A Systematic Review of Access to Rehabilitation for People with Disabilities in Low- and Middle-Income Countries

Rehabilitation seeks to optimize functioning of people with impairments and includes a range of specific health services—diagnosis, treatment, surgery, assistive devices, and therapy. Evidence on access to rehabilitation services for people with disabilities in low- and middle-income countries (LMICs) is limited. A systematic review was conducted to examine this in depth. In February 2017, six databases were searched for studies measuring access to rehabilitation among people with disabilities in LMICs. Eligible measures of access to rehabilitation included: use of assistive devices, use of specialist health services, and adherence to treatment. Two reviewers independently screened titles, abstracts, and full texts. Data was extracted by one reviewer and checked by a second. Of 13,048 screened studies, 77 were eligible for inclusion. These covered a broad geographic area. 17% of studies measured access to hearing-specific services; 22% vision-specific; 31% physical impairment-specific; and 44% measured access to mental impairment-specific services. A further 35% measured access to services for any disability. A diverse range of measures of disability and access were used across studies making comparability difficult. However, there was some evidence that access to rehabilitation is low among people with disabilities. No clear patterns were seen in access by equity measures such as age, locality, socioeconomic status, or country income group due to the limited number of studies measuring these indicators, and the range of measures used. Access to rehabilitation services was highly variable and poorly measured within the studies in the review, but generally shown to be low. Far better metrics are needed, including through clinical assessment, before we have a true appreciation of the population level need for and coverage of these services.


Introduction
The World Health Organization (WHO) estimates that over one billion people, or 15% of the global population, live with a disability, with 80% living in low-and middle-income countries (LMICs) [1]. Disability, defined by the International Classification of Functioning, Disability and Health (ICF), is an umbrella term for impairments, activity limitations, and participation restrictions [2]. People with disabilities experience an impairment (e.g., visual impairment) because of a health condition (e.g., glaucoma). Contextual factors, both at the individual (e.g., age, sex) and wider societal level (e.g., access to health services, attitudes towards disability), play a crucial role an individual's experience of the impairment.
People with disabilities often experience poorer levels of health than people without disabilities for various reasons [1]. By definition, people with disabilities have an underlying health condition

Eligibility Criteria
Studies were eligible if they met the following criteria: (1) quantitative research that included people with disabilities; (2) results reported access to rehabilitation for people with disabilities; and (3) research was undertaken in a LMIC as defined by the World Bank country classification 2017. No restrictions were placed on publication date, or language. Studies were excluded if the full text was not available after exhausting all possible sources. Duplicate reports from the same study were either combined if they reported different result or one was excluded if the results were the same.

Access to Rehabilitation Defined
For this review access was defined as use and coverage of services. Rehabilitation was defined in relation to the WHO definition as a "set of measures that assist individuals who experience or are likely to experience, disability to achieve and maintain optimal functioning in interaction with their environments" [1]. Using this definition, a broad range of interventions that may be required to maximize functioning were included: access to medical rehabilitation, access to therapy, coverage of assistive devices, and adherence to medication. Medical rehabilitation is defined as improving functioning through the diagnosis and treatment for health condition, reducing impairments and preventing or treating complications. Therapy is defined as restoring or compensating for loss of functioning, and preventing deterioration in functioning which may include physiotherapy, occupational therapy, and speech therapy. Assistive devices are defined as any equipment that is used to increase or maintain functional capabilities. We did not include studies measuring curative interventions, such as provision of spectacles, cataract surgery, hip replacement surgery, and similar treatments [12][13][14]. Whilst we recognize that rehabilitation extends beyond specialist health-related needs, this was beyond the scope of our review, which focused on health-related rehabilitation.

Types of Disability Measures
Studies defining disability using both the ICF definition (e.g., functioning, or activity limitations, and participation restrictions) and medical model definitions (i.e., specific impairments or disorders) were included.

Information Sources
Six databases (EMBASE, Global Health, CINAHL, Web of Science, MEDLINE, and PSYCINFO) were searched. The search strategy used key words for the following concepts: LMICs, people with disabilities, and access to health services. Terms were developed using MeSH or equivalent as well as from other reviews on similar topics. Boolean, truncation, and proximity operators were used to construct and combine searches for the key concepts as required for individual databases. An example of the search strategy is provided as Table S1. Systematic reviews identified through the search were reviewed for relevant included studies. If study protocols were identified, a search was made to determine whether the results of the study had been published. Furthermore, studies known to authors were included. No restrictions were made on language or time of publication.

Study Selection
All studies identified through the search process were exported to an EndNote database (version X7, Clarivate Analytics, Philadelphia, PA, USA) for removal of duplications and screening. Two reviewers (Tess Bright and Hannah Kuper) independently examined the titles, abstracts, and keywords of electronic records according to the eligibility criteria. Results were compared. The full texts were double screened (Tess Bright and Hannah Kuper) according to the eligibility criteria for final inclusion in the systematic review. Any disagreements in the selection of the full text for inclusion were resolved through discussion.

Data Collection Process
Data were extracted in to a Microsoft Excel database developed for the purposes of this review. The first author (Tess Bright) extracted all data and this was independently examined by a second reviewer to ensure accuracy (Sarah Wallace). Data were extracted on the following study components: • General study information, including author, year of publication • Study design, sampling, and recruitment methods • Study setting, and dates conducted • Population characteristics including age, sex, and sample size • Disability type/domain being studied, and means of assessing disability • Results: main findings related to access to rehabilitation and any disaggregation by age, sex, urban-rural status, or other variables. We extracted data on the proportion covered by rehabilitation services in the population. Where unmet need was presented, we calculated the met need as one minus the unmet need.
We conducted a narrative synthesis due to the variation in included study designs, measurement of disability and outcomes which made meta-analysis impossible.

Risk of Bias in Individual Studies
Quality assessments of all eligible studies were carried out independently by two reviewers (Tess Bright and Sarah Wallace). We evaluated studies based on a set of criteria according to the SIGN50 guidelines [15]. Table 1 outlines the criteria used to evaluate studies.

++
Low risk of bias: All or almost of the above criteria were fulfilled, and those that were not fulfilled were thought unlikely to alter the conclusions of the study + Medium risk of bias: Some of the above criteria were fulfilled, and those not fulfilled were thought unlikely to alter the conclusions of the study −− High risk of bias: Few or no criteria were fulfilled, and the conclusions of the study were thought likely or very likely to alter with their inclusion

Study Selection
8886 unique records were identified through electronic searches. 8609 studies were excluded during title and abstract screen, resulting in 278 for the full text screen. Following full text review, 201 studies were excluded, and the full text could not be identified for 14 articles ( Figure 1). Consequently, 77 studies were selected for inclusion and provided data for 106,462 people with disabilities across 64 countries.  Table 2 summarizes the characteristics of the studies eligible for inclusion. By region, most studies were conducted in sub-Saharan Africa (31%), followed by South Asia (18%), Latin America (16%), East Asia (16%), Middle East (9%), and Europe (3%). A further 8% were conducted in multiple countries. In terms of location, 49% were conducted in both urban and rural areas, with 18% in urban only and 13% in rural only (location unclear for 19% of studies). Most studies (73%) were conducted at subnational (e.g., district(s), or provincial level), with the remaining 27% carrying out national surveys. Over half of studies were conducted in 2010 or later (53%). The vast majority of studies were cross-sectional surveys (82%) with the remaining studies using cohort (5%), case control (10%) or retrospective longitudinal (3%) study designs. In terms of country income group, 33% of studies were conducted in low income, 28% in low-middle income, 29% in upper-middle income and 8% in countries of varying income levels.  Table 2 summarizes the characteristics of the studies eligible for inclusion. By region, most studies were conducted in sub-Saharan Africa (31%), followed by South Asia (18%), Latin America (16%), East Asia (16%), Middle East (9%), and Europe (3%). A further 8% were conducted in multiple countries. In terms of location, 49% were conducted in both urban and rural areas, with 18% in urban only and 13% in rural only (location unclear for 19% of studies). Most studies (73%) were conducted at subnational (e.g., district(s), or provincial level), with the remaining 27% carrying out national surveys. Over half of studies were conducted in 2010 or later (53%). The vast majority of studies were cross-sectional surveys (82%) with the remaining studies using cohort (5%), case control (10%) or retrospective longitudinal (3%) study designs. In terms of country income group, 33% of studies were conducted in low income, 28% in low-middle income, 29% in upper-middle income and 8% in countries of varying income levels.

Participants
Most studies included people of all ages (38%). 32% included adults only, 9% included older adults (>40 years), and 14% included children only (<18 years). In 6% of studies the age group was unclear. Considering disability domain, a large proportion of studies measured access outcomes related to mental impairment (44%), which we defined according to the International Classification of Diseases 10 (ICD10) "mental and behavioral disorders" included mental illnesses, intellectual impairment, and developmental delay. Epilepsy, although a neurological condition according to ICD10 was also grouped under mental impairment for simplicity. The remainder considered services related to hearing impairment (17%) visual impairment (22%), physical impairment (31%) or disability in general, across multiple domains (31%). The method of assessment of disability varied across studies, with 33 using self-reported measures (11 used the Washington Group short or extended set), 31 studies used clinical examination, four used a combination of reported and clinical measures, two used registry data, in two studies assessment methods were unclear, and the remaining three studies used alternative methods (e.g., community health worker report).

Outcome Types
Types of rehabilitation outcomes included: • In addition, data on barriers to accessing rehabilitation for people with disabilities were extracted as secondary outcomes in 23 studies (30%).

Description of Studies
Results of the 77 included studies are presented below by access to services specific to the following disability domains: hearing, mental health, physical, and visual. Where multiple domains were measured, and access outcomes were not disaggregated by domain, the results are presented in a separate section on rehabilitation for any disability.

Access to Rehabilitation for Hearing Impairment
In total, 13 studies measured access to hearing specific services in 12 LMIC countries, and four World Bank regions. The study populations used to assess access varied across studies, with the majority using population-based data; however, one sampled children from deaf schools, two from registries and one from a clinic. Most studies in this group (seven studies) were conducted among people of all ages. Five studies were conducted in children, and two among older adults. The method of assessment varied, with five using the Washington Group short or extended set, one using the WHO 'Ten Questions', three using a bespoke self-reported tool, two conducting clinical assessments, and the remaining two using other methods (registry, community health worker identification). The access results are thus not directly comparable. Results are outlined in Table 3. Overall, nine studies measured coverage of assistive devices, seven studies measured access to medical rehabilitation, and one measured adherence. Coverage of assistive devices ranged from 0-66% across studies. General rehabilitation coverage (i.e., access to hearing services) was between 3-62%. Finally, one study measured adherence/compliance with referral and estimated this to be 34%.
Across studies, no clear patterns of access were seen by country group, locality, or by age. Coverage of assistive devices tended to increase with country income group but was typically quite low. One national study by Malta et al. (2016) in Brazil measured association between locality (urban or rural) and access and found a higher proportion had assistive devices in urban areas compared to rural areas. In terms of the quality of the evidence across studies, most studies were judged to have low risk of bias (eight studies). Six studies were judged to have high or medium risk of bias due to small sample size (three studies), means of assessing disability unreliable (three studies), or poor response rate (two studies).

Access to Rehabilitation for Mental Impairment
In total, 34 studies measured access to specialist health services for people with mental impairments in 17 countries across six World Bank regions. Three studies were multi-country studies, for which it was possible to disaggregate results by country. For several countries, multiple studies were identified-three in China, three in Lebanon, four in Mexico, five in India, four in South Africa and four in Brazil. Considering age, the majority were conducted among adults (19 studies), among people of all ages, four among children, and one among older adults. Most studies sampled participants from the population (28 studies); the remaining sampled from schools (one study), clinic (three studies), or a variety of sources (two studies).
This category encompasses a broad range of conditions, from depression to intellectual impairment. Our search identified nine studies focusing on depression (or major depressive disorder), four studies on schizophrenia, three on epilepsy, five studies on psychiatric disorders, 14 measured general mental disorders with quite varied measures of assessment, two studies measured unspecified mental health conditions and the remaining two studies focused on intellectual impairment. In terms of method of assessment, a wide range of tools were used: five used a clinical diagnosis/examination, eight used the WHO composite international diagnostic interview, five used other validated questionnaires or tools (e.g., DSM-IV), two used the Washington Group short set, two used other validated self-reported tools, eight used bespoke self-reported tools (three of these combining with a clinical screen), one used household report, and one used global burden of disease data (see Table 4 for details).
In terms of outcomes, 28 measured access to medical rehabilitation, and five measured adherence to treatment. Access to medical rehabilitation for depression, which included treatment coverage and use of mental health services, most ranged from 0% for males in Mexico (subnational) to 54% in Brazil (national). El Sayed et al. (2015) found 65% of people with depression were in treatment across various LMIC using nationally representative data from the World Health Surveys. For schizophrenia, treatment coverage ranged from 50-71% in India (both subnational studies). Two multi-country studies were conducted, the first by Lora et al. (2012) found coverage of 11% (low income countries) to 31% (low-middle income countries) using the WHO Assessment Instrument for Mental Health Systems and the second by El Sayed et al. (2015) found coverage of 67% World Health Survey data. Coverage of epilepsy treatments ranged from 0% for older adults in Zimbabwe (subnational), to 52% among people of all ages in The Gambia (subnational). For children with intellectual disabilities coverage was higher: 73% in Ethiopia (subnational) and 87% in India (subnational) (two studies only). For other less specific conditions, coverage of medical rehabilitation ranged from 1% in China (national) (use of services, all ages) to 68% for adults in South Africa (subnational) (percent needing rehabilitation who received, all ages).
The broad range of conditions, source of participants, outcomes, and age groups mean that estimates within this group cannot be directly compared. However, it was clear that access for all outcomes was quite low across studies, except for children with intellectual impairments. There was considerable variation, even within studies conducted in the same country.
Across studies, no clear pattern was seen by country income level, locality or by age. One study by Lora et  In terms of the quality of the evidence, the vast majority of studies included in this group were judged to have low risk of bias (30 studies). Three studies had high or medium risk of bias due to small sample size (three studies), unclear or low response rate (four studies), or unreliable means of assessing disability (five studies).     3.5.3. Access to Rehabilitation for Physical Impairment Table 5 provides the results of 24 studies measuring access to rehabilitation for physical impairment. Studies were conducted across 17 countries and five World Bank regions. Types of physical impairments were varied, including rheumatoid or other arthritis (five studies), cerebral palsy (two studies), leprosy (two studies), difficulties walking (six studies), amputation (one study), musculoskeletal impairment (three studies), and unspecified physical impairment (eight studies). In terms of method of assessment, four used the Washington Group short or extended set questions (self-reported difficulties walking), eight used other self-reported tools, one used a chronic disorders checklist, five used a clinical diagnosis, four selected participants from a registry, one used community health worker report, and one study the method was unclear. Five studies were conducted among adults, 11 among people of all ages, six among children and in two studies the age group was not presented. Outcomes included access to physical therapy, assistive devices, medical rehabilitation, and adherence. The vast majority of studies were conducted on population-based samples; however, six sampled from clinic/hospital, and two from registries.
Access results for arthritis varied, with the highest coverage seen in Jordan (subnational) (76%) and lowest in India (subnational) (4%). Adherence to leprosy treatment was also quite high (71-75% in Nepal and Chad, both subnational studies); however, this may reflect the fact that these were both clinic-based studies. Results were more varied for less specific physical impairments such as "difficulties walking", musculoskeletal impairment, and physical impairment-with coverage of assistive devices ranging between 5-57% in Tanzania (subnational) and 41-93% in Cameroon (subnational) (depending on the type of assistive device). Coverage of medical rehabilitation in Brazil was 18%, while in South Africa this was 66%.
Coverage did not tend to increase with country income group or show a clear pattern by age or locality across studies. El Sayed et al. (2015) found higher coverage among those covered with insurance in a multi-country study [36].
Ten studies were judged to have low risk of bias. A further 14 studies were judged to have medium (ten studies) or high risk of bias (four studies) due to unclear or unreliable measure of disability or access (eight studies) or small sample size (four studies), or low response rate (three studies).

Access to Rehabilitation for Vision Impairment
In total, 17 studies measured access to rehabilitation for people with visual impairment across 13 countries in four World Bank regions. Table 6 outlines the results of these studies. The method of assessment varied across studies with seven using self-reported tools (of these four used Washington Group), seven using clinical examination, and three using other methods (registry, community leaders).
Thirteen studies measured medical rehabilitation, five studies measured access to assistive devices, and one study measured uptake of referral. Medical rehabilitation for people with visual impairment included consultation with specialist provider, and surgery uptake. All but two studies used a population-based sample. Access to medical rehabilitation was varied, from 5% among people of all ages in Brazil (national) to 82% among people of all ages in Nigeria (subnational). Similarly, results for assistive device coverage were highly variable, but typically low.
Across studies, a clear pattern was not observed by country income group, age, or urban-rural status. Higher coverage was identified for people with higher levels of education in several studies;  (Fletcher et al., 1999).
Considering the quality of studies in this category, 12 were judged as having low risk of bias. The remaining five studies had high or medium risk of bias due to low or unclear response rate (four studies), unclear measure of disability (two studies), or unclear measure of access (one study). Table 8 provides the results of 28 studies measuring access to rehabilitation for any disability (i.e., those studies that did not disaggregate by impairment type, or reported overall coverage results). These studies were conducted in 23 countries in six regions: the majority in sub-Saharan Africa (12 studies). Outcomes included access to assistive devices (18 studies), general rehabilitation (22 studies), and adherence (one study). Most studies sampled participants from the population, with one each using clinic or registry as a sampling frame. 21 studies measured disability using self-reported tools, including 12 using the Washington Group questions, two using the Rapid Assessment of Disability tool, and the remainder used bespoke tools. Four studies used a clinical examination. Two studies used registries to identify participants.

Access to Rehabilitation for Any Disability
Coverage of general rehabilitation varied across studies. Coverage was particularly low in India (subnational) and Bangladesh (subnational) at 5% and 7% respectively. In contrast studies in the Philippines, South Africa, Malaysia, and Brazil (all subnational studies) found higher coverage at 70%, 71%, 76%, and 80%. Substantial variation was also found for access to assistive devices, but generally coverage was low.
There did not appear to be a trend in coverage by country income group. The vast majority of these studies were conducted in both urban and rural areas and did not disaggregate results, thus examining patterns by locality was not possible. Furthermore, most studies were conducted among people of all ages, with no disaggregation of results by age group. Within studies, four studies examined coverage outcomes by indicators of equity. Three studies found lower coverage among females (Hosain et al. Considering the strength of evidence for access to any specialist services, eight studies were judged to have high or medium risk of bias, while the remaining were assessed as having low risk. The main risks were-unclear or unreliable measure of disability (five studies), or low or unclear response rate (five studies).

Barriers
Of the 77 included studies, 22 evaluated barriers to accessing rehabilitation as secondary outcomes. Commonly reported barriers included logistical factors (distance to service, lack or cost of transport), affordability (of services, treatment, lack of insurance), and knowledge and attitudinal factors (including perceived need, fear, and lack of awareness about the service) ( Table 9). Many of these barriers identified are not unique to disability. However, particular barriers were disability-related, including discrimination from the health provider, provider lacking skills, and communication barriers, or potentially enhanced among people with disabilities (e.g., lack of affordability). Table 9. Barriers to accessing rehabilitation reported across studies.

Review of Findings
This systematic review summarises the available evidence on access to rehabilitation services for hearing (13 studies), visual (17 studies), physical (24 studies) mental (34 studies), and any disability-related service (27 studies). The review captured studies a wide range of World Bank geographic regions, and over 60 countries.
Access results were varied across studies. Access to hearing specific services ranged from 0 to 66%. For visual impairment this was 0 to 82%, physical 0 to 93%, mental 0 to 97% and any disability-related services was 5 to 80%. Despite the variation, overall, access was low; however, there were some outlier studies showing high coverage. The review highlighted that outcomes used to measure access to rehabilitation, as well as measures of impairment/disability, are varied making comparisons and generalizability difficult. Coverage of services where disability is measured using self-reported tools such as the Washington Group short set of functioning, assumes that people who report difficulties are in need of rehabilitation. This may not be the most accurate measure of coverage (e.g., people blind from cataract may require surgery, not low vision aids) and further work is required to develop standard methods of measurement. Most studies used population-based, cross-sectional data, where the population in need in a particular region were identified (i.e., a prevalence study) and asked about access to services. However, we included studies where participants were sampled from clinics, or registries. These studies are very likely to overestimate coverage given these individuals have already been in touch with some type of service.
In terms of barriers to accessing rehabilitation, common themes across 22 studies in a diverse range of settings included lack of affordability of services, equipment, or medication as reasons for not accessing care. In addition, logistical or geographical factors such as distance to the service, transportation problems, and a lack of a chaperone. Several service-related barriers including discrimination from provider, communication barriers, and lack of provider skill were also common. These barriers may be specific to or greater for people with disabilities than those without disabilities. Further research is needed to examine particular barriers to access that people with disabilities face in greater depth.
The quality of included studies was generally high. There was limited evidence to support an association of coverage with country income group, age, urban-rural location, or other variables such as socioeconomic status. Included studies did not routinely disaggregate results by these variables-with less than a third of studies measuring variables related to equity of coverage.

Consistency with Previous Reviews
To our knowledge, this is the first systematic review that has attempted to summarize the available evidence on access to health-related rehabilitation for people with disabilities in LMIC. Thus, there are few similar examples from the literature to which the results can be compared.
Several previous reviews have focused on coverage of mental health services, evidence on assistive device coverage, and rehabilitation workforce literature. In a recent scoping review by Matter et al. (2017), authors identified a lack of publications on assistive devices from LMIC, in particular with respect to data on hearing, communication or cognition [96]. Similarly, a previous review by De Silva et al. (2014) on coverage of mental health programs highlighted that there was limited evidence on the topic [97]. They noted coverage estimations varied across studies, making comparisons difficult and called for coverage estimates to be stratified by age, gender, socioeconomic status to understand equity of coverage. These conclusions align with the findings of our review.
Jesus et al. (2017) conducted a review of rehabilitation workforce literature [98]. They found that substantial shortages of rehabilitation workers are documented in low income countries, particularly in sub-Saharan Africa and Latin America-with only six physicians specialized in rehabilitation in sub-Saharan Africa. Few programs exist for obtaining a qualification in rehabilitation, with several studies reporting alternative health worker cadres which could mitigate this; however, there is limited evidence on effectiveness. Although these findings have a health systems perspective on access to health services, they help to explain the reported low coverage of rehabilitation services in many studies in our review. Bruckner et al. (2010) also found that out of 58 LMIC involved in the WHO Assessment Instrument for Mental Health Systems surveys, that the vast majority did not meet expected health workforce targets for delivery of mental health services [99].
Several national surveys have been conducted in high-income countries such as the United Kingdom, the United States, and Korea. In the United States, a nationwide survey of people with cerebral palsy, multiple sclerosis, and spinal cord injury found that nearly one third of those who indicated a need did not receive assistive equipment every time it was needed. Over half of people had an unmet need for rehabilitative services [100]. In Korea, a 2009 nationally representative study (Korean National Health and Nutrition Examination Survey-KHANES) found that less than 10% of people with depressive mood had used mental health services [101]. In the United Kingdom, analysis of the European Health Interview Survey found that people with severe disability had higher odds of facing unmet need for health care, with the largest gap for mental health care [102]. Although these studies show high unmet need for services also exists in high-income contexts, access to rehabilitation is likely to be much poorer in LMIC.
The WHO have commonly cited statistics on coverage of assistive devices. For instance, it is estimated that hearing aid production meets less than 10% of the global need and less than 3% of people who need hearing aids in LMIC actually receive them. Furthermore, previous WHO estimates suggests that in many LMIC, 5-15% of people with disabilities have access to assistive devices [6]. Our review found wide variation in coverage of hearing aids and assistive devices but does agree that coverage is generally low. Again, the range of measurements of both disability and access limit comparability across studies.

Implications for Practice
This review has shown that in general, access to rehabilitation services is low in many LMIC. However, evidence is lacking from many countries of the world. To enable full implementation of the UNCRPD, member states must ensure that rehabilitation services are accessible to people with disabilities. Despite the UNCRPD providing a clear legal and regulatory framework, this review alongside key publications from the WHO, suggests that people with disabilities are not receiving a range of specific health services required to improve functioning. Evidence suggests that per capita income is linked to the level of implementation of the UNCRPD-underlining the major challenge for LMIC [103]. As outlined in the call to action in Rehabilitation 2030 there is an urgent need to address the unmet need for these services [5]. Although we have specifically focused on people with disabilities, rehabilitation has a broader scope, with some people needing rehabilitation temporarily at certain points in life (e.g., after a sports injury). Thus, addressing rehabilitation needs for people with disabilities has a wider benefit. Increasing life expectancy means the needs for rehabilitation will also increase, reinforcing the need to address this gap.
Rehabilitation should be integrated in to health systems at all levels to maximize access and achieve UHC. Rehabilitation in Health Systems guidance from the WHO provides recommendations for member states to strengthen and expand the availability of quality rehabilitation [104]. These, and other initiatives, include supply-side interventions, which attempt to address the dearth of services available to provide rehabilitation in LMIC. For instance, the GATE program of the WHO aims to improve access to affordable devices globally through various mechanisms [11]. Community-based models of health care delivery have been attempted for specific health services including: mental health, eye care, and ear and hearing care. These task shifting approaches are endorsed by the WHO as a mechanism to overcome skills shortages and reach underserved populations [105]. Telemedicine is a growing area for provision of rehabilitation and may help overcome the geographical barriers commonly reported in the literature. As an example, in the field of hearing impairment, telemedicine has been used for screening, diagnosis, and hearing aid fittings [106]. Furthermore, mobile technology has huge potential for improving access to rehabilitation. For example, in Kenya smartphone-based assistive technologies have been tested for students with visual impairment with positive impact on access to education, and participation in everyday life [107]. Sureshkumar et al. (2015) have tested a smartphone-based educational intervention for people with physical impairments following stroke in India [108].
Furthermore, demand-side interventions such as financial incentives and health promotion/education may help to improve uptake of available services. This includes strategies such as ensuring health insurance covers rehabilitation services, which will help to avoid catastrophic health expenditure. Two systematic reviews conducted by Bright et al. found that delivery of services at or close to home, text-message reminders, and vouchers may be beneficial for improving access to services for children in LMIC, but more evidence is needed on "what works" to improve access for people with disabilities [109,110].

Use Common Definitions of Disability and Coverage
To monitor progress towards the SDGs with respect to disability, and for program-planning purposes, key indicators of access to and coverage of rehabilitation should be developed, with a uniform method of measurement to allow comparability. This includes using clear definitions of what is meant by rehabilitation (e.g., medical rehabilitation, assistive technology, and therapy) and how coverage or access are measured. Access to health-related rehabilitation in this review was usually measured in terms of "coverage", that is the proportion of people needing a service who reported receiving it. However, this may overestimate coverage as the service may be inadequate and/or the full course of treatment may not be completed. Better measures of "access" are therefore needed. Furthermore, common definitions of disability should be adopted. Ideally, this should focus on clinical measurement of impairment, as these will also provide further information about the rehabilitation needs [111]. For instance, self-reported hearing difficulties does not give adequate information about service needs, which may range from basic wax removal to more complex surgeries or hearing aid fitting. Clinical assessment would provide the information needed to plan rehabilitation and specialist services. In addition, equity of service coverage should be assessed as part of any data collection to monitor access to rehabilitation. Sociodemographic information such as age, gender, socioeconomic status, locality, should be collected which can then allow data disaggregation. Monitoring the effectiveness and quality of rehabilitation care received is crucial for informing service delivery improvements, and ensuring functioning is maximized for people with disabilities.

Limitations and Strengths
This review has several limitations that need to be taken in to account. We focused on literature from peer-reviewed sources, and it is possible that some relevant data is available in grey literature sources, not captured in our search. Although we placed no restrictions on language, the electronic searches were conducted on six databases in the English language, and thus some literature may have been missed. Although our review encompassed a broad range of countries, and all the World Bank regions except for North America (high income), a third of studies came from sub-Saharan Africa. Our results may be slightly biased towards the conditions in these countries. However, the range of countries in sub-Saharan Africa included were limited to 15 of the 48 countries-suggesting that despite the largest proportion of data coming from this region, further research is required. Data was lacking from many parts of the world, with only 16% of included studies from Latin American countries, therefore included studies may not be representative of the level of access to rehabilitation in many LMICs. Studies may have been conducted in countries where stronger rehabilitation services exist, which may exaggerate the results found. The vast majority of studies were conducted at district level (73%), rather than national level, so making inferences about the situation of rehabilitation access in a whole country is limited. In the analysis we compared results by country income level (low, low-middle, and upper-middle). Ideally, a comparison between the results of studies by region (e.g., LMICs in Africa) would have been made, however the range of measurement types used limits comparability. Our review did not have a focus on the availability of services, which is an important dimension of access and may help to explain poor coverage of rehabilitation [112]. The scope of our review was on health-related rehabilitation and does not focus on broader needs such as education or work-related rehabilitation. We also did not include access to sign language education, rather than medical interventions for hearing impairment. Thus, we have not captured access to rehabilitation in its broadest sense as defined in Rehabilitation 2030. This warrants further attention. We did not assess the costs of accessing rehabilitation services, even though financial constraints were a major reason for not seeking care. Finally, we did not place any restrictions on publication date in our review, which means we have captured available literature to date; however, some studies may be outdated, and not reflective of the current level of access in the country studied.
There are also several strengths. This review was large, and adopted a systematic approach, following Cochrane guidelines. We used a comprehensive list of search terms to capture the literature available on this topic. It captured a broad range of disability types, and across a diverse range of countries and published in different languages.

Conclusions
This systematic review on access to rehabilitation for people with disabilities found wide variation in reported coverage across studies. In general, coverage appeared to be low for medical rehabilitation, assistive devices, therapy, and adherence. However, the review has identified a need to develop standard indicators for measuring coverage of rehabilitation to allow comparability. There is also a need to use comparable measures of disability. Common measures will contribute towards a greater understanding of the met and unmet needs for rehabilitation for people with disabilities and allow planning of appropriate services.  Funding: This research was funded by CBM International grant number ITCRZK1810. The funders had no role in the design of the study, data extraction, analysis, interpretation or writing of the report.