Vision impairment and traffic safety outcomes in low income and middle income countries: a systematic review and meta-analysis

Authors: Piyasena P, Olvera-Herrera VO, Chan VF, Clarke M, Wright DM, MacKenzie G, et al.

Geographical coverage: Low- and middle-income countries

Sector: Epidemiology

Sub-sector: Burden of disease

Equity focus: None

Study population: Road users

Review type: Other review

Quantitative synthesis method: Narrative review and meta-analysis

Qualitative synthesis method: Not applicable

Background: The burden of deaths from traffic accidents is especially great in low income and middle income countries (LMICs), with annual fatality rates per 100,000 population of 24.1 in low income countries, compared with 9.2 in high income countries. Despite the high burden of traffic crashes and associated mortality in LMICs, evidence for an association between vision function and traffic safety outcomes is scarce. This fact makes it difficult to advocate that policy makers should develop effective vision-based road crash prevention strategies.

Objectives: To systematically analyse the current evidence to establish whether any aspects of vision, and particularly interventions to improve vision function, are associated with traffic safety outcomes in LMICs.

Main findings:

Overall, this review showed a positive association between vision impairment and traffic crashes in LMICs. Findings provide support for mandatory vision function assessment before issuing a driving licence.

The review included 29 cross-sectional and observational studies, with no randomized trials or intervention studies. These studies were conducted in various countries, including one low-income country (Ethiopia), 22 lower-middle-income countries (11 in Nigeria, nine in India, two in Ghana), and six upper-middle-income countries (four in China, one each in Turkey and Iran). Out of these, 20 studies assessed the link between vision function and traffic safety, 26 reported the prevalence of vision disorders among drivers, and 14 described adherence to driving licensure requirements. Fourteen of the 20 studies reporting on vision and traffic safety outcomes qualified for a meta-analysis on the association between visual impairment and road traffic crashes. No studies were found assessing interventions to improve vision function. The studies enrolled a total of 15,394 participants (average sample size of 530) with an average age of 39.3 years. Only five studies included women. Most studies (22) focused on commercial drivers of trucks, buses, or taxis. Two studies assessed safety outcomes on driving simulators, one enrolled patient from a glaucoma clinic, one assessed self-reported crashes among schoolchildren riding bicycles in China, and one included motorcycle delivery drivers. Both Snellen and logarithm of the minimum angle of resolution charts were used in the studies, with most using WHO definitions of visual impairment.

The prevalence of vision impairment among road users ranged from 1.2% to 26.4% (26 studies), colour vision defects from 0.5% to 17.1% (15 studies), and visual field defects from 2.0% to 37.3% (10 studies). A substantial proportion (range 10.6-85.4%) received licences without undergoing mandatory vision testing. The meta-analysis revealed a 46% greater risk of having a road traffic crash among those with central acuity visual impairment (risk ratio [RR] 1.46 [95% CI 1.20-1.78]; p = 0.0002, 13 studies) and a greater risk among those with defects in colour vision (RR 1.36 [1.01-1.82]; p = 0.041, seven studies) or the visual field (RR 1.36 [1.25-1.48]; p = 0·0001, seven studies). The I² value for overall statistical heterogeneity was 63.4%.

The authors suggest that randomised trials, possibly using a step-wedge design to address ethical issues related to withholding known vision-enhancing interventions, could provide the most pertinent evidence to inform policy and practice in low-to-middle-income countries (LMICs). They also highlight that studies incorporating under-represented groups of drivers, especially women, could produce more widely applicable results. Furthermore, they advocate for a more transdisciplinary research approach, taking into account the Five Pillar Approach for road safety (which includes road safety management, safer roads and mobility, safer vehicles, safer road users, and post-crash response), as essential for devising strategies to reduce the mortality rate from road traffic accidents in LMICs.

Methodology:

The review included interventional or observational studies from LMICs that assessed the correlation between vision and traffic safety outcomes, the prevalence of poor vision among drivers, and adherence to licensure regulations. Eligible participants were road users in LMICs, such as drivers, cyclists, pedestrians, and public transit users, with a focus on individuals who earned their income from driving. Studies from high-income countries were excluded. The primary outcome was any traffic-crash-related injury that could potentially cause morbidity and mortality. This also encompassed surrogate outcomes like hard braking or accelerometer-measured events, primarily based on self-reported data. The main exposure of interest was poor vision functions of drivers.

The authors conducted a search on Medline, EMBASE, PsycInfo, CINAHL, Web of Science, Cochrane Database of Systematic Reviews, and the Cochrane Central Register of Controlled Trials in the Cochrane Library, from the inception of these databases until 2 April 2020. Data extraction was performed by two reviewers, with one extracting the data and the other verifying it. The risk of bias and the quality of the studies were evaluated using the appropriate tool for each study: the National Institutes of Health quality assessment tool for observational cross-sectional studies and the relevant tool from the Critical Appraisal Skills Programme for other study designs.

Authors first described study characteristics, such as study design, country, setting, type of driver, and category of vehicle, and then provided meta-analyses of the findings for reported outcomes, using odds ratios or risk ratios (RRs) for binary outcomes. They did a narrative synthesis of the prevalence of vision disorders among drivers and rates of compliance with vision-related licensure requirements.

Statistical heterogeneity was evaluated across the studies. A random-effects model was applied to estimate τ². Reported relative risks were used to compare the risk of crashes among drivers with poor versus good vision function. Separate meta-analyses were conducted for visual acuity, colour vision, and visual field outcomes. Subgroup analyses were performed for individuals whose income was derived from driving a vehicle in a LMIC, if sufficient data was available. Small-study bias was assessed using Harbord’s test and a contour-enhanced funnel-plot with a non-parametric trim-and-fill method for imputing potentially missing data from small studies. An attempt was made to assess confounding using multivariate meta-regression; however, due to an insufficient number of studies and a lack of necessary primary data, modelling was not successful.

Applicability/external validity: Authors note a number of limitations which may make their results less applicable/valid.

Geographic focus: Authors note that most of the conducted studies were located in African countries, and may not necessarily reflect trends in LMICs as a whole.

Summary of quality assessment:

There were several limitations in the approaches used to identify, include and critically appraise studies. While the key activities were all undertaken by at least two authors, there is no evidence of reference lists of included studies being reviewed, or of consideration of unpublished or non-English material for inclusion. In addition, information on the quality assessment of studies is not presented on a study-by-study basis. The data analysis was generally robust; however, as noted by the authors, it was not possible to undertake a meta-regression to analyse the impact of some of the characteristics on which studies varied (for example, type of participants). For these reasons, a low confidence was attributed in the findings of this review.