The association between disability and all-cause mortality in low and middle-income countries: a systematic review and meta-analysis
1International Centre for Evidence in Disability, London School of Hygiene & Tropical Medicine, Keppel st, WC1E7HT, London, UK
2Division of Physiotherapy, Department of Health and Rehabilitation Sciences, Stellenbosch University, Cape Town, South Africa
*corresponding author, tracey.smythe@lshtm.ac.ukAbstract
Background
There are at least one billion people with disabilities globally. On average they have poorer health, yet worse healthcare access. We aimed to systematically review the association between disability and mortality in low- and middle-income countries (LMICs).
Methods
We searched MEDLINE, Global Health, PsycINFO and EMBASE from 1st January 1990 to 14th November 2022. We included any longitudinal epidemiological study in any language with a comparator group that measured the association between disability and all-cause mortality in people of any age. Two reviewers independently assessed study eligibility, extracted data, and assessed risk of bias. We used a random-effects meta-analysis to calculate the pooled hazard ratio (HR) for all- cause mortality by disability status. We then conducted meta-analyses separately for different impairment and age groups.
Findings
We identified 6146 unique articles, of which 70 studies (81 cohorts) were included in the systematic review, from 22 countries. There was variability in the methods used to assess and report disability, as well as mortality. The meta-analysis included 53 studies, representing 62 cohorts (comprising 267,415people with disabilities). Pooled HRs for all-cause mortality were 2.06 (95%CI 1.80 - 2.34) for people with disabilities versus those without disabilities, with high heterogeneity between studies (τ²=0·24, I²=98%). This association varied by impairment type; from 1.32 (95%CI 1.13 – 1.55) for visual impairment to 3.39 (95%CI 0.90 – 12.76) for multiple impairments. The association was highest for children under 18 (4.46, 95%CI 3.01–6.59); and lower in people aged 15 – 49 (3·53, 95%CI 1·29–9·66); and older people over 60 years (1·97, 95%CI 1·63–2.38).
Conclusion
Disability increases the risk of all-cause mortality in LMICs, particularly in childhood. Interventions are needed to improve health of people with disabilities and reduce their risk of death.
Key messages
What is already known on this topic
Globally, people with different impairments have a higher risk of death than those without disabilities and many deaths will be avoidable.
What this study adds
People with disabilities in LMIC have two-fold higher mortality rates that those without disabilities. Disability is associated with a higher hazard of age-adjusted all-cause mortality across diverse LMIC settings and populations as well as impairment types. The risk of dying early is highest for children with disabilities.
How this study might affect research, practice or policy
Improved understanding of the association between disability and mortality will help to inform public health planning and policy, and the allocation of limited health-care resources to optimise healthy longevity for all populations worldwide. Without a focus on disability it may be difficult to reach SDG3 and other key global health targets.
Article notes
Competing Interest Statement
The authors have declared no competing interest.
Funding Statement
This project is funded by the National Institute for Health Research (NIHR) [NIHR Global Research Professorship (Grant Reference Number NIHR301621)] awarded to Prof. Hannah Kuper. Hannah Kuper has also received support from the Programme for Evidence to Inform Disability Action (PENDA) grant from FCDO
Introduction
There are at least one billion people with disabilities globally, more than 80% of whom live in low and middle-income countries (LMICs) (1). People with disabilities include those who have long-term physical, mental, intellectual or sensory impairments which, in interaction with various barriers, may hinder their full and effective participation in society on an equal basis with others (2). Implicit in this conceptualisation are different pathways linking disability to premature mortality. People with disabilities, by definition, have an impairment and underlying health condition, which may increase their risk of death. For instance, people blind from diabetic retinopathy will still experience diabetes, which is linked to stroke and cancer (3, 4) . The impairment may also lead to further health conditions, linked to mortality (e.g. pressure sores or respiratory disease among people with physical impairment). Additionally, people with disabilities face a range of exclusions, so that they are more likely to live in marginalised and poor positions in society (5), which are linked to higher mortality (6). They also frequently face widespread difficulties and barriers in accessing healthcare, including inaccessible transport and facilities, poor skills of healthcare providers around disability, and high costs (7, 8). Consequently, people with disabilities have worse health than others in the population (9), are at higher risk of morbidity and mortality (1).
Many deaths will be avoidable. For instance, the confidential inquiry into premature deaths of people with intellectual disabilities in the UK estimated that 37% of deaths were avoidable from causes amenable to change by good quality healthcare (10). Data from the USA showed that people with disabilities were diagnosed with cancer at similar or earlier stages, but experienced higher rates of cancer-related mortality, implying health system failures in care (11). An improved understanding of the association between disability and mortality will help to inform public health planning and policy, and the allocation of limited health-care resources. Different systematic reviews have shown the relationship between specific impairment types (e.g. hearing (12), vision (13), intellectual (14, 15)) and mortality, but not disability overall. Furthermore, many of the studies included in these reviews are from High Income Settings, although 80% of people with disabilities live in LMICs. Age and impairment type are strong common risk factors for both disability and mortality (16, 17) and the association between disability and mortality may vary by these factors, yet this has not been explored to date. A comprehensive systematic review of disability and mortality in LMICs is therefore needed.
We undertook a systematic review and meta-analysis on the association between disability and age-adjusted all-cause mortality in all ages in LMIC. This systematic review and meta-analysis aimed to answer the following questions:
- 1.What is the extent, strength and quality of the published evidence of the association of disability and risk of all-cause mortality in LMIC?
- 2.To what degree does disability affect the risk of all-cause mortality, and does this risk vary based on age and impairment type in LMIC?
Methods
Search strategy and selection criteria
We developed a protocol for this systematic review and meta-analysis that was registered with Prospero (ref CRD42022302557) (18) and followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (supplementary file 1). We systematically searched MEDLINE, Global Health, PsycINFO and EMBASE from 1st January 1990 to 14th November 2022 to identify available evidence in any language on disability and mortality in LMIC, including estimation of mortality rates and all-cause mortality.
Disability was conceptualised as arising from a combination of impairment and contextual barriers, as described in the following definitions: (i) UN Convention on the Rights of Persons with Disabilities (2006) (2) - People with disabilities include those who have long-term physical, mental, intellectual or sensory impairments which in interaction with various barriers may hinder their full and effective participation in society on an equal basis with others; (ii) World Report on Disability (2011) (1) - [Disability is] an umbrella term for impairments, activity limitations and participation restrictions, denoting the negative aspects of the interaction between an individual (with a health condition) and that individual’s contextual factors (environmental and personal factors). Search terms are outlined in supplementary file 2.
We included prospective and retrospective epidemiological studies (cohorts and randomised controlled trials) that measured the association of disability and all- cause mortality in any age. Conference abstracts were included if sufficient information was included for data extraction. There was no limitation of follow up period.
Inclusion criteria
- 1.Disability as defined in line with the International Classification of Functioning, Disability and Health Model and the United Nations Convention on the Rights of Persons with Disabilities (19). It includes people with specific conditions deemed likely to result in disability (e.g., spina bifida, schizophrenia), specific impairments (e.g. visual, hearing, physical) as well as disability measured through functioning/activity limitations/self-report (e.g., Washington Group questions, activities of daily living).
- 2.People of any age and sex, and any type of disability
- 3.LMIC, as defined by World Bank July 2021
- 4.Studies with a comparator group or that compared effect of disability on mortality to population statistics in a country as a whole (e.g. Standardised Mortality Ratio – SMR)
- 5.The outcome was all-cause mortality after baseline assessment of disability
Studies where all participants had a specific systemic disease (e.g. diabetes) were excluded due to issues with generalisability. Certain measures of disability were not eligible, including: 1) mild functional impairment (e.g. visual acuity ≥ 20/40, cognition as measured by MMSE ≥ 21, Washington group questions that reported ‘some difficulty’ or less, only one functional limitation/ limitation in activities of daily life (ADL)), 2) depression alone, 3) frailty alone, 4) disability as a continuous measure. Studies including both dementia and Alzheimer’s disease only had data extracted for dementia as Alzheimer’s was classified as a disease. Studies were excluded if there was no comparator group of people without disabilities or if they reported cause- specific mortality only (e.g. congenital heart disease, covid-19).
The outcome was all-cause mortality after baseline assessment of disability. Mortality could be reported using different measures of effect size and we included all measures. We also included cause of death and death rate.
Data search and selection
The electronic search was conducted by TS. Results were screened and stored in Rayyan. Study selection was conducted independently by TS and HK. Each reviewer independently and systematically screened all titles and abstracts against the eligibility criteria. From these preliminary lists, the reviewers independently assessed the full text of each article for inclusion. The reason for exclusion was recorded and discrepancies discussed at each stage.
Data extraction
To chart the data, we developed an extraction tool in Excel to systematically record information from included studies. The extraction form was piloted on four included studies to inform the final version. Each included study was charted by one reviewer and checked by the second reviewer.
Extracted information included: We extracted the crude estimates, and when estimates were reported with more than one level of adjustment, we extracted two additional estimates: (1) Minimally adjusted – usually the age and sex-adjusted estimate; and (2) Maximally adjusted - the greatest number of additional covariates, to elucidate potential pathways of the association of disability and mortality. When data on multiple cohorts were presented in a single publication, each cohort was separately eligible for inclusion. Where results from a study evaluated the association of disability and mortality over different population categories (e.g. non frail and frail people with disabilities) a weighted average of the effect measure was calculated for the entire cohort.
- 1.Publication characteristics: author, title, year of publication, setting/country
- 2.Study design: study design, sample size
- 3.Participant characteristics: age, sex, disability measure, and any other relevant descriptive data
- 4.Outcomes: Effect size for mortality outcomes (e.g. hazard ratio, cumulative incidence ratio, death rate, odds ratio) and cause of death.
Risk of bias assessment
The full texts of all eligible studies were assessed independently by both reviewers against quality assessment criteria adapted from Lund et al (20). The studies were evaluated for methodological quality and appropriateness for inclusion, without consideration of their results, based on a set of pre-determined criteria (Table 1). Disagreements on risk of bias ratings were resolved through discussion.
Data Analysis
Meta-analyses were performed using a random-effects model to generate a pooled effect estimate, reported as the hazard ratio (HR) with 95% CIs for the association between disability and age-adjusted, all-cause mortality. Between-study heterogeneity was assessed with I² and τ² statistics. We included age-adjusted estimates of the effect of disability on mortality risk in the meta-analysis since age is a strong confounder for the association of mortality and disability. Where only multivariable adjusted estimates were available, these were used. Studies that did not adjust for age were therefore not included in the meta-analysis; studies that reported odds ratios were not included in the meta-analysis as the estimates could not be pooled with HRs. There is the potential for effect modification of the association by impairment type and age and therefore we undertook separate meta-analyses for different age groups (<=18 years, 15 years – 49 years and >=60 years) and impairment types (cognitive, functioning, hearing, neurological, multiple, physical, psychosocial and visual). We assessed the risk of publication bias using Egger’s test (threshold for significance p<0·05) and by inspection of funnel plots. A meta- regression was performed to estimate the between study variance and to calculate uncertainty in the estimated overall effect size risk of bias. We performed all statistical analyses using RevMan 5.4.1 (Cochrane collaboration).
Patient and public involvement
Patients were not involved in this research.
Results
We identified 10,547 articles through electronic database searches. After removing 4,401 duplicate references, we screened the titles and abstracts of 6,146 articles. Of these, we identified 154 articles for full-text review. Eighty-four studies were excluded, leaving 70 studies that met the inclusion criteria (fig 1). Sixty-two cohorts from 53 studies were eligible for inclusion in the meta-analysis.
The characteristics of the studies are reported in table 2. Geographically, the largest number of studies were conducted in the Western Pacific Region (n=33, 44%; China n= 32, Hong Kong n = 1) followed by the Region of the Americas (n=22, 29%; Brazil n=15, Cuba n=1, Dominican republic n=1, Mexico n=2, Peru n=2, Venezuela n=1), and the Africa Region (n=17, 22%; Cameroon n=1, Congo n=1, Ethiopia n=2, Kenya n=2, Malawi n=2, Nigeria n=2, South Africa n=1, Tanzania n=2, Togo n=1, Uganda n=1, Zimbabwe n=1, Zambia n=1). There were no studies from the Eastern Mediterranean region. The South-East Asian region was only represented by India (n=3), and the European region by Turkey (n=1). There is an increase in the number of studies collecting data on disability and death since 2010. The duration of follow up in included studies ranged from 28 days to 17 years. Sample sizes ranged from 14 (people with neurological impairment in China) (21) to 749,720 (people with neurological impairment in Brazil)(22).
Studies used a wide variety of measures to assess disability (Table 3). Fifty-nine studies (62 cohorts) used clinical assessments and 14 studies (19 cohorts) used self- report to measure disability. The most commonly used assessment for visual impairment was visual acuity testing with a standardised chart (23), and the mini- mental state examination (MMSE) questionnaire (24) was most commonly used to determine cognitive impairment. The Katz Index of Independence in Activities of Daily Living (ADL) (25) was the tool used most to evaluate functioning.
Studies used various strategies to assess mortality, and were included regardless of the methods used because official death registries might not have been available or provided high-quality data in many LMIC (26). Most cohorts (n=41) were followed up by methods such as verbal autopsy, interviews with key informants, or both, with some studies (n=18) searching official vital records. Table 3 summarises the characteristics of included studies.
All 70 studies (comprising 81 cohorts) calculated the association of mortality in participants with disabilities compared to those without disabilities (Table 4). There was high variability in effect estimates used for the association of disability and mortality, including HR (38 studies, 46 cohorts), OR (7 studies, 8 cohorts), RR (12 studies, 15 cohorts) and SMR (12 studies, 12 cohorts). Studies produced a combination of estimates; unadjusted (34 studies, 40 cohorts), age-sex adjusted (38 studies, 43 cohorts) and multivariable adjusted (40 studies, 47 cohorts). For risk of bias assessment (20), fifty-one (73%) studies (59 cohorts) received an assessment of low risk of bias across all nine domains. Six (9%) studies (8 cohorts) were assessed as having a high risk of bias, meaning that few criteria were fulfilled and the conclusions of the study are likely to alter with their inclusion.
Sixty-two cohorts from 53 studies were eligible for inclusion in the meta-analysis, comprising 267,415 participants with disabilities. The remaining studies identified in the systematic review were not included due to reporting measures of effect that could not be pooled with HRs (31, 34, 39, 47, 56, 73, 93, 95), and reporting of unadjusted measures of effect (21, 22, 29, 35, 43, 55, 61, 71, 88).
The pooled maximally adjusted HR for mortality in people with disabilities compared with those without disabilities was 2.06 (95%CI 1.80 - 2.34) and heterogeneity between studies was high (τ²=0·24, I²=98%) suggesting differing effects across studies (Supplementary file 3). This association varied by impairment type; 1.91 (95%CI: 1.61 – 2.26) for cognitive impairment, 1.95 (1.67 – 2.27) for functional impairment, 1.40 (95%CI 1.30 – 1.51) for hearing impairment, 2.39 (95%CI 1.86 – 3.07) for neurological impairment, 3.39 (0.90 – 12.76) for multiple impairments, 1.77 (95%CI 1.25 – 2.51) for psychosocial impairment and 1.32 (95%CI 1.13 – 1.55) for visual impairment (Figure 2).
Four studies compared the hazard for mortality among children with disabilities versus children without disabilities. Pooled HRs for all-cause mortality was 4.46 (95%CI 3.01–6.59) for children under 18 with disabilities versus those without disabilities, with low heterogeneity between studies (τ²=0·05, I²=34%) that suggests a consistent effect across studies (Figure 3). Two studies compared the hazard for mortality in participants with disabilities aged 15 – 49 versus those without disabilities, with a HR for mortality of 3·53 (95%CI 1·29–9·66, τ²=0·50, I²=96%). Twenty-seven studies (35 cohorts) compared the hazard for mortality in participants over the age of 60 with disabilities to those without disabilities; the adjusted HR for mortality was 1·97 (95%CI 1·63–2.38, τ²=0·28, I²=99%).
Meta-regression analysis of studies comparing all-cause mortality between the participants with disabilities and those without revealed that the association between disability and mortality was not influenced by risk of bias (Chi2 3.03, p=0.22). Funnel plots were reviewed for studies comparing all-cause mortality in participants with disabilities with participants who did not have a disability, and no evidence of publication bias was identified (Supplementary file 4).
Discussion
Review of findings
There is limited knowledge about all-cause mortality for people with disability in general, and within LMIC settings specifically. This systematic review and meta- analysis synthesizes the available data from 22 countries and 5 WHO regions on the association between disability and the risk of mortality in LMICs. We found an increased risk of early death for people with disabilities across impairment types and in all ages. The association varied by impairment type as it was strongest for people with multiple impairments, followed by those with neurological impairment. The risk of early death is greatest for children with disabilities and remains increased for adults aged 15 – 49 as well as adults over 60 years of age. All 17 heterogenous studies that were identified in the systematic review and that were not included in the meta-analysis reported an association between disability and increased mortality.
There was little evidence that the quality of the studies or publication bias impacted on the results. Whilst the quality of the included studies was generally high, this review highlighted that there is variability in the methods used to assess and report disability, as well as mortality. Studies predominantly used clinical diagnosis followed by self-report to evaluate disability status and verbal autopsy and population registers to establish death.
Consistency with other reviews
Our results are consistent with meta-analyses of specific impairments and mortality. Globally, Ehrlich et al (13) found that people with visual impairment have a higher risk of early death compared to those with normal vision and the magnitude of the effect increased with severity of impairment. O’Leary et al (15) established that people with intellectual impairments are missing 20 years of life experience globally, and that age of death is lower for people with more severe intellectual disabilities and for people with additional comorbidities such as epilepsy, genetic syndromes and functional impairments. Abuga et al (96) identified that the risk of early death is increased in children with neurological impairments and adults with child onset neurological impairments compared with the general population (median SMR was 2.9 (range 2.0–11.6)), with few studies eligible for inclusion from LMIC.
Limitations and strengths
This systematic review and meta-analysis has limitations that need to be taken into account. Whilst we placed no restrictions on language, the electronic searches were conducted in the English language, and some literature may have been missed. Data was missing from the Eastern Mediterranean region, with only four studies representing the South-East Asia and European regions. Therefore, included studies may not be representative of all LMIC. Nevertheless, 62 cohorts from 22 countries were included in the meta-analysis. Included studies demonstrate a significant association between disability and all-cause mortality in both multivariable adjusted and age-adjusted pooled effect estimates. We included multivariable adjusted studies where the age-adjusted estimates were unavailable. These studies adjusted for other important factors that could explain the association between disability and mortality, such as socio-economic status and comorbidities, which are common risk factors for disability and mortality (3–6). They may therefore have over-adjusted the statistical models and included variables that lie on the causal pathway between disability and mortality. This adjustment could have biased study results toward no effect. Underreporting of disability in studies using self-report is a potential concern, as it may lead to misclassification and bias of the association towards the null . There were also inconsistencies in the assessment of mortality, disability and adjustments made in the analyses, which make comparisons difficult and may explain the high heterogeneity in the meta-analyses.
This study has several key strengths. We included studies from regions in the world that are not represented in any other meta-analyses for disability and mortality. We also included a risk of bias assessment for the included studies. The quality of included studies was generally high and nearly three quarters of studies were estimated to have a low risk of bias across all domains. There was no evidence from the meta-regression analyses that the estimated association was affected by risk of bias.
Implications for research, policy and practice
Our findings that all people with disabilities are at risk of early death in LMIC have important implications for research, policy and practice. Consistent methods of measurement and disaggregation of results by disability are imperative to monitor progress toward the SDGs with respect to disability. Our study highlights the need for widespread adoption of standard definitions and protocols to promote comparability across cohorts. There was notably more information on conditions, such as schizophrenia and dementia, which are frequently recorded in medical records, while other groups (e.g. physical impairment, hearing impairment) were lacking. Attention is therefore needed on how to record disability and impairment type more consistently in medical records to allow use of this resource to measure mortality trends. Research is also needed to develop and test interventions to reduce mortality among people with disabilities, which may include health system strengthening (e.g. healthcare worker training on disability), improving affordability of care for people with disabilities (e.g. provision of health insurance) and tackling the social determinants of poor health in this group.
A key implication of our review is that it will be difficult to reach SDG3, in particular targets related to reducing mortality (e.g. SDG3.2 - “end all preventable deaths under five years of age”), if the one billion people with disabilities globally continue to experience higher mortality. The global community must therefore design disability- inclusive health systems to promote health, wellbeing and rehabilitation for people with disabilities. These systems must “expect, accept and connect” people with disabilities, meaning that they are able to reach accessible services, experience positive attitudes from healthcare workers, and are connected to further services that they need (97). Services for people with disabilities should be included in national health systems at all levels to maximise access and targeted healthcare improvements required for people with disabilities are available when needed (e.g. health promotion programs that focus on lifestyle behaviours and wellness (98)). Reasonable accommodation must be made within provision of quality health care. For example, for people with intellectual impairment, inclusion of anticipatory care in the form of health checks has been shown to improve the management of long-term conditions and quality of life, in addition to having greater clinical and cost- effectiveness when compared with standard care (99, 100).
Conclusion
Disability increases the risk of all-cause mortality in LMICs, particularly in childhood. Interventions are needed to improve health of people with disabilities and reduce their risk of death. Without a focus on disability, it may be difficult to reach SDG3 and other key global health targets.
Contributors
TS and HK conceived the study, developed the research protocol and all data collection instruments and co-led data collection. TS led analysis. TS wrote the first draft of the manuscript with substantial input from HK. Both authors approved of the final version for publication.
Funding
This project is funded by the National Institute for Health Research (NIHR) [NIHR Global Research Professorship (Grant Reference Number NIHR301621)] awarded to Prof. Hannah Kuper. Hannah Kuper has also received support from the Programme for Evidence to Inform Disability Action (PENDA) grant from FCDO.
Disclaimer
The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR, NHS, the UK Department of Health and Social Care or FCDO.
Competing interests
The authors declare no competing interests
Patient consent for publication
Not required.
Provenance and peer review
Not commissioned; externally peer reviewed.
Data sharing statement
No additional data are available
Data Availability
All data produced in the present work are contained in the manuscript
Untitled section