Results

PREVALENCE OF EYE DISEASES

A total of 1386 (94%) of the study children between the ages of 7 and 19 years were examined by the eye team; 141 pupils (10.2%) were found to have abnormalities or poor eyesight (VA<6/12) in at least one eye. An additional 73 (5.3%) pupils reported having night blindness (Table 1).

Bilateral poor eyesight (VA<6/12 in the better eye) was found in 10 pupils and unilateral poor eyesight (VA<6/12 in either eye) in an additional 14 pupils (Table 2). None had bilateral blindness (VA<3/60 in the better eye), but one child had visual acuity of 3/24 in the better eye (severe visual impairment), and five of the 24 pupils with poor eyesight (5/1386 = 0.36%) were blind in one eye (VA<3/60). The causes of unilateral blindness were cataract (two), macular scars (two), and evisceration (one). Significant refractive errors causing visual acuity less than 6/12, strabismus and amblyopia were uncommon; each had a prevalence of 1% or less (Table 1). All the 14 pupils with refractive errors were myopic. Refractive errors were responsible for nine of the 10 cases of poor bilateral eyesight (VA<6/12 in both eyes) (Table 2). In all 10 pupils with bilateral poor eyesight, visual acuity was improved to normal (VA⩾6/12) in at least one eye by either providing glasses for refractive errors (nine pupils) or by operating on the one pupil with bilateral cataracts. Refractive errors (n=5; 0.36%) and macular disease (n=4; 0.29%) were the commonest problems in the 14 additional pupils with unilateral visual impairment (VA <6/12 in one eye) (Table 2).

Seventy six children (5.5%) were diagnosed as having active trachoma (follicular trachoma with or without inflammation). In addition, one child had trachomatous scarring (Table 1). Half of the children with trachoma (38/77) came from one school (Sagani) (Table 3; prevalence by school p<0.001), which is in a poorer area with less access to water than the other two schools. The majority of pupils with trachoma (63/77=82%) were attending school year 1 to 3 (Table 3; χ² test for heterogeneity p=0.001), which was paralleled by 85% of the children with trachoma being less than 13 years old (Table 4). There was no difference in the prevalence of trachoma between boys and girls (p=0.8; Table 4). Children with trachoma usually came from poorer households (Table 4). For example, they were less likely to have a father in salaried employment (p=0.015) or to have literate parents (p=0.004), and the households they lived in less commonly owned a bicycle (p=0.016) or a radio (p<0.001). Children with trachoma were less likely to live in houses built with cement as opposed to mud blocks (p=0.002) or in houses roofed with iron sheets as opposed to thatch (p=0.010). All these differences were still significant after adjusting for age and sex. After adjusting for age, sex, and the other significant variables, children with trachoma were still less likely to come from a household that owned a radio or to live in a house built from cement blocks. Their household was more likely to own more than 10 cows. Death of the child's mother was of borderline significance for a higher prevalence of trachoma (p=0.05; 95%CI 1.01–13.76) (Table 4).

A relatively high proportion (5.3%) of pupils reported night blindness (inability to see well in low levels of light), and Bitot's spots were found in 0.6% of pupils (Table 1). However, none of the pupils who reported night blindness had Bitot's spots. As for trachoma, night blindness reports were commoner in Sagani school (11.0%) than in either of the other schools (3.5% and 4.1% respectively) (p<0.001). The prevalence of Bitot's spots was similar and always less than 1.0% in each of the three schools (p=0.381). Although corneal scars were found in 11 children (0.8%) (Table 1), only one pupil had unilateral impaired eyesight (VA=6/18) due to a corneal scar (Table 2), and bilateral corneal scars were rare (two pupils).

Two of the three pupils who had cataracts had visual deficit as a result (Table 2); one pupil had bilateral VA<6/12, and one unilateral VA <6/12. One girl was referred for further examination because of suspected glaucoma. Only five pupils (0.4%) were found to suffer from vernal keratoconjunctivitis (Table 1).

EVALUATION OF THE VALIDITY OF THE TEACHERS' EYE SCREENING

The teachers screened the eyesight of 1438 pupils; 1242 before the eye team's visit and another 196 pupils during the team's visit. The eye team rechecked visual acuity in 1386 of these children (96.4%).

If the results of the teachers' VA test and their question about the child's distance eyesight are combined, and if all pupils whom the teacher reported as having either unilateral or bilateral VA<6/12 are included, the screening by the teachers identified the great majority of pupils with bilateral poor eyesight (sensitivity: 80%, specificity: 91%). Using the VA test on its own, the sensitivity decreased to 70% (Table 5).

DATA QUALITY

In all, 168 pupils were randomly chosen for re-interview and repeat VA testing by a different VA tester. The extent of discrepancies between the two interviewers varied considerably from question to question, from 1/168 (0.6%; language spoken by parents/guardians) to 26/168 (15.5%; exact number of siblings). However, out of the 14 questions in the sociodemographic part of the questionnaire, a discrepancy of 10.0% or more was only found in one question (exact number of siblings). The interinterviewer discrepancies for the questions assessing subjective quality of eyesight (near/distance/in low levels of light) were all present in less than 5.0%.

When the ophthalmologist (SW) rechecked discrepant results with the pupils concerned, the answers sometimes agreed with those of one interviewer and sometimes with the other interviewer and, rarely, were changed again.

A discrepancy for visual acuity tests was defined as:

• any child having good eyesight in one test (6/6, 6/9, 6/12) but poor eyesight (<6/12) in the other test, or

• two lines or more difference when VA was tested by both testers as being less than 6/12.

 Using this definition there was only one discrepant result on the 168 repeat VA tests; one child with VA of 6/60 in one eye in the first test was reported to have VA=6/18 in that eye in the repeat test.

 A total of 130 children who had been examined by the ophthalmic nursing officer (ONO) were randomly chosen for re-examination by the ophthalmologist. Thirteen (10.0%) of these children had eye abnormalities and 117 had normal eyes (90.0%) diagnosed by the ophthalmologist. The ONO misdiagnosed only one child (who had trachoma) as having normal eyes (sensitivity: 12/13=92.3%; specificity: 117/117=100%) (Table 5).

 Out of the 1386 pupils examined, the eye abnormalities (not including visual impairment) found by the ONO were confirmed by the ophthalmologist in 90% of the children (129/143).

Discussion

DATA QUALITY

Pupils were reliable respondents for most variables. However, some of the pupils were unreliable respondents for some of the questions about their sociodemographic background. This probably reflected genuine problems with the questions. Detailed investigation by the ophthalmologist did not reveal any clear tendency for one interviewer to have received the “correct” answer more often than the other interviewers. It also revealed conceptual difficulties with the one question that had the highest proportion of discrepancies (exact number of siblings). Children genuinely did not know who to count as their “brother” or “sister” in their culture where these terms are used very loosely. The extensive quality control did not reveal any evidence of fabrication of results by the interviewers or visual acuity testers.

Quality control by the ophthalmologist (SW) of the assessments made by the ONO revealed a very high level of agreement.

REFRACTIVE ERRORS

In our survey sample of 1386 rural primary school children, the prevalence of bilateral impaired visual acuity (VA<6/12 in the better eye) was relatively low at 0.7%. Only four pupils (0.3%) had VA<6/18 in the better eye. Only one pupil had severe visual impairment (SVI), defined as VA<6/60 in the better eye, and no pupil was blind (VA<3/60 in the better eye). The school enrolment rate in Mwanza Region has been estimated between 53% (net enrolment rate) and 72% (gross enrolment rate).16 The prevalence of significant refractive errors may have been higher (or lower) in children who were not attending school. However, these results in primary school children were similar to those found in a population based survey of 7043 children 5–14 years of age in Botswana,17 where the prevalence of bilateral VA<6/18 was 0.2%.

In our study sample, refractive errors caused almost all of the cases of bilateral visual impairment (9/10), and a substantial proportion of those with unilateral impairment (5/14) (Table 2). Only pupils with VA<6/12 in either eye were refracted because the compliance with prescribed glasses was expected to be low in children with minor refractive errors.18 This may also explain the absence of hypermetropia, since only high degrees of hypermetropia lead to a decrease in visual acuity in children. In all our study children with refractive errors, visual acuity increased to within the normal range (6/12 or better) in at least one eye with glasses. Again these findings were similar to these from the study in Botswana where 15 of the 17 cases of bilateral poor eyesight were due to refractive errors and after correction all of these children had good eyesight, defined in that study as VA⩾6/18.17

Although differences in definitions complicate comparisons, the prevalences of impaired visual acuity and of refractive errors found in these two African surveys were lower than those in surveys reported from non-African countries.5 19-21 Bilateral visual impairment (VA<6/18) was found in 3.1% and SVI (VA<6/60) in 0.5% of urban Indian schoolchildren.19 Similarly, the prevalence of refractive errors was high; using a definition of refractive errors of ⩾0.5 dioptres, the prevalence of myopia was 8.6%, hyperopia 22.6%, and astigmatism 10.3%.19 Ten per cent of 12 853 ten year old children in the UK were found to wear glasses, but only 80% of the children who wore glasses had uncorrected VA of less than 6/9.20 In a study of Chinese schoolchildren in two primary and two secondary schools, myopia of ⩾0.5 dioptres was detected in up to 50%, and 24% had unaided visual acuity of 6/12 or worse.21

These differences may be due to racial/ethnic variations in the prevalence of refractive errors, and/or due to different lifestyles or living conditions. The prevalence of myopia in African American children in an inner city school in the USA (preschool to second grade) was found to be as high as in children in a predominantly white community, whereas a national survey done 20 years earlier had found that the prevalence in African American pupils was almost half that observed in their white peers.5 This finding implies that differences in lifestyles (for example, reading, watching TV, or computer visual display units), living conditions (for example, nutrition) or medical care (for example, unnecessary or overcorrection of refractive errors which may worsen the refractive error by inhibiting natural “emmetropisation”) may be more important than racial/ethnic differences.2 Studies on children from families of high socioeconomic status in African cities might shed further light on this issue.

TRACHOMA

Trachoma was classified as being a problem of public health importance in only one of our study schools (Sagani) where the prevalence of active trachoma was 15.5% (Table 3). Sagani is the poorest of the three communities and is furthest from Lake Victoria. As expected from previous studies,22-24 active trachoma was commonest in young children attending the first three school years (Tables 3 and 4), in children from the poorest households, and in those coming from households with large numbers of cattle (Table4).25 The prevalence of trachoma was similar in girls and in boys (Table 4).

VITAMIN A DEFICIENCY

Night blindness was reported by 5.3% of pupils in our study sample (Table 1).

However, further investigations are needed to confirm whether these reports were valid and the result of vitamin A deficiency, especially since no specific term seems to exist for night blindness in Sukuma, the main local language. However, the fact that eight pupils had Bitot's spots (0.6%) indicates that xerophthalmia must be present either in this age group or in younger children. The frequency of lesions that are unresponsive to vitamin A and therefore not due to recent vitamin A deficiency increases in school age children.26 27 This may explain why the pupils with Bitot's spots had no other evidence of xerophthalmia. In school age children the extent of active xerophthalmia and its clinical importance are poorly understood. Studies of the prevalence of xerophthalmia in school age children that have been conducted in areas where it was known to be present in younger children have reported prevalences of Bitot's spots ranging from 1.0% to 9.2% of children and night blindness prevalences ranging from 0.0% to 3.6%.13 28-31 Further studies are planned in Mwanza Region of the validity of reported night blindness using a scotopic vision tester32 and on the prevalence of active xerophthalmia and subclinical vitamin A deficiency in this age group and in younger children.

TEACHERS' EYE SCREENING

Teachers have been used in India to screen visual acuity in schoolchildren.2 33 In our survey, selected teachers were trained to test visual acuity and then asked to screen all the pupils in their schools. However, some of the teachers delegated part of the screening to untrained colleagues. Also in one school about 40% of the children had not been screened before the arrival of the eye team, while in another school data were available for all the pupils, even the ones who had left the school long before the survey, so data must have been fabricated for these pupils. With strict supervision to make sure that only trained teachers do the screening and to try to eliminate fabricated data, the sensitivity for the teachers' screening (VA test and one question on distance eyesight) could well be higher than 80%. Even VA testing on its own may then be a sensitive enough screening test. However, any screening by teachers or others will depend not only on the accuracy of the screening, but also on the prevalence and the public health importance of poor visual acuity in the population screened.

POLICY AND PROGRAMME IMPLICATIONS

Many factors need to be considered when deciding whether or not to introduce an eye screening programme in primary schools. These include the prevalence and health, educational, or work impact of poor eyesight within this population, the human and financial resources available for screening, the cost and effectiveness of the screening and the treatment given, and the availability and compliance with any treatment offered. No studies of the cost effectiveness of primary school screening for eye diseases have been reported from developing countries. Furthermore, there are no data on the impact of visual impairment, as opposed to blindness, on school or work performance in the context of rural African villages similar to the ones in this survey. Given this lack of objective data, any judgments related to whether or not to introduce a primary school eye screening programme are necessarily subjective. Several African countries have a policy that ophthalmic nurses should conduct regular eye screening in primary schools, though this policy is rarely acted on. In our opinion, the prevalence of significant refractive errors that caused bilateral visual impairment (VA<6/12 in the better eye) was too low at 0.6% to be the sole justification for a screening programme. In one of the three schools, trachoma was common enough at 15.5% to justify regular screening and treatment. However, since the prevalence was much lower than this in the other two schools (3.8% and 2.8%, respectively), a further survey including a larger number of schools is required before a definite recommendation can be made as to whether trachoma is sufficiently prevalent to justify a large scale school eye screening programme in this area of Tanzania. As discussed above, further studies are needed to establish whether vitamin A deficiency is a problem of public health importance in primary school children in Mwanza Region. Even if this is so, this would not necessarily be a justification for routine eye screening, since the likely intervention would not be targeted at individual pupils but would be universal vitamin A supplementation to all pupils. In areas where a school screening programme is introduced, teachers could do the VA screening as long as they are well trained and supervised.