Sensitivity and specificity of teachers for vision screening among primary school children in South India.

AIMS: This study aims to determine the sensitivity and specificity of vision screening by school teachers among primary school children. SETTINGS AND
DESIGN: Prospective nonrandomized clinical trial. SUBJECTS AND METHODS: The study was carried in primary school children of Madurai, Tamil Nadu from April 2007 to October 2007. Sixty-five primary school teachers from 57 schools around Madurai, Tamil Nadu, India, underwent a half-day training programme at the hospital. Each teacher on return to the school screened around a hundred children using the tumbling E. All screened children were subsequently screened by a team from the hospital. The optometrists used tumbling E (Snellen's 20/30 and Early Treatment of Diabetic Retinopathy Study (ETDRS) 20/32) to check the vision of all children. Any child complaining of defective vision or noted to have defective vision on screening underwent a comprehensive ophthalmic evaluation. The sensitivity and specificity of teacher screening was assessed based on the data obtained.
RESULTS: Screening was completed for 5150 children. The prevalence of vision <20/30 was 2.82%. Teacher screening was noted to have a sensitivity of 24.8% and a specificity of 98.65%. Tumbling E ETDRS screening by the optometrist had the highest sensitivity of 94.48% and specificity of 97.09%.
CONCLUSIONS: Our study shows that vision screening by trained optometrists is a very useful tool to identify visual impairment in primary school children. Measures need to be taken to improve the sensitivity of teacher screening before recommending its generalized use.

Introduction

India is home to 407 million children below the age of 16 years, which accounts for about 40% of the population.[1] India has the largest burden of childhood blindness in the world. It is estimated that there are 6,80,000 blind children in India including 2,26,440 with uncorrected refractive error (RE).[2] Between 16.3% and 37% of this blindness is treatable or avoidable.[3] Most studies have identified uncorrected REs as the major etiology for visual impairment in children.[4] This stresses the need for a screening program, preferably at schools. While many school screening programs are in place, there is little data on their impact.[5] The problem is further compounded by the inequitable distribution and shortage of pediatric ophthalmology units in India.[6]

Our hospital has been running a successful school screening program for 20 years. Screening is done by trained school teachers. The defective children identified in this preliminary screening are shortlisted. On a mutually agreed date, a visit by the hospital team comprising of an optometrist, pediatric ophthalmologist, assisting nurses and support staff is arranged. Children identified by the school teachers as defective are subjected to a vision screening (by Snellen's Chart), refraction (cycloplegic refraction when required) and a comprehensive ophthalmic examination. A decision on glass prescription is made by a pediatric ophthalmologist. The concept of teachers screening for diseases is an attractive one, because teachers are closely involved with children, and can help in disseminating information about visual health in the community surrounding the school. This model is low cost and has been successfully tried in parts of India, China, and Brazil among others.[578] However, before this approach is recommended for wider use, it is important to know accurately the sensitivity and specificity of the screening carried out by teachers. The present study was designed for this purpose.

Subjects and Methods

The study was carried out from April 2007 to October 2007 in primary schools of rural Madurai, Tamil Nadu, India. The study adhered to the tenets of Declaration of Helsinki. As a preparatory step, a school screening team comprising of a project director, pediatric ophthalmologist, project coordinator, organizer for outreach activities, and school eye health coordinator was formed.

After necessary planning, the school eye health coordinator cum ophthalmic assistant visited each school to brief the principal/headmaster about the program. Following the visit, teachers willing and able to participate in the program were identified for training.

Minimum sample size

To detect a 5% difference between the proportion of children deemed “passed” among each pair of the three arms (Teacher vision screening, Vision screening with Early Treatment of Diabetic Retinopathy Study [ETDRS], and Snellen acuity charts) with 95% confidence and 80% power, we needed to study 1000 children. The “Phase 2,” where we intend validating screening for REs against objective cycloplegic refraction requires 655 children to be examined to detect a 5% difference in “passes” with 95% confidence and 80% power.

Training of school teachers

Sixty-five primary school teachers from 57 schools around Madurai, Tamil Nadu, India, were invited to a half-day training program at our hospital. The training was given by consultants and optometrists from the pediatric ophthalmology department along with the school eye health coordinator. The first half of the training program was devoted to an introduction to the anatomy of the eye, its functioning, common eye problems in school children, and the importance of recognizing these problems at an early stage. Slides, posters, videos, and actual cases were presented in addition to didactic lectures. The school teachers were also briefed about the aims and objectives of the study. The second part of the program was devoted to practical training, on how to measure visual acuity using Snellen's chart. This was first practiced by the teachers among themselves and then on consenting patients from the outpatient clinics under the supervision of a senior orthoptist. The school teachers were provided with a tumbling E chart and a rope of 6 meters length (instead of an inch tape) to measure the distance between the chart and the child being screened. They were provided with a form to fill in the details of the child. Each teacher was expected to screen one hundred children.

Teacher screening

From a distance of 6 meters (measured with the rope provided), the child was presented with a vision chart exhibiting four black “E” of a standard size (20/30 of Snellen's chart). They were instructed to wear glasses if he/she had one. With each eye, the child was asked to indicate the direction of the open end of the “E” pointed out by the teacher. Students with 20/30 or better in both eyes were marked as “Good Vision;” those demonstrating <20/30 even in one eye, as well as doubtful cases were marked as “Not Good Vision” in a separate format. Children from classes 1–5 were selected for the study. The respective school authorities were responsible for informing the coordinator once the entire school was screened. A form was provided to the teachers for marking as normal or abnormal.

Screening by the hospital team

On another day, no later than a month after teacher screening was completed, the hospital team-comprised of a pediatric ophthalmologist, refractionist and a patient counselor-screened the entire batch of children. The team carried direct ophthalmoscopes, streak retinoscopes, and all the necessary instruments and equipment for a comprehensive ophthalmic examination. A request was communicated to the parents to be present on the day this screening was carried out.

Four optometrists with >5 years of experience were included for the study. Before commencing the study, a pilot study was done to test the agreement of the optometrists on a sample of 30 school children. All of them tested the vision of 30 school children with the Snellen's and ETDRS charts. There was excellent agreement among them for in their ability to screen RE using Snellen's chart (Kappa 0.8; P < 0.001).

All children underwent vision screening by the optometrists using a tumbling Snellen's E 20/30 and ETDRS E optotype 20/32 here. Any child complaining of defective vision or noted to have a defective vision by either the teacher or the optometrist (either the Snellen's or the Tumbling E) underwent dynamic refraction and vision testing with a Snellen Letter chart. If the child's parent's consented, cycloplegic refraction was carried out on the children suspected to have a RE. Those requiring postmydiatric tests were asked to come to the main hospital. A final decision on spectacle prescription or other treatment was made by the pediatric ophthalmologist. Children having Hyperopia >3.0 D, Myopia >0.5 D and Astigmatism >0.5 D were labeled as having significant RE and were prescribed glasses. The term “abnormal” used in the present study refers only to defective vision in either eye with or without a RE on a comprehensive ophthalmic examination. All data were entered into the school study form [Supplement 1] by the optometrists. The marking as normal or abnormal on the school study form was done by the authors both of whom were pediatric ophthalmologists.

During the visit by the hospital team, an optical dispensing unit from the hospital also accompanied the team. Spectacle orders were taken, lenses ground, fitted with frames, and dispensed at the screening site at a nominal cost of Rs. 50/-per spectacle. Free spectacles were provided to children who could not afford them. Minor problems such as allergic conjunctivitis, convergence insufficiency were treated in the camp itself. Children requiring further treatment for other ocular ailments were asked to come to the main hospital.

Results

The screening was completed for 5150 children in 56 schools. Thirteen children were excluded because of incomplete forms, and another seven were excluded for having missed one of the two screenings. A total of 145 children (2.82%) were noted to have abnormal vision. Of these, 14 children were myopes, 84 hyperopes, and 47 had astigmatism. Among these 145, one child had optic disc pallor in both the eyes, one had strabismsmic amblyopia, and yet another had ammetropic amblyopia with high myopia.

The agreement between teacher screening, optometrist screening, and examination result is depicted in Tables 1–3. Teachers identified 5027 children as normal (109 actually abnormal) and 103 as abnormal (only 36 actually abnormal and 67 normal). The sensitivity was thus 24.8%. The specificity worked out to be 98.65% [Table 1]. ETDRS screening by the optometrist was noted to have the highest sensitivity of 94.48% and a specificity of 97.09% [Table 3]. Optometrist screening with the Snellen chart had a sensitivity of 82.76% and a specificity of 97.9% [Table 2]. The specificity was high across all the three groups. Other disorders (n = 10) picked up during the camp that required referral to the main hospital included six squints (two Duane's retraction syndrome, two exotropias, and two esotropias), one high myopia, one pale disc, and two patients with high cup-disc ratio.

Table 1

Agreement between teacher’s screening and examination result

Teacher’s screening		Examination normal	Examination abnormal		Total
Normal		4918	109		5027
Abnormal		67	36		103
Total		4985	145		5130
Prevalence				0.028265 (95% CI: 0.023981-0.03327)
Sensitivity				0.248276 (95% CI: 0.182022-0.328149)
Specificity				0.98656 (95% CI: 0.982856, 0.989491)
Predictive value positive				0.349515 (95% CI: 0.259956, 0.450454)
Predictive value negative				0.978317 (95% CI: 0.973802-0.982083)
Kappa teacher screen result
Expected (%)		κ	SE	Z	Probability >Z
Agreement	Agreement
96.57	95.28	0.2733	0.0138	19.87	0.0000

Kappa statistic: 0.2733. CI: Confidence interval, SE: Standard error

Table 3

Agreement between Early Treatment of Diabetic Retinopathy Study chart and refraction result

ETDRS by refractionist		Examination normal	Examination abnormal		Total
Normal		4840	8		4848
Abnormal		145	137		282
Total		4985	145		5130
Prevalence				0.028265 (95% CI: 0.023981-0.03327)
Sensitivity				0.944828 (95% CI: 0.890567-0.97412)
Specificity				0.970913 (95% CI: 0.965765-0.97532)
Predictive value positive				0.485816 (95% CI: 0.426323-0.545701)
Predictive value negative				0.99835 (95% CI: 0.996611-0.999233)
Kappa ETDRS result
Expected (%)		κ	SE	Z	Probability >Z
Agreement	Agreement
97.02	91.99	0.6278	0.0132	47.69	0.0000

Kappa statistic: 0.6278. ETDRS: Early Treatment of Diabetic Retinopathy Study, CI: Confidence interval, SE: Standard error

Table 2

Agreement between Snellen chart and examination result

Optometrist screening with Snellen tumbling E chart		Examination normal	Examination abnormal		Total
Normal		4881	25		4906
Abnormal		104	120		224
Total		4985	145		5130
Prevalence				0.028265 (95% CI: 0.023981-0.03327)
Sensitivity				0.827586 (95% CI: 0.754042-0.883294)
Specificity				0.979137 (95% CI: 0.974676-0.982843)
Predictive value positive				0.535714 (95% CI: 0.468138-0.602046)
Predictive value negative				0.994904 (95% CI: 0.992367-0.996628)
Kappa Snellen result
Expected (%)		κ	SE	Z	Probability >Z
Agreement	Agreement
97.49	93.05	0.6380	0.0136	46.86	0.0000

Kappa statistic: 0.6380. CI: Confidence interval, SE: Standard error

Discussion

The overall prevalence of visual acuity <20/30 was 2.85% in our study. Uncorrected REs accounted for most of the visual acuity impairment. There are a limited number studies that have addressed the prevalence of REs in rural India. Dandona et al. reported a prevalence of 2.7%. REs were the major cause of visual impairment.[9] Another study in a rural population of south India showed a much low prevalence of 0.8% of vision <20/40.[4]

There are many school screening programs in India. School eye screening programs are part of the activities of the district blindness control society since 1996. In addition, many hospitals in the private sector run school eye screening programs. Screening of children in schools is most commonly done by trained school teachers, although some programs utilize ophthalmic assistants/ophthalmologists for primary screening. Screening school children is arguably the second largest program of the national program for control of blindness in India after cataract surgery and is currently a priority of the Sarva Shiksha Abhiyan education initiative of the government. There is no data however on the sensitivity and specificity of these programs and their impact on ocular morbidity in children. A study from Satna inferred the efficacy of teacher screening (for students from 5th–12th standard) by indirectly studying the referral rate, false positives and false negative (arrived at by screening a small sample of students inferred as normal in the teacher screening).[5] To the best of our knowledge, no other studies have addressed the efficacy of teacher screening in India. Another study in rural south India used community workers to assess visual impairment in children noted a sensitivity of 50% in identifying visual impairment in the 5–10 years of age group. Our study showed that teacher vision screening had a low sensitivity of 24.8% in primary school children. This is also reflected in the poor agreement (k = 0.2733), [Table 1] of teacher screening with the comprehensive eye examination. Possible reasons for the same could be a lack of motivation among teachers, difficulty in getting cooperation in this age group and inadequate spare time among the teachers to accomplish this activity. It is also difficult to standardize the time allotted by the teachers toward screening or to build in a quality control. Screening using the ETDRS chart by the optometrist showed a good sensitivity. It is reasonable to assume that the better sensitivity is because of greater experience.

Studies in other parts of the world have reported higher sensitivities. The Xichang pediatric RE study reported a sensitivity of 85.2% for teacher screening in middle school children (age - 13–17 years) presenting with visual impairment. The government-run program had a sensitivity of 86.7% in the same group of children but a very poor specificity of 28.7%. This study in China was on older children (middle school). Furthermore, the teachers were compensated $ 13 for their participation.[8] Another study from Tanzania on teacher vision screening for primary school children reported a sensitivity of 80% and a specificity of 91%. The teacher training employed in the study had rigorous protocols including written guidelines, detailing the testing conditions, procedures, and pitfalls. Both of these studies included a smaller number of children than ours.[10] It is possible therefore that reducing the number of students screened by the teacher, having strict protocols and incentivizing the screening process would make school vision screening programs more effective. This, however, would require a strong commitment from school management and policymakers. Screening in older children would obviously be simpler but would be less effective in the early detection of amblyopia. Saxena et al. had reported a sensitivity of 79.2% and a specificity of 93.3% for teacher vision screening using the 6/9.5 optotype.[11] The study is not directly comparable to ours for two reasons. First, the study included older children up to standard IX. Testing vision in primary school children is expected to be more difficult. Second, the study used a primary eye care worker as the gold standard. Our study had employed a comprehensive eye examination, and the pediatric ophthalmologist took the final decision to label a child as defective or not.

The gold standard in detecting vision problems in children would, of course, be a comprehensive eye examination by a team comprising of a Pediatric optometrist and Pediatric Ophthalmologist. However, this may not be practical even in developed countries. Our study suggests that vision screening done by experienced optometrists is a powerful tool to identify ocular morbidity in school children. The poor sensitivity of vision screening by the teachers below the age of 10 years is a cause for concern and possibly highlights the need to standardize teacher training and screening. The limitations of our study include limited time devoted to teacher training, failure to standardize various factors like time devoted per student by the teacher and lack of written protocols for teachers. We also could not measure the agreement between the teachers for vision screening using the tumbling E chart because of the large numbers involved (65 teachers in 57 schools). It was also logistically difficult to get teachers from different schools on the same day to the hospital. Adequate time was given to the teachers to practice during the training program. Having the teachers to practice at school under the supervision of an optometrist before the study may have improved the quality of teacher screening. However, we feel that our study mirrors most school screening programs running in the country and the findings of our study call for a systematic evaluation of the school screening programs in the country. Improving the sensitivity of teacher/community worker screening by standardized testing protocols, providing adequate spare time for the teachers, having supervised test runs at school and possibly compensating them for their effort would go a long way in reducing the burden of ocular morbidity in school children. Alternate approaches like supplementing the efforts of teachers with trained optometrists also need to be explored.

Conclusion

Vision screening for primary school children by trained optometrists with the ETDRS chart has a high sensitivity and specificity, while teachers were noted to have poor sensitivity in our study. Hence trained optometrists may be employed for school vision screening in this age group whenever possible. When not feasible, measures need to be taken to improve the sensitivity and specificity of teachers.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.