8. Photoscreening

Chapter editor: Anna Horwood

a. Introduction

Vision screening can be broadly divided into testing for the primary signs of amblyopia and low vision (by testing visual acuity and eye alignment), which is a skilled task and imprecise in children under four years of age¹; or looking for the risk factors for these conditions (significant refractive errors, strabismus and media opacities) by objective and semi-automated methods such as autorefraction or photoscreening. Screening using stereotests to look for reduced stereopsis which can indicate amblyopia, is another form of screening (see chapter 7 section a-ii). The rationale behind early screening for risk factors is that by correcting them early, amblyopia can be prevented or mitigated. Semi-automated risk factor screening can be done earlier than VA testing and sometimes more easily.

Autorefraction takes images of a child’s eyes, which are then analysed by software to estimate refractive error and sometimes evaluatee eye alignment. Autorefraction frequently arrives at an actual measure of refraction i.e. a glasses prescription, but in the context of this manual we will be discussing photo- or auto-refraction only as used in the screening context and will be referred to as “photoscreening”. Photoscreening can be used on much younger children than visual acuity measurement (which requires much more cooperation from the child), because the only cooperation required of the child is to briefly look at the camera.

A major decision for commissioners of services is whether vision screening is targeting low vision, or for the larger number of children with[popup_anything id=”3321″]for low vision, or both. Although early and automated testing can seem an attractive option for those funding public healthcare, relative costs and benefits over the whole patient journey may be less clear cut. It is very important for funders and planners of services to understand the controversies.

As mentioned in the previous chapter, the EUSCREEN Country Reports suggest that in many cases photoscreening is being added to some existing local screening services for the younger children for whom VA testing is imprecise. Thus within a country, some services will be looking for low vision, while others will be looking for not only low vision, but its risk factors as well. Where private providers (e.g. pediatricians) are screeners, it is frequently offered as a ‘billable extra’ to mandated tests, so children whose parents chose to pay for this option are screened and treated for risk factors, while other children will be screened for low vision only. This makes any local or regional audit very complex, and reduces the equity of opportunity that are WHO and EU aims, and also risks introducing a profit motive for providers of screening or care. Both photoscreening and visual acuity measurement have merits, so decision makers must decide which is most appropriate for their situation. Setting up either type of screening service will involve many of the stages outlined in chapter 9, but this chapter outlines some of the issues that may influence which modality to choose. The EUSCREEN model will help compare relative costs, particularly the significant effects of adding photoscreening to existing services.

b. Photorefractors and autorefractors

Most automated screeners use the principle of photorefraction. Autorefractors usually test one eye at a time, while photoscreeners generally test both eyes at the same time, so can also detect some types of strabismus. Early studies with equipment no longer available began in the early 1980’s² but more devices have been commercially available since the mid-1990s. The child simply looks at some form of sensor (usually infra-red) and an algorithm calculates an estimate of refractive error from the characteristics of the retinal reflex produced from an infra-red light source. Photoscreeners can be set up to give an estimate of actual refractive error (i.e. used as an autorefractor and the refraction reading used by the screener to decide whether referral thresholds are exceeded) or as a pure screening tool to be administered by lay screeners with a simple ‘pass/refer/untestable’ result. Most are supplied with factory-set referral criteria, based on evidence-based guidelines such as those recommended by American Association of Pediatric Ophthalmology and Strabismus (AAPOS)³. These can result in unacceptably high referral rates and some services have adjusted these to optimise results, for example in Germany and Flanders. In most cases these settings can be adjusted if different levels of sensitivity and specificity are required, or if the target condition is not only amblyopia risk factors, but also specific levels of refractive error, for example developing myopia.

Photoscreening can sometimes be carried out with additional lenses or filters to extend the equipment operating ranges or compensate for sub-optimal light levels in the testing environment.

A newer method of automated screening – birefringence scanning – uses a different principle based on detection of foveal fixation, which is generally defective or eccentric in an amblyopic eye⁴. This test has potential to be more specific for amblyopia and strabismus, but at the time of writing there still are limited published data. It should be noted that new technologies are being developed such as eye tracking systems, that could possibly be used for vision screening in the future, although at this point little is known about the public health advances these technologies offer.

All photoscreening methods are designed to be child-friendly, portable and to be operated by minimally trained personnel. Testing only requires a child to look steadily for long enough for the reading to be obtained. This usually takes a few seconds, so it is often possible to get an accurate estimate of refraction even in infancy.

There is a large literature on photoscreening and autorefraction, but it is important to note that sensitivity, specificity data are generally reported in terms of success in detecting the risk factors for amblyopia, not low vision or amblyopia itself. Visual acuity screening literature, on the other hand, reports on detecting the conditions themselves. Therefore, direct comparisons between photoscreening and visual acuity measurement are frequently difficult.

c. Early versus later detection of amblyopia

There is strong evidence to show that amblyopia can be prevented, and is more easily treated, earlier in the critical period of visual development, and certainly before seven years of age⁵. There is little doubt that all screening for amblyopia before this age reduces preventable and generally permanent loss of vision.

Detection of risk factors is often advocated in order to detect and treat amblyopia earlier, before the amblyopia is so firmly established, or where skilled screeners are not available. Both visual acuity screening and earlier photoscreening will detect amblyopia before it is too late to treat, but the argument between early versus late (for example two versus four years) screening is much less clear than the argument between screening and no screening at all. There is some evidence that earlier detection has some advantages in terms of somewhat better outcomes, more rapid or easier treatment of amblyopia, and prevention of a few cases of strabismus, but these relative advantages are more modest⁶ (see previous chapter). For example early detection may result in a line better on a vision chart, a shorter period of wearing an eye patch or a small number of cases of strabismus prevented. These advantages could also be counteracted by more hospital visits or family difficulties caused by years of enforcing many reluctant young children to wear glasses or patches.

The wider availability of photoscreeners has highlighted the debate between earlier versus later detection of amblyopia and refractive errors.

More very young children will have amblyopia [popup_anything id=”3321″]than will go on to develop reduced visual acuity, amblyopia or significant refractive errors. This is because some young children have refractive errors which will resolve spontaneously due to[popup_anything id=”3534″]in the first years of life⁷. This is particularly the case for hypermetropia (long sight) and astigmatism in infants which may not persist into later childhood. Emmetropisation is still very active in the second year of life so many infants have moderate refractive errors that will resolve. Beyond two years of age fewer children with a significant refractive error will grow out of it completely, but it may reduce to levels below a screening threshold⁸. The degree of emmetropisation varies among children and some children may even have increasing hypermetropia and these in particular would be prone to develop accommodative strabismus and amblyopia⁹. It is possible that children with increasing rather than decreasing hyperopia or anisometropia are particularly at risk of amblyopia and strabismus.

Some refractive errors are more ‘amblyogenic’ than others, such as hypermetropia over approximately +3.00D, hypermetropic anisometropia (one eye more long sighted than the other), and significant astigmatism. Mild hypermetropia or myopia (short sight) more rarely lead to amblyopia. Myopia is rarely present in early infancy, typically develops in later childhood and adolescence, and myopic children have clearer vision for near so rarely develop amblyopia. Hypermetropia is a particular screening problem because not all even significantly hypermetropic children (e.g. +5.00D), will have low vision¹⁰ and it can also be missed by photoscreening¹¹; although they might struggle with prolonged close work.

At the time of writing, we do not know the relationship between the presence and level of early refractive risk factors and the likelihood of developing amblyopia in an individual child. We also do not know how much glasses for mild refractive error in the pre-school years help general development or lead to better long-term outcomes; or conversely, lead to more stress, cost, or social stigma.

d. Limitations of the evidence base

i. Photoscreening to detect refractive error

The availability of earlier detection by photoscreening has highlighted some deficiencies in the evidence upon which decisions must be made.

The World Health Organisation offers guidance for conditions for which screening is recommended (see chapter 2a). While amblyopia fulfils most of these, because it is preventable and must be treated in the critical period of visual development, it is less clear that these criteria apply for early detection of refractive error which does not cause amblyopia, such as mild[popup_anything id=”3366″]or[popup_anything id=”3362″].

Vision concerns may differ across the world. The distribution of refractive errors varies between ethnic groups and national priorities may vary. For example, myopia is very common in East Asian populations¹²; hypermetropia and amblyopia more common in Caucasian populations; and astigmatism common in South Asian and some native American populations. Decision-makers considering what form of screening to adopt should consider their local distribution of refractive errors and thus their prevalence of amblyopia.

Refractive error is largely unpreventable, and there is only weak evidence at the moment that mild uncorrected refractive errors degrade school performance in the very early years of schooling (although future research may change this opinion in due course).

ii. Stand-alone photoscreening or combined with visual acuity testing

Much of the published literature reports the use of photoscreening as a stand-alone test, often only administered at one time point in a child’s life¹³. Most reports do not consider uptake of follow up for accurate diagnosis or treatment outcomes. Consideration of photoscreening versus other screening modalities such as visual acuity testing is rare. In particular, the total cost of a patient journey, or lifetime costs, in relation to differences in outcome for early vs later screening, is rarely considered.

In contrast to the impression of photoscreening use reported in the literature, the EUSCREEN data show that, in specific regions in many countries, photoscreening is used as an additional test within a combination of tests, or carried out on very young children before subsequent visual acuity testing carried out when older, or is repeated. Adding photoscreening rarely replaces tests already being done, and is often done by more expensive testers trained to also test vision. If paid for as an additional private test, there can be an unfortunate incentive to repeat photoscreening in some practices. This may lead to higher costs for both screening and treatment. Adding photoscreening is likely to lead to more, and earlier referrals. More children are given glasses, both to correct refractive error and to try to prevent or mitigate amblyopia development. These children are then often kept under observation and treatment until their VA can be tested later, and are often screened again with a VA test, though many of these children now are already undergoing treatment. Early referral of children with amblyopia risk factors means that children also need more specialist visits because they may still be followed to visual maturity, whatever the age they are referred. Many more glasses will be prescribed to young children: after five years of photoscreening in Flanders, Belgium the number of 4-year-old children wearing glasses had risen from 4.7% to 6.4%¹⁴. Although treatment outcomes may be slightly better, this is by a smaller margin than might be expected¹⁵ ¹⁶.

Decision makers must weigh the relative advantages and disadvantages of visual acuity screening which is only accurate in children over 4 years of age versus earlier or concurrent photoscreening. Some of the costs and disadvantages of photoscreening may be long term or more hidden because long term treatment and outcome data lies outside public health databases or control. The EUSCREEN model may help inform these decisions.

Think box

If you photoscreen at age 3, many children will be referred because of risk factors (and some due to untestability). Many of these children will then be given glasses, or observed until their VA is testable at around 5 years of age. Some amblyopia will probably be prevented, but many children move early from the community into the secondary referral system and start incurring treatment costs.

At age 5, most of the severe cases (those with refractive risk factors already screened at 3, plus those obvious strabismus, low bilateral vision which present to healthcare due to parental concern) will now already be under treatment, so additional VA screening will now only pick up small numbers of additional mild problems, and from a population with (now) a lower prevalence of amblyopia because the early treatment prevented a proportion of cases.. The cost of this second testing, for a relatively small number of additional cases detected, may make this “double screen” the most costly combination of all. Although re-screening children already under treatment is unnecessary, it can be more administratively challenging to find out who NOT to screen in a class, especially if children are not wearing their prescribed glasses, so many children are screened twice. Some tests and even referrals may be superfluous, but still incur costs (extra administration, same staff, travel, equipment).

e. Advantages and disadvantages of photoscreening

i. Advantages of photoscreening

Testing is often possible at any age from infancy, so this is a clear advantage over visual acuity testing which is only accurate from around 4 years of age (see discussion in previous chapter).
Earlier referral of at risk children, so amblyopia and strabismus outcomes may be better.
Each test takes only a minute or two so many children can be tested in one session. However, some highly efficient VA screening services delivered by orthoptists can also test many children in a session¹⁷.
Modern commercially available photoscreeners are designed to be administered by minimally trained screening personnel so staff and initial training costs are much lower. Where skilled testers who can gain experience in testing many hundreds of children are not available, photoscreening may be the only viable option.
Costs per screen are reported to be low because of rapid testing time, and being delivered by lower paid operatives, especially if equipment costs are not considered¹⁸.
Automated, so greater consistency between screeners.
Photoscreening will detect refractive error as well as risk factors for amblyopia. This may be a consideration if refractive error is included as one of the target conditions for the screening
Estimates are possible of refractive errors in anisometropia, myopia and astigmatism without needing dilating eye drops, which are the gold standard requirement for accurate refraction in children, but not possible in a screening situation.
An oft-heard argument for early detection and correction of refractive error is that it will aid general development and educational attainment. While this seems likely for the few children with severe vision problems who cannot access the size of print and close work tasks they need to do, it has not been established if a delay of a year or two in correction of milder refractive error carries any proven long-term harm if they can still engage in age-appropriate toys and print, even if it is slightly blurred.

ii. Disadvantages of photoscreening

Equipment, maintenance and replacement costs are higher than for visual acuity screening (a photoscreen device can cost up to €7,500 and may need replacing every few years, many times more expensive than a VA test). Some companies offer rental or lease arrangements for equipment. If many screening sites can share one device, this cost may be acceptable, but the device needs to be carried, and set up every time, from site to site, increasing wear and tear. If one device per site is needed (for example if screening is carried out by paediatricians, every paediatrician in a city will need one), costs increase dramatically and many hundreds of devices will be required.
Photoscreeners are not as accurate at measuring hypermetropia compared to the other refractive errors. It can be missed or underestimated if a child compensates for their refractive error by accommodating (focusing) briefly during the test. Thus, one of the major risk factors for strabismus and amblyopia can still be missed¹⁹.
Other conditions known to cause low vision will not be detected by photoscreening such as nystagmus, microstrabismus, retinal abnormalities.
There will be many referrals per target case of amblyopia²⁰. Referral rates from photoscreening in young children are frequently reported to be close to 20% or more²¹, compared with around 5-8% for good visual acuity screening services, in populations where the prevalence of amblyopia is around 3%.
A substantial amount of young children will be untestable because they will not look at the sensors for long enough for a reading to be obtained. The ‘untestable’ rate, if tested before 2 years of age, can exceed 12%²². Most screening programmes would repeat screen or refer these children for diagnostic assessment, leading to additional costs.
Some children with risk factors will not develop amblyopia and will eventually be discharged without treatment. This is particularly significant if babies are tested, because a larger proportion of them will emmetropise. They will need comprehensive testing and often follow up before decisions to treat or discharge are made. A significant proportion of children will never receive glasses or only get them at a later date²³. Follow up and treatment costs may therefore be high in the longer term.
Children referred from photoscreening with genuine amblyopia or low vision enter healthcare services earlier. Because, whatever the age of referral, children with amblyopia need supervision until the[popup_anything id=”3338″]of visual development is over, they will need more visits to more expensive services, for possibly only modest improvement in outcome.
Services need to be willing and able to receive, treat and monitor the larger number of referrals. In countries where skilled paediatric ophthalmic services are limited, very high false positive rates and the many children with milder refractive errors could be a burden on already stretched services. Public health thresholds for concern may differ from professional standards. Once referred, ophthalmologists may apply lower treatment criteria than those used for screening referral²⁴, so children may eventually get glasses for a condition that would not have failed screening. For example a child referred as an untestable baby might be given glasses for -1.00D of myopia at age four, which is not an amblyopia risk factor and would be unlikely to hold them back at school at this age.
Equipment limitations.
- Data capture may be inaccurate or impossible in the case of ambient lighting levels being too high/low or if the child has small/large pupils
- Most photoscreeners have a limited operating range to detect large refractive errors if an estimate of refraction is required
- False readings can occur due to certain eyelid features or curly eyelashes obscuring the pupil margins
- Some specific pigmentation of the iris and retina may make the pupil margins undetectable to the camera and the estimate of refraction unreliable
- The photoscreeners use internal mean calibration factors informed by validation studies. There may be individual differences in calibration factors, which may differ between ethnicities with lighter and ethnicities with darker eyes
- These latter points all increase false positives or repeat screens. The last three points may mean that some photoscreening is less reliable in some non-Caucasian populations
Parents may be less likely to follow up the screening visit if told the child “might have a problem”, rather than if they are told (or see) that the child actually cannot see. This is particularly important in countries with low awareness of the importance of children’s eyesight, low trust in healthcare providers, poor cultural acceptance of children in glasses, suspicion of a profit motive or logistical difficulties in accessing specialist follow up where parents might need to be persuaded to make a long expensive journey to a specialist.

f. Current controversies

Do the visual, social and developmental benefits of early detection and referral for a few children outweigh the additional equipment, referral and treatment costs associated with lower[popup_anything id=”3349″]and longer treatment/observation times for many others?
How much does earlier detection reduce the long term prevalence, incidence, severity and treatment times of amblyopia cases?
In particular, does one good visual acuity assessment which does not depend on parents keeping screening appointments, at an age when it is accurate lead to worse public health outcomes than more frequent, additional or earlier photoscreening?
Should photoscreening be repeated, and if not, what is the optimal testing age? Anecdotally, many experts in screening consider that multiple photoscreenings improve sensitivity and specificity, but more visits mean higher costs.
Photoscreening may be an option when no skilled testers are available, but could more investment in training a dedicated group to be skilled VA testers be an alternative option? This could lead to lower long-term costs by reducing false referrals?
If refractive error and amblyopia are the target conditions for vision screening, how important is it to correct modest refractive errors which make only a line or two difference on a vision chart, but would not prevent a child’s activity in the preschool years, compared to on school entry?
When children enter formal schooling at different ages, does the age that glasses are prescribed make a visual and/or educational difference?
There is some evidence that the trajectory of emmetropisation could be a stronger predictor of amblyopia than a single photoscreening result. In other words, a child who is growing out of their hyperopia in their first years is less at risk of both strabismus and amblyopia than a child who is not, or where it is even increasing. This implies that two measures of refractive error one or two years apart might help identify these children and would argue for repeated photoscreening from infancy into childhood. However, there is limited data to indicate the cut-off points for the increased risk, and photoscreening is particularly bad at detecting precise measures of hyperopia. An “increase” of 1.00D in hyperopia over a year could be genuine, but it could equally be test-retest variability as the child accommodated more for the first test than the second. Costs of the additional screenings would be incurred, and many more children would need full diagnostic assessment and follow up, so at the moment this does not appear to be a cost effective approach to screening. More data is required in this area as it is likely to be a high cost-option and a clear benefit of such an approach would need to be demonstrated.

g. Questions a potential buyer of a photoscreener should ask if considering purchase

What is/are the target condition(s)? Is the priority to detect amblyopia or to detect refractive error?
What referral criteria are the factory settings? Can they be tailored for your population, the age of the children being tested, and screening requirements? In the case of Flanders in Belgium, the referral criteria needed to be adjusted following excessive referrals.
Is the local population likely to be willing or able to access follow-up and treatment if their very young child is referred?
Are there robust procedures in place to evaluate follow-up, prescribing practices and long term outcomes in terms of uptake, improvement in vision, acceptability and costs post-referral?
Will adding photoscreening to your existing screening programme improve long term outcomes? Who will bear the costs of additional equipment and higher referral rates, and are increased costs justified?
Are local care providers willing to see a high proportion of false referrals in young children which may occur?
Which photoscreener? What is the operating range for accurate estimates of refractive error? How tolerant of different pupil size, light levels, child eye colouring is the equipment?

Think box

Photoscreening

– Expensive equipment

– Lower cost staff and training

– Earlier referral

– Many more referrals

– More false positves and equivocal responses

– Marginally better outcomes overall (particularly for a few severe cases)

– Increased treatment and monitoring costs

– Resource intensive for healthcare post-referral

Later VA screening

– Cheaper equipment

– Needs skilled testers with more intensive training

– Later referral and treatment

– Lower referral rates

– Fewer false positives or equivocal referrals

– Shorter treatment period and lower treatment costs

– Treatment still possible, but starts later so marginally worse overall outcomes than from earlier referral (especially for a few severe cases)

h. Conclusions

Photoscreening is an alternative approach to vision screening and appears to be a low-cost and attractive option to decision-makers and commissioners of screening services. The available literature suggests, however, that visual acuity screening beyond four years of age is more cost effective overall, especially if public funds subsidise not only the screening but also subsequent treatment costs. Photoscreening referral rates are high, many referred children will not be amblyopic, or not be given glasses immediately; and some amblyopic and hypermetropic children will still be missed²⁵. Higher referral rates, poorer positive predictive value and longer treatment times for early photoscreening referrals, may load higher costs onto the whole patient journey in the long-term; currently without clear evidence of significantly better outcomes at a public health level. The EUSCREEN interactive model will help in this decision-making. By using different combinations of visual acuity measurement and photoscreener episodes it will quickly become clear that its costs are always higher, primarily because the machine costs 100x more than a VA chart and secondarily because treatment costs are higher due to longer treatment times and more children being given more glasses.

Adding photoscreening for infants and children to existing services testing VA when older may be the most costly of all, because more children will be referred earlier, falsely referred and treated and observed for longer, while the community costs of VA screening later are still incurred.

next chapter