Journal of Pediatric Ophthalmology and Strabismus

Original Article 

The Positive Predictive Value of Smartphone Photoscreening in Pediatric Practices

Robert W. Arnold, MD, FAAP; Andrew W. Arnold, BA, COA; Taryn T. Hunt-Smith; Robin L. Grendahl, MD, FAAP; R. Kevin Winkle, MD, FAAP

Abstract

Purpose:

To compare smartphone photoscreening with other commercial objective screeners for amblyopia screening for young children.

Methods:

Ten pediatricians in four practices employed Nokia 1020 smartphones (Espoo, Finland) with single-axis Gobiquity software (Scottsdale, AZ) during well-child visits. Outcomes of confirmatory pediatric ophthalmology examinations were prospectively compared using American Association for Pediatric Ophthalmology and Strabismus uniform standards.

Results:

Five percent of 6,310 in-office screenings were referred: 25% for high anisometropia, 31% for hyperopia, and 15% for myopia. The positive predictive value (PPV) in 217 follow-up examinations was 68% (95% confidence interval: 62% to 74%) by 2013 age-stratified standards and 77% (confidence interval: 71% to 83%) by 2003 American Association for Pediatric Ophthalmology and Strabismus standards. The follow-up rate was 65%.

Conclusions:

Smartphone photoscreening had PPV comparable with other commercial objective screeners. Simple, valid photoscreeners should help pediatricians achieve widespread compliance with screening guidelines to reduce the burden of pediatric amblyopia vision impairment.

[J Pediatr Ophthalmol Strabismus. 2018;55(6):393–396.]

Abstract

Purpose:

To compare smartphone photoscreening with other commercial objective screeners for amblyopia screening for young children.

Methods:

Ten pediatricians in four practices employed Nokia 1020 smartphones (Espoo, Finland) with single-axis Gobiquity software (Scottsdale, AZ) during well-child visits. Outcomes of confirmatory pediatric ophthalmology examinations were prospectively compared using American Association for Pediatric Ophthalmology and Strabismus uniform standards.

Results:

Five percent of 6,310 in-office screenings were referred: 25% for high anisometropia, 31% for hyperopia, and 15% for myopia. The positive predictive value (PPV) in 217 follow-up examinations was 68% (95% confidence interval: 62% to 74%) by 2013 age-stratified standards and 77% (confidence interval: 71% to 83%) by 2003 American Association for Pediatric Ophthalmology and Strabismus standards. The follow-up rate was 65%.

Conclusions:

Smartphone photoscreening had PPV comparable with other commercial objective screeners. Simple, valid photoscreeners should help pediatricians achieve widespread compliance with screening guidelines to reduce the burden of pediatric amblyopia vision impairment.

[J Pediatr Ophthalmol Strabismus. 2018;55(6):393–396.]

Introduction

Amblyopia is a developmental disorder of cerebral vision leading to decreased visual acuity. It affects 2% to 4% of children and is most effectively treated when detected early.1,2 The American Academy of Pediatrics (AAP) endorses a series of screening tests for amblyopia including “conventional” methods of having cooperative, older children monocularly identify visual acuity optotypes.3 Photoscreening is a technique recently appended to AAP guidelines4 that identifies the risk factors for amblyopia: high levels of anisometropia, hyperopia, astigmatism, and manifest strabismus. Photoscreeners are flash cameras usually characterized by a close flash-to-lens distance so they produce red reflex in the pupils. Photoscreening involves analysis of light crescents in the pupillary red reflex related to defocus. Photo-screening has promise to reliably detect a child's amblyopia 2 to 3 years earlier than conventional visual acuity screening. The Alaska Blind Child Discovery (ABCD) vision study has adapted several consumer digital cameras as photoscreeners,5 eventually demonstrating that an iPhone 4 (Apple, Cupertino, CA) with its proximal LED flash near the lens can detect amblyopia risk factors.6 Gobiquity (Scottsdale, AZ) then developed a series of applications called “GoCheckKids” customized to various models of iPhone and Nokia (Espoo, Finland) smartphones with internal, onsite, computer interpretation of the captured, red-reflex images.7 In a pediatric eye clinic validation of high-risk children, GoCheckKids with the Nokia model 1020 smartphone had a sensitivity of 76%, a specificity of 67%, and a positive predictive value (PPV) of 57% by 2013 age-stratified American Association for Pediatric Ophthalmology and Strabismus (AAPOS) guidelines.8

The predictive value of a screening test is the proportion of referred cases found to have true disease. A high PPV has been shown for central-interpreted MTI Polaroid photoscreening (73% to 94%),9,10 and internally interpreted, infrared photo-screening (51% to 80%).11–15 PPV for photoscreening outperforms visual acuity screening in pediatric practice.16,17 Some pediatricians in our area have adopted Gobiquity instrument-based vision screening with modified Nokia model 1020 smartphones since 2014. We prospectively have sought to determine the PPV of patients thus referred to ours, the only pediatric ophthalmology clinic in Alaska.

Patients and Methods

This ABCD study was compliant with the Health Insurance Portability and Accountability Act of 1996 and consistent with the tenets of the Declaration of Helsinki, and was conducted with institutional review board approval from Providence Hospital. The study employed AAPOS 2003 and 2013 revision18 guidelines for uniform validation of vision screening.

As a part of well-child visits with 10 pediatricians representing four different private practices, young children received photoscreening using the Gobiquity app on Nokia model 1020 smartphones. Software versions for instrument referral criteria during the duration of this study were 4.2.xx to 4.6.xx and image processing software was version R2 from December 2014 through November 2016, version R4b from December 2016 through March 2017, and version R4d from April through June 2017.

Pediatric clinical staff were trained by Gobiquity consultants with ongoing support through ABCD. ABCD has prospectively monitored referral rates and predictive values for various photoscreeners charitably deployed throughout Alaska since 1996.19

Alaska does not yet employ a statewide follow-up coordinator for amblyopia screening, but office staff made multiple attempts to contact and reschedule patients who failed to show up for appointments. The only pediatric ophthalmology practice in Alaska does not own an optical shop and therefore is not on the provider panels for vision “insurance” plans. Confirmatory examinations by pediatric ophthalmologists included age-appropriate attempts at visual acuity, then pupil examination and motility, with assessment of accommodation, cycloplegic retinoscopy, and anterior segment and retinal examination.

Results

From December 2014 through May 2017, 217 patients referred by pediatrician Gobiquity screening had a confirmatory eye examination at a subspecialty pediatric ophthalmology practice. The time from photoscreening to examination ranged from 0 to 190 days, with a mean of 39 days.

The 10 pediatricians screened a total of 6,310 children; 332 did not pass, resulting in a referral rate of 5%. A consistent 10 to 14 patients were referred per month. The age at referral was younger than 35 months in 112, 36 to 59 months in 111, 60 to 85 months in 79, and older than 7 years in 30. Our examination of 217 patients represents a 65% follow-up rate. The number of children who had follow-up eye examinations with other local eye physicians or who never followed up is not known.

Reasons for Gobiquity referral are broken down by instrument estimation of risk factors: 83 for presumed anisometropia greater than 1.50 diopters (D) range, 104 for presumed hyperopia greater than 2.00 D range, and 49 for presumed myopia greater than 3.00 D range. There was overlap in comorbid referral indications; 52 had referral indications for hyperopia and anisometropia and 10 for combined myopia and anisometropia. Referral reasons less than these ranges, pupil asymmetry, and alignment anomalies constituted the other 172 cases.

Table 1 gives PPV calculations for the 217 referred children broken down by age ranges stratified using the 2013 AAPOS guidelines.20 The mean ± standard deviation PPV for all 217 referred children using 2013 guidelines was 68% ± 3% (95% confidence interval [CI]: 62% to 74%). We also compared the whole group using the less-specific 2003 AAPOS guidelines, with a PPV of 77% ± 3% (95% CI: 71% to 83%).18

PPV From Pediatrician Gobiquity Photoscreening Referrals Using AAPOS 2003 and 2013 Age-Based Validation Standards

Table 1:

PPV From Pediatrician Gobiquity Photoscreening Referrals Using AAPOS 2003 and 2013 Age-Based Validation Standards

The mean sphere was −1.40 ± 1.30 D in right eyes and −1.20 ± 1.50 D in left eyes in myopic patients, whereas the hyperopic patients had mean meridional hyperopia of +2.60 ± 2.10 D in right eyes and +2.80 ± 2.20 D in left eyes. The mean cylinder was 1.40 ± 1.10 D in right eyes and 1.40 ± 1.20 D in left eyes.

Discussion

Smartphone pediatrician photoscreening using the Gobiquity app yielded PPV consistent with other automated instrument-based vision screening technology.

Gobiquity has developed the smartphone vision screening app balancing accuracy with ease of use and speed in the pediatric office. Prior validation of single image (faster) was comparable to sequential, orthogonal flash (slower) and therefore the current referral paradigm uses only a single image.21 The flash-to-lens orientation is designed to detect suprathreshold astigmatism that is with the rule (plus axis near 90°). Suprathreshold astigmatism can be missed by this version of the Gobiquity app if it is against the rule or if it is mixed myopic-hyperopic.

Gobiquity has ongoing development of applications for most of the Apple smartphones since model 4, each of which has a light-emitting diode (LED) instead of a true flash. By mid-2018, iPhone models 6, 6s, 7, and 7+ have been calibrated (video link: https://vimeo.com/robertarnold/gobiquity2017). Patient movement and image blur is a limitation with the LED. Gobiquity has also modified the aperture of the true flash on Nokia smartphones to produce the preferred focus with higher speed imaging (video link: https://vimeo.com/robertarnold/gobiquitynokialumina). All of the Anchorage pediatricians employed Nokia model 1020 smartphones with the GoCheck-Kids app.

Validation statistics including PPV are influenced by the gold-standard examination criteria, the instrument referral criteria, the prevalence and severity of disease, and the screening environment. Screening environment includes pediatric eye office patients with enriched prescreening prevalence, community screening, and pediatrician office screening. GoCheckKids had pediatric eye office validation8 but lacked real pediatrician validation until we found a PPV of 77% ± 3% by the 2003 guidelines to compare to other prior objective screeners. The PPVs of commercially available photoscreeners are comparable. The Welch-Allyn SPOT (Skaneateles Falls, NY) had a PPV from 58% to 78%.12,13 Plusoptix (Nuremberg, Germany) models S04, S08, S09, and S12 had a PPV from 51%11 to 96%.22 The 2WIN (Adaptica, Padova, Italy) in the pediatric eye office had a PPV of 80% compared to 74% for the SPOT and 81% for the Plusoptix S12.14 The iScreen (iScreen Vision, Cordova, TN) centrally interpreted had a PPV of 87%.15 The Welch-Allyn Suresight had a PPV of 30%23 to 70%24 and the Retinomax (Nikon, Mellville, NY) had a PPV of approximately 56%.25

This study also allows some degree of comparison of the different AAPOS gold standard criteria; the PPV impact from the 2003 AAPOS guidelines compares most closely with the older age triad from the 2013 guidelines, whereas the PPV for younger age groups was approximately 10% lower.

We experienced a fair follow-up rate (65%) fostered by the close relationship between our pediatric ophthalmology practice and these referring pediatric practices. Prior experience before 2006 in a different practice found a 49% follow-up rate without a coordinator compared to a 72% follow-up rate with a paid coordinator.9 Before 2010, Iowa had an 81% follow-up rate with their coordinator.10 We suspect that our lack of a dedicated follow-up coordinator plays a partial role in suboptimal follow-up.

Strengths of our study are familiarity with the process and technology and the prospective design adhering to industry-standard, uniform validation guidelines. Weaknesses include the lack of follow-up on one-third of referrals and that pediatric ophthalmologists were not blinded to referral indications at the time of confirmatory examination.

Gobiquity smartphone photoscreening is a viable alternative so more pediatricians can comply with early, specific, objective, scheduled, well-child vision screening within the medical home.

References

  1. Repka MX, Kraker RT, Holmes JM, et al. Atropine vs patching for treatment of moderate amblyopia: follow-up at 15 years of age of a randomized clinical trial. JAMA Ophthalmology. 2014;132:799–805. doi:10.1001/jamaophthalmol.2014.392 [CrossRef]
  2. Kirk VG, Clausen MM, Armitage MD, Arnold RW. Preverbal photoscreening for amblyogenic factors and outcomes in amblyopia treatment: early objective screening and visual acuities. Arch Ophthalmol. 2008;126:489–492. doi:10.1001/archopht.126.4.489 [CrossRef]
  3. Donahue SP, Baker CNCommittee on Practice and Ambulatory Medicine, American Academy of PediatricsSection on Ophthalmology, American Academy of PediatricsAmerican Association of Certified OrthoptistsAmerican Association for Pediatric Ophthalmology and StrabismusAmerican Academy of Ophthalmology. Procedures for the evaluation of the visual system by pediatricians. Pediatrics. 2016;137. doi:10.1542/peds.2015-3597 [CrossRef]
  4. Miller JM, Lessin HR, American Academy of Pediatrics Section on Ophthalmology et al. Instrument-based pediatric vision screening policy statement. Pediatrics. 2012;130:983–986. doi:10.1542/peds.2012-2548. [CrossRef]
  5. Arnold RW, Clausen M, Ryan H, Leman RE, Armitage D. Predictive value of inexpensive digital eye and vision photoscreening: “PPV of ABCD.”Binocul Vis Strabismus Q. 2007;22:148–152.
  6. Arnold RW, Davis B, Arnold LE, Rowe KS, Davis JM. Calibration and validation of nine objective vision screeners with contact lens-induced anisometropia. J Pediatr Ophthalmol Strabismus. 2013;50:184–190. doi:10.3928/01913913-20130402-02 [CrossRef]
  7. Arnold RW, Armitage MD. Performance of four new photoscreeners on pediatric patients with high risk amblyopia. J Pediatr Ophthalmol Strabismus. 2014;51:46–52. doi:10.3928/01913913-20131223-02 [CrossRef]
  8. Peterseim MMW, Rhodes RS, Patel RN, et al. Effectiveness of the GoCheck Kids vision screener in detecting amblyopia risk factors. Am J Ophthalmol. 2018;187:87–91. doi:10.1016/j.ajo.2017.12.020 [CrossRef]
  9. Arnold RW, Donahue SP. The yield and challenges of charitable statewide photoscreening. Binocul Vis Strabismus Q. 2006;21:93–100.
  10. Longmuir SQ, Pfeifer W, Leon A, Olson RJ, Short L, Scott WE. Nine-year results of a volunteer lay network photoscreening program of 147 809 children using a photoscreener in Iowa. Ophthalmology. 2010;117:1869–1875. doi:10.1016/j.ophtha.2010.03.036 [CrossRef]
  11. Lowry EA, Wang W, Nyong'o O. Objective vision screening in 3-year-old children at a multispecialty practice. J AAPOS. 2015;19:16–20. doi:10.1016/j.jaapos.2014.09.008 [CrossRef]
  12. Bregman J, Donahue SP. Validation of photoscreening technology in the general pediatrics office: a prospective study. J AAPOS. 2016;20:153–158. doi:10.1016/j.jaapos.2016.01.004 [CrossRef]
  13. Arana Mendez M, Arguello L, Martinez J, et al. Evaluation of the Spot Vision Screener in young children in Costa Rica. J AAPOS. 2015;19:441–444. doi:10.1016/j.jaapos.2015.08.002 [CrossRef]
  14. Kirk S, Armitage MD, Dunn S, Arnold RW. Calibration and validation of the 2WIN photoscreener compared to the PlusoptiX S12 and the SPOT. J Pediatr Ophthalmol Strabismus. 2014;51:289–292. doi:10.3928/01913913-20140701-01 [CrossRef]
  15. Silbert DI, Arnold RW, Matta NS. Comparison of the iScreen and the MTI photoscreeners for the detection of amblyopia risk factors in children. J AAPOS. 2013;17:34–37. doi:10.1016/j.jaapos.2012.09.015 [CrossRef]
  16. Arnold RW, Stange CA, Ryan C. The compared predictive value of Bruckner, acuity and strabismus from pediatric referrals. Am Orthopt J. 2006;56:15–21. doi:10.3368/aoj.56.1.15 [CrossRef]
  17. Salcido AA, Bradley J, Donahue SP. Predictive value of photoscreening and traditional screening of preschool children. J AAPOS. 2005;9:114–120. doi:10.1016/j.jaapos.2003.10.011 [CrossRef]
  18. Donahue SP. The relationship between anisometropia, patient age, and the development of amblyopia. Trans Am Ophthalmol Soc. 2005;103:313–336.
  19. Arnold RW, Gionet E, Jastrzebski A, et al. The Alaska Blind Child Discovery project: rationale, methods and results of 4000 screenings. Alaska Med. 2000;42:58–72.
  20. Donahue SP, Arthur B, Neely DE, Arnold RW, Silbert D, Ruben JB. Guidelines for automated preschool vision screening: a 10-year, evidence-based update. J AAPOS. 2013;17:4–8. doi:10.1016/j.jaapos.2012.09.012 [CrossRef]
  21. Silbert DI, Arnold RW. Do we need to directly detect astigmatism when photoscreening for amblyopia risk factors (ARFs)?J AAPOS. 2015;19:e61. doi:10.1016/j.jaapos.2015.07.196 [CrossRef]
  22. Arnold RW, Tulip D, McArthur E, et al. Predictive value from pediatrician Plusoptix screening: impact of refraction and binocular alignment. Binoc Vis Strabismus Q. 2012;27:227–232.
  23. Rowatt AJ, Donahue SP, Crosby C, Hudson AC, Simon S, Emmons K. Field evaluation of the Welch Allyn SureSight vision screener: incorporating the vision in preschoolers study recommendations. J AAPOS. 2007;11:243–248. doi:10.1016/j.jaapos.2006.09.008 [CrossRef]
  24. Silbert DI, Matta NS, Ely AL. Comparison of SureSight autorefractor and plusoptiX A09 photoscreener for vision screening in rural Honduras. J AAPOS. 2014;18:42–44. doi:10.1016/j.jaapos.2013.09.006 [CrossRef]
  25. Lowry EA, de Alba Campomanes AG. Efficient referral thresholds in autorefraction-based preschool screening. Am J Ophthalmol. 2015;159:1180–1187. doi:10.1016/j.ajo.2015.02.012 [CrossRef]

PPV From Pediatrician Gobiquity Photoscreening Referrals Using AAPOS 2003 and 2013 Age-Based Validation Standards

GuidelineTrueFalsePPV ± 95% CIMonthsAnisometropia (D)Hyperopia (D)Astigmatism (D)Myopia (D)
2003
  All1684977% ± 6%12 to 84> 1.50> 3.50> 1.50a> 3.00
  Myopia191066% ± 17%> 3.00
  Anisometropia461082% ± 10%> 1.50
  Hyperopia601086% ± 8%> 3.50
2013
  Toddlers593563% ± 10%12 to 30> 2.50> 4.50> 2.00> 3.00
  Preschool412166% ± 12%31 to 48> 2.00> 4.00> 2.00> 3.00
  Kindergarten451180% ± 10%> 49> 1.50> 3.50> 1.50> 3.00
  Total1456768% ± 6%12 to 84
Authors

From Alaska Blind Child Discovery, Alaska Children's Eye & Strabismus, Anchorage, Alaska.

Dr. Arnold is a non-paid member of the Medical Advisory Board for Gobiquity and a board member of Glacier Medical Software that markets neonatal intensive care unit cloud-based vision monitoring software, ROP-Check. Mr. Arnold has been a calibration study site coordinator for Gobiquity. The remaining authors have no financial or proprietary interest in the materials presented herein.

Supported by Alaska Blind Child Discovery.

Correspondence: Robert W. Arnold, MD, FAAP, Alaska Children's Eye & Strabismus, 3500 Latouche #280, Anchorage, AK 99508. E-mail: eyedoc@alaska.net

Received: March 20, 2018
Accepted: May 25, 2018
Posted Online: August 29, 2018

10.3928/01913913-20180710-01

Sign up to receive

Journal E-contents