Comparison of preschool vision screening tests as administered by licensed eye care professionals in the vision in preschoolers study☆

Participants 

Subjects were children who were enrolled in Head Start programs25 in the vicinity of the 5 VIP clinical centers (Berkeley, California; Boston, Massachusetts; Columbus, Ohio; Philadelphia, Pennsylvania; Tahlequah, Oklahoma). Head Start is a national, comprehensive child development program that serves preschool children and their families. The goal of Head Start is to increase the school readiness of children from low income families. Vision in Preschoolers Study subjects were ≥3 and <5 years on September 1 of the academic year in which they were tested. To obtain a sample that was enriched with children who had vision problems, recruitment of children was based on the results of a regular vision screening conducted by local Head Start personnel. Screening procedures varied by site. In each year, the main goal was to recruit approximately 350 children who had ≥1 targeted conditions. To accomplish this goal, all children at participating Head Start centers who had failed the Head Start vision screening were asked to participate in the VIP Study, as were a randomly selected subset of children who had not failed the screening. The project was approved by the appropriate institutional board(s) associated with each center.

Screeners and examiners 

All testing was conducted by LEPs who had experience in eye care of young children and who had completed VIP Study–specific training and certification. In year 1 (October 2001 through June 2002), 25 LEPs conducted screening tests, and 35 LEPs conducted gold standard examinations (GSEs). In year 2 (October 2002 through June 2003), 29 LEPs served as screeners, and 45 LEPs served as gold standard examiners. Some examiners were certified to conduct both screening tests and GSEs; however, care was taken to ensure that examiners did not conduct both screening and a GSE on the same child.

Screening procedures and instruments 

Instruments 

Details concerning the screening instruments are provided in Table 1. Additional details are provided below. The manufacturer of one additional instrument, the VisiScreen (Vision Research, Inc., Birmingham, AL), was invited to participate in the study in year 2 but declined due to ongoing equipment modifications.

Table 1. Key Characteristics of Screening Tests Used in the Vision in Preschoolers Study

	Test/Instrument*	Testing Distance	Stopping Rules for Each Eye	Maximum Number of Measurements per Eye	Possible Results
	Lea Symbols distance visual acuity test	3 m	2 wrong on 1 optotype size, or 3 correct on smallest optotype size	4 optotypes per optotype size	10/100 3 yrs: 10/32, 10/25, 10/20 4 yrs: 10/25, 10/20, 10/16
	HOTV distance visual acuity test	3 m	2 wrong on 1 optotype size, or 3 correct on smallest optotype size	4 optotypes per optotype size	10/100 3 yrs: 10/32, 10/25, 10/20 4 yrs: 10/25, 10/20, 10/16
	Random Dot E stereoacuity test	0.5 m, 1 m, 1.5 m	2 wrong at a distance, or 4 correct at 1.5 m	5 presentations per distance	Nonstereo card only: 504, 252, 168 arc sec
	Cover–uncover test	3 m, 40 cm	≥3 cover–uncover strokes	1	Strabismus, no strabismus
	Noncycloplegic retinoscopy	3 m	1 measurement	1	Refractive error values
	Retinomax Autorefractor	∼4 cm	Reliability score ≥8	3	Refractive error values
	Stereo Smile II stereoacuity test	40 cm	2 wrong at a disparity, or 4 correct at 120 arc sec	5 presentations per disparity	Nonstereo card only: 480, 240, 120 arc sec
	Power Refractor II	∼1 m	Value shown in green (valid reading)	3	Refractive error values in red or green
	iScreen Photoscreener	∼68 cm	Pupils in focus and diameter of ≥4 mm, good fixation in ≥1 eye in each image	3	Unreadable, abnormal, normal
	MTI Photoscreener	∼1 m	4 pupils in focus, pupils ≥4 mm in diameter, good fixation in ≥1 eye in each image	3	Unreadable, fail, pass
	SureSight Vision Screener	∼36 cm	Reliability score ≥6, no asterisk printed or asterisks on 2 readings	3	Refractive error values

Lea symbols distance visual acuity test 

Visual acuity is tested monocularly with a modification of the MassVAT (Optom Vis Sci 74:S174, 1997) form of the Lea Symbols,26 consisting of cards with linear arrays of either 4 (10/100 size) or 5 (other sizes) picture optotypes (square, circle, house, and apple). Optotypes are spaced at 1 optotype width and have a crowding bar surrounding the line at a distance of 0.5 optotype width. During a binocular pretest, the child must identify verbally, or by matching, each optotype presented singly at 1 m. Upon successful completion, the screener patches the child's left eye, moves to 3 m, and presents the 10/100 and age-specific cards. If the child correctly identifies 3 of 3 or 3 of 4 optotypes, the screener presents the next card. The procedure is repeated for the left eye, beginning with the 10/100 card. Vision in Preschoolers Study staff asked Head Start teachers to review the Lea Symbols with the children before the screening day.

HOTV distance visual acuity test 

The procedures for the HOTV test are identical to those used for the Lea Symbols test, except that the optotypes are the letters H, O, T, and V.

Random dot E stereoacuity test 

The RDE test consists of a demonstration plate (a non–stereo raised E), a blank plate (a random dot pattern), and a plate displaying a random dot stereo E. The child wears polarizing glasses. The screener familiarizes the child with the raised E plate and then conducts a pretest by presenting the raised E plate paired with the blank plate at a distance of 50 cm. If the child correctly identifies the raised E plate on 4 of 4 or 4 of 5 presentations, the screener successively presents the stereo E plate paired with the blank test plate at 50 cm, 1 m, and 1.5 m as long as the child correctly identifies the E plate on 4 of 4 or 4 of 5 presentations. The screener varies the left–right and/or up–down position of the plate in a nonsystematic manner.

Cover–uncover test 

Eye alignment is assessed using a cover–uncover test at both distance (3 m) and near (40 cm). The screener asks the child to look at a detailed, standardized fixation target and places a cover paddle over the child's left eye. The paddle is kept in front of the eye for approximately 3 seconds. The screener observes the unoccluded right eye to determine if refixation occurs. The cover–uncover stroke is repeated at least 3 times. The procedure is repeated, covering the right eye.

Noncycloplegic retinoscopy 

Noncycloplegic retinoscopy is used by eye care professionals to measure refractive error. The screener uses a streak retinoscope and a retinoscopy lens rack or handheld trial lenses. The child wears retinoscopy spectacles corresponding to the screener's working distance to control accommodation. Measurements, corrected for the screener's working distance, are obtained along the 2 principal meridia of each eye. The screener instructs the child to fixate on an animated target on a video screen.

Retinomax autorefractor 

The Retinomax is a handheld autorefractor that measures refractive error monocularly along 2 meridia. The screener places the instrument's headrest on the child's forehead, encourages the child to fixate the internal target, and focuses the mire in the center of the right pupil while ≥8 readings are taken automatically. The screener repeats the process for the left eye. The screener prints the refractive error and reliability rating for each eye. If the reliability reading is <8 (the manufacturer's recommended minimum value), the measurement is repeated.

Stereo smile II stereoacuity test 

The Stereo Smile II test consists of a demonstration plate (a nonstereo smile face on a background of random dots), a blank plate (a random dot pattern), and 3 plates each displaying a random dot stereo smile face of successively finer levels of stereopsis. The child wears Polaroid glasses. The screener conducts a pretest, by presenting the demonstration plate paired with the blank plate. If the child correctly identifies the demonstration plate on 4 of 4 or 4 of 5 presentations, the screener successively presents the blank plate paired with the plates of finer disparity as long as the child correctly identifies 4 of 4 or 4 of 5 presentations. The screener varies the left–right and/or up–down position of the plate in a nonsystematic manner.

Power refractor II 

The Power Refractor II (version 3.11.01.24.00) is a tabletop video/photorefractor that binocularly measures refractive error in 8 meridia and measures eye alignment. When the child fixates on the red and green lights on the camera, the screener begins the measurement and continues until the refractive error in each eye and gaze deviation appears in green on a display or until the instrument times out. The screener prints the display image. If the refractive error displayed for either eye is red, the measurement of the highlighted eye(s) is repeated. If the output for either eye is again red, measurement may be made monocularly.

iScreen photoscreener 

The iScreen consists of an off-axis photorefractor connected to a laptop computer that binocularly measures refractive error in one meridian and measures eye alignment. The child sits in front of the camera unit, with his or her head against a headrest, and fixates on a red light. The screener positions the child so that the eyes are centered horizontally along a crosshair that is visible on the display and takes a photograph. If the image is not adequate, the screener repeats the process. The last image collected is transmitted to the iScreen scoring center in Memphis, Tennessee.

MTI photoscreener 

The MTI consists of an off-axis photorefractor in which the eccentric flash rotates 90° to allow binocular measurement of refractive error along 2 orthogonal axes and measurement of eye alignment. The screener holds the camera approximately 1 m from the child, positions 2 pointed target lights on the child's forehead, tells the child to watch the red lights on the camera, and takes a flash photograph. The flash automatically rotates 90°, and the screener takes a second flash photograph. The images are developed in approximately 30 seconds on instant film. If any of the criteria in Table 1 are not met, the screener takes additional photographs. All photographs for each child are sent to the Vanderbilt Ophthalmic Imaging Center for interpretation.

SureSight vision screener 

The SureSight Vision Screener (version 2.12) is a handheld autorefractor that measures refractive error monocularly in 2 meridia based on wavefront technology.27 The screener looks through the viewfinder, instructs the child to fixate the red lights on the front of the machine, and centers a crosshair on the pupil of the right eye. Auditory cues signal the screener to adjust the testing distance and signal the completion of measurements for the right eye. The screener repeats the process for the left eye and prints a refractive error and reliability rating for each eye. If the reliability reading is <6 (the manufacturer's recommended minimum value) or an asterisk is printed to denote a refractive error that exceeds the SureSight normal bounds, the measurement is repeated. If 2 consecutive asterisks are obtained for the same eye, then the measurement is not repeated.

Training of screeners 

Each year, a team of VIP Study personnel conducted a daylong training program for screeners at each local clinical center. During year 2, the training program included instruction by representatives from the manufacturers of the Power Refractor II, the iScreen Photoscreener, and the SureSight Vision Screener, as well as trainers on the MTI Photoscreener who were based at the Vanderbilt Ophthalmic Imaging Center. Screeners were observed by the local principal investigator or coinvestigator while testing ≥3 children on each screening test. Screeners completed written knowledge assessments and, in year 2, had to submit ≥3 readable images to the respective reading center for the iScreen Photoscreener and the MTI Photoscreener. During year 2, the Coordinating Center provided alerts on individual screeners with relatively high percentages of images judged as unreadable. These screeners were asked to review the corresponding screening test protocol.

Gold standard examination procedures 

Monocular distance VA assessment, cover testing, and cycloplegic retinoscopy are used to determine if a child had amblyopia or reduced VA, strabismus, and/or significant refractive error. Evaluation of anterior segment and binocular indirect ophthalmoscopy is also performed to detect other possible causes of reduced VA. Other examination procedures such as color vision testing, NCR, stereoacuity testing, and ductions and versions are performed to provide a complete ocular assessment of the child. Gold standard examinations are conducted in the VIP van, which was reconfigured into 2 examination rooms. Examiners did not have access to the Head Start or VIP screening results.

Monocular distance visual acuity testing 

Monocular distance threshold VA testing is conducted using the Electronic Visual Acuity tester (Jaeb Center for Health Research, Tampa, FL) at 3 m,28 according to the protocol established by the Amblyopia Treatment Study.29 The letters H, O, T, and V are presented as isolated optotypes with crowding bars surrounding the letter. Visual acuity between 20/16 and 20/800 can be measured. Children who wear spectacles are tested while wearing the spectacles. Children are retested on the same day after cycloplegic retinoscopy with their full cycloplegic refractive correction if their initial VA scores place them in a category other than normal (Table 2) and they have a refractive error in either eye of ≥0.5 diopters (D) of myopia, ≥2.0 D of hyperopia, or ≥1.0 D of astigmatism.

Table 2. Definitions of Targeted Disorders in the Vision in Preschoolers Study

	Targeted Disorder	Definitions*
	Amblyopia
	Presumed unilateral	≥3-line interocular difference in VA and a unilateral amblyogenic factor†
	Suspected unilateral	2-line interocular difference in VA and a unilateral amblyogenic factor†
	Suspected bilateral
	3-year-olds	Worse than 20/50 in one eye, worse than 20/40 in the contralateral eye, and a bilateral amblyogenic factor‡
	4- and 5-year-olds	Worse than 20/40 in one eye, worse than 20/30 in the contralateral eye, and a bilateral amblyogenic factor‡
	Reduced VA§
	Bilateral
	3-year-olds	Worse than 20/50 in one eye, worse than 20/40 in the contralateral eye; no bilateral amblyogenic factor‡
	4- and 5-year-olds	Worse than 20/40 in one eye, worse than 20/30 in the contralateral eye; no bilateral amblyogenic factor‡
	Unilateral
	3-year-olds	Worse than 20/50 in only one eye or ≥2-line difference between the eyes (except 20/16 and 20/25); no unilateral amblyogenic factor†
	4- and 5-year-olds	Worse than 20/40 in only one eye or ≥2-line difference between the eyes (except 20/16 and 20/25); no unilateral amblyogenic factor†
	Strabismus	Any heterotropia in primary gaze
	Significant refractive error	Cycloplegic refraction
	Astigmatism	>1.50 D between principal meridians
	Hyperopia	>3.25 D in any meridian
	Myopia	>2.00 D in any meridian
	Anisometropia	>1.00-D interocular difference in hyperopia; >3.00-D interocular difference in myopia; >1.50-D interocular difference in astigmatism; antimetropic∥ difference >1.00 D and one eye >1.00 D of hyperopia; antimetropic∥ difference >3.00 D and one eye >2.00 D of myopia

D = diopters; VA = visual acuity.

* Applied sequentially for amblyopia and reduced VA. † Strabismus, anisometropia (as defined in Table 2), and a difference in spherical equivalent of ≥0.50 D when ≥1 eye had >3.50 D of hyperiopia were considered unilateral amblyogenic factors. ‡ Astigmatism of >2.50 D, hyperopia of >5.00 D, or myopia of >8.00 D in each eye were considered bilateral amblyogenic factors. § Reduced VA due to a cause other than amblyopia or refractive error; cause identified or not. ∥ One eye hyperopic, 1 eye myopic.

Training of examiners 

The examiners were trained by the VIP training team during a daylong program at the local clinical center. Examiners were observed performing GSE procedures on ≥2 preschool children and completed written knowledge assessments. During year 2, alerts were provided by the Coordinating Center to the clinical centers concerning individual examiners with a relatively high percentage of children with incomplete results on the GSE. The local clinical center principal investigator was asked to review the protocol procedures with the examiner.

Classification of children 

Definitions for the targeted conditions of amblyopia, strabismus, or significant refractive error are provided in Table 2, along with definitions for a fourth category, unexplained reduced VA. This fourth category was added to account for children who have VAs below age-specific thresholds that are not attributable to an amblyogenic factor.

The targeted disorders were grouped by severity of the condition into a hierarchy consisting of 3 groups. The groups were defined by a consensus of the VIP Executive Committee, with input from the VIP Advisory Committee and without reference to study data. The conditions included in each group of the hierarchy are described in the lefthand column of Table 3. Group 1 conditions are considered very important to detect and treat early. Group 2 conditions are considered important to detect early (but with less urgency than group 1). Group 3 conditions are considered less urgent, but nonetheless are clinically useful to detect. Children with multiple conditions were classified on the basis of their worst condition.

Table 3. Hierarchy of Vision in Preschoolers Targeted Disorders

	Condition	n	%
	Group 1: very important to detect and treat early	311	12.0
	Amblyopia	92	3.6
	Presumed unilateral and worse eye VA of ≤20/64	36	1.4
	Suspected bilateral	56	2.2
	Strabismus—constant	67	2.6
	Refractive error	269	10.4
	Severe anisometropia (interocular difference >2 D of hyperopia, >3 D of astigmatism, or >6 D of myopia)	43	1.7
	Hyperopia ≥5.0 D	123	4.8
	Astigmatism ≥2.5 D	146	5.6
	Myopia ≥6.0 D	15	0.6
	Group 2: important to detect early	229	8.8
	Amblyopia	27	1.0
	Suspected unilateral	20	0.8
	Presumed unilateral and worse eye VA of >20/64	7	0.3
	Strabismus—intermittent	29	1.1
	Refractive error	202	7.8
	Anisometropia, but not severe	64	2.5
	Hyperopia of >3.25 D and <5.0 D and interocular difference in SE of ≥0.5 D	64	2.5
	Astigmatism of >1.5 D and <2.5 D	113	4.4
	Myopia of ≥4.0 D and <6.0 D	3	0.1
	Group 3: detection clinically useful	215	8.3
	Reduced VA	173	6.7
	Bilateral	54	2.1
	Unilateral	119	4.6
	Refractive error	51	2.0
	Hyperopia of >3.25 D and <5.0 D and interocular difference in SE of <0.5 D	45	1.7
	Myopia of >2.0 D and <4.0 D	6	0.2
	Normal	1833	70.9
	Total no. of children	2588

D = diopters; SE = spherical equivalent; VA = visual acuity.

Children who did not have any conditions on their GSE that would place them in groups 1, 2, or 3 were classified as normal or indeterminant. Results for a child were considered indeterminant if their GSE results included incomplete or missing information that prevented them from being assigned to a group or to the normal category.

Data analysis 

Data collection forms from the screening and GSE sessions and the interpretation of iScreen and MTI photoscreener images were data entered, edited, and analyzed at the Coordinating Center. Statistical computations were performed using SAS 8.2 (SAS Institute, Cary, NC).

Only data from children who were both screened and examined are included in this report. Children who did not have a complete VA test, cover test, or cycloplegic refraction on the GSE are excluded from the analyses of sensitivity and specificity. Sample sizes for individual screening tests in each year vary because children left the van before administration of all screening tests and because of data-recording errors. When multiple readings of refractive error were obtained, the first measurement with a reliability score or color code considered reliable by the manufacturer was used. If no reading met the criteria, the reading with the highest score was used.

Study population 

In years 1 and 2, 2211 children who had failed the Head Start vision screening and 1772 children who had not failed the screening were selected for enrollment. Consent was obtained and eligibility criteria fulfilled for 3121 of the 3983 children (78.3%). Screening was completed for 2780 (89.1%) of the children enrolled. Gold standard examinations were performed on 2666 (95.6%) of screened children, of whom 2588 (97.1%) completed VA testing, cover testing, and cycloplegic refraction. All children were 3, 4, or 5 years old when screened; however, few young 3-year-olds or old 5-year-olds were tested (Table 4). There was substantial diversity in ethnicity and race.

Table 4. Characteristics of Children

	Characteristic	n	%
	Age at screening (mos)
	36–41	35	(1.4)
	42–47	473	(18.3)
	48–53	645	(24.9)
	54–59	725	(28.0)
	60–65	637	(24.6)
	66–71	73	(2.8)
	Gender
	Female	1291	(49.9)
	Male	1297	(50.1)
	Hispanic or Latino
	Yes	585	(22.6)
	No	1876	(72.5)
	Unknown	127	(4.9)
	Racial category
	Black	1239	(47.9)
	White	464	(17.9)
	American Indian	199	(7.7)
	Asian	94	(3.6)
	Native Hawaiian/Pacific Islander	20	(0.8)
	Other	112	(4.3)
	Mixed race	101	(3.9)
	Unknown	359	(13.9)
	Total no. of children	2588

The distribution of children according to the hierarchy of VIP targeted disorders is displayed in Table 3. Significant refractive error, as defined in Table 2, was present in 284 (91.3%) of the 311 children with group 1 conditions and in 204 (89.1%) of the 229 children with a group 2 condition in the absence of a group 1 condition. Among the 2588 children, 755 (29.2%) had ≥1 targeted disorders: 163 (6.30%) had amblyopia, 110 (4.25%) had strabismus, 539 (20.8%) had significant refractive error, and 246 (9.5%) had reduced VA as defined in Table 2. Significant refractive error was present in 58 (52.7%) of the 110 children with strabismus and in 30 (79.0%) of the 38 children with strabismic amblyopia.

Testability, readability, and number of repeated procedures 

The proportion of children who were not testable was <1% for the majority of the screening tests (Table 5). The proportion untestable was highest (9.7%) for the RDE test, with 24 of 215 3-year-olds (11.2%), 63 of 608 4-year-olds (10.4 %), and 24 of 319 5-year-olds (7.5%) not testable.

Table 5. Children Not Testable on Screening Tests

	Screening Test	n	Children Not Testable [n (%)]
	Year 1
	Lea Symbols VA	1142	6 (0.5)
	HOTV VA	1141	7 (0.6)
	Random Dot E	1142	111 (9.7)
	Cover–uncover	1141	24 (2.1)
	Noncycloplegic retinoscopy	1142	9 (0.8)
	Retinomax Autorefractor	1142	6 (0.5)
	Year 2
	Stereo Smile II	1446	27 (1.9)
	Power Refractor II*	1438	22 (1.5)
	iScreen Photoscreener†	1439	2 (0.1)
	MTI Photoscreener‡	1444	0 (0.0)
	SureSight Vision Screener	1442	11 (0.8)
	Retinomax Autorefractor	1435	2 (0.1)

VA = visual acuity.

* Thirty-two (2.2%) readings displayed red; invalid measurement. † Thirty-seven (2.6%) images ungradable. ‡ Eighty-four (5.8%) photographic sets ungradable.

Although Power Refractor II, iScreen, and MTI images were obtained for nearly all children, 32 of 1438 (2.2%) children had invalid Power Refractor II readings, 37 of 1439 (2.6%) children had unreadable iScreen Photoscreener images, and 84 of 1444 (5.8%) children had unreadable MTI Photoscreener images. One image set was obtained for 1301 of the 1438 (90.5%) children with the Power Refractor II, for 1011 of 1439 (70.3%) children with the iScreen Photoscreener, and for 919 of 1444 (63.6%) children with the MTI Photoscreener. Mean numbers of image sets per child were 1.14, 1.37, and 1.48, respectively, for the 3 instruments. In year 2, only 1 Retinomax Autorefractor reading was taken for 2523 of 2870 (87.9%) eyes (mean, 1.15 readings). Only 1 SureSight Vision Screener reading was taken for 1726 of 2884 (59.8%) eyes (mean, 1.64 readings).

Sensitivity of screening tests 

Overview 

Failure criteria when specificity was set at 0.90 are summarized in Table 6 for tests of VA and stereoacuity and in Table 7 for tests involving refractive error. Table 8 displays the sensitivity of the tests for detection of children who have any targeted condition, and for detection of children with group 1, group 2, or group 3 conditions. The P values for pairwise comparisons between screening tests when specificity is set at 0.90 are provided in Table 9 for detection of any targeted condition, and for detection of children with group 1 conditions. For example, the P value (0.70) for the comparison of the sensitivity for detecting ≥1 targeted conditions between the SureSight Vision Screener (0.63) and the NCR (0.64) is found in the SureSight row and the NCR column, whereas the P value (0.04) for the comparison of the sensitivity for detecting a group 1 condition between the 2 screening tests (0.81 vs. 0.90) is in the NCR row and the SureSight column. Table 10 provides sensitivity and specificity for each screening test for detection of amblyopia, reduced VA, strabismus, and significant refractive error. Table 11 displays the sensitivity of the iScreen and MTI photoscreeners and of the screening tests when failure criteria were determined with specificity set at 0.94. The sensitivity of the cover–uncover test was significantly (P<0.0001) lower than for all other tests for detecting children with any targeted condition or with a group 1 condition; these comparisons do not account for the higher specificity (0.98) of the cover–uncover test.

Table 6. Failure Criteria for Visual Acuity (VA) and Stereoacuity Tests to Maximize Sensitivity When Specificity Was Set at 0.90

	Test	Age (yrs)	Failure Criterion* (Inability to Pass)
	Lea Symbols VA	3	10/32 line
		4	10/20 line
		5	10/20 line
	HOTV VA	3	10/25 line
		4	10/25 line
		5	10/20 line
	Random Dot E stereoacuity	3	Nonstereo card
		4	Stereo card at 50 cm (550 arc sec)
		5	Stereo card at 100 cm (252 arc sec)
	Stereo Smile II stereoacuity	3	240–arc sec card
		4	240–arc sec card
		5	120–arc sec card

* Chosen to maximize overall sensitivity for detecting any targeted condition when specificity was set to 0.90.

Table 7. Failure Criteria for Retinoscopy, Photorefraction, and Autorefractor Screening Tests to Maximize Sensitivity

	Instrument	Hyperopia	Myopia	Astigmatism	Anisometropia*
	Noncycloplegic retinoscopy	≥2.75 D	≥2.75 D	≥1.25 D	≥1.50 D
	Power Refractor II	≥3.50 D	≥3.00 D	≥2.00 D	≥1.50 D
	SureSight Vision Screener†
	Manufacturer	>2.00 D	>1.00 D	>1.00 D	>1.00 D SE
	VIP Study	≥4.00 D	≥1.00 D	≥1.50 D	≥3.00 D
	Retinomax Autorefractor
	Year 1	≥1.50 D	≥2.75 D	≥1.50 D	≥2.00 D
	Year 2	≥1.50 D	≥2.75 D	≥1.50 D	≥1.75 D

D = diopters; SE = spherical equivalent; VIP = Vision in Preschoolers. With the exception of criteria set by the manufacturer, failure criteria were chosen to maximize overall sensitivity for detecting any targeted condition when specificity was set to 0.90 or above.

* The maximum of intereye differences in the power of the most positive meridian, the most negative meridian, and the magnitude of cylinder was used to determine presence of anisometropia for all tests, except in the manufacturer's criteria for SureSight. † Used in child mode, which adds a correction for accommodation.

Table 8. Sensitivity by Vision in Preschoolers Hierarchy of Conditions with Specificity Set to 0.90 for Tests without Established Failure Criteria

	Screening Test	Any condition (n = 346)	Sensitivity			Specificity (n = 796)
			Group 1 (n = 139)	Group 2 (n = 108)	Group 3 (n = 99)
	Year 1
	Lea Symbols VA	0.61	0.77	0.57	0.41	0.90
	HOTV VA	0.54	0.72	0.41	0.44	0.89
	Random Dot E	0.42	0.59	0.33	0.27	0.90
	Cover–uncover	0.16	0.24	0.13	0.06	0.98
	Noncycloplegic retinoscopy	0.64	0.90	0.63	0.29	0.90
	Retinomax Autorefractor	0.63	0.87	0.63	0.30	0.90
	Year 2	(n = 409)	(n = 172)	(n = 121)	(n = 116)	(n = 1037)
	Stereo Smile II	0.44	0.72	0.30	0.20	0.91
	Power Refractor II	0.54	0.72	0.43	0.39	0.90
	iScreen Photoscreener	0.37	0.57	0.24	0.20	0.94
	MTI Photoscreener	0.37	0.55	0.27	0.19	0.94
	SureSight Vision Screener*	0.85	0.96	0.90	0.63	0.62
	SureSight Vision Screener	0.63	0.81	0.68	0.29	0.90
	Retinomax Autorefractor	0.64	0.88	0.55	0.37	0.90

VA = visual acuity.

* Screening failure defined as recommended by the manufacturer.

Table 9. P Values Associated with Pairwise Comparisons of Sensitivity between Screening Tests with Specificity Set at 0.90

	Sensitivity	NCR	Retinomax, Year 2	Retinomax, Year 1	SureSight	Lea VA	HOTV VA	Power Refractor II	Stereo Smile	Random Dot E
	Any condition	0.64	0.64	0.63	0.63	0.61	0.54	0.54	0.44	0.42
	Group 1	0.90	0.88	0.87	0.81	0.77	0.72	0.72	0.72	0.59
	NCR		0.72	0.37	0.04	0.0015	<0.0001	<0.0001	<0.0001	<0.0001
	Retinomax, year 2	1.00		0.86	0.01	0.0093	0.0004	<0.0001	<0.0001	<0.0001
	Retinomax, year 1	0.73	0.88		0.21	0.02	0.0004	0.0019	0.0009	<0.0001
	SureSight	0.70	0.50	0.88		0.40	0.06	0.03	0.02	<0.0001
	Lea VA	0.23	0.36	0.37	0.65		0.27	0.36	0.30	0.0007
	HOTV VA	0.0039	0.0090	0.0027	0.03	0.03		1.00	1.00	0.02
	Power Refractor II	0.0060	0.0007	0.01	0.0048	0.08	0.94		0.90	0.02
	Stereo Smile	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	0.0068	0.0015		0.02
	Random Dot E	<0.0001	<0.0001	<0.0001	<0.0001	<0.0001	0.0002	0.0010	0.51

NCR = noncycloplegic retinoscopy; VA = visual acuity. P values in lower left triangle are for comparisons of sensitivity for detection of all conditions; those in upper right triangle are for comparison of sensitivity for detection of group 1 conditions only. If the conservative Bonferroni approach to multiple comparisons is applied and all pairwise comparisons are considered of interest, then comparisons associated with a P value of 0.05/66 = 0.0008 would be considered statistically significant.

Table 10. Sensitivity by Condition Type∗ with Specificity Set to 0.90 for Tests without Established Failure Criteria

	Screening Test	Sensitivity				Specificity (n = 796)
		Amblyopia (n = 75)	Reduced VA (n = 132)	Strabismus (n = 48)	Refractive Error (n = 240)
	Year 1
	Lea Symbols VA	0.76	0.58	0.56	0.70	0.90
	HOTV VA	0.73	0.48	0.65	0.59	0.89
	Random Dot E	0.63	0.38	0.60	0.47	0.90
	Cover–uncover	0.27	0.06	0.60	0.16	0.98
	Noncycloplegic retinoscopy	0.85	0.47	0.56	0.81	0.90
	Retinomax Autorefractor	0.85	0.50	0.65	0.78	0.90
	Year 2	(n = 88)	(n = 114)	(n = 62)	(n = 299)	(n = 1037)
	Stereo Smile II	0.77	0.30	0.68	0.51	0.91
	Power Refractor II	0.80	0.43	0.55	0.61	0.90
	iScreen Photoscreener	0.62	0.27	0.50	0.43	0.94
	MTI Photoscreener	0.63	0.24	0.65	0.42	0.94
	SureSight Vision Screener†	0.98	0.70	0.92	0.92	0.62
	SureSight Vision Screener	0.89	0.43	0.59	0.75	0.90
	Retinomax Autorefractor	0.85	0.45	0.69	0.76	0.90

VA = visual acuity.

∗ Children may have more than 1 condition. † Screening failure defined as recommended by the manufacturer.

Table 11. Comparison of Sensitivity for the iScreen and MTI Photoscreeners with Failure Criteria for Other Tests Set to Provide Specificity of 0.94

		Any Condition			Group 1 Conditions
		Sensitivity	P Values		Sensitivity	P Values
			iScreen	MTI		iScreen	MTI
	Year 1
	Lea Symbols VA	0.49	0.0004	0.0004	0.65	0.10	0.06
	HOTV VA	0.36	0.94	0.94	0.48	0.14	0.25
	Random Dot E	0.22	<0.0001	<0.0001	0.30	<0.0001	<0.0001
	Noncycloplegic retinoscopy	0.57	<0.0001	<0.0001	0.87	<0.0001	<0.0001
	Retinomax Autorefractor	0.52	<0.0001	<0.0001	0.81	<0.0001	<0.0001
	Year 2
	Stereo Smile II	0.33	0.17	0.17	0.57	0.91	0.73
	Power Refractor II	0.36	0.93	0.90	0.56	0.87	0.90
	iScreen Photoscreener	0.37	—	0.93	0.57	—	0.71
	MTI Photoscreener	0.37	0.93	—	0.55	0.71	—
	SureSight Vision Screener	0.51	<0.0001	<0.0001	0.75	0.0003	<0.0001
	Retinomax Autorefractor	0.52	<0.0001	<0.0001	0.81	<0.0001	<0.0001

VA = visual acuity.

Detection of children with targeted conditions 

The results showed that NCR, autorefraction (Retinomax Autorefractor and SureSight Vision Screener), and monocular Lea Symbols VA testing, when administered by LEPs, were the most accurate in distinguishing children who had ≥1 targeted vision disorders from children who did not have the conditions (Table 8).

When we examined sensitivity (the ability to detect children who have the targeted conditions) and specificity (the ability to identify children who do not have any of the targeted conditions) of each screening test for detection of each of the targeted conditions separately, the most accurate tests for detection of children with amblyopia were NCR, the Retinomax Autorefractor, and the SureSight Vision Screener (Table 10). The most accurate tests for detection of strabismus were monocular HOTV VA testing, the Retinomax Autorefractor, and the Stereo Smile II test. For detection of significant refractive error, the most accurate tests were the same as the best tests for detection of children with amblyopia: NCR, the Retinomax Autorefractor, and the SureSight Vision Screener.

Detection of children by severity of the condition 

Although amblyopia, strabismus, and significant refractive error all merit detection and further evaluation by an eye care professional, each condition can range in severity. To take this into account, the VIP Executive Committee developed a hierarchy that ranked the importance of detection of different levels of severity of each of the targeted conditions (Table 3). When each screening test was evaluated with respect to the hierarchy of severity of the targeted conditions, the tests that were best in detecting group 1 conditions (very important to detect and treat early) were NCR, the Retinomax Autorefractor, SureSight Vision Screener, and monocular Lea Symbols test (Table 9). For all screening tests, the sensitivity was lower for detection of group 2 (important to detect early) and group 3 (detection clinically useful) conditions than for detection of group 1 conditions (Table 9). However, the best tests for detecting group 2 conditions were NCR, the Retinomax Autorefractor, SureSight Vision Screener, and monocular Lea Symbols VA test. The best tests for detecting group 3 conditions were the 2 VA tests, Power Refractor II, and Retinomax Autorefractor.

In some screening settings, very high sensitivity may be preferred over high specificity, particularly for group 1 conditions. Differences among tests may be enhanced under this requirement. For example, if detecting 95% of children with group 1 conditions is the criterion, the specificities for NCR, the Retinomax Autorefractor, SureSight Vision Screener, and Lea Symbols VA test are 0.81, 0.70, 0.65, and 0.38, respectively. This ranking of the tests is the same ranking as determined when specificity was set at 0.90 and 0.94.

Refractive error and the targeted conditions 

A striking finding of this initial phase of the VIP Study is the accuracy of NCR, the Retinomax Autorefractor, and SureSight Vision Screener in detecting children who have ≥1 targeted conditions, as well as the most severe of these conditions, defined as group 1 in the hierarchy of targeted disorders. It is not surprising that these tests are accurate at picking up high refractive error; however, it is not obvious that these would be appropriate choices for detection of strabismus and strabismic amblyopia. In agreement with previous reports based on other populations of children,31, 32 examination of the VIP study population revealed that both strabismus and strabismic amblyopia were frequently associated with the presence of significant refractive error.

Screening tests assessing visual acuity and stereopsis 

Traditionally, VA testing has been an integral part of vision screening programs. In this phase of the VIP Study, monocular VA testing with Lea Symbols performed well overall and reasonably well in the detection of amblyopia, strabismus, and significant refractive error, as well as in the detection of group 1 conditions. However, testing VA was no better than NCR, the Retinomax Autorefractor, and SureSight Vision Screener in detecting any of the targeted conditions or in detecting group 1 conditions.

Another integral part of many vision screening programs is assessment of binocularity through cover testing and/or assessment of stereopsis. The cover–uncover test, which is used specifically to detect manifest strabismus, did not perform better in a screening setting than any other test in identifying children with strabismus. The longer period of observation of the child by the GSE examiner, the monocular assessment of VA before cover testing, and the addition of the alternating cover test to the examination sequence may have enhanced the detection of tropias, especially intermittent tropias, during the GSE.

The tests of stereoacuity were not as accurate as NCR, the Retinomax Autorefractor, and SureSight Vision Screener in detecting children with ≥1 targeted conditions or children with group 1 conditions. However, no other screening test was more accurate than the Stereo Smile II test in detection of strabismus. Comparison of the results of the 2 tests of stereoacuity showed that the Stereo Smile II test had higher testability than the RDE test, especially for younger children. However, when children who were unable to perform the test were classified as failing, the 2 tests performed similarly in detecting any targeted condition. The sensitivity of the Stereo Smile II test for detecting children with group 1 conditions was higher than the sensitivity of the RDE test (72% vs. 59%), but not to a statistically significant degree (P = 0.02).

Summary 

Overall, the results suggest that personnel skilled in NCR can detect approximately two thirds of children with ≥1 targeted disorders, and 90% of those with group 1 conditions, while referring 10% of normal children for an eye examination (90% specificity). In addition, the performance of the SureSight Vision Screener and Retinomax Autorefractor, when used by highly skilled personnel, is similar to that of NCR. However, additional research is needed to provide an evaluation of the autorefractors in the hands of individuals less experienced in evaluating ocular conditions and in more realistic screening environments.

Results of this study indicate that, in the hands of eye care professionals, the Lea Symbols test was nearly as accurate as NCR and the 2 autorefractors in detecting children who have ≥1 targeted conditions. The accuracy of the HOTV VA test was not as high as that of the Lea Symbols test, but the difference was not statistically significant. An important consideration with respect to screening with VA tests is the length of time needed to obtain monocular VA results in preschool children, which is considerably greater than that needed to obtain results from NCR or autorefraction.

Since the introduction of the first photorefractors in the late 1970s, there has been considerable interest in the use of this technology for large-scale screening of children.14, 33, 34, 35, 36 The results of the present study show that the iScreen and MTI photorefractors are not as accurate as NCR, the SureSight and Retinomax autorefractors, or monocular Lea Symbols VA testing in detecting whether a child has ≥1 targeted conditions.

Although the results of this study provide insight into which vision screening tests are most accurate in detecting the most prevalent vision disorders in preschool-aged children, caution should be used in generalizing these results. In this initial phase of the VIP Study, all testing was conducted in the controlled environment of a mobile medical van, and tests were administered by highly skilled personnel (optometrists and pediatric ophthalmologists) using a standardized protocol. In addition, all personnel underwent a certification procedure that included extensive didactic and hands-on instruction by an experienced training team that traveled to all sites and that, in year 2, included representatives of the manufacturers of the Power Refractor II, iScreen and MTI photoscreeners, and SureSight Vision Screener. These features enhance the comparison of the performance of the screening tests performed within years 1 and 2 but may not provide accurate estimates of the sensitivity and specificity of the screening tests when administered in more realistic screening environments by less trained personnel.

The positive predictive value (proportion of children having a targeted condition among those referred for a comprehensive examination) is also an important feature in evaluating screening tests. However, in addition to the above caveats on generalizability of results, this initial phase of the VIP Study tested a sample of children that, by design, had an over-representation of children with vision problems. Nevertheless, if estimates of the prevalence of groups 1, 2, and 3 conditions are extrapolated from VIP Study data, the positive predictive value is approximately 50% to 60% for each of the tests evaluated.

The VIP Study results presented here are relevant to the recommendations on preschool screening by professional, governmental, and private organizations concerned with vision in children.1, 4, 5, 6, 9, 10, 11, 12 Phase I of the VIP Study has provided evaluative data from which policy makers can begin to determine for their constituents the most reasonable approach to identifying preschool children with amblyopia, strabismus, and significant refractive error. More data will be provided as phases II (pediatric nurse screeners and lay screeners in realistic screening environments) and III (primary screening tests in a more general population) are completed. Furthermore, information on prevalence of the targeted conditions and on costs and benefits of early detection needs to be considered in determining the role of preschool vision screening in eliminating preventable loss of vision in our children.