Abstract
Background
Individuals who do not perceive they have a health problem will not seek care, risking a deterioration of health status. We sought to understand how visually impaired individuals aged ≥50 y from 12 studies across sub-Saharan Africa self-reported their difficulties seeing, and to identify the factors associated with under-reporting.
Methods
We used R software to assess overall agreement between self-report and clinical assessment with a weighted kappa, and described the proportion of individuals accurately reporting, over-reporting and under-reporting their difficulties seeing. We used logistic regression to explore factors associated with under-reporting.
Results
In total, 5777/32 548 participants had moderate vision impairment (VI) or worse. Overall, agreement between clinically diagnosed VI and self-reported difficulties seeing was 0.41; 74.7% of participants self-reported a level of difficulty seeing that corresponded with their clinical VI but this varied significantly: individuals with no VI reported most accurately (83.8%), and blind individuals least accurately (18.7%). Individuals were less likely to under-report if they had other self-reported functional difficulties, ‘other’ primary causes of VI (compared with cataract), or were from certain locations. Individuals with unaddressed refractive error were more likely to under-report.
Conclusions
Health planners need to design programmes to reach individuals who are less likely to perceive they have eye health problems.
Contexte
Les personnes qui ne se rendent pas compte qu'elles ont un problème de santé ne consultent pas, ce qui peut entraîner une détérioration de leur état de santé. Nous avons cherché à comprendre comment les personnes malvoyantes âgées de 50 ans et plus inclues dans 12 études menées en Afrique subsaharienne ont auto-déclaré leurs difficultés à voir, et à identifier les facteurs associés à la sous-déclaration.
Keywords: disability, measurement, vision impairment
Abstract
Méthodes
Nous avons utilisé le logiciel R pour évaluer la concordance globale entre l'auto-déclaration de la difficulté à voir et l'examen clinique à l'aide d'un kappa pondéré. Nous avons également décrit la proportion de personnes ayant déclaré correctement, surestimé ou sous-estimé leurs difficultés à voir. A l'aide d'une régression logistique, nous avons exploré les facteurs associés à la sous-déclaration.
Résultats
Parmi les participants, 5 777/32 548 présentaient une déficience visuelle (DV) modérée ou plus grave. Dans l'ensemble, la concordance entre la DV diagnostiquée cliniquement et les difficultés visuelles autodéclarées était de 0,41. 74,7% des participants ont auto-déclaré un niveau de difficulté visuelle correspondant à leur DV clinique, mais avec des variations considérables: les personnes sans DV ont déclaré leur acuité visuelle le plus adéquatement (83,8%), tandis que les personnes aveugles l'ont fait le moins adéquatement (18,7%). Les personnes étaient moins susceptibles de sous-déclarer leur DV si elles déclaraient d'autres difficultés fonctionnelles, d'autres causes principales de DV (par rapport à la cataracte) ou si elles provenaient de certaines régions. Les personnes présentant un vice de réfraction non corrigé étaient plus susceptibles de sous-déclarer leur déficience.
Conclusions
La planification sanitaire doit reposer sur la conception de programmes destinés aux personnes qui sont moins susceptibles de percevoir qu'elles ont des problèmes de santé oculaire.
Introduction
Access to eye health services that prevent and correct blinding and visually impairing conditions remains variable between and within countries.1 The world's leading causes of blindness and vision impairment (VI), unoperated cataract and uncorrected refractive error, are relatively simple and cheap to treat, and yet coverage of these services varies hugely.2 A systematic review, published in 2022, reported that effective cataract surgical coverage, the number of people operated for cataract who can see well expressed as a percentage of all those in need of cataract surgery, ranged from 3.8% in Guinea Bissau to 70.3% in Hungary, and effective refractive error coverage ranged from 5.0% in Durban, South Africa, to 81.8% in Los Angeles, USA.3,4 Reasons for low access to eye health services include issues related to the availability and affordability of eye health services and the patient's personal, social and economic circumstances.5,6
Levesque et al. conceptualised a framework that describes the pathway to accessing healthcare, from acquiring a health condition through to enjoying the consequences of good quality care.7 According to the framework, ‘Ability to perceive’ the need is the first step that determines the demand for services. An individual's ability to perceive that they have a health condition which may benefit from accessing healthcare is determined by issues including their knowledge and beliefs about health and their trust and expectations of available health services. Individuals who do not perceive that they have an eye problem may not seek eye care, or may take longer to seek care, increasing the likelihood of deterioration and poorer outcomes when they do reach a service provider. For health services to plan and deliver health interventions effectively, they need to understand which groups of individuals are at risk of not accessing care, including those who do not recognise that they have a problem in the first place.
Not perceiving a need for care is consistently cited as a reason for not having been operated for cataract in VI surveys across Africa.8 Studies that compared clinical diagnoses of eye conditions and self-reported difficulties in seeing reported, almost universally, that sensitivity—the ability of the individual with a clinically diagnosed condition to self-report it—was lower than the specificity—the ability of an individual with no clinical diagnosis to confirm that they had no problem with seeing.9,10
Understanding the factors that may influence individuals’ perceptions of their health needs can help design better health education campaigns and more effective patient-centred services. This information can also help us understand where self-reported measures may be used as proxies for assessing the burden of VI, and where they will not be appropriate. This is an important question, as self-reported measures of functional difficulty are increasingly used to estimate the prevalence of disability in global populations.
This paper draws on 12 comparable surveys of VI among people aged ≥50 y that measured both clinically assessed visual acuity (VA), and self-reported difficulties seeing. This paper seeks to investigate how people with clinically assessed distance VI self-reported their difficulties seeing, and to identify the factors associated with under-reporting.
Materials and methods
Twelve population-based surveys from eight countries in sub-Saharan Africa were included in this analysis. The study areas were defined by Ministry of Health planning areas and covered populations ranging in size from 523 100 in Karamoja to 7 092 113 in Sierra Leone. All surveys used the standardised Rapid Assessment of Avoidable Blindness (RAAB) methodology with added questions on self-reported functional difficulties.11
The study locations and survey specifics are described in Table 1, and studies will be referred to by the study area described in this table throughout the text.
Table 1.
Survey overview: study characteristics
| Country | Study area | Estimated study area population | Year | Sample size | Response rate |
|---|---|---|---|---|---|
| Nigeria | Kogi | 4 473 500 | 2020 | 4150 | 94.6% |
| Plateau | 4 717 305 | 2023 | 4200 | 94.2% | |
| Sierra Leone | National | 7 092 113 | 2021 | 2650 | 94.3% |
| Senegal | Fatick | 932 629 | 2022 | 2000 | 89.2% |
| Kaolack | 1 228 746 | 2022 | 2000 | 88.6% | |
| Mali | Segou | 3 492 074 | 2023 | 2200 | 89.6% |
| Sikasso | 3 947 305 | 2023 | 2200 | 88.9% | |
| Koulikoro | 3 716 584 | 2024 | 2040 | 90.3% | |
| Uganda | Karamoja | 523 100 | 2023 | 3450 | 91.6% |
| Malawi | South-West zone | 2 662 643 | 2023 | 3850 | 81.6% |
| Tanzania | Dodoma | 3 085 625 | 2024 | 3500 | 94.3% |
| Liberia | National | 5 248 621 | 2024 | 4000 | 96.7% |
The surveys were conducted in sites where the international non-governmental organisation, Sightsavers, worked with the ministries of health to deliver eye health services and supported routine progress monitoring. The study population for each survey consisted of people aged ≥50 y who had lived in the area for at least 6 mo. Participants were selected through two stage sampling: first, enumeration units were selected using probability proportionate to size sampling based on data provided by the national statistical offices. Second, within each unit, 50 participants (40 in Mali) were enumerated and those present and consenting were examined by a team including an ophthalmologist and an assistant. For each survey, data collectors underwent a standardised training course and passed an inter-observer variation test to ensure consistency between assessors. Sample sizes were calculated to measure a precise estimate of the prevalence of VI and ranged from 2000 to 4200. Individual survey reports can be found at bit.ly/4hIGAKS.
Distance vision impairment
All examined participants underwent a basic vision examination to assess their VA and were categorised as being blind (presenting VA<3/60), having severe VI (SVI) (3/60≤VA<6/60), moderate VI (MVI) (6/60≤VA<6/18) or normal vision (VA≥6/18). While mild VI is defined as 6/18≤VA<6/12, this was not measured in Kogi, and so for consistency, individuals in this category have been pooled with those with normal vision.
Ophthalmologists assessed corneal and lens health using a torch and, if necessary, used a direct ophthalmoscope to assess the posterior segment. Ophthalmologists followed the standard RAAB protocol to assign a principal cause of VI to each eye, and to each person if the eyes had VI with different causes. Eyes with presenting VA<6/12 that improved to 6/12 with pinhole (6/18 in Kogi), were assigned unaddressed refractive error as the principal cause of VI. Eyes with VA<6/12 that did not improve with pinhole, and where the lens was observed to have significant occlusion, were assigned unoperated cataract as a principal cause. While other causes were assigned, the numbers were generally small and, for the purposes of this manuscript, were grouped together.
Self-reported difficulties seeing
All studies included the Washington Group Set of Questions on Disability, which requires individuals to self-report their level of difficulty (none; some; a lot; and, cannot do at all) in different functional domains due to a health condition.12 The questions were translated into the most common local language(s) in each setting and pretested to ensure the translations were appropriate and faithful to the original intention. These questions were part of the data collector training, and the teams asked these questions immediately after the vision examination.
The agreement between an individual's VA and their response to the question, ‘Do you have any difficulties seeing?’, is the main outcome of interest in this study, and our construction of that agreement is illustrated in Table 2. As some under- and over-reporting may be expected for a number of reasons, for the purpose of this analysis, we focused only on under- and over-reporting that we considered to be ‘significant’, that is, that varied by more than one step.
Table 2.
Agreement between clinical and self-reported measures
| No difficulties | Some difficulties | A lot of difficulties | Cannot do at all | |
|---|---|---|---|---|
| No VI | Agree (=) | Over-report (+) | Significant over-report (++) | Significant over-report (++) |
| Moderate VI | Under-report (-) | Agree (=) | Over-report (+) | Significant over-report (++) |
| Severe VI | Significant under-report (–) | Under-report (-) | Agree (=) | Over-report (+) |
| Blind | Significant under-report (–) | Significant under-report (–) | Under-report (-) | Agree (=) |
VI: vision impairment.
Therefore, significant under-reporting included blind individuals reporting some or no difficulties and individuals with SVI reporting no difficulties. Significant over-reporting included individuals with no VI reporting not being able to see at all, or a lot of difficulties, and individuals with MVI reporting not being able to see at all.
Other self-reported difficulties and Affect
The explanatory variables ‘Any other self-reported difficulty’ and ‘Affect’ were also constructed from responses to Washington Group questions. ‘Any other self-reported difficulty’ was identified when an individual reported a lot of difficulty or more in one of five other functional domains (hearing, walking, self-care, communication and remembering/concentrating). ‘Affect’ was identified when an individual reported severe anxiety and/or depression on a daily basis as a response to four questions that ask about the frequency and severity of symptoms related to each condition.
Poverty status
Household relative wealth was measured in all surveys. Eight of the surveys used the Equity Tool, a short set of questions about household asset ownership referenced against national standards as derived from a recent large national household survey. In four settings, more direct questions about whether the household has either a regular source of income, yes/no (Karamoja), or whether it has sufficient income for its needs: Not enough to cover our needs, we must borrow; Not enough to cover our needs, we use savings; Just enough to cover our needs; Enough to cover our needs, we are able to save a little; Enough to cover our needs, we are building savings (South-West zone, Dodoma, Liberia).
To construct a binary indicator of wealth across all studies, we used the raw scores from the Equity Tool to construct within-study internal quintiles. Individuals belonging to the two poorest quintiles in the studies that had used the Equity Tool, those who responded ‘no’ in Karamoja, or ‘not enough to cover our needs’ in the South-West zone, Dodoma or Liberia, were considered as belonging to households that were relatively poor. Other individuals were considered relatively wealthier.
Individuals’ age, sex and the main cause of VI were included as explanatory variables, and the study location was also included in all models to account for location-specific differences.
Data analysis
Data were captured in mobile-based apps and exported into R v4.4.0 (R; Vienna, Austria) for analysis.13
We retained, for analyses, participants who had data available for both clinically measured VI and self-reported difficulties seeing. Participants’ characteristics were described for the overall pooled sample. Characteristics of patients with any level of VI (moderate or worse) were described overall and by study.
The overall level of agreement between clinically measured VI and self-reported difficulties seeing across all studies was described using the weighted kappa.
The proportion of agreement and significant under-reporting (respectively over-reporting) were described among participants with SVI or blindness (respectively normal vision or MVI).
Patterns of association between significant under-reporting and covariates of interest were analysed using univariate and multivariable logistic regression models, accounting for the clustered nature of the data. Results are presented as ORs and associated 95% CIs.
The sensitivity of self-reported vision difficulties was calculated as the proportion of those with moderate or worse VI who accurately reported some difficulty seeing or more among all those with moderate or worse VI. The specificity of self-reported vision difficulties was calculated as the proportion of those with no VI, who accurately reported no difficulties seeing among all those with no VI.
All surveys were approved by local ethical boards and local health authorities (see the Ethical Approval statement for full details). The informed consent processes were thorough and inclusive of people with different communication needs. Written consent was provided by all participants, and individuals unable to sign their name provided a thumbprint which was witnessed by an individual, independent of the study team.
Results
Characteristics of participants in the overall sample
The pooled dataset comprised 35 683 individuals, of whom 32 548 had available data for both clinical VI and self-reported difficulties seeing (Table 3).
Table 3.
Characteristics of all survey participants: overall and by VI level
| Clinically assessed VI | |||||
|---|---|---|---|---|---|
| None/early | Moderate | Severe | Blindness | Overall | |
| Age group (y) | n (%) | n (%) | n (%) | n (%) | n (%) |
| 50–59 | 13 330 (49.8%) | 461 (13.8%) | 105 (11.0%) | 154 (10.3%) | 14 050 (43.2%) |
| 60–69 | 7989 (29.8%) | 917 (27.5%) | 206 (21.5%) | 266 (17.9%) | 9378 (28.8%) |
| 70–79 | 3935 (14.7%) | 1063 (31.9%) | 316 (33.0%) | 421 (28.3%) | 5735 (17.6%) |
| ≥80 | 1517 (5.7%) | 890 (26.7%) | 330 (34.5%) | 648 (43.5%) | 3385 (10.4%) |
| Sex | |||||
| Female | 15 524 (58.0%) | 1985 (59.6%) | 588 (61.4%) | 880 (59.1%) | 18 977 (58.3%) |
| Male | 11 247 (42.0%) | 1345 (40.4%) | 369 (38.6%) | 609 (40.9%) | 13 570 (41.7%) |
| Wealth group | |||||
| Relatively poorer | 13 444 (50.4%) | 1850 (55.8%) | 574 (60.7%) | 906 (61.2%) | 16 774 (51.8%) |
| Relatively wealthier | 13 213 (49.6%) | 1464 (44.2%) | 371 (39.3%) | 574 (38.8%) | 15 622 (48.2%) |
| Self-reported difficulties in seeing | 444 (1.7%) | 494 (14.8%) | 289 (30.2%) | 870 (58.4%) | 2097 (6.4%) |
| Self-reported difficulties other than seeing | 1605 (6.0%) | 430 (12.9%) | 164 (17.2%) | 578 (38.9%) | 2777 (8.6%) |
| Self-reported Affect | 957 (3.6%) | 145 (4.4%) | 53 (5.6%) | 120 (8.2%) | 1275 (3.9%) |
VI: vision impairment.
The median age of participants was 61 y; age increased with levels of VI; 58.3% of participants were female. Approximately one-half (51.8%) were in the relatively poorer group.
Overall, 82.3% of participants were assessed as having no VI, 10.2% had MVI, 2.9% had SVI and 4.6% were blind (Supplementary Table S1).
Overall, 93.6% of participants self-reported no difficulties seeing, while 6.4% self-reported some difficulty seeing or worse (Table 3). This proportion varied substantially across VI levels, ranging from 1.7% among people with normal vision to 58.4% among blind participants.
Additionally, 8.6% self-reported other functional difficulties, ranging from 6.0% among those with normal vision to 38.9% among blind participants; and 3.9% reported difficulties in Affect, ranging from 3.6% among people with normal vision to 8.2% among blind participants.
Characteristics of participants with clinically diagnosed VI
Among 5777 participants with moderate VI or worse, 59.8% were female; the median age of these participants was 72 y, and the proportion of those considered to be relatively poorer was 58%.
More than one-quarter of these participants were blind. The proportion of VI due to unoperated cataract was 61.3%, ranging from 45.4% in Kogi to 75.3% in Fatick.
The proportion of those who self-reported any difficulty seeing was 28.6%; the proportion of those who self-reported other difficulties was 20.4%, and the proportion of those who reported Affect was 5.6% (Table 4).
Table 4.
Characteristics of participants with moderate VI or worse: overall and by survey
| Blind, n (%) | Severe VI, n (%) | Moderate VI, n (%) | Unoperated cataract principal cause, n (%) | Female, n (%) | Age, median (IQR) | Relatively poor, n (%) | Any SR difficulties seeing, n (%) | Any other SR difficulty, n (%) | SR Affect, n (%) | |
|---|---|---|---|---|---|---|---|---|---|---|
| Kogi | 132 (24.7%) | 77 (14.4%) | 326 (60.9%) | 243 (45.4%) | 296 (55.3%) | 71 [60, 80] | 238 (44.5%) | 134 (25.0%) | 118 (22.1%) | 5 (0.9%) |
| Plateau | 175 (38.7%) | 97 (21.5%) | 180 (39.8%) | 313 (69.2%) | 261 (57.7%) | 76 [70, 82] | 260 (57.5%) | 164 (36.3%) | 132 (29.3%) | 27 (6.1%) |
| Sierra Leone | 116 (36.0%) | 55 (17.1%) | 151 (46.9%) | 205 (63.7%) | 189 (58.7%) | 69 [61, 78] | 142 (44.1%) | 154 (47.8%) | 91 (28.3%) | 21 (6.5%) |
| Fatick | 92 (23.4%) | 61 (15.5%) | 240 (61.1%) | 296 (75.3%) | 238 (60.6%) | 74 [65, 81] | 177 (45.0%) | 112 (28.5%) | 56 (14.2%) | 0 (0.0%) |
| Kaolack | 64 (20.6%) | 53 (17.1%) | 193 (62.3%) | 219 (70.6%) | 201 (64.8%) | 72 [65, 81] | 117 (37.7%) | 101 (32.6%) | 48 (15.5%) | 0 (0.0%) |
| Segou | 85 (23.7%) | 77 (21.4%) | 197 (54.9%) | 259 (72.1%) | 209 (58.2%) | 71 [65, 78] | 162 (45.1%) | 103 (28.7%) | 76 (21.2%) | 8 (2.2%) |
| Sikasso | 83 (22.9%) | 66 (18.2%) | 213 (58.8%) | 242 (66.9%) | 212 (58.6%) | 73 [66, 80] | 157 (43.4%) | 114 (31.5%) | 74 (20.4%) | 4 (1.1%) |
| Koulikoro | 101 (27.6%) | 65 (17.8%) | 200 (54.6%) | 208 (56.8%) | 200 (54.6%) | 71 [66, 79] | 154 (42.1%) | 122 (33.3%) | 71 (19.4%) | 3 (0.8%) |
| Karamoja | 196 (30.5%) | 125 (19.5%) | 321 (50.0%) | 325 (50.6%) | 455 (70.9%) | 76 [69, 81] | 449 (69.9%) | 169 (26.3%) | 129 (20.1%) | 127 (19.9%) |
| South-West zone | 107 (22.9%) | 94 (20.1%) | 266 (57.0%) | 321 (68.7%) | 297 (63.6%) | 75 [65, 83] | 438 (93.8%) | 152 (32.5%) | 144 (30.8%) | 104 (22.3%) |
| Dodoma | 183 (25.3%) | 126 (17.5%) | 413 (57.2%) | 501 (69.4%) | 395 (54.8%) | 77 [69, 84] | 404 (59.1%) | 115 (15.9%) | 114 (16.2%) | 14 (2.0%) |
| Liberia | 155 (18.3%) | 61 (7.2%) | 631 (74.5%) | 409 (48.3%) | 500 (59.0%) | 68 [60, 75] | 632 (74.6%) | 213 (25.1%) | 119 (14.0%) | 5 (0.6%) |
| Total | 1489 (25.8%) | 957 (16.6%) | 3331 (57.7%) | 3541 (61.3%) | 3453 (59.8%) | 72 [65, 80] | 3330 (58.0%) | 1653 (28.6%) | 1172 (20.4%) | 318 (5.6%) |
SR: self-reported; VI: vision impairment.
Agreement between clinically assessed VI and self-reported difficulties seeing
Overall, 74.7% of study participants self-reported a level of difficulty seeing that corresponded with the clinical assessment of VI as described in Table 2 (Table 5). However, there were substantial differences in the level of agreement by severity of VI. The highest proportion of agreement was for individuals with no VI (83.8%), followed by those with MVI (39.1%), SVI (29.4%) and, finally, blindness (18.7%) (Figure 1, and Supplementary Table S1-Supplementary Table S13).
Table 5.
SR difficulties seeing by levels of clinical VI
| SR difficulties in seeing | ||||
|---|---|---|---|---|
| Clinical VI | No difficulty | Some difficulty | A lot of difficulty | Cannot do at all |
| None | 22 444 (83.8%) | 3883 (14.5%) | 426 (1.6%) | 18 (0.1%) |
| Moderate | 1535 (46.1%) | 1302 (39.1%) | 484 (14.5%) | 10 (0.3%) |
| Severe | 344 (35.9%) | 324 (33.9%) | 281 (29.4%) | 8 (0.8%) |
| Blindness | 346 (23.2%) | 273 (18.3%) | 592 (39.8%) | 278 (18.7%) |
SR: self-reported; VI: vision impairment.
Figure 1.
Proportion of self-reported difficulties in seeing within levels of clinical VI: by survey. SWZ: South-West zone; VI: vision impairment.
There was overall moderate agreement between clinically diagnosed VI and self-reported difficulties seeing (weighted kappa=0.41) (Table 6).
Table 6.
Agreement between SR difficulties in seeing and clinical VI: survey-specific weighted kappas
| Survey | Weighted kappa |
|---|---|
| Kaolack | 0.49 |
| Sierra Leone | 0.49 |
| Kogi | 0.47 |
| Koulikoro | 0.46 |
| Sikasso | 0.42 |
| Fatick | 0.41 |
| Karamoja | 0.4 |
| Segou | 0.39 |
| Plateau | 0.38 |
| Liberia | 0.37 |
| Dodoma | 0.36 |
| South-West zone | 0.31 |
| Overall | 0.41 |
SR: self-reported; VI: vision impairment.
Significant under- and over-reporting
The proportion of individuals who significantly under-reported their difficulty seeing was high, 41.5% among blind individuals and 35.9% among people with SVI. The proportion of individuals who significantly over-reported their difficulty was low, 1.7% for individuals with no VI and 0.3% for those with MVI (Table 7).
Table 7.
Summary of agreement, significant over-reporting and significant under-reporting: by study and overall
| Normal to moderate VI | Severe VI or blindness | |||
|---|---|---|---|---|
| Agreement (=) | Significant over-reporting (++) | Agreement (=) | Significant under-reporting (–) | |
| Kogi | 87.6% | 0.2% | 18.2% | 33.0% |
| Plateau | 82.6% | 1.4% | 21.3% | 45.2% |
| Sierra Leone | 79.6% | 2.7% | 19.3% | 19.9% |
| Fatick | 78.8% | 1.0% | 23.5% | 41.2% |
| Kaolack | 84.9% | 0.8% | 35.9% | 39.3% |
| Segou | 76.4% | 1.0% | 25.3% | 34.6% |
| Sikasso | 74.7% | 1.0% | 30.2% | 30.9% |
| Koulikoro | 74.4% | 1.4% | 29.5% | 21.7% |
| Karamoja | 76.1% | 1.5% | 22.1% | 46.4% |
| South-West zone | 75.8% | 5.5% | 32.3% | 38.8% |
| Dodoma | 82.6% | 0.3% | 12.9% | 57.9% |
| Liberia | 70.2% | 1.3% | 19.0% | 38.9% |
| Overall | 78.9% | 1.5% | 22.9% | 39.4% |
VI: vision impairment.
Among those with normal to moderate VI, the proportion of participants self-reporting difficulties seeing that corresponded to their clinical VI ranged from 70.2% in Liberia to 87.6% in Kogi, while significant over-reporting ranged from 0.2% in Kogi to 5.5% in South-West zone. Variations across studies were more marked among those with SVI or blindness. Agreement ranged from 12.9% in Dodoma to 35.9% in Kaolack, while significant under-reporting ranged from 19.9% in Sierra Leone to 57.9% in Dodoma.
Factors associated with significant under-reporting
Univariate model results showed that four factors were significantly associated with under-reporting difficulties seeing, consisting of the principal cause of VI, other functional difficulties, Affect and study site. Compared with those with unoperated cataract, people with refractive error were more likely to significantly under-report their difficulty seeing, whereas people with another cause of VI were less likely to under-report it. Those who reported other functional difficulties or Affect were less likely to under-report. Finally, compared with those from Karamoja, participants from Koulikoro, Sikasso, Kogi and Sierra Leone were less likely to under-report. No statistically significant associations were observed between significant under-reporting and sex, age or poverty status (Table 8).
Table 8.
Factors associated with significant under-reporting among participants with severe VI or blindness: results from logistic regression models
| Univariate | Multivariate | ||||
|---|---|---|---|---|---|
| n (%) | OR | 95% CI | AOR | 95% CI | |
| Sex | |||||
| Male (ref.) | 978 (40.0%) | ||||
| Female | 1468 (60.0%) | 1.04 | [0.87, 1.23] | 1.02 | [0.79, 1.32] |
| Age group (y) | |||||
| 50–59 (ref.) | 259 (10.6%) | ||||
| 60–69 | 472 (19.3%) | 1.04 | [0.77, 1.39] | 1.28 | [0.80, 2.05] |
| 70–79 | 737 (30.1%) | 1.05 | [0.78, 1.42] | 1.23 | [0.79, 1.91] |
| ≥80 | 978 (40.0%) | 0.94 | [0.71, 1.26] | 1.36 | [0.87, 2.12] |
| Poverty status | |||||
| Relatively poorer (ref.) | 1480 (61.0%) | ||||
| Relatively wealthier | 945 (39.0%) | 1.04 | [0.83, 1.31] | 1.05 | [0.76, 1.43] |
| Other functional difficulties (sensory and physical) | |||||
| None (ref.) | 1696 (69.6%) | ||||
| Yes | 742 (30.4%) | 0.15 | [0.11, 0.20] | 0.12 | [0.08, 0.16] |
| Affect | |||||
| None (ref.) | 2245 (92.8%) |
||||
| Yes | 173 (7.2%) | 0.51 | [0.34, 0.75] | 0.51 | [0.30, 0.87] |
| Principal cause of VI | |||||
| Unoperated cataract (ref.) | 1564 (63.9%) | ||||
| Refractive error | 61 (2.5%) | 3.81 | [2.13, 6.83] | 5.68 | [2.08,15.48] |
| Other | 821 (33.6%) | 0.61 | [0.50, 0.74] | 0.53 | [0.40, 0.68] |
| Study location | |||||
| Karamoja (ref.) | 321 (13.1%) | ||||
| Liberia | 216 (8.8%) | 0.73 | [0.45, 1.20] | 0.58 | [0.30, 1.10] |
| Malawi SWZ | 201 (8.2%) | 0.73 | [0.42, 1.27] | 0.79 | [0.39, 1.59] |
| Mali Koulikoro | 166 (6.8%) | 0.32 | [0.19, 0.55] | 0.25 | [0.12, 0.53] |
| Mali Segou | 162 (6.6%) | 0.61 | [0.33, 1.14] | 0.45 | [0.19, 1.02] |
| Mali Sikasso | 149 (6.1%) | 0.52 | [0.27, 0.99] | 0.40 | [0.16, 0.99] |
| Nigeria Kogi | 209 (8.5%) | 0.57 | [0.35, 0.93] | 0.33 | [0.17, 0.66] |
| Nigeria Plateau | 272 (11.1%) | 0.95 | [0.55, 1.65] | 0.64 | [0.35, 1.16] |
| Senegal Fatick | 153 (6.3%) | 0.81 | [0.45, 1.45] | 0.33 | [0.16, 0.71] |
| Senegal Kaolack | 117 (4.8%) | 0.75 | [0.42, 1.33] | 0.51 | [0.24, 1.07] |
| Sierra Leone | 171 (7.0%) | 0.29 | [0.16, 0.52] | 0.18 | [0.07, 0.46] |
| Tanzania Dodoma | 309 (12.6%) | 1.59 | [0.99, 2.55] | 1.25 | [0.68, 2.29] |
AOR: adjusted OR; SWZ, South-West zone; VI: vision impairment.
Bold text indicates p<0.05.
Results from multivariable models were very similar, indicating that these factors were independently associated with significant under-reporting (Table 8).
Self-reported difficulty in seeing as a predictor of VI
Among 7879 individuals who self-reported some difficulty seeing or worse, 45.1% (n=3552) had VI categorised as moderate or worse. This included 14.5% (n=1143) who were blind, 7.8% (n=613) who had SVI and 22.8% (n=1796) who had MVI. More than one-half of these participants (54.9%, n=4327) had no VI.
Among 24 669 participants who said they had no difficulty in seeing, 91.0% (n=22 444) had no VI, 6.2% (n=1535) had MVI, 1.4% (n=344) had SVI and 1.4% (n=346) were blind.
The sensitivity of self-reported vision difficulties was 61.5%, ranging from 42.1% in Dodoma to 84.2% in Sierra Leone. The specificity of self-reported vision difficulties was 83.8%, ranging from 74.3% in Liberia to 93.4% in Kogi. The positive predictive value was 45.1% overall, ranging from 33.6% in South-West zone to 60.6% in Kaolack. The negative predictive value was 91.0% overall, ranging from 84.6% in Dodoma to 96.3% in Sierra Leone (Table 9).
Table 9.
Self-reporting some difficulties in seeing or more—sensitivity, specificity, PPV and NPV in terms of moderate VI or worse—by survey
| Survey | Sensitivity (%) | Specificity (%) | PPV (%) | NPV (%) |
|---|---|---|---|---|
| Kogi | 49.72 | 93.36 | 54.18 | 92.17 |
| Plateau | 61.95 | 85.05 | 34.83 | 94.54 |
| Sierra Leone | 84.16 | 81.86 | 47.88 | 96.31 |
| Fatick | 51.40 | 87.56 | 53.87 | 86.44 |
| Kaolack | 56.13 | 92.27 | 60.63 | 90.84 |
| Segou | 61.00 | 80.97 | 41.63 | 90.32 |
| Sikasso | 70.17 | 78.34 | 42.4 | 92.04 |
| Koulikoro | 75.14 | 77.42 | 45.23 | 92.62 |
| Karamoja | 66.51 | 80.33 | 46.31 | 90.39 |
| South-West zone | 57.39 | 80.22 | 33.63 | 91.51 |
| Dodoma | 42.11 | 91.57 | 59.03 | 84.58 |
| Liberia | 72.26 | 74.28 | 44.06 | 90.52 |
| Overall | 61.49 | 83.84 | 45.08 | 90.98 |
NPV: negative predictive value; PPV: positive predictive value; VI: vision impairment.
Discussion
This study explored the concordance between perceived difficulty seeing and objectively verified distance VI. This analysis illustrates that, while the majority of participants self-reported their difficulties seeing in a manner consistent with their clinically assessed VI, a significant proportion underestimated their difficulties. The disagreement was particularly striking among participants with severe levels of vision loss. Nearly one-quarter of participants who were blind and more than one-third of those who had SVI said they had no difficulty seeing. This has important service delivery implications as people who do not think they have a health problem are unlikely to seek care, irrespective of health information or available services.
There was a substantial proportion of participants with severe vision loss who reported only some difficulty seeing. There is little research on the topic, but in low-income contexts where eye care services are limited and patients incur significant costs through user fees or travel expenditure, this proportion of patients willing to travel to seek care may be small. Combining people who report either no or just some difficulties seeing, >40% of people who are blind and approximately 70% of those who have SVI are at risk of delaying seeking care or forgoing it completely. It is difficult to place our findings in the context of other research, which comes predominantly from high-income countries where eye care services are more widely available and affordable. The few studies available in low- and middle-income countries show a similar trend. For example, Boggs et al. reported that, in their three-country study, the majority of people with VI (70–85%) reported only some difficulty seeing; the proportion of those who reported a lot of difficulty or worse was 11% in the Gambia, 30% in Cameroon and 39% in India.9
Three individual characteristics were associated with significant under-reporting in our study. First, people who self-reported difficulties functioning in other domains including anxiety and/or depression were less likely to under-report their difficulty seeing. People who experience and perceive difficulties in other domains may feel the effect of VI more acutely than individuals who do not experience other functional limitations. Individuals with more experience of health problems may be able to recognise and express difficulties more comfortably.
Second, compared with people with unoperated cataract, people with unaddressed refractive error were more likely to under-report their difficulties seeing, and people with ‘other’ causes were less likely. While all these individuals experienced SVI or blindness, the principal causes may have onset at different ages and may present with different symptoms in addition to distance VI, which affects how individuals perceive their difficulties. Studies from Nigeria and Kenya found that individuals were more likely to seek care for conditions that were either painful or disfiguring.14,15 It is difficult to say whether patients who significantly under-reported difficulty in seeing in our study did it because they did not perceive the difficulty or because they did not want to report it. Either way our findings show that there is a considerable proportion of people who may not acknowledge the severity of the vision problem they have and may not seek care. Eye health programmes seeking to improve coverage of services must consider how they can identify and motivate such individuals to take up treatment.
Some of the individuals with VI due to refractive error had been operated for cataract and they will have experienced an improvement in their VI compared with the presurgery levels. This improvement may have influenced their perception of having no, or not much difficulty seeing. The relationship between self-perception of vision quality and postcataract surgery correction and other comorbidities has been recognised in other studies.16,17
Finally, the differences by location were difficult to interpret. All settings where participants were less likely to under-report difficulties in seeing compared with the reference group were in West Africa, but this could be a coincidence. We have no reason to suspect that people in these settings are more likely to recognise vision loss.
Other studies showed a greater variety of factors associated with being unaware of having different health conditions, including older age, indigenous status, years since last eye test and the type of provider seen at that test (optometrist or ophthalmologist), which our study did not measure.10,18 It is important to note that most of these are from high-income settings, and the relationship may be mediated by the availability, affordability and perceived quality of services, which are context specific and difficult to compare. However, one similarity is that self-reported difficulties seeing seem to be affected by both vision and non-vision related factors and the difficulties in the domain of Affect play an important role.19 For example, the US National Health and Aging Trends Study showed that adults with VI were less likely to not report vision difficulties if they had multiple comorbidities or depressive symptoms.20 The same study also reported the relationship between self-perceived difficulties with vision and gender and income. The impact of vision loss is less likely to be perceived if it does not impact daily activities. For example, Boggs et al. highlight that if someone is illiterate, they may not perceive a need for glasses to read a newspaper.21 Further research that explores the impact of vision loss on daily activities and social and economic outcomes, particularly from low-income settings, is needed.
Our second question examined the use of self-reported difficulties seeing as an indicator of the need for eye care services. In many countries, disability is legally defined following the United Nations Convention on the Rights of Persons with Disability, which describes individuals with 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, as having a disability.22 The search for a single or small number of questions that can accurately measure this state is ongoing, with no existing measures being able to adequately capture the state of an individual and their environment in a way that translates appropriately across contexts.23
From a policy planning perspective, our study results highlight the need for caution in using self-reported vision difficulties as the sole metric for estimating the burden of VI. While it is established that binary yes/no questions such as ‘Do you have a disability?’ significantly underestimate the prevalence of disability, an understanding is required of how increasingly popular measures of self-reported functional difficulties, including but not limited to those used in this study, relate to clinically assessed measures of impairment.24
We found that self-reported difficulty seeing had higher specificity than sensitivity, which means that this self-reported measure was more accurate in identifying people without VI than those with VI. These findings are similar to other studies. A study in Australia reported sensitivity of self-reported vision difficulties of between 7.6% and 52.2% and specificity of between 63.9% and 99.2%.10 It is also important to note that in our analysis we used some difficulty or worse as a cut-off point for vision-related problems. With the cut-off point of a lot of difficulty or worse, sensitivity could be lower.
This study is strengthened by its use of large, population-based datasets from multiple sub-Saharan African countries. The standardised methodology ensures comparability across study sites and this makes it one of the most comprehensive analyses of self-reported vision difficulties in the region. However, there are a number of limitations that should be considered when interpreting the findings. While the RAAB itself is a well-established and robust methodology, its vision examination is rapid and relies on basic equipment. It is possible that there may be inaccuracies in the clinical assessment of VI and the principal cause. Furthermore, RAAB only recognises the ‘principal’ cause of VI, and we cannot assess how multiple ocular morbidities may influence an individual's perception of their difficulties seeing. Factors identified by other studies as influencing individuals’ perceptions, such as time since last eye examination, were not measured.
The vision assessment included in RAAB is simple with only distance vision measured, and a more complete series of tests among the same participants would likely result in a much higher prevalence of VI and ocular problems than what is reported here. Therefore, it is likely that our data underestimate the dissonance between clinically assessed and self-reported measures, although it is not possible to estimate by what degree, given the finding that results differ by cause of vision loss. Self-reported difficulties seeing were measured using a well-validated tool,12 however, despite efforts to standardise across settings, languages and data collectors, it is possible that there were variations in how the questions were both asked and understood. Diverse results from Senegal and Mali where study areas were geographically, culturally and linguistically similar and where the fieldwork was conducted by the same teams, suggests that this is not the case and other factors must play a role.
The measurement of poverty differed between surveys, and while we have attempted to standardise the analysis in a useful way, there are limitations to the approach taken in each individual dataset, and the creation of the compound indicator, which may have led to non-differential mis-categorisation of some individuals, rendering the results related to poverty inaccurate.
Finally, there are challenges when making such comparisons between countries, particularly using subnational datasets that represent only a small proportion of the country. We cannot suggest that results may be applicable to other parts of those countries, or to individuals aged <50 y who were not surveyed. Qualitative research to explore the sociocultural reasons behind self-reporting discrepancies would be valuable to explore the decision-making processes engaged in answering these questions.
Conclusions
Individuals with worse levels of VI, refractive errors, and those with physical, sensory or mental comorbidities, may be less likely to recognise vision difficulties and seek or access care for eye conditions that need clinical attention. Health managers involved in planning services need to consider different approaches for reaching this group of individuals and may require further evidence on the drivers of eye health-seeking behaviour in these contexts to do so effectively. Relying on self-reported functional difficulties may lead to an underestimation of the prevalence of disability, and this should be accounted for when planning data collection and interpreting data.
Supplementary Material
Contributor Information
Emma Jolley, Evidence, Research & Innovations, Sightsavers, 35 Perrymount Road, Haywards Heath, West Sussex, RH16 3BW, UK.
Anita Jeyam, Evidence, Research & Innovations, Sightsavers, 35 Perrymount Road, Haywards Heath, West Sussex, RH16 3BW, UK.
Stephen Pye, Evidence, Research & Innovations, Sightsavers, 35 Perrymount Road, Haywards Heath, West Sussex, RH16 3BW, UK.
Stevens Bechange, Evidence, Research & Innovations, Sightsavers, Kampala, Uganda.
Shaffi Mdala, Ophthalmology Unit, Department of Surgery, Kamuzu University of Health Sciences, Kamuzu, Malawi; International Centre for Eye Health, London School of Hygiene and Tropical Medicine, London WC1E 7HT, UK.
Mouctar Badiane, Ministère de la Santé et de l'Action Sociale, Dakar, Senegal.
Joshua Owoeye, University of Ilorin, Ilorin, Nigeria.
Moses Kashadakawo, Mulago National Referral Hospital, Kampala, Uganda.
Alice Ramyil, Department of Ophthalmology, College of Health Sciences, University of Jos, Jos, Nigeria.
Lamine Traoré, Ministère de la Santé et du Développement Social, Bamako, Mali.
Jalikatu Mustapha, Ministry of Health and Sanitation, Freetown, Sierra Leone.
Japhet Boniface, Dodoma Regional Referral Hospital, Dodoma, Tanzania.
Joseph Kerkula, Ministry of Health, Monrovia, Liberia.
Elena Schmidt, Evidence, Research & Innovations, Sightsavers, 35 Perrymount Road, Haywards Heath, West Sussex, RH16 3BW, UK.
Authors’ contributions
EJ, AJ, SP, SB and ES conceived the study. EJ, SB, SM, MB, JO, MK, LT, JM, JB, JK and AR acquired the data. EJ, AJ and SP analysed the data. All the authors were involved in interpreting the data. EJ, AJ and ES drafted the manuscript. SP, SB, SM, MB, JO, MK, LT, JM, JB, JK and AR critically revised the manuscript for important intellectual content. All the authors read and approved the final version of the manuscript. EJ is the guarantor of the paper.
Funding
This work was supported by:
Kogi: UK Foreign, Commonwealth & Development Office [contract award DFID 8219]; Plateau: Sightsavers and CBM; Sierra Leone: Irish Aid [PG 10–2017]; Fatick: Irish Aid [PG 10–2017]; Kaolack: Irish Aid [PG 10–2017]; Segou: Sightsavers; Sikasso: Sightsavers; Koulikoro: Sightsavers; Karamoja: UK Foreign, Commonwealth and Development Office [UKAM grant number: 205210-265]; South-West zone: UK Foreign, Commonwealth and Development Office [UKAM grant number: 205210-265]; Dodoma: Sightsavers; Liberia: Irish Aid [A Better World].
Competing interests
None declared.
Ethical approval
Ethical approval was provided by the following bodies:
Kogi: Research Ethics Committee of Kogi State Ministry of Health, Nigeria; Plateau: Research Ethics Committee of Plateau State Ministry of Health, Nigeria; Sierra Leone: Office of the Sierra Leone Ethics and Scientific Review Committee; Fatick: Comité National d'Ethique pour la Recherche en Santé, Senegal (SEN22/10–0044); Kaolack: Comité National d'Ethique pour la Recherche en Santé, Senegal (SEN22/10–0044); Segou: Comité National d'Ethique pour la Santé et les sciences de la vie, Mali (No. 2022135); Sikasso: Comité National d'Ethique pour la Santé et les sciences de la vie, Mali (No. 2022135); Koulikoro: Comité National d'Ethique pour la Santé et les sciences de la vie, Mali (No. 2024007); Karamoja: Mulago Hospital REC (MHREC-2022-83) and Uganda National Council for Science and Technology (221558ES); South-West zone: Malawi National Health Sciences Research Committee; Dodoma: Tanzania National Institute for Medical Research; Liberia: Liberia National Research Ethics Board (NREB-004-24).
Data availability
The data underlying this article are available in the article and in its online supplementary material.
References
- 1. Burton MJ, Ramke J, Marques AP et al. The Lancet Global Health Commission on Global Eye Health: vision beyond 2020. Lancet Glob Health. 2021;9(4):e489–551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. World Health Organization . Report of the 2030 targets on effective coverage of eye care. Geneva, Switzerland: WHO, 2022, 48. [Google Scholar]
- 3. McCormick I, Butcher R, Evans JR et al. Effective cataract surgical coverage in adults aged 50 years and older: estimates from population-based surveys in 55 countries. Lancet Global Health. 2022;10(12):e1744–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Bourne RRA, Cicinelli MV, Sedighi T et al. Effective refractive error coverage in adults aged 50 years and older: estimates from population-based surveys in 61 countries. Lancet Glob Health. 2022;10(12):e1754–63. [DOI] [PubMed] [Google Scholar]
- 5. Mailu EW, Virendrakumar B, Bechange S et al. Factors associated with the uptake of cataract surgery and interventions to improve uptake in low- and middle-income countries: a systematic review. PLoS One. 2020;15(7):e0235699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Lewallen S, Schmidt E, Jolley E et al. Factors affecting cataract surgical coverage and outcomes: a retrospective cross-sectional study of eye health systems in sub-Saharan Africa. BMC Ophthalmology. 2015;15(1):67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Levesque J-F, Harris MF, Russell G. Patient-centred access to health care: conceptualising access at the interface of health systems and populations. Int J Equity Health. 2013;12(1):18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Aboobaker S, Courtright P. Barriers to cataract surgery in Africa: a systematic review. Middle East Afr J Ophthalmol. 2016;23(1):145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Boggs D, Kuper H, Mactaggart I et al. Exploring the use of Washington Group Questions to identify people with clinical impairments who need services including assistive products: results from five population-based surveys. Int J Environ Res Public Health. 2022;19(7):4304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Foreman J, Xie J, Keel S et al. The validity of self-report of eye diseases in participants with vision loss in the National Eye Health Survey. Sci Rep. 2017;7(1):8757. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Kuper H, Polack S, Limburg H. Rapid assessment of avoidable blindness. Community Eye Health. 2006;19(60):68–9. [PMC free article] [PubMed] [Google Scholar]
- 12. Madans JH, Loeb ME, Altman BM. Measuring disability and monitoring the UN Convention on the Rights of Persons with Disabilities: the work of the Washington Group on Disability Statistics. BMC Public Health. 2011;11(4):S4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. R Core Team . R: a Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2024. [Google Scholar]
- 14. Senyonjo L, Lindfield R, Mahmoud A et al. Ocular morbidity and health seeking behaviour in Kwara State, Nigeria: implications for delivery of eye care services. PLoS One. 2014;9(8):e104128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Kimani K, Lindfield R, Senyonjo L et al. Prevalence and causes of ocular morbidity in Mbeere District, Kenya. Results of a population-based survey. PLoS One. 2013;8(8):e70009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Grimfors M, Mollazadegan K, Lundström M et al. Ocular comorbidity and self-assessed visual function after cataract surgery. J Cataract Refract Surg. 2014;40(7):1163–9. [DOI] [PubMed] [Google Scholar]
- 17. Rönbeck M, Lundström M, Kugelberg M. Study of possible predictors associated with self-assessed visual function after cataract surgery. Ophthalmology. 2011;118(9):1732–8. [DOI] [PubMed] [Google Scholar]
- 18. Patty L, Wu C, Torres M et al. Validity of self-reported eye disease and treatment in a population-based study: the Los Angeles Latino Eye study. Ophthalmology. 2012;119(9):1725–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Tabrett DR, Latham K. Factors influencing self-reported vision-related activity limitation in the visually impaired. Invest Ophthalmol Vis Sci. 2011;52(8):5293–302. [DOI] [PubMed] [Google Scholar]
- 20. Potter T, Almidani L, Diaz M et al. Concordance between self-reported visual difficulty and objective visual impairment: the national health and aging trends study. Ophthalmology. 2024;131(12):1447–56. [DOI] [PubMed] [Google Scholar]
- 21. Boggs D, Hydara A, Faal Y et al. Estimating need for glasses and hearing aids in the Gambia: results from a national survey and comparison of clinical impairment and self-report assessment approaches. Int J Environ Res Public Health. 2021;18(12):6302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. United Nations Department of Field Suport: Geographic Information Section; CRPD and Optional Protocol Signatures and Ratifications . 2017, Map No. 4496 Rev. 8.
- 23. Groce NE, Mont D. Counting disability: emerging consensus on the Washington Group questionnaire. Lancet Glob Health. 2017;5(7):e649–50. [DOI] [PubMed] [Google Scholar]
- 24. Mont D. Measuring health and disability. Lancet. 2007;369(9573):1658–63. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article are available in the article and in its online supplementary material.