Sensitivity and Specificity of Visual Acuity Screening

0 downloads 0 Views 350KB Size Report
WEI-HAN CHUA, BMedSc(Hons), BMed(Hons), and CHING-YE HONG, MBBS. Singapore ..... The visual acuity measurements taken from young Singapore.
1040-5488/02/7910-0650/0 VOL. 79, NO. 10, PP. 650–657 OPTOMETRY AND VISION SCIENCE Copyright © 2002 American Academy of Optometry

ORIGINAL ARTICLE

Sensitivity and Specificity of Visual Acuity Screening for Refractive Errors in School Children LOUIS TONG, MBBS, FRCS, SEANG-MEI SAW, MBBS, MPH, PhD, DONALD TAN, MBBS, FRCS, FRCOphth, FAMS, KEE-SENG CHIA, MBBS, PhD, WAI-YING CHAN, BSc(Hons) Optom, GD Comp Sci, FIACLE, ANDREW CARKEET, PhD, WEI-HAN CHUA, BMedSc(Hons), BMed(Hons), and CHING-YE HONG, MBBS Singapore National Eye Centre, Singapore (LT, DT), Department of Community, Occupational and Family Medicine, National University of Singapore, Singapore (SMS, KSC, CYH), Singapore Eye Research Institute, Singapore, Singapore (DT, WYC, AAC, WHC), Department of Ophthalmology, National University of Singapore, Singapore (DT)

ABSTRACT: Purpose. To examine the optimal cutoff point for the use of the visual acuity test to screen for refractive errors in schoolchildren. Methods. In a sample of schoolchildren between 7 and 9 years old, visual acuity testing was performed using modified ETDRS charts monocularly without optical aids by trained personnel. Cycloplegic autorefraction was performed in each eye. The screening efficacy of using various cutoff points for referring children for further optometric/ ophthalmic assessment was studied. Myopia was defined as a spherical equivalent of at least ⴚ0.5 D, hyperopia a spherical equivalent of at least ⴙ2.0 D, and astigmatism a cylinder of at least ⴚ1.0 D in at least one eye. The sensitivity, specificity, and predictive values were calculated using each patient as a case; a receiver operator curve was plotted. Results. A total of 1028 children were tested. A satisfactory sensitivity/specificity profile was obtained using a referral criterion of visual acuity worse than or equal to 0.28 logarithm of the minimum angle of resolution in at least one eye. In this scenario, the sensitivity and specificity of this screening test were 72% (95% confidence interval [CI], 68 to 76) and 97% (95%CI, 95 to 98), respectively. The positive and negative predictive values were 96% (95%CI, 93 to 98) and 78% (95%CI, 75 to 82), respectively. Conclusions. The modified ETDRS visual acuity chart can be used to predict refractive errors in schoolchildren in Singapore in a sensitive and specific manner using a referral criterion of worse than or equal to 0.28 logarithm of the minimum angle of resolution. (Optom Vis Sci 2002;79:650–657) Key Words: screening, visual acuity, cross-sectional study, epidemiology, myopia

T

he prevalence of refractive errors in children and adults is an increasing public health problem in Singapore and many Asian countries.1, 2 The Orinda study, an epidemiological study of myopia conducted in America, has identified refractive errors (myopia, hyperopia, and astigmatism) as important vision disorders requiring screening.3 Subsequently, the importance of screening refractive errors in children using visual acuity tests has been noted4, 5 Screening in children is important because, unlike adults, some younger children may not complain of blurred distance vision. The refractive errors in such children may remain undiagnosed for a period of time in the absence of screening. Early detection of such cases allows timely interventions in the form of spectacle correction. In Singapore, a multilingual society with a high rate of English literacy, Snellen vision charts with English

alphabets have been used for many years to screen for eye problems in schools. Children with visual acuity of 6/12 or worse are referred for further assessment. The validity of using this referral criterion has not been proven. Various studies6 –10 have reported screening children using the arbitrary criterion of “6/12 or worse on the Snellen chart” or its equivalent. One such study8 reported the positive predictive value of visual acuity tests, but the true positives included those found to be “abnormal” on subjective refraction or “retinoscopic observations” or in other components of a more comprehensive examination. In addition, the more comprehensive examination was only offered to those who failed some part of a screening battery of tests, and hence the inability to assess the predictive value of test negative findings. Two studies11, 12 used the equivalent of

Optometry and Vision Science, Vol. 79, No. 10, October 2002

Visual Acuity Screening in Children—Tong et al.

worse than 6/9.5 as a criteria, but the true negatives and false positives were those who had a normal result on a full eye examination. In four studies,6, 7, 9, 10 only the children that failed the “6/12 or worse criteria” were refracted, but no information about the refractive status of the test negative cases was provided. A further study13 on visual acuity that did not report refractive errors used an equivalent of 6/8.5 as a cutoff, citing the reason that this is the minimum acuity required to see letters on the blackboard.14 There is no reason to presume that the threshold used in these studies is the optimal one without examining the sensitivity and specificity profiles of various thresholds. One study15 utilized thresholds from 6/6 to 6/60 and reported the number of cases of refractive errors found in each scenario, but all the subjects in the study were referred cases or cases failing the visual acuity criteria, precluding the calculation of sensitivity and specificity because of the unknown true and false negative rates. Conversely, a study has documented acuity levels from 6/6 to 6/60 or worse but only in the children with refractive errors.16 The acuity levels of those without refractive errors (i.e., true negative rate and false positive rate) were not reported. A further study examined various acuity levels in children but did not include refraction as part of the study protocol.5 A study is now warranted because none of the previous studies have investigated the most suitable logarithm of the minimum angle of resolution (logMAR) threshold for referring patients. Furthermore, none of the previous studies has documented the sensitivity and specificity of using an accurate visual acuity test chart against the findings of a reliable refraction assessment in the context of schoolchildren. The objective of this study was to determine the optimal visual acuity cutoff value for the prediction of the presence of refractive errors (presence of myopia, hyperopia, or astigmatism) in Singapore schoolchildren. As a secondary objective, the study aimed to estimate the efficiency of using the modified ETDRS charts to conduct screening in Singapore children by determining the predictive values of this test using various referral thresholds.

drops, leaving 1003 children for data analysis. The age of the child was computed from the date of birth, and the gender and ethnic group of the children were obtained from mandatory birth certificates. Table 1 shows the proportion and number of eyes with refractive errors. For each subject, the eye with the worse visual acuity was used for analysis. The same eye may have more than one type of refractive error, e.g., astigmatism and myopia. In the event that the right and left eyes had the same visual acuity, the refractive condition or conditions present in either of the eyes were counted. There were no cases of hyperopia in one eye and myopia in the other. The proportion of children testing positive for refractive error (i.e., being classified with myopia, hyperopia, and/or astigmatism in at least one eye) was 47.7%. In any screening tests advocated for a large population, economic and logistic considerations require an estimate of the number of subjects likely to be test-positive. This would have a lot of impact on the service load of the referral center. The prevalence rates of logMAR visual acuity being worse than or equal to 0.2, 0.3, and 0.4 were calculated based on the monocular acuity of the worse eye.

Measurement Visual acuity was performed monocularly in each eye without optical aids. Although the study also included the visual acuity measurement with the preexisting glasses or pin hole after the measurement of unaided visual acuity, this data was not used in the analysis. The testers were not masked as to whether the children wore glasses. LogMAR visual acuity was recorded using a modified nonilluminated ETDRS20 chart with Sloan letters (Lighthouse distance visual acuity test second edition, Cat. No. C105) under room lighting at a distance of 4 m at eye level. This chart is designed to be used at 4 m, it has five letters per line, and the acuity that can be measured on it ranges from 1.10 to ⫺0.30 logMAR. TABLE 1. The study population.a

METHODS Subjects The initial cross-sectional results of an ongoing longitudinal observational study which commenced in 1999, Singapore Cohort Study of the Risk Factors of Myopia (SCORM),17–19 are reported. All schoolchildren in grades I to III, aged 7 to 9 years, attending two Singapore schools were recruited for this study. Subjects with established eye pathology detected before the commencement of the study or known allergy to eyedrops were excluded. Participation rate was 62%. The proportion of children who reported myopia before the school eye examination was not different between participants (27.3%) and nonparticipants (26.8%). This study was approved by the ethics committee of the Singapore Eye Research Institute, and all procedures adhered to the Declaration of Helsinki. Informed written consent was obtained from parents of the children who chose to participate. A total of 1028 schoolchildren (49.6% males) with mean age of 7.4 years (SD ⫽ 0.5) were recruited. However, 25 children could not complete the necessary measurements due to allergy to eye-

651

Total Gender Male Female Ethnic Group Chinese Malay Indian Others Age (yr) 7 8 9

N

Myopia N (%)

Hyperopia N (%)

Astigmatism N (%)

1003

339 (33.8)

17 (1.7)

225 (22.4)

497 506

170 (34.2) 169 (33.4)

5 (1.0) 12 (2.4)

114 (22.9) 111 (21.9)

729 193 56 25

283 (38.8) 38 (19.7) 14 (25.0) 4 (16.0)

8 (1.1) 7 (3.6) 2 (3.6) 0 (0.0)

166 (22.8) 48 (24.9) 10 (17.9) 1 (4.0)

522 319 162

154 (29.5) 113 (35.4) 72 (44.4)

8 (1.5) 8 (2.5) 1 (0.6)

125 (23.9) 68 (21.3) 32 (19.8)

a The number of eyes and the proportion are shown for each age, gender, and ethnic group. The refractive error of the eye with the worse visual acuity was used to calculate prevalence rates of each type of refractive error. The same eye may have more than one refractive error type, e.g., astigmatism and myopia.

Optometry and Vision Science, Vol. 79, No. 10, October 2002

652

Visual Acuity Screening in Children—Tong et al.

Chart 1 was used to measure the visual acuity of the right eyes, whereas chart 2 was used for the left eyes. The luminance or light reflected from the chart under testing conditions was measured using a Minolta CS100 Colorimeter. The luminance of the white portion of the chart was 65.7 (SD ⫽ 7.7) cd/m2, whereas the corresponding value for the black lettering was 3.16 (SD ⫽ 0.42) cd/m2, giving an average Weber contrast of 0.952. The testing procedure and scoring was formulated based on the ETDRS-Fast Procedure21; the unique feature of this procedure is that when the subject is far from the threshold, the number of stimulus presentations is greatly reduced. The reproducibility and other details of the ETDRS-Fast Procedure have been published.21 This method reduces testing time and yet retains the accuracy of the standard test procedure.21 Please refer to the appendix of this paper for a description of the steps of the testing and a scoring example. Briefly, in any line of the chart, each letter read correctly would reduce the logMAR by 0.02 from the identifying logMAR level of the row above. The advantage of letter counting to determine visual acuity in this way has been published.22 After instillation of 0.5% proparacaine, cycloplegia was accomplished with three drops of topical 1% cyclopentolate in each eye, each drop at 5-min intervals. Cycloplegic measurements were performed 30 min after the last drop instillation. Autorefraction was performed using a Canon RK-5 autorefractor (Canon, Tochigiken, Japan). The mean of five refractive errors was calculated,23 with all refractive errors expressed as negative cylinders whenever a cylinder was measured.

Data Analysis Cycloplegic autorefraction, as performed in our study, was reported to give a reliable measure of refractive error.24 Using such a reference standard, positive cases were defined as myopia, hyperopia, or astigmatism in at least one eye. Myopia in this study was defined as a spherical equivalent (sphere ⫹ 0.5 ⫻ cylinder) of at least ⫺0.5 D. Hyperopia was defined as a spherical equivalent of at least ⫹2.0 D. Astigmatism was defined as a cylinder of at least 1.0 D. A common definition of myopia of ⫺0.5 D or worse was adopted, although we are aware that definitions of myopia vary from study to study.25 The above practical definitions would likely include all children who require spectacle correction to optimize their vision for daily activities, although not all of these children, especially the hyperopes, may require spectacle correction on further clinical evaluation. The definition of hyperopia in this study was purely arbitrary. As clinicians, we are not keen to investigate or treat minor degrees of asymptomatic hyperopia in children. The ideal functional definition of significant hyperopia is problematic, but hyperopia does not play a great role in the context of screening using visual acuity charts. This is explained further in the Discussion. Large amounts of hyperopia and astigmatism are more closely associated with amblyopia than smaller degrees of spherical myopia. Nevertheless, we used a combined category of refractive errors rather than separate refractive error categories because in a real screening scenario in schools, only one visual acuity criterion is used for referral. We also simplified the study by not including anisometropia as a target for screening. Because most of the children were likely to have myopia as the refractive error, the main

type of anisometropia expected in this population was myopic anisometropias. We hoped that our visual acuity referral criterion for the worse eye would capture these children. For the sensitivity and specificity calculations, the visual acuity referral criteria examined were logMAR of 0.20, increasing in steps of 0.02, including 0.30 (corresponding to Snellen acuity of 6/12), and up to 0.48 (corresponding to Snellen acuity of 6/18) or worse in at least one eye. The positive and negative predictive values were also calculated. All indices were expressed as percentages with 95% confidence intervals. Using the specificity and sensitivity ratios at the above referral cutoff points, a receiver operator curve was plotted. We felt it would be interesting to know the proportion of children of each refractive category that could be captured from a best compromise cutoff referral visual acuity, and hence this was also analyzed.

RESULTS Fig. 1 shows the distribution of the logMAR acuity in the worse eye of the children. This is heavily skewed to the right, with fewer eyes having poorer vision. The median acuity was 0.14 (range, ⫺0.20 to ⫹1.10). Table 2 shows the prevalence rates of “poor visual acuity” using the three predetermined thresholds. Fig. 2 compares the visual acuity of eyes with and without refractive error. The mean acuity for eyes with refractive errors was 0.50 (SD ⫽ 0.34) and those without any of these conditions was 0.07 (SD ⫽ 0.11). The former was worse than the latter, and the difference was statistically significant (t ⫽ ⫺29.6, df ⫽ 572.9, p ⬍ 0.001). This analysis was performed using the independent samples t-test without assuming equal variances (Levene’s test for equality of variances showed F ⫽ 830 and p ⬍ 0.001). Fig. 3 shows the receiver operating characteristics of this method of screening. As expected, a referral criterion at a better visual acuity would improve the sensitivity but reduces the specificity of

FIGURE 1. Histogram showing the percentage of children with the visual acuity indicated on the horizontal axis; the worse eye of the two was used in the plot. The smaller peak at the right-hand side of the histogram is from a ceiling effect due to the worst possible acuity that the visual acuity chart can measure.

Optometry and Vision Science, Vol. 79, No. 10, October 2002

Visual Acuity Screening in Children—Tong et al.

653

TABLE 2. logMAR visual acuity of the study population in each age, gender, and ethnic group.

Total Gender Male Female Ethnic Group Chinese Malay Indian Others Age (yr) 7 8 9 a b

N

Myopia and/or Hyperopia and/or Astigmatismb N (%)

logMAR VA ⱖ0.2a N (%)

logMAR VA ⱖ0.3 N (%)

logMAR VA ⱖ0.4 N (%)

1003

478 (47.7)

466 (46.5)

360 (35.9)

309 (30.8)

497 506

237 (47.7) 241 (47.6)

224 (45.1) 242 (47.8)

181 (36.4) 179 (35.4)

156 (31.4) 153 (30.2)

729 193 56 25

373 (51.2) 75 (38.9) 24 (42.9) 6 (24.0)

358 (49.1) 78 (40.4) 22 (39.3) 8 (32.0)

284 (39.0) 56 (29.0) 14 (25.0) 6 (24.0)

248 (34.0) 43 (22.3) 13 (23.2) 5 (20.0)

522 319 162

238 (45.6) 155 (48.6) 85 (52.5)

239 (45.8) 147 (46.1) 80 (49.4)

174 (33.3) 119 (37.3) 67 (41.4)

145 (27.8) 104 (32.6) 60 (37.0)

VA, visual acuity. Myopia ⱖ⫺0.5 D, hyperopia ⱖ2.0 D, and astigmatism ⱖ1.0 D in at least one eye.

FIGURE 2. Box plots showing the visual acuity (logMAR) of eyes having myopia (worse than or equal to ⫺0.5 D), hyperopia (worse than ⫹2 D), or astigmatism (worse than or equal to ⫺1.0 D) and those without these refractive errors. The worse eye of the two was used. The boxes indicate the interquartile ranges, the dark horizontal lines represent the median, and the 2.5th and 97.5th percentile values have been marked by lighter horizontal lines. Other points represent outliers.

the screening. Table 3 shows the sensitivity, specificity, and predictive values using the various visual acuity cutoff points. If a higher logMAR score is chosen as a cutoff point, this increases the specificity but reduces the sensitivity.

We feel that the profile of the curve shows an arbitrary optimal sensitivity/specificity profile corresponding to the visual acuity cutoff of 0.28 logMAR. For this cutoff point, the sensitivity and specificity of this screening test were 72% (95% confidence interval [CI], 68 to 76) and 97% (95%CI, 95 to 98), respectively. The positive and negative predictive values were 96% (95%CI, 93 to 98) and 78% (95%CI, 75 to 82), respectively. Selecting a visual acuity level that is better than 0.20 logMAR as the referral criterion will result in better sensitivity, but the resulting poor specificity makes it impractical in a screening scenario. For this reason, the receiver operating characteristic (ROC) in Fig. 3 has not been extended upwards (for visual acuity better than 0.20 logMAR). Conversely, a grossly nonsensitive test is equally useless, and hence the ROC curve was not projected downwards. Table 4 shows the proportions of children of each refractive category that would be referred using the “best compromise referral threshold” of 0.28 or worse logMAR. Because 87.6% of the myopia was detected by this visual acuity threshold, the test would be best for the detection of myopia. The test detected also a fairly large proportion of cases of astigmatism (67.3%). However, the test only referred a small proportion of hyperopia cases. We accept that the number of cases of hyperopia (as defined in this study) was small and hence the wide confidence interval of the proportion.

DISCUSSION The visual acuity measurements taken from young Singapore schoolchildren predicted the presence of refractive errors in an accurate fashion. A threshold set at 0.28 or worse logMAR represents the optimal referral criterion. No other large study has investigated the best referral cutoff point for this method of visual acuity assessment (logMAR acuity) using the ROC curve.

Optometry and Vision Science, Vol. 79, No. 10, October 2002

654

Visual Acuity Screening in Children—Tong et al.

FIGURE 3. Receiver operating curve. Sensitivity was plotted against 1-specificity. The desired level of specificity and sensitivity must be compromised so that optimal screening can be achieved. From left to right, the logMAR acuity cutoff points used were 0.477 or worse, 0.44 or worse, 0.42 or worse, and reducing in steps of logMAR 0.02 to 0.20 or worse. Selected data points were labeled with the logMAR cutoff points.

There is more than one way of interpreting ROC curves. Kraemer26 presented a more objective method of determining the ideal point for ROC and quality ROC plots. Using his method, the optimal threshold obtained would be a logMAR of 0.22 (␹2 ⫽ 546.45, E k(5,0) of 0.74). Such a threshold would produce a very

good sensitivity of 81% but a specificity of only 92%. Clinical costs and benefits, although difficult to measure,27 must be considered for any recommendation. We are somewhat unwilling to recommend any screening test of specificity below 95% to screen for refractive errors. Unnecessary referrals will increase cost and cause inconvenience.28 In this scenario, where “poor visual acuity” is highly prevalent and the condition to be screened is not life threatening, we preferred a more stringent referral threshold. Our study confirms that a high proportion of schoolchildren have poor unaided visual acuity (Table 2: 30 to 50%, depending on the specific criteria of poor visual acuity). Our study shows that in a large sample of Asian schoolchildren with a high degree of English literacy, visual acuity assessment may be used to accurately screen for refractive errors. The majority of the cases detected by this screening strategy consisted of cases of myopia and astigmatism. As mentioned above, the number of children defined as hyperopic in this study was too small to enable us to evaluate the effectiveness of our visual acuity test for the detection of hyperopia. It is worth noting that the mean acuity of the hyperopic eyes was fairly good at 0.22 (SD ⫽ 0.13). This was not significantly different from the mean acuity of emmetropic children (p ⬎ 0.05) and suggests that our acuity test criterion may be unsuitable for the detection of hyperopia. Young hyperopes have good vision because of accommodation, hence vision screening is ineffective (and also unnecessary) to detect hyperopia. For these reasons, we did not investigate the issue of hyperopia further by sampling a greater number of hyperopes and redefining hyperopia to be a spherical equivalent with a greater plus dioptric power. The strengths of this study include a large sample size drawn from a prereferral population, uniformity of assessment, and objectivity of autorefraction. In addition, the use of cycloplegia excluded pseudomyopia or accommodative spasm.

TABLE 3. Sensitivity, specificity, positive predictive values (PPV), and negative predictive (NPV) in percentages (95% confidence interval in brackets). logMAR 0.20a 0.22 0.24 0.26 0.28 0.30b 0.32 0.34 0.36 0.38 0.40 0.42 0.44 0.46 0.48c

Sensitivity% (95% CI)

Specificity% (95% CI)

PPV% (95% CI)

NPV% (95% CI)

84.0 (80.4–87.0) 80.8 (77.0–84.1) 77.6 (73.6–81.1) 75.2 (71.1–78.8) 71.7 (67.5–75.6) 70.5 (66.3–74.5) 67.7 (63.4–71.8) 66.5 (62.2–70.6) 63.5 (59.1–67.7) 61.7 (57.3–66.0) 60.7 (56.3–65.0) 57.7 (53.2–62.1) 55.9 (51.4–60.3) 54.7 (50.2–59.1) 52.9 (48.4–57.3)

88.8 (85.8–91.3) 92.2 (89.5–94.3) 93.6 (91.0–95.4) 95.3 (93.0–96.9) 96.8 (94.8–98.1) 96.8 (94.8–98.1) 97.2 (95.3–98.3) 97.5 (95.7–98.6) 97.5 (95.7–98.6) 97.5 (95.7–98.6) 97.5 (95.7–98.6) 98.3 (96.7–99.2) 98.3 (96.7–99.2) 98.3 (96.7–99.2) 98.3 (96.7–99.2)

87.7 (84.3–90.4) 90.8 (87.6–93.2) 91.9 (88.8–94.3) 93.8 (98.0–95.8) 95.5 (92.7–97.3) 95.4 (92.6–97.2) 95.8 (92.9–97.5) 96.2 (93.5–97.9) 96.1 (93.2–97.8) 96.0 (93.0–97.7) 95.9 (92.9–97.7) 97.0 (94.1–98.5) 96.9 (93.9–98.5) 96.8 (93.8–98.4) 96.7 (93.6–98.4)

85.5 (82.2–88.2) 83.6 (80.2–86.4) 81.5 (78.2–84.5) 80.3 (76.9–83.3) 78.4 (75.0–81.5) 77.7 (74.3–80.8) 76.1 (72.2–79.3) 75.5 (72.1–78.7) 73.9 (70.5–77.1) 73.0 (69.5–76.2) 72.5 (69.0–75.7) 71.1 (67.7–74.4) 70.3 (66.8–73.5) 69.7 (66.2–73.0) 68.9 (65.4–72.1)

a

This level corresponds to “Snellen equivalent” of 6/9. This level corresponds to Snellen equivalent of 6/12. c This level corresponds to Snellen equivalent of 6/18. It should be emphasized that there is strictly speaking no Snellen equivalent for any level of logMAR because the latter acuity is obtained by letter counting rather than the number of lines correctly read. b

Optometry and Vision Science, Vol. 79, No. 10, October 2002

Visual Acuity Screening in Children—Tong et al.

655

TABLE 4. Refractive error subtypes captured or referred by the criteria of ⱕ0.28 logMARa

Refractive Category

No. of Children Referred at ⱕ0.28 logMAR

Total No. in This Refractive Group

Proportion (Percent) Referred and 95% Confidence Interval of Proportion in Brackets

Myopia Myopia with astigmatism Myopia with no astigmatism Astigmatism Astigmatism with no myopia Hyperopia Hyperopia with astigmatism Hyperopia with no astigmatism

340 114 226 167 53 6 4 2

388 129 259 248 119 17 6 11

87.6 (83.8–90.7) 88.4 (81.2–93.1) 87.3 (82.4–90.9) 67.3 (61.1–73.1) 44.5 (35.5–53.9) 35.3 (15.3–61.4) 66.7 (24.1–94.0) 18.2 (3.2–52.2)

a The visual acuity of the worse eye is used to classify whether the child would be referred or not because the referral could be triggered by at least one eye reaching the referral threshold. The definitions of myopia, astigmatism, or hyperopia are defined in the Methods section of this paper, applying them to the eye with a worse visual acuity. In the instance where the visual acuity is identical in the two eyes, the refractive category is based on the eye with the more severe or more negative spherical equivalent. Note that some categories overlap, e.g., the all-myopia category and the myopia and astigmatism category.

In this study, a complete ophthalmic examination was not performed in all patients to exclude symptomless eye pathology. However the aim of this study was to detect refractive error likely to require optical correction and not other problems such as squints. Because subjective refractions could not be performed as part of the study protocol, all subjects requiring refraction were referred to a specific center. The referral optometry center was required to provide feedback to the study investigators if a corrected visual acuity better than 0.30 logMAR in each eye could not be achieved. Because such communication did not occur, one can reasonably infer that all subjects referred had good corrected visual acuity. This suggested that the visual loss was primarily from refractive errors rather than other pathology. The study sample was not randomly selected from the school population of Singapore. Sampling is relevant in this study because the sensitivity and specificity profile may change with differing disease prevalence, i.e., Berkson’s fallacy,29 which essentially dictates that in a sample obtained from a high-risk and low-risk populations, a biased sensitivity estimate is obtained in the high-risk population and a biased estimate of specificity in the low-risk population. These limitations of the study make it possibly harder to generalize our findings to the entire country. The information obtained by cycloplegic refraction is valuable for a specific study of this nature, but is unnecessary and not feasible for the wide-scale screening of children. Having established the best cutoff point for the visual acuity for screening, how can this be applied in an actual population? Large-scale visual acuity testing may be conducted by nurses because it is a noninvasive procedure. Currently, screening in Singapore has been performed using Snellen visual acuity charts. The present criterion for referral is an unaided visual acuity of 6/12 or worse in children who are not wearing glasses or an aided acuity of worse than the same level in children with glasses correction. Visual acuity recorded on the Snellen chart is not directly comparable to logMAR visual acuity, but our threshold of 0.28 appears to be close to the presently used threshold. It may be advisable to use the logMAR charts for screening rather than the Snellen charts because of the unifor-

mity of visual acuity sampling and other advantages associated with a logMAR visual acuity.30 –32 We are aware that our study has not been designed to compare the efficacy of screening using different types of visual acuity charts. The clinicians and health personnel involved in visual acuity screening using the Snellen charts may have the impression that the 0.30 logMAR threshold may be preferable to 0.28 logMAR in the modified ETDRS chart. This impression of “convenience” arises because 0.30 logMAR (but not 0.28) appears to correspond to a complete line on the modified ETDRS charts. This impression may not be entirely true, as explained in the appendix to this paper. Having studied the referral criteria for screening, further studies should address the optimal frequency of screening in children. For now, it suffices to conclude that visual acuity is an accurate screening tool for myopia and astigmatism, and a 0.28 logMAR cutoff referral is a compromise to achieve satisfactory sensitivity and specificity. Should there not be any logistic problems, we intend to propose that a nationwide school screening program in Singapore adopt this method.

APPENDIX Steps for Determination of logMAR Visual Acuity 1. The testing begins with the right eye. 2. The left eye is occluded with an opaque occluder. 3. The subject is requested to read the first letter on the right hand side of each row and then proceed to the next row below after successful identification of the letter. 4. The above step is repeated until the subject shows hesitation, in which case, we proceed to the next step. 5. The line immediately above is tested, and the letters are read by the subject from left to right. 6. The number of mistakes made is recorded. 7. Steps 5 and 6 have to be repeated upward every time the subject makes two or more errors. 8. If there is only one error or no errors in this line, proceed to the next row below.

Optometry and Vision Science, Vol. 79, No. 10, October 2002

656

Visual Acuity Screening in Children—Tong et al.

TABLE 5. Scenarios in visual acuity testing.a

Scenario A

Scenario B

Scenario C

0.28 logMAR

0.30 logMAR

0.3 line: o o o o o 0.2 line: o x x x x 0.1 line: x x x x x 0.3 line: o o o o x 0.2 line: o o x x x 0.1 line: x x x x x 0.3 line: o o o x x 0.2 line: o o o x x 0.1 line: x x x x x

0.3 line: o o o o o 0.2 line: x x x x x 0.1 line: x x x x x 0.3 line: o o o o x 0.2 line: o x x x x 0.1 line: x x x x x 0.3 line: o o o x x 0.2 line: o o x x x 0.1 line: x x x x x

a

The table shows three hypothetical scenarios on the left when the acuity would be recorded as 0.28 logMAR and another three scenarios on the right when the acuity would be recorded as 0.30 logMAR. Circles indicate correct answers to the letters of chart, and crosses indicate incorrect answers to the letters. We assume that correct answers were obtained for all letters above the line labeled as 0.30 logMAR from the commencement of the test.

9. If the subject can read more than half of the letters (i.e., three or more), the testing proceeds to the line below; otherwise, the test is terminated. 10. In the situation when the subject reads all the letters correctly down to the bottom row, the bottom row must be read completely.

Scoring Example 1. The line labeled logMAR 0.4 has the letters V, B, K, L, and N; the line labeled logMAR 0.3 has the letters Z, G, H, C, and E; the line labeled logMAR 0.2 has the letters D, P, T, S, and X; and the line labeled logMAR 0.1 has the letters L, R, U, V, and Q. 2. The subject reads N, E from the logMAR 0.4 and logMAR 0.3 rows and shows hesitation when asked to read the letter X. 3. The subject is requested to read the entire line labeled logMAR 0.3. 4. The letters Z, G and E are read correctly, but H and C are read incorrectly. 5. All letters in the line labeled logMAR 0.4 can be read correctly. 6. Testing proceeds to the line labeled logMAR 0.2. D and S are read correctly, but P, T, and X are read incorrectly. The test is terminated because the subject is unable to read at least three of five letters of this line. 7. The last attempted line is the logMAR 0.2 line (provisional scoring 0.20). On the logMAR 0.3 line, two mistakes were made (add ⫹0.02 ⫻ 2 to the score). On the logMAR 0.2 line, three mistakes were made (add ⫹0.02 ⫻ 3 to the score). 8. The total (final) score is 0.20 ⫹ 0.04 ⫹ 0.06, or 0.30.

Threshold of 0.28 or 0.30 logMAR The ROC curve in Fig. 3 shows a minimal difference in sensitivity/specificity profile for these two thresholds. However, we ar-

gue that in a letter counting type of visual acuity measurement, the extra precision is useful. For this reason, 0.28 logMAR is recommended. Table 5 shows the various scenarios that could theoretically be equivalent to an acuity of 0.28 or 0.30 logMAR. Only one such scenario pertains to the child correctly reading a complete line and no further. Because there is no evidence from psychophysical studies that this scenario is more common than the other scenarios, the logMAR criterion of 0.30 is not necessarily more convenient than that of 0.28.

ACKNOWLEDGMENT This study was supported by a National Medical Research Council Grant: SERI/MG/97-04/0005. Received October 10, 2001; revision received June 12, 2002.

REFERENCES 1. Wong TY, Foster PJ, Hee J, Ng TP, Tielsch JM, Chew SJ, Johnson GJ, Seah SK. Prevalence and risk factors for refractive errors in adult Chinese in Singapore. Invest Ophthalmol Vis Sci 2000;41:2486–94. 2. Saw SM, Katz J, Schein OD, Chew SJ, Chan TK. Epidemiology of myopia. Epidemiol Rev 1996;18:175–87. 3. Blum HL, Peters HB, Bettman JW. Vision Screening for Elementary Schools: the Orinda Study. Berkeley: University of California Press, 1959. 4. Stewart-Brown SL, Brewer R. The significance of minor defects of visual acuity in school children: implications for screening and treatment. Trans Ophthalmol Soc UK 1986;105:287–95. 5. Stewart-Brown S, Butler N. Visual acuity in a national sample of 10 year old children. J Epidemiol Community Health 1985;39:107–12. 6. Wedner SH, Ross DA, Balira R, Kaji L, Foster A. Prevalence of eye diseases in primary school children in a rural area of Tanzania. Br J Ophthalmol 2000;84:1291–7. 7. Lithander J, Al Kindi H, Tonjum AM. Loss of visual acuity due to eye injuries among 6292 school children in the Sultanate of Oman. Acta Ophthalmol Scand 1999;77:697–9. 8. Bailey RN. Assessing the predictive ability of the test-positive findings of an elementary school vision screening. Optom Vis Sci 1998;75: 682–91. 9. Yang YF, Cole MD. Visual acuity testing in schools: what needs to be done. BMJ 1996;313:1053. 10. Bremner MH. Visual acuity in the primary school child aged four to twelve years: a review of amblyopia treatment in this age group at Princess Margaret Hospital. Aust J Ophthalmol 1984;12:395–9. 11. Thomson WD, Evans B. A new approach to vision screening in schools. Ophthalmic Physiol Opt 1999;19:196–209. 12. Hatch SW. Computerized vision screening: validity and reliability of the VTA/VERA vision screener. J Behavioural Optom 1993;4: 143–8. 13. Nishi M, Miyake H, Shikai T, Takeuchi M, Tanaka H, Minagawa N, Morimoto Y, Wada M. Factors influencing the visual acuity of primary school pupils. J Epidemiol 2000;10:179–82. 14. Hiwarashi M. Problem of visual acuity in the school. Acta Soc Ophthalmol Jpn 1963;67:930–9. 15. Ingram RM. Review of children referred from the school vision screening programme in Kettering during 1976–8. BMJ 1989;298: 935–6. 16. Jensen H, Goldschmidt E. Visual acuity in Danish school children. Acta Ophthalmol (Copenh) 1986;64:187–91. 17. Saw S, Chua WH, Wu HM, Hong CY, Chan WM, Chi KS, Tan D. Design and initial results of the Singapore Myopia Cohort Study. In:

Optometry and Vision Science, Vol. 79, No. 10, October 2002

Visual Acuity Screening in Children—Tong et al.

18. 19.

20.

21.

22.

23.

24.

25.

Thorn F, Troilo D, Gwiazda J, eds. Myopia 2000: Proceedings of the VIII International Conference on Myopia. Boston: New England College of Optometry, Schepens Eye Research Institute, 2000:4–10. Saw SM, Hong CY, Chia KS, Stone RA, Tan D. Nearwork and myopia in young children. Lancet 2001;357:390. Saw SM, Wu HM, Hong CY, Chua WH, Chia KS, Tan D. Myopia and night lighting in children in Singapore. Br J Ophthalmol 2001; 85:527–8. Early Treatment Diabetic Retinopathy Study Research Group. Photocoagulation for diabetic macular edema: Early Treatment Diabetic Retinopathy Study report number 1. Arch Ophthalmol 1985;103: 1796–806. Camparini M, Cassinari P, Ferrigno L, Macaluso C. ETDRS-fast: implementing psychophysical adaptive methods to standardized visual acuity measurement with ETDRS charts. Invest Ophthalmol Vis Sci 2001;42:1226–31. Bailey IL, Bullimore MA, Raasch TW, Taylor HR. Clinical grading and the effects of scaling. Invest Ophthalmol Vis Sci 1991;32: 422–32. Thibos LN, Wheeler W, Horner D. Power vectors: an application of Fourier analysis to the description and statistical analysis of refractive error. Optom Vis Sci 1997;74:367–75. Zadnik K, Mutti DO, Adams AJ. The repeatability of measurement of the ocular components. Invest Ophthalmol Vis Sci 1992;33: 2325–33. Zadnik K, Mutti D. Let’s define myopia: a need for consensus? In: Thorn F, Troilo D, Gwiazda J, eds. Myopia 2000: Proceedings of the

26. 27.

28.

29. 30.

31. 32.

657

VIII International Conference on Myopia. Boston: New England College of Optometry, Schepens Eye Research Institute, 2000: 320–4. Kraemer HC. Evaluating Medical Tests: Objective and Quantitative Guidelines. Newbury Park, CA: Sage Publications, 1992. McNeil BJ, Varady PD, Burrows BA, Adelstein SJ. Measures of clinical efficacy: cost-effectiveness calculations in the diagnosis and treatment of hypertensive renovascular disease. N Engl J Med 1975;293: 216–21. Wilson JMG, Jungner G. Principles and Practice of Screening for Disease. World Health Organization Public Health Papers, No 34. Geneva: World Health Organization, 1968. Berkson J. Limitations of the application of four-fold table analysis to hospital data. Biometrics Bull 1946;2:47–53. Morad Y, Werker E, Nemet P. Visual acuity tests using chart, line, and single optotype in healthy and amblyopic children. J AAPOS 1999;3:94–7. Giaschi DE, Regan D, Kraft SP, Kothe AC. Crowding and contrast in amblyopia. Optom Vis Sci 1993;70:192–7. Hohmann A, Haase W. Development of visual line acuity in humans. Ophthalmic Res 1982;14:107–12.

Optometry and Vision Science, Vol. 79, No. 10, October 2002

Louis Tong Singapore National Eye Centre 11 Third Hospital Avenue Singapore 168751 e-mail: [email protected]