Development and Diagnostic Accuracy of a Novel Screening Tool for Early Detection of Pediatric Visual Impairment in Indonesian School-Aged Children

Response: We thank the reviewer for this important observation and fully agree. We have added an explicit statement early in the Discussion section to frame CIPSEL unambiguously as a preliminary triage instrument rather than a diagnostic tool. The following text has been added (Discussion, Paragraph 1):

"It is essential to emphasize that a positive CIPSEL screen does not constitute a diagnosis. The tool is intended as a preliminary triage layer, identifying children who warrant referral for formal clinical evaluation, including best-corrected visual acuity testing and, where indicated, cycloplegic refraction, rather than as a replacement for ophthalmological assessment."

This framing is also reinforced in the Conclusions section, which now consistently refers to CIPSEL as a first-line screening and triage tool.

Comment 2

Since the study focused on children between aged 8 to 12 years, this tool functions more accurately as a refractive error screener rather than a comprehensive amblyopia prevention tool. Given that effective amblyopia treatment is most successful in younger cohorts, please include a discussion on the necessity of developing a separate version of CIPSEL tailored for the 4–6 age group to truly address preventable vision loss early in the critical period of visual development.

Response: We agree with this clinically important observation. A dedicated paragraph has been added to the Discussion section addressing the critical period of visual development and the necessity of developing a CIPSEL version adapted for the 4–6 age group. This paragraph is located immediately following the community ophthalmology implications discussion, before the Limitations paragraph:

"The current study was conducted among children aged 8–12 years (grades 3–5), a population in whom the critical period for amblyopia development has largely concluded. While CIPSEL demonstrates strong screening performance in this age group, the greatest opportunity for preventing permanent vision loss lies in earlier detection, ideally before the age of 7–8 years, when neuroplasticity of the visual cortex allows for effective amblyopia treatment [33]. Future development should therefore prioritize the adaptation of CIPSEL for younger cohorts, particularly children aged 4–6 years. A version tailored for this age group would require significant modification: the questionnaire format would need to shift from self-report to a proxy-report instrument completed by parents or caregivers, and items would need to be redesigned around observable behaviors accessible to non-clinical observers, given the limited verbal and cognitive capacity of preschool-aged children."

The supporting reference cited is: Thompson B et al. Harnessing Brain Plasticity to Improve Binocular Vision in Amblyopia: An Evidence-Based Update. Eur J Ophthalmol. 2024;34:901–912 [Ref. 33].

Comment 3

Please provide specific examples of how the "medium risk" category should be managed in real world clinic settings such as implementing a formal 3-6 month monitoring interval for scheduled re-screening.

Response: We thank the reviewer for this practical and clinically relevant suggestion. We have added a dedicated paragraph within the tiered stratification discussion section that provides concrete, real-world management guidance for the Medium Risk category:

"For children in the Medium Risk category (score = 3), a structured monitoring approach is recommended rather than immediate referral or discharge. A re-screening interval of 3–6 months is proposed, allowing sufficient time to observe whether symptoms stabilize or progress, while avoiding unnecessary burden on specialist services. If symptoms worsen before the scheduled re-screening — such as increased squinting, declining academic performance, or new visual complaints — prompt referral to an ophthalmologist is warranted."

Additionally, Table 6 (CIPSEL Risk Stratification and Recommended Clinical Actions) has been updated to specify this interval explicitly in the Suggested Action column for Medium Risk, replacing the previous non-specific phrase 'Re-screen within a few months.'

Comment 4

Typo in Figure 2. Correct Students with normal "Visus" to Vision.

Response: We thank the reviewer for identifying this typographical error. The label in Figure 2 has been corrected from "Students with Normal Visus" to "Students with Normal Vision" in the revised figure file submitted alongside this manuscript.

Comment 1

The study included students in grades 3-5, and the age range is stated in the manuscript as "approximately aged 8–12 years" (line 270). This suggests that the authors did not directly record the age, but rather used an estimated value. However, the Abstract states "mean age 8.6 years" (line 23). The Methods section states that demographic data (including age) were not individually recorded due to an anonymity protocol (line 190). So how was the average age of 8.6 years arrived at? If the estimated age was calculated based on the grade level of the participating children and the number of children in each grade, is this truly a solid enough statistical calculation to justify specifying an average value in such a study?

Response: We thank the reviewer for identifying this important inconsistency. Upon reflection, we fully agree that specifying a mean age value of 8.6 years was methodologically unjustifiable given that individual age data were not recorded. The value was retrospectively estimated from grade-level distribution, which does not constitute a robust enough basis to report as a precise statistical mean.

Accordingly, we have removed the mean age of 8.6 years from the Abstract and all other sections of the manuscript. All references to participant age now consistently use the estimated range of 8–12 years based on grade level, with explicit justification. The following clarification has been added to the Methods section:

"The age range of 8–12 years was estimated based on the standard Indonesian primary school entry age (7 years) and the grade levels of participating students (grades 3–5), consistent with the Ministry of Education's national schooling guidelines. Individual age data were not recorded in accordance with the anonymization protocol."

The Highlights section has also been updated to read: "131 students from grades 3 to 5, estimated to be between 8 and 12 years of age based on the Indonesian standard school entry age system," replacing the previous incorrect reference to a mean age.

Comment 2

Similarly, gender information was not recorded. Since neither gender nor age data is available, it is not possible to perform subgroup analyses.

Response: We agree with this observation and acknowledge it as a meaningful limitation. The absence of individually recorded sex and age data was a consequence of the anonymization protocol applied during the government-mandated school health screening program under which this study was embedded. While this protocol is ethically justified and consistent with the KEPPKN 2021 exemption criteria, the reviewer is correct that it precludes sex-stratified and age-stratified subgroup analyses.

We have added an explicit acknowledgement and methodological rationale for this limitation in the Limitations section:

"The absence of individually recorded sex and age data, while consistent with the anonymization protocol of the government-mandated screening program, precludes sex-stratified or age-stratified subgroup analyses. Given that the prevalence and symptomatology of refractive errors may differ between sexes and across specific age cohorts within the 8–12 year range, future studies should incorporate these variables under appropriate data protection frameworks to enable more granular analysis."

We also note this explicitly in the Future Research section as a priority direction for subsequent validation studies.

Comment 3

The study used the Snellen test. This test does not replace cycloplegic refraction. The authors, in a very honest approach, acknowledged this fact and did not claim a definitive diagnosis. However, the article highlights a fundamental drawback of this choice. Using a Snellen test without cycloplegia can lead to diagnostic difficulties. For example, hyperopia can easily be overlooked. Or children with eye strain, attention deficit, or cooperation problems might be mistakenly classified as having "visual impairment." In short, AUC values are compared with a flawed reference standard rather than a true gold standard. In this case, if the reference standard itself is flawed, how can we be sure how well or poorly the instrument accurately reflects its true diagnostic performance?

Response: We thank the reviewer for raising this fundamental methodological point with such clarity. This is perhaps the most critical scientific concern in the review, and we have addressed it substantively rather than defensively.

We agree that the use of non-cycloplegic Snellen visual acuity as the reference standard represents an important limitation that requires transparent and specific acknowledgement. We have substantially expanded the discussion of this limitation with a dedicated paragraph supported by three peer-reviewed references, now included in the Discussion section (following the initial Snellen acknowledgement):

"A fundamental methodological consideration concerns the use of uncorrected Snellen visual acuity as the reference standard, rather than cycloplegic refraction, which is widely regarded as the gold standard for diagnosing refractive errors in children [18]. This limitation was acknowledged a priori and the study outcome variable was consequently framed as 'visual impairment consistent with possible URE' rather than 'confirmed refractive error.' It is recognized that non-cycloplegic Snellen testing may underestimate accommodative hyperopia, as children with significant hyperopia may compensate through active accommodation and achieve normal visual acuity despite clinically relevant refractive error [20], and may conversely overestimate visual impairment in children with accommodative excess or spasm, where transiently reduced distance acuity reflects a functional rather than structural deficit [21], or in those with limited cooperation, attention, or comprehension of the test task during mass screening [19]. However, this pragmatic choice reflects real-world primary care constraints in Indonesia, where cycloplegic refraction is not feasible in mass school-based screening. Importantly, this same limitation is shared by virtually all large-scale school vision screening programs globally, including those endorsed by the WHO, which rely on visual acuity as the primary screening metric because of its practicality [22]. The AUC of 0.887 therefore reflects CIPSEL's performance against the same reference standard used operationally in the target setting, a programmatically valid benchmark even if not the biological gold standard."

Regarding the reviewer's specific concern about whether the AUC can be trusted given an imperfect reference standard: we wish to clarify that the vast majority of school vision screening validation studies, including those cited in our manuscript, use non-cycloplegic visual acuity as their reference standard precisely because it represents the operational benchmark against which any screening tool must perform in real-world settings. Cycloplegic refraction is not available at the point of primary-care screening anywhere in Indonesia's public health system. The CIPSEL tool is designed to work within this operational reality, and its AUC reflects its accuracy relative to the same standard it would be used alongside in practice. Future studies should indeed compare CIPSEL performance against cycloplegic refraction as a confirmatory gold standard, a recommendation now explicitly stated in the Future Research section.

New references have been added to support the specific claims in this paragraph:

18. Morgan, I.G.; Iribarren, R.; Fotouhi, A.; Grzybowski, A. Cycloplegic Refraction Is the Gold Standard for Epidemiological Studies. Acta Ophthalmol. (Copenh.) 2015, 93, 581–585, doi:10.1111/aos.12642.

19. Arnold, R.W.; Donahue, S.P.; Silbert, D.I.; Longmuir, S.Q.; Bradford, G.E.; Peterseim, M.M.W.; Hutchinson, A.K.; O’Neil, J.W.; De Alba Campomanes, A.G.; Pineles, S.L. AAPOS Uniform Guidelines for Instrument-Based Pediatric Vision Screen Validation 2021. J. Am. Assoc. Pediatr. Ophthalmol. Strabismus 2022, 26, 1.e1-1.e6, doi:10.1016/j.jaapos.2021.09.009.

20. O’Donoghue, L.; Rudnicka, A.R.; McClelland, J.F.; Logan, N.S.; Saunders, K.J. Visual Acuity Measures Do Not Reliably Detect Childhood Refractive Error - an Epidemiological Study. PLoS ONE 2012, 7, e34441, doi:10.1371/journal.pone.0034441.

21. Monsa Hilora; Koushik Tripathy Accommodative Excess; StatPearls [Internet].; StatPearls Publishing: Treasure Island (FL), 2025;

Comment 4

The Buderer formula indicated that the study's sample size should be at least 130 people. 131 students were included in the study. However, all the children were selected from a single school in the South Jakarta region. How can we say that these 131 students represent all of Indonesia or the WTO countries, including the region where the study was initially conducted? Furthermore, the 19.8% prevalence found in the study seriously contradicts the 40% reference used in the methodology section.

Response: We thank the reviewer for raising both the generalizability concern and the prevalence discrepancy. We address each in turn.

On generalizability: We agree that a single-school sample from South Jakarta cannot and should not be presented as representative of all of Indonesia, let alone other LMIC settings. The study was explicitly designed as an initial pilot validation study, not a population survey, and the sample size was calculated for diagnostic accuracy purposes (i.e., to ensure reliable sensitivity and specificity estimates), not for population representativeness. We have strengthened the Limitations section with the following explicit statement:

"As a single-center pilot study conducted in one urban private school in South Jakarta, generalizability to rural, public school, or lower socioeconomic settings remains to be established. The study was designed as an initial validation study, not a population survey, and the sample size was calculated for diagnostic accuracy studies specifically, not for prevalence estimation."

On the prevalence discrepancy (19.8% vs. 40%): The reviewer identifies that the observed prevalence of 19.8% is substantially lower than the 40% figure used in the Buderer sample size calculation. We wish to clarify that the 40% figure was deliberately selected as a conservative, worst-case prevalence estimate derived from a post-pandemic Jakarta-wide study (Darusman et al., 2023), which assessed a broader and more heterogeneous school population. In diagnostic accuracy sample size calculations, using a higher prevalence estimate is standard practice to ensure that the number of 'positive' cases (i.e., children with visual impairment) is sufficient to produce reliable sensitivity estimates. Using the conservative estimate of 40% ensured that our sample would contain an adequate number of true positive cases even if the actual prevalence was lower. The observed prevalence of 19.8% in this private school sample is consistent with the urban private school setting of this study, where children may have already undergone prior eye care, and should not be interpreted as contradicting the broader Jakarta prevalence data. We have added the following clarifying sentence to the Methods section, immediately following the sample size paragraph:

"The 40% prevalence figure was selected as a conservative worst-case estimate to ensure adequate sample size for diagnostic accuracy analysis. The actual observed prevalence of 19.8% in this single-school pilot study reflects the specific characteristics of the study setting and should not be interpreted as a population-level prevalence estimate."

We appreciate the reviewer's attention to this detail, which has allowed us to provide a more rigorous and transparent explanation of the sample size rationale.

A

No — never worn glasses

0

No — does not wear glasses at all

0

0

No prior diagnosis or correction; vision has not been evaluated or corrected

B

Yes — worn glasses before

1

No — does not wear them consistently

0

1

Prior diagnosis but non-compliant with correction; potential ongoing uncorrected impairment

C

Yes — worn glasses before

1

Yes — wears them throughout the day

1

2

Prior diagnosis and fully compliant with correction; most likely managed