This study recruited 465 amblyopic children aged 4–17 years from the China Amblyopia Behavioral Cohort (CABC), a multicenter initiative conducted between April 2024 and May 2026 across Guangxi, Guangdong, and Sichuan Provinces. The sample size was determined via an empirical approach (29), calculating 5–10 times the number of items in the SNAP-IV scale (26 items), yielding a minimum target of 130–260 participants.

The inclusion criteria for the study participants were as follows: The subject met the diagnostic criteria for amblyopia of the Chinese Medical Association Ophthalmology Branch (30), including anisometropic amblyopia (binocular spherical anisometropia ≥ 1.5D or cylindrical anisometropia ≥ 1.0D), ametropia amblyopia (binocular equivalent spherical anisometropia < 0.75D), and strabismic amblyopia; In addition, the subject had no other eye diseases or related systemic diseases; and no history of eye surgery. The exclusion criteria were as follows: The subject was found to be (1) unable to cooperate with the examination and (2) had other diseases that may affect the development of the central nervous system.

The severity classification criteria for amblyopia were as follows: (1) mild-to-moderate amblyopia, with best-corrected visual acuity (BCVA) lower than the age-specific normal value but ≤ 0.6 logMAR; and (2) severe amblyopia, with BCVA > 0.6 logMAR.

This study received ethical approval from the Ethics Committee of the Guangxi Zhuang Autonomous Region People’s Hospital (Approval No. KY-KJT-2023-285) and complied with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants and their guardians.

The data were analyzed via IBM SPSS Statistics 28.0 and AMOS 26.0. Normality was assessed via the Shapiro–Wilk test. Descriptive statistics (mean ± SDs) were used for normally distributed data; medians and interquartile ranges (IQR) were used for skewed data. Categorical variables are reported as frequencies/percentages (n, %). All the statistical tests were two-tailed, and a p-value < 0.05 was considered statistically significant. For comparisons involving multiple groups, Bonferroni-adjusted post hoc tests were applied to control for Type I error. To comprehensively convey the magnitude and precision of the findings, 95% confidence intervals (CIs) and specific effect sizes were reported for all key results. The specific effect size metrics used for each type of statistical test were listed as follow and were interpreted according to conventional thresholds.

A total of 465 amblyopic children (99.1% valid questionnaires) participated in this study, and their demographics and clinical features are summarized in Table 1 . The sample had a mean age of 6.28 ± 1.87 years, with 67.96% of subjects being aged ≤ 6 years and 51.0% being boys. Amblyopia severity was predominantly mild-to-moderate (94.6%), with refractive amblyopia being the most prevalent subtype (63.6%). Stratified analyses ( Supplementary Table S1 ) revealed no significant gender- or age-related differences in amblyopia severity (all p > 0.05, φ = 0.005–0.041, indicating negligible effects). However, severe amblyopia was associated with poorer BCVA in the worse eye (U=0, p<0.001, r_rβ = 0.398, medium effect) and was more prevalent in the anisometropic and strabismic subtypes (χ² = 17.2, p<0.001, Cramér’s V=0.192, small effect). Maternal smoking/alcohol history was low (< 5%), and 58.4% of the caregivers had a college-level education or higher. These results indicate that the sample was representative of typical clinical amblyopia populations.

As shown in Table 2 , all the factors were significantly positively correlated after Bonferroni correction (p < 0.001), with correlations ranging from moderate to strong (r_s = 0.724–0.923, large effect). The factor–total correlations (r_s = 0.887–0.923, large effect) were consistently greater than the correlations between factors (r_s = 0.724–0.768, large effect), supporting strong internal consistency. The strongest link was observed between inattention and hyperactivity/impulsivity (r_s = 0.768, 95% CI: 0.715–0.795, large effect), followed by moderate associations between inattention and oppositional behavior (r_s = 0.724, 95% CI: 0.677–0.766, large effect) and between hyperactivity/impulsivity and oppositional behavior (r_s = 0.739, 95% CI: 0.693–0.779, large effect). These results confirm the internal homogeneity of each factor and highlight meaningful relationships among core behavioral constructs.

As demonstrated in Table 3 , the reliability results for the SNAP-IV scale are presented. Cronbach’s α and split-half coefficients are retained as reliability metrics, which inherently reflect effect size-like properties for internal consistency. The Cronbach’s α for the total scale is 0.965 (95% CI: 0.958–0.972), and the Cronbach’s α values of each factor are greater than 0.90. The split-half reliability for the total scale is 0.891, and the individual factor coefficients range from 0.844–0.904. These metrics exceed the conventional psychometric thresholds (α ≥ 0.70), confirming that the SNAP-IV yields stable and reproducible scores in amblyopic children.

The correlations with the CPRS revealed strong convergent validity ( Table 4 ). Overall, the oppositional of the SNAP-IV exhibited the strongest associations, particularly with the conduct problem of the CPRS (r_s = 0.837, 95% CI: 0.807–0.863, p<0.001, large effect), indicating a close alignment between behavioral measures of oppositionality across the two instruments. Inattention was strongly correlated with both learning problems (r_s = 0.808, 95% CI: 0.767–0.834, p<0.001, large effect) and conduct problems (r_s = 0.719, 95% CI: 0.675–0.765, p<0.001, large effect), In contrast, correlations between the SNAP-IV subscales and the psychosomatic and anxiety factors of the CPRS were moderate (r_s = 0.346–0.478, p < 0.001, medium effects), indicating that these dimensions are less directly related to core ADHD symptom domains but still significantly overlap. Collectively, these results provide robust evidence that the SNAP-IV captures behavioral dimensions that are meaningfully associated with established CPRS constructs.

The initial three-factor model demonstrated suboptimal fit (χ²/(296) = 1225.6, χ²/df = 4.141, p < 0.05, RMSEA=0.082, CFI=0.90, IFI=0.90). Model refinement resulted in indices that met to acceptable levels (χ²/(291) = 1033.4, χ²/df = 3.551, p<0.05, RMSEA=0.074, CFI=0.92, IFI=0.92; Figure 1 ). The improved fit supports the scale’s structural validity in amblyopic populations.

The results of the convergent validity study are presented in Table 5 . Standardized factor loadings exceeded 0.5 (λ = 0.619–0.847, indicating large effects), with composite reliability (CR=0.924–0.928) and average variance explained (AVE=0.50–0.57) both meeting psychometric standards. The discriminant validity results in Table 6 show that the correlation coefficients between the factors range from 0.767 to 0.851, with some exceeding the square root of their respective AVEs. The correlation coefficient between inattention and hyperactivity/impulsivity (r =0.851) is greater than the square root value of the AVE (0.769), and the correlation coefficient between hyperactivity/impulsivity and the oppositional factor (r =0.811) is greater than the square root value of the AVE (0.762), indicating that some factors are not independent enough and that the discriminant validity needs to be improved. Furthermore, we now include maximum shared variance (MSV=0.724) and average shared variance (ASV=0.657) as supplemental effect size metrics, confirming moderate-to-high shared variance but acceptable discriminant validity per Fornell-Larcker criterion. These indices collectively suggest that while the factors are robust, discriminant validity between them requires improvement.

Measurement invariance for gender and age groups was evaluated via the three-factor model, with the results summarized in Table 7 . ΔCFI and ΔRMSEA are reported as effect sizes for model differences (e.g., for gender scalar invariance: ΔCFI=-0.002, small effect; ΔRMSEA=-0.001, negligible effect). Both groups showed an acceptable fit in the configural model (RMSEA=0.063, CFI=0.885, RMR=0.028 for gender; RMSEA=0.063, CFI=0.886, RMR=0.331 for age). The fit of the metric, scalar, and strict factor models was similar to that of the configural model (ΔRMSEA < 0.015, ΔCFI < 0.01), thereby supporting measurement invariance across gender and age groups.

This study represents the first systematic psychometric evaluation of the SNAP-IV rating scale in amblyopic children at high AD/HD risk, demonstrating exceptional reliability and validity. The scale exhibited outstanding internal consistency, surpassing previous cross-cultural adaptations in Brazil (α > 0.91) and China (α =0.88–0.90) (24, 28). These findings serve to reinforce the stability of the SNAP-IV in capturing homogenous behavioral constructs across diverse populations. The findings of the factorial analysis provided further support for the construct validity of the scale, with strong correlations between factors, indicating that all items reliably discriminate their respective dimensions (40, 41).

Importantly, the observed effect sizes across the main psychometric indices ranged from medium to large, highlighting that the associations observed are not only statistically significant but also of practical and clinical importance. Convergent validity with the CPRS was robust, particularly with respect to the oppositional factor of the SNAP-IV scale, which showed a large effect size correlation with the conduct problem factor of the CPRS (r_s = 0.837, p <0.001), which is consistent with evidence that children with unilateral visual impairment frequently present with heightened behavioral difficulties, including externalizing symptoms (23). Notably, large correlations between inattention and behavior problems (with learning problems and conduct problems, all p <0.001) underscore that attentional deficits exert a substantial behavioral impact in visually impaired children. Such magnitudes suggest that these relationships reflect meaningful behavioral mechanisms rather than trivial statistical findings. The large effect sizes observed support the hypothesis that attentional resource allocation deficits secondary to visual impairment, manifesting as cascading behavioral sequelae (42), thereby supporting theories of symptom overlap in neurodevelopmental disorders (43, 44). Together, these results demonstrate both statistical robustness and real-world relevance, underscoring the need for integrated interventions that addressing both visual symptoms and behavioral comorbidities in children with amblyopia.

The three-factor model of the SNAP-IV (inattention, hyperactivity/impulsivity, oppositional) is illustrated in Figure 1 . This model demonstrated strong cross-cultural adaptability with acceptable fit indices (p < 0.05) (28, 41, 45). However, initial model misfit (RMSEA=0.082) highlighted insufficient factor independence, a finding that is probably attributable to the clinical co-occurrence of AD/HD symptoms (44) and nonspecific behavioral responses to visual dysfunction (46). As depicted in Figure 1 , the standardized path coefficients ranged from 0.62 to 0.85, corresponding to large effect sizes, which affirm both the statistical robustness and the practical interpretability of the model. These magnitudes indicate that items have substantial loading strength on their latent constructs, advancing SNAP-IV understanding by confirming its factor structure in amblyopic children-a previously untested domain-and supporting its use for nuanced ADHD subtyping in visual impairment contexts. A high proportion of individuals ≤6 years of age (67.96%) may attenuate age norms, but invariance tests confirmed their applicability. Similarly, the measurement invariance across gender groups further corroborated the scale’s applicability across demographic subgroups.

However, configural and metric invariance were maintained (Δχ² = 26.782, Δdf = 23, p =0.265), and the small-to-medium effect size differences observed at the intercept level (ΔRMSEA = 0.001, ΔCFI = 0.006), which may reflect variations in behavioral expression due to age and environmental factors. This aligns with developmental trends showing declining hyperactivity with increasing age (20, 24). These effects are modest yet informative, emphasizing the need for age-adjusted thresholds in clinical practice. Furthermore, small effect size differences were detected across gender in scalar and strict invariance models (both p<0.05, ΔRMSEA < 0.005, ΔCFI < 0.002), with behavioral patterns consistent with clinical features of AD/HD (47). The minimal magnitude of these differences suggests they represent true behavioral variation rather than measurement bias. The small magnitude of these differences indicates they reflect genuine behavioral variation rather than measurement bias. Cross-cultural evidence suggests that these gender differences cannot be attributed to reporter bias or differences in assessment tools (48). These insights underscore the necessity of gender-sensitive assessment protocols in pediatric populations.

The SNAP-IV scale has demonstrated strong reliability and validity across various cultural adaptations (28, 49, 50), underscoring its clinical value in assessing behavioral issues in amblyopic children. The validated reliability and validity of the SNAP-IV in this understudied population provide a critical tool for the early identification of AD/HD comorbidities in amblyopic children. Given that a high percentage of amblyopic children develop AD/HD-related behaviors (51) and that visual treatment windows is typically limited to age 7 (52), this scale offers timely clinical utility. Our findings align with transactional developmental theory (42, 53), suggesting that visual impairment disrupts attentional processes, which in turn exacerbate behavioral outcomes. The overlap between attentional difficulties and visual impairment may further complicate diagnostic processes, underscoring the need for precise, population-specific assessment protocols. This mechanistic framework justifies multidisciplinary interventions targeting both sensory and behavioral domains in children with amblyopia. From the perspective of multidisciplinary management of amblyopia, the results of this study provide a practical tool for the comprehensive assessment and intervention of children with amblyopia, ultimately leading to better long-term outcomes for these children.

While groundbreaking, this study has several limitations. First, parent-reported data introduced potential reporter bias, as both the SNAP-IV and the CPRS were completed by caregivers (54). Second, the absence of a clinical AD/HD diagnosis group precluded sensitivity/specificity analysis. Third, the cross-sectional design provided limited insights into the temporal stability of the scale. Fourth, both the authors and the participants in this study belong to the same ethnic and cultural group, potentially influencing the results through in-group familiarity with subtle behavioral norms (42). Caution is warranted when generalizing findings to other populations. Last but still important, while the SNAP-IV/CPRS lack age/gender norms (unlike Conners-4), they were chosen for validated Chinese adaptations in pediatric cohorts and brevity for clinical feasibility. Future research should incorporate teacher and self-reports to triangulate data, longitudinal tracking to assess developmental trajectories (55, 56), and validation against gold-standard diagnostic interviews (e.g., Conners-4) (57). Additionally, optimizing items for age-specific expression (e.g., preschool vs. school-age) and incorporating neurocognitive measures (e.g., eye-tracking, fMRI) could enhance discriminative validity (58).