Correlates of Physical Activity among Adults with Sight Loss in High-Income-Countries: A Systematic Review

Background: Physical activity (PA) is essential for almost all facets of health; however, research suggests that PA levels among populations with sight loss are critically low. The aim of this review was to identify the correlates of PA among people with sight loss in high income countries, to inform future interventions and policies. Methods: MEDLINE, Web of Science, PsycINFO, SPORTDiscus, The British Journal of Visual Impairment, The Journal of Visual Impairment and grey literature were searched for studies which reported correlates of PA among adults with sight loss. The protocol is available from PROSPERO (CRD42020215596). Results: A total of 29 articles were eligible for review. Evidence from multiple studies reported that the vision impairment category, worse visual acuity, bilateral visual field loss, worse contrast sensitivity, those of the female gender, low self-efficacy, and environmental barriers were associated with lower levels of PA among populations with sight loss. Conclusions: Overall, correlates of PA among people with sight loss in high income countries are complex and vary across different population groups. Health professionals, eye care, and sight loss services should work together to identify people at risk of low PA, and provide a range of services and interventions to influence the modifiable factors that are associated with low PA.


Introduction
Physical activity (PA) is defined as 'any movement produced by skeletal muscles that requires energy expenditure-including activities undertaken while working, playing, carrying out household chores, travelling, and engaging in recreational pursuits' [1]. The World Health Organisation (WHO) recommends that adults should engage in at least 150 min of moderate PA or 75 min of vigorous PA, or an equivalent combination of both intensities throughout the week, these recommendations are the same for adults with disabilities [2]. Moderate PA can be defined as PA performed at 3-6 times the intensity of rest, whilst vigorous PA can be defined as PA performed at >6 times the intensity of rest and moderate-vigorous PA (MVPA) is PA performed at >3 the intensity of rest [2]. Regular and sustained participation in MVPA is beneficial for almost every aspect of physical

Study Selection
Studies were included if they were cross-sectional or longitudinal observational studies that met the following inclusion criteria: (1) included adults aged 18 and over, or the study performed a sub-group analysis of adults (age 18 and over). (2) Studies were conducted among participants with sight loss whilst wearing corrective devices, either measured by clinical assessment or self-reported sight loss. (3) Studies conducted in HICs (defined as countries identified by the World Bank as high income in 2020). (4) The dependent variable was a measure of PA engagement (e.g., how often and for how long a person goes to a sports club or walks for transport). (5) The significance level of the association between PA and an independent variable was examined, and a p-value was reported. The review excluded studies conducted among populations under the age of 18, as correlates of PA among people under the age of 18 with sight loss is the focus of another review [17]. In addition, the review focused on correlates among HICs because these countries were more likely to be similar in terms of economic and social characteristics, therefore the results could be synthesized to inform interventions in HICs.
The exclusion criteria were: (1) Studies not written in the English language. (2) Studies which examined the relationship between PA and risk factors for sight loss or eye disease e.g., the relationship between ocular perfusion pressure and PA. (3) Studies which compared PA between populations with sight loss and populations without sight loss and did not report the factors associated with PA within the population with sight loss. There were no further restrictions applied to the population of study i.e., studies were included regardless of the gender, ethnicity and reported comorbidities of the population group.
The data were independently extracted by one reviewer (R.K.L), and then discussed with a second reviewer (L.S). The data extracted included: age (years), the gender of participants, the sample size, the PA measurement tool, the vision measurement tool, the eye disease examined (if applicable), the statistical test used, the country in which the participants were recruited from, the confounders controlled for when examining correlates of PA, and the main findings.

Results
The database search yielded 2854 results; of these citations, 792 duplicates were removed. The grey literature was also searched, and no eligible studies were found. The primary reason no eligible studies were found in the grey literature search was that no studies reported the significance level of the variables which were associated with PA participation. A search of key terms related to physical activity in the British Journal of Visual Impairment, and The Journal of Visual Impairment yielded 112 and 71 articles, respectively.
Following the abstract and title screening, and the subsequent full text screening, 29 studies were retained for the final review [7, (Figure 1). The data from eligible studies was extracted and reported in Table 1. participation. A search of key terms related to physical activity in the British Journal of Visual Impairment, and The Journal of Visual Impairment yielded 112 and 71 articles, respectively.
Following the abstract and title screening, and the subsequent full text screening, 29 studies were retained for the final review [7, (Figure 1). The data from eligible studies was extracted and reported in Table 1.  The Critical Appraisal Skills Programme (CASP) checklist [46] was completed for each study and informed the final quality assessment; the results are presented in Table 2. Out of all eligible studies, 4/29 were considered to be of a high quality, 14/29 studies were of a medium quality, and 11/29 of the studies were considered to be of a low quality. Studies were more likely to be considered of a higher quality if they had minimised the risk of selection bias, had recruited a sample size which allowed for reliable conclusions to be made regarding the statistical significance of associations, and controlled for confounding factors, such as age and gender, in the analysis. Studies which relied on objective measures of PA and included a clinical assessment of vision parameters to categorise participants were also more likely to be considered of a higher quality than studies which used self-reported methods of PA and sight loss. Gender was the only significant predictor of MET minutes per week (β = 0.25, p < 0.05), with men reporting more MET minutes than women.       In multivariable models bilateral VF loss but not unilateral VF loss was associated with fewer daily steps (p < 0.05) and less MVPA (p < 0.01). Post-refraction VA was associated with fewer 15% daily steps (p = 0.04) and 36% less MVPA (p = 0.04). There was a significant negative correlation between objectively measured MVPA and worse VA, VF and CS. However, when self-reported MVPA and step count was examined, the associations were weaker and in the opposite direction, compared to when PA levels were objectively measured.

Measures of Vision
A range of different vision parameters and their association with MVPA were examined (Table 3), including self-reported VI classification (5/29), visual acuity in the better eye (3/29), visual field (5/29), contrast sensitivity (5/29) and colour vision (1/29). For studies which examined the association between self-reported VI classification and MVPA, there was evidence that being classified as blind was associated with lower levels of PA when compared to being classified as VI [21,31]. However, there was no evidence across studies of a dose-response interaction between the severity of sight loss, based on the international blind sports classification of VI (B1 vs. B2 vs. B3 vs. B4), and MVPA [24,27,42]. All of the studies which examined the relationship between MVPA and VI classification relied on self-reported MVPA measurement tools. In addition, it is possible that three of the studies may have selected the same participants for multiple studies, due to the similarities in the methods used in the recruitment of participants [21,24,25].  [40] In contrast, among studies which used objective measures of sight loss and of MVPA, there was stronger evidence from one high-quality study, two medium-quality studies and one low-quality study that visual field loss in the better eye, or in both eyes, was significantly associated with lower levels of MVPA [37,38,44,45] However, unilateral visual field loss was not associated with MVPA [44]. Another study used a principal component analysis with varimax rotation to establish three independent factors; the first factor, which loaded on to superior visual field measures, and the second factor, which loaded on to inferior visual field measures, were also not associated with self-reported MVPA [19]. One study reported that the binocular visual field was associated with MVPA when MVPA was objectively measured, but not when MVPA was self-reported in the same group of participants [45].
The most frequently studied vision parameters that were associated with walking were visual acuity in the better eye (3/29), visual field (3/29) and contrast sensitivity (3/29) ( Table 4). Worse measures of visual field in the better eye [38], bilateral visual field [44] and integrated visual field (IVF) sensitivity, which was defined as a summary of the average overall and inferior field sensitivities [39], with a lower IVF representing a worse visual field, were associated with lower levels of walking. Visual field (worse) [39] (Integrated visual field sensitivity) [38] (visual field loss in better eye) [45] [45] (unilateral visual field loss was not associated with less steps) [44] Vision parameters [41] Glaucoma (present) [38] Glaucoma (severe) (present) [38] AMD (present) [40] Stage of AMD [43] Sig. Cataract/PCO [40] * Participants classified as B2 spent significantly more minutes of walking than participants classified as B1. ** Strakoff, B.E. (2017) reported a significant difference in walking between participants who self-reported VI categories B1 vs. B2 vs. B3 vs. B4. The mean min/day of participants classified as B1 was 46.8 min, com-pared to 95.8 min for participants in B2. However, there was no dose response identified as participants in the B3 category engaged in a mean of 62.6 min per day of walking. *** In regression analysis self-reported VA diagnosis, was a discriminating factor in walking for those over the age of 77 with a breaking point for those under and over 87.5 years.
Studies which explored the association between visual acuity in the better eye and walking were all considered medium-quality studies and reported mixed results. One study, which used objective measures of walking, reported a significant negative relationship between visual acuity and daily steps taken [40]. In contrast, two studies which examined walking reported different outcomes which were dependent on the nature of the walking, or the group, examined. One study that reported walking was not associated with visual acuity in the better eye when the weekly time walked was examined, however, there was a negative association between visual acuity in the better eye and daily stairs taken [43]. Another study reported that the self-reported visual acuity status was a discriminating factor in walking distance for those over the age of 77, with a breaking point for those under and over 87.5 years [36].
The studies in the review reported mixed results for the association between PA and contrast sensitivity. Worse contrast sensitivity was found to be associated with lower levels of objectively measured MVPA in two studies, [40,45] and self-reported MVPA in one study [19]. However, contrast sensitivity was also found to have no association with objectively measured MVPA in two studies [37,38], and self-reported MVPA in one study [45]. For walking, there was no significant association found in any of the three studies between contrast sensitivity and objectively measured walking [38,40,45].

Personal Correlates
In terms of the variables classified as non-modifiable personal correlates, twelve variables were tested for their association with MVPA (Table 5), and six variables were tested for their association with walking ( Table 6). The most frequently reported correlation was between PA and gender. Overall, three studies reported that male participants selfreported engaging in significantly more MVPA than female participants [23,27,42], and five studies reported that there was no association between self-reported MVPA and gender [21,22,24,29,32]. One low-quality study examined the relationship between objectively measured MVPA and gender and reported no association between gender and MVPA [35]. For walking, two low-quality studies, one that relied on self-reported walking and the other that relied on objectively measured walking, both reported no association between walking and gender [35,42].  Older age was not found to be associated with self-reported MVPA in any of the four studies which examined MVPA [21,22,24,29]. One study reported that older age was associated with participants who self-reported less walking, however, the influence of age on walking was dependent on the sub-group examined. Among younger participants with poorer sight and who were more active, age was reported to have a greater impact on walking than sight loss [36]. No studies used objective measures of MVPA or walking to examine the association between PA and age.
Comorbidities, BMI, health related quality of life, depression, and level of independence were also included as non-modifiable personal correlates. These factors could have a bidirectional relationship with PA, which could also make them modifiable correlates. For example, improvements in depressive symptoms could help an individual feel more energised, and thus they are more likely to engage in PA.

Personal Correlates
A range of psychosocial factors and their association with MVPA were examined, including self-efficacy [24,25] social support [21], self-regulation [21], the perceived barriers to PA [33], the theory of planned behaviour constructs (attitude towards PA, subjective norm, perceived behaviour control, intention to engage in PA) [22], use of a mobility aid [20], and levels of self-reported independence [23] (Table 7). Whilst only two studies, which were both conducted in similar cohorts, explored the relationship between MVPA and self-efficacy, the results from other studies provided evidence of factors which may also influence self-efficacy, defined as 'the belief an individual has in their ability to perform a task and to obtain the desired results'. For example, a fear of falling was found to mediate the relationship between sight loss and PA in one study that was included in our review [37]. Perceived barriers to PA [33], and a lower self-reported level of independence [23] also had a negative and significant association with MVPA; it is plausible that these factors would influence an individual's self-efficacy for PA. Social support was also reported to be positively associated with MVPA in one low-quality study [21], and peer/buddy support was also found to have a significant positive association with sports participation in a medium-quality study [30]. No studies explored psychosocial factors that were associated with walking.

Environmental Correlates
When compared to the non-modifiable factors that were associated with MVPA, the modifiable environmental factors were less researched (Table 8). One low-quality study reported a negative correlation coefficient between the logit of the perceived barriers to PA and self-reported PA; included among the most severe barriers to PA which were cited by participants was the environmental barrier: 'lack of transportation to get to places to exercise' [33]. One medium-quality study found a positive association between access to services and non-walking PA, and a negative association between physical barriers to walking and non-walking PA [18]. [18] Physical barriers to walking (1 unit increase) [18] Three studies examined environmental variables and their association with walking (Table 9). There was evidence from one medium-quality study, which used self-reported measures of walking, that the number of years lived at the same address was associated with increased walking, whilst feeling unsafe while walking around the neighbourhood was associated with less walking [36]. Another medium-quality study, which relied on selfreported walking measures, found no association between walking and neighbourhood aesthetics [18]. Only one low-quality study used objective measures of walking and found that there was no association between day of the week and walking [28]. Feeling of safety when walking in the neighbourhood (worse) [36] Years lived at the same address (i.e., neighbourhood familiarity) [36] Neighbourhood aesthetics [18] Due to the variations between PA measurement tools and study design, it was not appropriate to conduct a meta-analysis for this review.

Discussion
Our review aimed to identify modifiable and non-modifiable correlates of PA among people with sight loss. Evidence from multiple studies reported that the VI category, worse visual acuity, bilateral visual field loss, worse contrast sensitivity, individuals of female gender, lower self-efficacy, and environmental barriers were associated with lower levels of PA among populations with sight loss.
Visual field and visual acuity are common measurements taken during a routine eye examination that is carried out by an optometrist or ophthalmologist. The measures of visual field and visual acuity are also used to classify people as blind or vision impaired. Our findings that worse visual acuity, bilateral visual field loss, and being classified as blind versus visually impaired could be associated with lower PA highlight that optometrists or ophthalmologists may be important for identifying populations at risk of low PA.
Optometrists and ophthalmologists could therefore work with community groups and low vision services, to refer populations identified as being at risk of low PA to PA opportunities and support services. However, the vision parameters did not fully explain the variances in PA. Our review found additional non-modifiable and modifiable correlates of PA, which may have important implications for future PA interventions.
In terms of non-modifiable personal factors which correlate with PA, we found evidence that male participants were more likely to engage in higher levels of PA than female participants. These findings are in line with previous research in sighted populations. Globally, and particularly in HIC Western countries, men are, on average, more physically active than women [47]. However, one study in our review found that the relationship between gender and PA was mediated by social support and self-regulation [21]. Although these findings are limited to one study, which had a small sample size, the results highlight the importance of understanding the mechanisms which may lead to gender differences in PA. An intervention aimed at targeting low PA among women may not be effective if it does not target the mechanisms which result in women engaging in less PA than men.
Mixed results were found regarding the relationship between age and PA among people with sight loss. In sighted populations, research has consistently reported lower levels of MVPA that are associated with older age [48]. It is possible that because sight loss is associated with older age, the studies did not have a large enough sample of young adults with sight loss to be able to identify a negative association between older age and PA levels. Given that all age groups were identified as being at a possible risk of low PA, the findings highlight that multiple age groups could benefit from being targeted in PA interventions.
Importantly, our review also highlighted several areas where there is a lack of research. Firstly, there was limited research that explored the association between ethnicity and PA levels among people with sight loss. Among sighted populations, there is a large body of evidence that PA varies between ethnic groups within the UK [49,50] It is important to understand the PA differences between ethnic groups, to ensure that PA interventions do not compound the existing ethnic inequalities regarding the access to sight loss services [51]. In addition, there was a lack of research that explored the association between additional disabilities, as well as sight loss, and PA. Research suggests that additional disabilities are common among people with sight loss. For example, one in three people in the UK with a learning disability is estimated to be affected by a sight problem [52]. It is plausible that having multiple disabilities, including sight loss, may limit an individual's opportunities and ability to engage in PA, thus increasing that individual's risk of low PA. Therefore, it is important that interventions can be adapted to accommodate for people with additional disabilities, as well as sight loss. Further to this, there was a lack of research that explored the association between mobility measures and PA levels in populations with sight loss. Mobility measures may be modifiable by PA interventions, as research suggests PA programs could attenuate differences in gait and functional parameters between populations with sight loss and sighted populations [53]. Improvements in mobility among VI populations could also improve mental health outcomes, as previous research has found gait speed to be a significant predictor of depressive symptoms in VI populations [54]. Therefore, there is a need to understand the bidirectional relationship between PA and mobility measures within populations with sight loss, to determine how mobility measures should be targeted in an intervention to increase PA, and to determine target population groups (e.g., individuals with a slower gait speed).
The review also identified a range of psychosocial factors that are associated with low PA, including lower levels of social support, self-efficacy, intention to engage in PA, perceived barriers to PA, and lower levels of self-reported independence. There are a range of interventions which could be used to target these psychosocial factors; for example, group-based PA may encourage social support, whilst PA that is prescribed by a health professional could promote an intention to engage in PA. In addition, sight loss services may be able to support people in becoming more independent, and reduce the impact of barriers to PA, by supporting people in maximising their residual vision and improve daily functioning.
When compared to non-modifiable variables that were associated with PA, modifiable factors that were associated with the environment were less researched. Our review reported that an access to services, physical barriers, fears of safety, and perceived barriers, including 'lack of transportation to get to places to exercise', as well as familiarity with the neighbourhood, were associated with lower PA among people with sight loss. These could be considered barriers to PA which could be addressed by sight loss services; for example, orientation and mobility training could reduce individuals' fears of safety in the neighbourhood. However, it is important that policies and planning also ensure that environments are designed to be accessible, and that the interventions to promote PA should consider environmental changes which can facilitate PA, as well as individual support to increase PA. Future research should also explore accessibility and mobility barriers which exist in low-income countries. It is plausible that the environmental barriers to PA that were identified in HIC countries in this review may be more pronounced in low-income countries, due to a lack of investment in public transport and infrastructure which make the streets more accessible, such as tactile paving, signal controlled pedestrian crossings, maintained pavements and detectable kerbs that separate traffic from pedestrians.
However, there were limitations to this review. Firstly, there were a limited number of studies which used objective measures of both sight loss and PA, which are considered more reliable and valid tools than self-reporting measurement tools. One study reported that in a group of patients with AMD, lower MVPA measured by objective tools was associated with worse visual acuity, visual field and contrast sensitivity, whereas when the same participants' self-reported measures of PA were examined, the study did not report a significant negative correlation [45]. Although this study included a small sample size, the findings indicated the influence that a measurement tool can have on the outcome results. In addition, sight loss was defined and measured using a range of methods, and the studies used different criteria to define visual impairment and blindness, thus limiting the comparability of results. Future research that is conducted among populations with sight loss should adopt a standard definition of sight loss, to ensure the future comparability between studies. We suggest that sight loss can be defined as a self-reported 'sight loss whilst wearing corrective devices', however, a follow-up, standardised eye test should be used, if available, to describe the degree and type of sight loss of the population being studied. In the absence of eye testing equipment or expertise, then follow-up questions should be asked to understand the nature of the participants sight loss. Although some studies objectively measured vision and PA using validated tools, these studies were often limited by a smaller sample size than those which relied on self-reported measures. There were also limitations among studies as a result of the recruitment procedures. Studies which recruited participants via online channels that were distributed by a VI organisation risked selection bias, which resulted in younger and more active participants than the overall population of people living with sight loss. In addition, people with additional disabilities, in particular, cognitive impairments such as dementia and learning difficulties, may be excluded from studies if they are unable to understand the PA survey or provide informed consent. Therefore, the correlates of PA may not be representative of the barriers experienced by populations with the lowest levels of PA.

Conclusions
Overall, our findings have highlighted the complexities of the factors which are associated with PA behaviour among people with sight loss in HICs. Optometrists and ophthalmologists are well positioned to identify patients with sight loss who may be at risk of low PA, and collaborate with sight loss services (e.g., charities, community groups, and council services) to refer people to PA advice and support. In addition, people working in the delivery of sight loss services may be able to support people by addressing the barriers to PA, and promoting greater independence, which could facilitate PA. However, PA is complex and our review highlighted the need for PA interventions to meet the needs of a range of population groups with sight loss. We suggest that future research aims to understand how different sectors and services could identify people at risk of low PA, and work together to provide individualised support to promote PA.

•
Worse visual acuity and visual field may indicate that an individual is at risk of lower PA [29,[36][37][38]40,[43][44][45]. These measures are examined in routine eye tests, thus optometrists and ophthalmologists could identify people at risk of low PA, and play a key role in referring people to PA groups and opportunities.

•
It is important to understand how factors including gender, age, ethnicity, and additional disabilities influence PA in the context in which interventions are being delivered. Interventions should work with communities to understand local needs, develop appropriate interventions, and target different sociodemographic groups, when appropriate [55,56].

•
Interventions should consider the environmental factors, such as unsafe streets [41], and a limited access to services [18] which influence PA and make adjustments to minimise these barriers.

•
Future studies, with larger, representative sample sizes, and objectively measured PA, are required, to explore the findings in studies which are currently limited to a small evidence base.

Conflicts of Interest:
The authors declare no conflict of interest.

Appendix A
("activities of daily living" OR "physical activity" OR "lifestyle activity" OR "inactive" OR "insufficientactivity" OR "mobility" OR "incidental activity" OR "walking" OR "active transport" OR "non exercise activity thermogenesis" OR "lipa" OR "lpa" OR "neat" OR "light exercise" OR "moderate exercise" OR "nepa" OR "sport" OR "exercise" OR "walking") AND (Barrier* OR facilitator* OR modifier* OR motivator* OR influences OR uptake OR engagement OR correlate* OR encourage OR obstacle OR prevents OR participation OR predictor OR mediator OR moderator) AND ("Visually impaired" OR "Sight loss" OR "Sight impairment" OR "macular degeneration" OR AMD OR "Uncorrected refractive error" OR "Glaucoma" OR "Diabetes retinopathy" OR "Eye impairment*" OR "Visual impairment" OR "Vision disorder *" OR "Vision impaired" OR "low vision" OR "ocular pathology").