Introduction

The introduction of the sight test fee in 1990 increased concern about undetected visual problems in older people,¹ and visual screening for older people in general practice was advocated.² Renewed concern has been expressed recently by the Royal National Institute for the Blind³ and the Department of Health.⁴

Since 1990 general practitioners have been required to offer an annual screening assessment to all patients aged 75 and over,⁵ specifically including an assessment of vision. While multiphasic screening of older people has been shown to be beneficial overall,⁶ exactly which procedures are effective is uncertain. The 75 and over programme is currently under review.⁷

Rationale for screening older people for visual problems
Visual impairment is common among older people. In community based surveys of unselected older people undertaken in the United Kingdom, visual acuity of less than 6/12 has been found in around 2% of those aged 65 to 74 and around 20% of those aged 75 and over. ^{8 9} This level of visual acuity is below the requirements for driving in the United Kingdom.¹⁰ Larger surveys have been performed in the United States,^11-13 and, though the prevalence of reduced visual acuity is lower in one of these,¹¹ similar overall trends are seen. Various adverse factors have been reported in association with visual impairment. These include reduced functional status and quality of life,^14-16 depression,^16-18 and falls.^19-21

	Methods

The research question was defined as, what is the effectiveness of population screening for visual impairment in improving vision among unselected older people (aged 65 or over) in a community setting, either alone or as part of a multiphasic assessment programme?

Search methods
A systematic Medline search for randomised controlled trials evaluating screening in older people was performed for the period 1966 to December 1996 by using a published optimally sensitive search strategy.²⁷ The exploded (that is, including all sub-branches) MeSH terms "mass screening," "preventive health services," "eye diseases," and "diagnosis, eye," and the non-exploded MeSH terms "health promotion" and "geriatric assessment" were used. Titles and abstracts were searched for the words "geriatric" or "elderly" combined with any of "screening," "assessment," "health," "function," or "surveillance." Other textword searches used the words "macular degeneration," "cataract," and "presbyopia." Wild card characters were used to ensure all forms of words were included. At all stages articles about animals and children were excluded. The October 1997 Cochrane Library (which includes the Cochrane Controlled Trials Register and the Database of Abstracts of Reviews of Effectiveness) was also searched for both trials and relevant reviews.²⁸ Other review articles and books were consulted. Bibliographies of all relevant articles were scanned. Experts in screening for older people and in ophthalmology were consulted.

Inclusion criteria
The inclusion criteria were in two stages. In the first stage, articles were included only if they were randomised controlled trials of either visual or multiphasic screening of unselected subjects in a community setting that included patients aged 65 and over. Trials of screening undertaken on selected groups of patients were excluded on the grounds that the results would not answer the question under review. Studies including adults aged under 65 only were also excluded. Because visual screening may have been only one small part of a multiphasic screening programme and data about visual outcomes may not have been included in published reports of trials, therefore, the named author for correspondence for all trials identified in the first stage was contacted (at their current addresses verified by telephone) to ask for any further unpublished data about visual screening tests used and visual outcomes. Non-responders were sent two reminders and were telephoned.

Study selection and data extraction
The full texts of articles selected for retrieval and review from the abstracts and citations were obtained. A checklist of the first stage inclusion criteria was then applied by both authors independently. All criteria had to be met. After contact with the trial authors a structured data extraction form was used, again by both authors independently. Data was extracted as follows: setting; subjects; overall intervention; visual screening test(s) used; duration of follow up; and visual outcome measures and visual outcome data. Data about randomisation process, the degree of allocation concealment, and blinding of outcome assessors was also extracted. Authors were contacted for clarification if required. Disagreements between reviewers were discussed and a consensus reached on all articles.

Critical appraisal
There is empirical evidence that inadequate randomisation to intervention and control groups can affect the outcome of a trial.²⁹ In particular, it is important that the treatment assignment is known only after the decision has been taken for the patient to enter the trial. This is known as concealed allocation. Studies were therefore classified for allocation concealment by using previously developed categories: A=trials in which allocation appeared to be adequately concealed (for example, a central randomisation procedure); B=trials in which allocation concealment was unclear, either because the approach was not reported or in which the randomisation process did not match categories A or C; and C=trials in which allocation concealment was inadequate (such as alphabetical use of surname).³⁰ In addition, data were extracted about the degree of blinding of outcome assessors in the trials.

Outcome measures
The aims of multiphasic screening of older people are broadfor example, to improve quality of life and reduce admission to hospital. Any benefit arising from the inclusion of a visual assessment, however, will necessarily be dependent on improved vision, which was therefore used as the outcome measure for this review. Any method of visual assessment (such as acuity chart, questions about vision, or measures of visual function) was accepted as an outcome measure. We felt that stricter inclusion would excessively limit the data included in the review.

Analysis
We used RevMan software.³¹ Heterogeneity between trials was tested for with ² test. Odds ratios were combined with the fixed effects Mantel-Haenszel method. The relative risk is presented as well as the odds ratio because it is more easily understood by some.³²

	Results

In total 2246 citations and abstracts were screened, and 147 full text articles were reviewed in detail. Seventeen trials met the first stage inclusion criteria, all of which were trials of multiphasic screening. The reference details are given on the web.^w1-17 There were no trials that primarily assessed visual screening. Requests for further information led to replies from authors of 16 of the 17 trials. Five trials met the final inclusion criteriathat is, visual outcome data with follow up of at least 6 months.^{w3 w9 w13 w14 w16} These trials are summarised in table 1.

All five trials adequately concealed randomisation. Regarding blinding of outcome assessors, the trial participants were aware of whether they had received a screening assessment. Thus, despite attempts to blind the outcome assessors, which arm of the trial subjects were in could clearly emerge during the face to face assessments of outcome. This phenomenon was noted by Vetter in two trials^{w3 w13} and led McEwan to consider such blinding impossible.^w9

Individuals in the trial by Wagner and colleagues were recruited from a health maintenance organisation and may have differed in their overall baseline health from the general population.^w16 This trial had three arms. For the purposes of our review subjects in the two arms who did not receive any form of visual screening assessment were analysed together as control patients.

Table 2 shows the visual assessment methods and outcome measures. All the trials used self reported measures to assess impaired vision, both as the screening assessment and as the outcome measure.

Effects of screening on visual impairment
Statistical testing showed no evidence of heterogeneity of effect in the five trials (²=0.96, df=4, P=0.92). When odds ratios were pooled there was no reduction in the proportion of individuals with visual impairment as a result of screening (pooled odds ratio 1.04; 95% confidence interval 0.89 to 1.22). The figure shows the outcome data, and individual and pooled odds ratios. The pooled relative risk for visual impairment was 1.03 (0.92 to 1.15).

View larger version (20K): [in this window] [in a new window]   Outcome data: individual and pooled odds ratios for visual screening as part of multiphasic screening package versus standard care. Outcome was not seeing well, as defined by each trial

	Discussion

The evidence from randomised controlled trials does not support the inclusion of an assessment of vision in regular multiphasic assessment programmes for unselected older people in a community setting or visual screening in primary care.² Although a reduction of 11% in the number of older people with visual impairment cannot be excluded, even this figure is disappointingly low.

Results from community surveys in the over 75 age group suggest that over half the visual impairment in this age group could potentially be reduced with treatment, notably by cataract surgery or refractive correction. ^{8 13} In light of this the lack of improvement seen in these trials is somewhat surprising and cannot be explained from the data presented. Several factors, however, may have contributed. Firstly, the visual assessment was only one component of the screening package in all five trials, and visual screening performed in isolation may have produced a greater effect. This hypothesis was previously suggested as an explanation of the lack of effectiveness of screening for visual impairment seen in a trial of a multiphasic screening assessment among middle aged men. ^{33 34} In clinical practice, however, screening for visual impairment is highly likely to be one part of a broader screening package, and therefore an assessment of effectiveness within a broader package is the most pragmatically useful measure. Secondly, the people who reported visual problems when prompted to do so in a screening programme may not have perceived their previously unreported visual impairment as a "need" for intervention. Lastly, there is evidence that considerable obstacles exist to improving visual impairment among older people. In the United Kingdom fear of costs has been repeatedly cited by a proportion of older people as a barrier to attending an optometrist and obtaining glasses. ^{3 14 24 35} In addition, ophthalmic services are unable to meet current demands for treatment, with long waiting lists for cataract surgery in many areas.³⁶

The aim of population screening of older people for visual impairment is presumably to discover visual impairment in those who are not presenting to the health services and to offer them interventions to improve their vision. Further work is required, however, to determine the needs of older people with unreported visual problems. Before population screening can be effective, existing obstacles to the reduction of visual impairment among older people must be overcome.