which reported ND-relevant comorbidities (e.g. ASD with underlying ADHD).

Sample demographics of ND individuals 4
There was a wide range of participant ages, from 18 to 80 years, with most samples having a mean age from the mid-20s to early 40s. Most studies reported on participants' gender, educational level and occupation. The proportion of males and females was almost equal (male n = 380, female n = 408, others n = 19, no information n = 102). Further, 43% of studies that provided information on educational level reported their samples to have a university degree. Participants worked in various sectors, such as higher education, public sector, health care and financial services. Occupations included administrative workers, teachers, assistants, sales workers, employment specialists, graphic designers, chemical technicians and communication specialists. The most commonly held jobs by autistic participants (nine studies) were information technology-related jobs (e.g. software test analyst, software developer, computer system analyst or data entry specialist).

Work context
The majority of studies included varied work and office settings. Only four studies (Hedley et al., 2018;Kendall, 2013;Morris et al., 2015;Pierce, 2018) appeared to solely focus on office settings and one study was unclear (Kirchner & Dziobek, 2014). We inferred this by participants' occupations, case study descriptions or by contacting the studies' authors who were not always responsive.

Outcomes classified
None of the studies specifically focused on occupational longevity, performance and health/ well-being but reported on components of these concepts. Twelve studies reported on aspects related to occupational longevity (such as maintaining employment, enablers and barriers to success at work), 11 studies on aspects related to performance (such as work ability or concentration) and 6 studies on aspects related to health/well-being (such as work satisfaction, adjustment fatigue or overstimulation symptoms).

Quality assessment
In answer to the research question on the quality of the available evidence, study quality was overall heterogeneous, meeting 20%-100% of the quality criteria across assessment tools. The assessment scores for the qualitative papers ranged from 60% to 100% (average 88%). Most qualitative methods were justified; however, some lacked robust data collection methods (e.g. conducting open-ended interviews via email) or recruited a sample that did not conform with the specified inclusion criteria or research aims (e.g. unemployed workers and students despite inclusion criteria being minimum 1 year in the workforce). A third of the studies did not adequately consider the relationship between researcher and participants (e.g. clear accounts on reflexivity); some provided limited detail in the presentation of the analysis or interpretations. The assessment scores for the mixed-method papers ranged from 27% to 73% (average 47%). Predominant weaknesses of the mixed-methods studies included an unsatisfactory integration of the different study component-methods (qualitative/quantitative) and non-adherence to the quality criteria of each component-method involved. For the quantitative study, the assessment score was 20%. Recruitment methods were detailed but did not provide a clear justification of the arguably non-representative sample size, used non-validated measures and failed to account for confounding variables in their design or analysis. The quality assessments for the quantitative and mixed-methods studies were the lowest and need to be interpreted with caution. The detailed quality appraisal for all studies can be found in the supporting information (Tables S1 and S2).

Study synthesis
Regarding the nature and extent of the available evidence, we present a summary of the results in Table 4 (adjustments) and Table 5 (outcomes). We synthesised findings with the five dimensions of the analysis framework (managing environmental stimulation and social stimulation, environmental control and sensory-unspecific adjustments, and outcomes) as particularly relevant to ND participants. This proved more difficult when evidence was provided for mixed groups in qualitative studies (e.g. Black et al., 2019). We have included the full results of the studies (neurotypical and ND evidence) in the data extraction table (Table 2). Where studies did not include any ND individuals but company directors, line managers, leaders, frontline supervisors or employers of ND workers (Dreaver et al., 2020;Pierce, 2018;Waisman-Nitzan et al., 2019), we embedded these with findings from studies that included ND participants. When results were solely retrieved from non-ND accounts, we have made this clear in the text.

Visual
Fifteen studies reported on available, desired or recommended adjustments to reduce visual stimulation and environmental stress (hypersensitivity) for workers with autism (Baldwin et al., 2014;Black et al., 2019Black et al., , 2020Diener et al., 2020;Dreaver et al., 2020;Harvery et al., 2021;Hayward et al., 2019;Hedley et al., 2018;Kendall, 2013;Lorenz et al., 2016;Müller et al., 2003;North, 2021;Pfeiffer et al., 2017;Pierce, 2018;Waisman-Nitzan et al., 2021). No other ND conditions were covered, nor did any study report on the provision of visual stimulation to accommodate hyposensitivity. Adjustments that specifically target light sensitivities of autistic workers included adjusting the type of light and reducing light levels. Six studies reported that natural, incandescent light is preferred/suggested and fluorescent light should be avoided (Diener et al., 2020;Harvery et al., 2021;Kendall, 2013;Müller et al., 2003); desk lamps were preferred over overhead lighting (North, 2021), and special lighting was suggested (Baldwin et al., 2014). Ten studies suggested giving autistic employees control to reduce the light levels by dimming lights, switching the light off, removing light bulb over work area or generally not working in a cubicle environment where the light is bright and not controllable (Black et al., 2019Dreaver et al., 2020;Harvery et al., 2021;Hedley et al., 2018;Kendall, 2013;Lorenz et al., 2016;Pfeiffer et al., 2017;Pierce, 2018) or by using tools such as blinds or wearing sunglasses (Dreaver et al., 2020;Hayward et al., 2019). Adjustments that specifically target visual distractions (e.g. movement of other people) were reported by two studies and included visual screens (Pfeiffer et al., 2017;Waisman-Nitzan et al., 2021) and clear space design (Waisman-Nitzan et al., 2021).

Tactile
One study reported on requested adjustments to address undesired tactile stimulation and environmental stress (hypersensitivity) in an autistic sample, specifically adjusting the office temperature (Pfeiffer et al., 2017). Another study focused on study leaders/managers/ supervisors' suggestions to provide tactile sensory reduction items, such as fidget toys, and a private decompression room to reduce overstimulation symptoms in autistic workers (Pierce, 2018).

Olfactory
No study mentioned specific adjustments to reduce olfactory stimulation, although one did suggest that the use of personal fragrances can assist autistic workers and that employers should consider this as a potential adjustment (Hayward et al., 2019).

Modality unspecific sensory accommodations
Several authors made general suggestions, such as providing a sensory-friendly environment (Hayward et al., 2019) and accommodating sensory sensitivity (Schreuer & Dorot, 2017;Waisman-Nitzan et al., 2019). The facilitation of a decompression room was one more specific adjustment suggested for autistic workers experiencing symptoms of overstimulation (Pierce, 2018).

Dimension 2: Adjustments for managing social stimulation
Nine studies reported on available, desired or recommended adjustments to reduce social stimulation, which concerned all included neurodiversity conditions (Black et al., 2019;Diener et al., 2020;Harvery et al., 2021;Hayward et al., 2019;Kendall, 2013;Kirchner & Dziobek, 2014;Morris et al., 2015;Müller et al., 2003;Soeker, 2020). Adjustments included avoiding open-plan office/shared office spaces and providing private work areas. These could encompass a private office, workspace away from others, office in a quiet corridor with less 'human traffic' passing, offering withdrawal options, relocating noisy co-workers or other means of reducing social crowding (Black et al., 2019;Diener et al., 2020;Harvery et al., 2021;Hayward et al., 2019;Kendall, 2013;Kirchner & Dziobek, 2014;Morris et al., 2015;Müller et al., 2003;Soeker, 2020). Three studies reported on desired and available adjustments to facilitate social interaction in autistic samples (Hedley et al., 2018;North, 2021;Pierce, 2018). Adjustments included social break rooms and locating workstations close to co-workers. One study highlighted that, contrary to stereotypical assumptions, many participants who were provided with a separate workspace indicated that they did not want to work in complete isolation but with social interaction (North, 2021).

Dimension 3: Adjustments for providing environmental control
Twelve studies reported on available, desired or recommended adjustments to allow for environment control at work, which concerned all ND conditions relevant to our review. Of those, seven studies raised unspecified general physical modification to the work environment to suit sensory and general needs (Black et al., 2019;Diener et al., 2020;Dreaver et al., 2020;Hayward et al., 2019;Hedley et al., 2018;Pfeiffer et al., 2017;Pierce, 2018); four studies specified the control over physical stimulation (e.g. lighting, sounds, smells or temperature; Black et al., 2019;Hayward et al., 2019;Kendall, 2013;Pfeiffer et al., 2017). Seven studies specified the necessity of providing a private office to reduce sensory overload and distraction across neurodiversity conditions (Black et al., 2019;Hayward et al., 2019;Kendall, 2013;Kirchner & Dziobek, 2014;Lorenz et al., 2016;Morris et al., 2015;North, 2021). Furthermore, providing owned/private space ('his own corner' and 'my own space, my personal space') was not only related to increased control over the sensory and socio-spatial environment but also facilitated an increased sense of self and identification in two studies of autistic employees (Pfeiffer et al., 2017;Waisman-Nitzan et al., 2019). Interestingly, one study reported that workers with autism were significantly more likely to identify control over the physical environment as an employment enabler than neurotypical workers (Hayward et al., 2019).

Dimension 4: Non-stimulation specific adjustments
Three studies described sensory-unspecific adjustments, which we considered resources. These concerned workers with autism and ADD and included adequate furniture (suggested by leaders/managers/supervisors of individuals with ASD; Pierce, 2018), 'a good seat' (Morris et al., 2015), a treadmill desk to alleviate anxiety during the workday (Morris et al., 2015) and adequate/extended desk size (Pfeiffer et al., 2017). Nine studies highlighted being allowed to work from home as an environment-related and policy-related adjustment for managing sensory and socio-environmental needs (Black et al., 2019;Diener et al., 2020;Harvery et al., 2021;Hayward et al., 2019;Kendall, 2013;Lorenz et al., 2016;Morris et al., 2015;Pfeiffer et al., 2017;Schreuer & Dorot, 2017).

Dimension 5: Related outcomes
Regarding the available evidence for physical workplace adjustment to support occupational longevity, performance and health/well-being, we present a summary of the results in Table 6. Due to the heterogeneous quality, study designs, aims and outcomes explored, related outcomes are merely indicative.

Occupational longevity
Twelve studies indicated a relationship between adjustments and occupational longevity across ND conditions. Specifically, nine studies suggested that sensory adjustments, socio-spatial adjustments, such as private work areas, environmental control including enabling individuals to alter physical environment conditions, allowing working from home, understanding individual needs and provision of a sensory-friendly environment at baseline can facilitate obtaining appropriate employment (Harvery et al., 2021), maintaining employment ( (Kendall, 2013;Pfeiffer et al., 2017). Acoustic isolation through workstations in quiet areas/ private office, using noise-cancelling headphones or working at home enhanced attention and organisation skills at work by mitigating environmental distraction (Morris et al., 2015;Schreuer & Dorot, 2017). Two studies reported that participants used fidget toys, noisecancelling headphones, and screens to block sight to improve concentration (Pfeiffer et al., 2017;Pierce, 2018). There were also suggestions that performance can be affected by sensory disturbances. These included smell, temperature, background noise, loud conversations, crowding, visual distraction (e.g. people passing by or cluttered desks; Black et al., 2020;Kirchner & Dziobek, 2014;Pfeiffer et al., 2017;Pierce, 2018;Soeker, 2020;Waisman-Nitzan et al., 2019) and a lack of environmental adjustments (Morris et al., 2015;Pfeiffer et al., 2017;Waisman-Nitzan et al., 2019).
Health/well-being Six studies indicated a relationship between environmental factors or sensory adjustments and various health/well-being outcomes (Baldwin et al., 2014;Morris et al., 2015;North, 2021;Pfeiffer et al., 2017;Pierce, 2018;Soeker, 2020) Specifically, too much sensory information in the environment, such as large and loud work environments, left participants feeling overwhelmed. In contrast, control over sensory adaptations to the work environment were reported to reduce sensory and physical aversions (Pfeiffer et al., 2017;Soeker, 2020) and increase satisfaction (Pfeiffer et al., 2017). Lighting control was crucial when participants suffered from headaches (Pfeiffer et al., 2017). Working in a soundproofed private office and having the opportunity to use a home office relieved stress and improve work satisfaction (Morris et al., 2015;Pfeiffer et al., 2017). Access to breakout rooms that are quiet and to breakout rooms that are social was critical to general workplace well-being. Providing a decompression room (sensory break room with sensory reduction item) was reported by leaders/managers/supervisors of individuals with ASD to be helpful in alleviating stress in workers with ASD (Pierce, 2018). Regarding special equipment, fidget toys alleviated anxiety, relieved stress and reduced overstimulation symptoms, such as irritability and outbursts (Pierce, 2018). The provision of a treadmill desk was suggested for mitigating anxiety (Morris et al., 2015). Furthermore, working from home reduced stress from interpersonal interaction (Morris et al., 2015) and providing an expanding desk improved satisfaction (Pfeiffer et al., 2017). In contrast, a poor physical work environment and inappropriate/insufficient reasonable adjustments were linked to an overall negative work environment (Baldwin et al., 2014), leading to stigmatisation and isolation, which have a negative impact on career development (North, 2021).

DISCUSSION
It was our intent to examine the evidence regarding extent, robustness and quality of physical workplace adjustments to support occupational longevity, performance and health/well-being in ND workers with specific sensory needs.

Extent, nature and quality of evidence found
The review identified a number of adjustments available to ND workers and the perceived benefit that these adjustments afford. However, the review revealed an underdeveloped evidence base, comprising studies of varying methodological quality without any robust intervention study designs. Such lack of rigourous experimental work means that we do not yet know which types of adjustments yield robustly evidenced positive outcomes, neither do we know clearly under which circumstances or for whom. Such a vacuum is preventing the development of evidence-based practice, including quality control and return on investment specifications (Briner & Rousseau, 2011;Doyle, 2020;Doyle & McDowall, 2021). None of the studies focused solely on the efficacy of physical workplace adjustments but explored a mix of psychosocial and physical adjustments in an unsystematic fashion. Therefore, it is difficult to draw firm conclusions for four reasons. First, most studies used mixed work settings (only four studies used office-only settings). Second, most of the evidence concerns workers with a primary condition of ASD (although NDrelevant comorbidities existed, e.g. ASD with underlying ADHD), leaving other ND groups underrepresented, which is a common issue in ND research (cf. Doyle & McDowall, 2021). Third, studies (N = 5) used mixed-diagnoses-type samples that were formally, informally and/or self-diagnosed. This reflects a current debate in the field, where some recommend the exclusion of self-diagnosed individuals to ensure common diagnostic framing, whereas others point to the necessity of inclusion as the high number of under-diagnosis due to various barriers of receiving a formal diagnosis (e.g. access to clinicians, cost and potential stigma; Huang et al., 2020) and changes, for example in autism diagnostic criteria (Lai & Baron-Cohen, 2015). Although the present study differentiated self-from formally diagnosed individuals (cf. Harvery et al., 2021) in the data extraction table, it cannot offer this level of differentiation in the results. Fourth, validity of the study findings might be threatened by an overrepresentation of females in comparison to latest estimated prevalence in society. On the one hand, this could be due to underdiagnosed ASD conditions in females, which is currently widely discussed (Loomes et al., 2017). On the other hand, females might find their way into employment easier as they internalise more (Bauminger et al., 2010). Similarly, females might be more interested to report their experiences and participate in ND research due to their higher tendency for social camouflaging to increase their social acceptance (Cook et al., 2021;Dean et al., 2017;Nagib & Wilton, 2020). Hence, it is not entirely clear whether the results are a valid representation for both genders with ASD or whether any gender-differentiated adjustment needs exist in the first place.

Extent and nature of sensory adjustments
With regard to the extent and nature of adjustments to support ND workers with specific sensory needs, visual and acoustic adjustments were most prevalent. However, included studies addressed sensory modalities in an unsystematic way, and some modalities were not addressed at all, such as gustatory or vestibular modalities. This leaves a variety of potential sensory issues that could be perceived as barriers to working well unaddressed (e.g. affecting health/well-being and impeding performance). It remains unclear, whether, for example, vestibular adjustments are not required because those issues are less pertinent in an office context or not addressed due to an unsystematic approach. In contrast, generic charity and design guidelines list various vestibular adjustments to prevent accidents, such as stair lifts, special furniture and appropriately configured workspaces (Dyspraxia Foundation, 2016;NDA, 2015;Standifer, 2009).
Similarly, most studies addressed hypersensitivity by reducing environmental stimulation whereas none of the studies directly addressed hyposensitivity, yet this is a critical sensory response for ND individuals and a potential reason for impeded health and performance at work. Hypersensitivity and hyposensitivity are both arousal states where too much and too little can impede performance (cf. Yerkes-Dodson Law of Arousal; Cohen, 2011) if not addressed by through alternative workplace adjustments. This was also evident in participants' accounts of understimulation, which hampered concentration (Morris et al., 2015). The included studies did not address this directly; for example, stimulation items were only positioned to relieve overstimulation. In contrast, hyposensitivity has been considered in generic guidance documents (Cassidy, 2018 This review elicits indicative benefits of specific sensory adjustments to support occupational longevity, performance and health/well-being in ND workers. However, due to the methodological quality of available evidence, it is not possible to draw any firm conclusions about the cause and effect of specific adjustments. Such lack of evidence is surprising given the openly addressed social exclusion of ND individuals from the labour market (Australian Bureau of Statistics, 2017; Organisation for Economic Co-operation and Development, 2010;Roux et al., 2017;Shattuck et al., 2012), the well-documented problems of ND individuals in gaining and maintaining employment and the legal rights for adjustments, which vary by geographic location (cf. Doyle, 2020).

Indicative benefits of sensory adjustments
We derive indicative findings on the merit of adjustments from our review. The following sections discuss the research findings to signpost the need for future research framed in our person-environment fit perspective and include benchmarking against existing generic guidelines.

Environmental stimulation
Predominantly, adjustments targeted the reduction of sensory stress by reducing acoustic and visual stimulation. Very few studies touched on tactile and olfactory adjustments. Acoustic adjustments for various ND conditions included headphones or earplugs and private rooms or quiet locations. Only one study touched on pleasant acoustical stimulation. Visual adjustments, for ASD only, targeted light levels and types as well as screening out visual distractions (screens). One study pointed to the use of stimulation objects to reduce overstimulation. Overall, adjustments targeting too much environmental stimulation were indicatively related to occupational longevity, performance and health/well-being. Heuristic guidelines typically target hypersensitivity, considering all sensory modalities, by recommending the elicited adjustments too, but go further by additionally proposing, for example, white noise machines (JAN, 2020;NDA, 2015;Standifer, 2009), non-stimulating colour schemes (HOK, 2019), air purifiers for odour removal (JAN, 2019), being mindful of textures (HOK, 2019) and ensuring ample space exists to move between furniture and machinery that could be knocked over (NDA, 2015). To target hyposensitivity and/or creating desired environmental stimulation, guidelines propose natural stimuli (e.g. greenery, nature material, sound or imagery; Group GSA, 2020;HOK, 2019HOK, , 2020, sensory stimuli objects to help to focus when understimulated (HOK, 2019; La Trobe University, 2020), providing strongly scented planting (e.g. herb gardens; Group GSA, 2020), stimulating colour schemes (HOK, 2019) or activity areas (e.g. balancing boards or bouncy balls; Group GSA, 2020HOK, 2019).

Social stimulation
Reduction of social stimulation across all ND conditions included private offices, areas with less human traffic, as well as options to withdraw and limit access to oneself. Very few studies addressed facilitation of social interaction. Adjustments for excessive social stimulation are indicatively related to occupational longevity and some with performance and health/well-being. It is widely acknowledged that social demands at work can be challenging, particularly for workers with ASD and ADHD/ADD (e.g. Black et al., 2019;Diener et al., 2020;Morris et al., 2015;Pfeiffer et al., 2017). It is no surprise that guidelines also strongly recommend private spaces, preventing crowding, and single offices to reduce stress from social stimulation and demands (ACAS, 2016;Group GSA, 2021;HOK, 2019;Scottish ADHD Coalition, 2018;TUC, 2014b). However, as pointed out by North (2021), there appears to be an inherent bias and stereotypical assumption that certain groups, particularly those with ASD, do not want or require any social interaction, leading to inadequate adjustments 'serve[ing] to increase stigma and isolation, leading to reduced opportunities which limit their career development and perpetuate their marginalisation' (North, 2021, p.16). Heuristic guidelines from the design industry addressed this by creating spaces that facilitate social interaction (Cassidy, 2018;Group GSA, 2021).

Environmental control
Over half the included studies suggest provision of environmental control across neurodiversity conditions by specific (e.g. private office) and unspecific (e.g. providing control) means. One study pointed out that for workers with ASD, control over the physical environment was more important than for neurotypical workers (Hayward et al., 2019). Adjustments for providing environmental control were indicatively related to occupational longevity and performance; fewer health/well-being relationships were found. Environmental control (actual or perceived) is a well-researched moderator on the relationship between environmental stressors (including social stimulation) (cf. Evans & Cohen, 1987). As such, it is no surprise that heuristic guidelines also suggest sensory stimulation control by providing activity-based working environments that offer choice of settings or zones that vary in their intensity of sensory stimulation (Cassidy, 2018;Group GSA, 2020HOK, 2019;Resource Architecture et al., 2018).

Other adjustments
Few studies listed sensory-unspecific adjustments, which we positioned as resources rather than stress reduction adjustments, such as adequate seating, treadmill desk or adequate desk size. Further, almost half of the studies indicated the value of a home office as a policy-and environment-related adjustment. Those were similarly related to occupational longevity and performance and health/well-being. Heuristic guidelines also listed sensory-unspecific adjustments, such as designing spaces that are intuitive to navigate through and which have good way-finding properties (Heasman et al., 2020;HOK, 2019;NDA, 2015;Standifer, 2009).

Considerations when implementing sensory adjustments
Several authors highlighted points of caution when implementing physical workplace adjustments, which we consider important for future research in the field and the academic discussion. These points included being cautious of one-size-fits-all approaches of using measures in isolation, of potential mismatch between employee and employer perspective/lack of education of employers, and unawareness of adjustment fatigue by taking an ableist perspective.
Firstly, Pierce (2018) argued that single facets do not universally apply to all (autistic) employees; thus, there is the need for adequate consultation of ND workers. Further, case-by-case consideration of adjustment should prevent compliance-based adjustment by employers taking a blanket approach. Such failure to consider individual need carries the risk of increasing stigma and perpetuating marginalisation, leaving NDs 'stuck in a deskilling cycle of underemployment' (North, 2021, p. 16). However, North (2021, p. 16) also points out that 'universal [inclusive] design [in addition to case-by-case adjustments] … may be regarded as too idealistic and vague for resource poor … organisations', thus hampering social commitment to employment of ND individuals and adjustment despite appropriate government policies.
Second, the risk of applying isolated measures, such as only adjusting the physical environment without addressing psychosocial factors or having appropriate workplace and personnel policies in place (Pierce, 2018), carries the risk of ineffective smoke smoke-and-mirrors inclusion efforts (e.g., Diener et al., 2020).
Third, studies indicated a mismatch between employee and employer perspectives, which can result in unrealistic expectations from employers once adjustments are in place (Diener et al., 2020). This highlights the need for 'training employer stakeholders (support personnel) to recognize various facets of disabilities and to provide training related to acceptable support and adjustments for individual employees' (Pierce, 2018, p. 225) and manage expectations.
Fourth, closely related to unrealistic expectations and the need for employer education is an ableist perspective from management and/or co-workers and resulting unawareness of adjustment fatigue (Soeker, 2020). Such accommodative fatigue has been described in physiological research as 'reduced performance of the accommodative system due to prolonged and/or repeated effort ' (Vilupuru et al., 2005, p. 191). This is summarised in this quote from Soeker's (2020) study: There's also the issue of adjustment fatigue from an autistic person's point of view … you can make a person who's wearing a prosthetic leg … walk up and down stairs, they can do it and they will do it well for a while but it takes more effort for them to do it than it does for someone with two normally functioning legs. So you have to recognize that it takes more effort to accommodate for that … if you do not put in a lift then at least give them a bit more time to walk up and down the stairs.
Education and the development of a psychologically safe environment is a necessity. Participants frequently revealed a desire for cultural flexibility, greater awareness and sensitivity from neurotypical colleagues, which was balanced with perceived risks of disclosure and stigma (Morris et al., 2015;North, 2021).
Overall, this review highlights a lack of robust evidence on the efficacy of physical workplace adjustments. This is a significant shortcoming of the field. Studies in this review point to indicative findings that broadly accord with existing guidance documents.

Limitations
The study has limitations relating to three key points: the search strategy (search strings and search terms); the strict inclusion/exclusion criteria and the outcome domain categorisation.
First, the combination of four search strings may have limited the search as studies were only retrieved that matched all four search strings. Further, the specification of possible types of workplace adjustments (e.g. 'Light') or their sensory target (e.g. 'Auditory'; see Table 1) by S4 may also have constituted a reductionist approach as studies were only retrieved that included one of the listed search terms within S4 (see Table 1). However, piloting without the S4 string elicited too many papers, which were logistically impossible to review given the resources available for this work. Furthermore, relevant papers may have been excluded if they used different wording from MeSH words or in the title, abstract or text. Specifically, there were concerns that the two search terms concerning olfactory and vestibular adjustments ('Air quality' and 'Lift*') might have limited the retrieval. However, post hoc trials that replaced the terms 'Air quality' and 'Lift*' with general terms 'olfactor*' and 'vestib*' did not render an improved search. Further, extensive handpicking aimed to retrieve any empirical evidence available.
Second, strict inclusion criteria excluded studies that only listed technological adjustments (excluded during screening); used adult and adolescent ND populations (excluded during screening); used populations that neither identified as ND via self-nor formal diagnosis (excluded during eligibility check) and non-empirical works (excluded during eligibility check).
Third, the outcome domain categorisation for occupational longevity, performance and health/well-being rested on the positioning of the evidence by the title authors. If authors were unclear or did not use any terminology that related to the outcome domains (e.g. health/wellbeing), the evidence was classified based on the review authors' interpretation of participant quotes (e.g. headache à health/well-being). As such, there is a risk that evidence might have been incorrectly classified. However, we have retained the title author's wording of the evidence in the manuscript and in the tables as much as possible to maintain transparency and minimise risk of wrong interpretations and categorisations. Further, including two researchers at each decision point aimed to reduce possible classification error.
Limitations recognised, we also note the contribution of our approach regarding the comprehensive database search, specific focus on sensory modalities, workplace setting, and physical adjustments, which address limitations in previous studies that lacked differentiation.

Implications for further research
Future studies will benefit from attending to four areas.
First, there is a need to develop a clear understanding of the sensory needs by ND condition/type, differentiating sensory modality and hypo-and hypersensitivity. Robust approaches should also control for potentially confounding individual characteristics and contextual conditions. These include self-diagnosis and gender. For this, we suggest the incorporation of Dunn's Sensory Process Model of sensitivity-environment interaction, which takes into account all sensory processing modalities and specifies processing patterns including seeking, avoiding, sensitivity, and low registration.
Second, we recognise that individuals with certain ND conditions, in particular ADHD and ASD, are found in the highest or lowest percentiles with respect to these sensory processing patterns (Dunn, 2001;Schreuer & Dorot, 2017). However, there is a need to develop contextsensitive outcome measures, which encompass employer and employees' perspectives. In other words, further research is needed to understand the impact of the sensory physical environment from multiple perspectives.
Third, we need to implement rigourous experimental designs to develop and test effective sensory interventions/adjustments to improve person and environment fit. This is a necessary baseline for the development of future evidence-based guidance.
Fourth, there is a need for a nuanced understanding of the factors influencing the implementation of adjustments employing process evaluation methodologies to enable a granular understanding of what works, for whom and under which circumstances. This review examined the extent, nature and quality of the current empirical evidence for physical workplace adjustment to support occupational longevity, performance and health/wellbeing in ND workers because of their specific sensory needs. Our results indicate that the research is generally not well-developed, is methodologically weak and therefore confined to offering indicative effects. The majority of studies addressed a mixture of work settings, a mixture of adjustments/barriers and facilitators, and focused predominantly on workers with ASD. This means that the effects of required sensory adjustments have been researched without due consideration of specific environments and inclusive sampling. Given the necessity of physical workplace adjustments to support inclusion and occupational outcomes for ND individuals, this review highlights the need for more methodologically sound research to inform evidence-based (design) guidelines for ND office environments.