Arabic child oral-health scale shows provisional four-factor structure in Saudi mothers

A cross-sectional study of 602 Saudi mothers with children aged 2–6 years found that the Arabic Early Childhood Oral Health Impact Scale, or A-ECOHIS, had a four-factor structure in this population: Family Impact, Child Self-Image, Child Psychology, and Child Function and Symptoms. Exploratory factor analysis identified these dimensions, and confirmatory analysis in an independent subsample supported the same arrangement with acceptable model fit. The investigators described an 11-item version excluding difficulty pronouncing words and missed daycare or preschool as promising but provisional, not yet a validated replacement for the original 13-item scale. The findings indicate that the Arabic questionnaire may not work as a single undifferentiated measure, although external validation is still required.

The researchers distributed an Arabic Google Forms questionnaire by email to educational institutions and through WhatsApp groups for schools, parents, and parent associations. Recruitment combined convenience and snowball sampling across several Saudi regions, including Riyadh, Makkah, Madinah, Aseer, the Eastern region, Al-Qassim, and Tabuk. Of 730 questionnaires received, 128 were excluded, leaving 602 eligible mother–child pairs. The children’s mean age was 4.94 years, ranging from 2 to 6 years, and the sample was almost evenly divided between girls and boys. Most mothers were aged 26–35 years, and 64.8% held a bachelor’s degree. The sampling method and educational profile indicate that the results may not represent all Saudi families with preschool-aged children.

The 13-item A-ECOHIS asks caregivers how often a child experienced oral-health-related effects. Nine items concern the child and four concern the family; responses range from “never” to “very often,” while “I don’t know” is treated as missing. The questionnaire covers pain, drinking and eating difficulties, pronunciation, missed daycare or preschool, sleep disturbance, irritability, avoided smiling or talking, family distress, work absence, and financial impact. Complete responses from 567 participants yielded a mean total score of 12.25, with a standard deviation of 7.17; lower scores represent better perceived oral health-related quality of life. Missingness was low, ranging from 0% to 1.3% for individual items, and no values were imputed for score-based analyses.

The original scale showed good internal consistency across the complete cases, with Cronbach’s alpha of 0.825 and McDonald’s omega of 0.827. The provisional 11-item version had nearly identical coefficients, alpha 0.828 and omega 0.828. Reliability differed across domains: Child Self-Image was strong at alpha 0.831 and omega 0.832; Family Impact was acceptable at 0.794 and 0.796; and the revised Child Function and Symptoms factor was acceptable at 0.762 and 0.771. Child Psychology remained questionable, with alpha 0.614 and omega 0.611, indicating that its two retained items may not fully capture the intended construct. The original four-item Child Function domain performed poorly, with alpha 0.590 and omega 0.645, whereas the full 13-item scale and the revised 11-item scale both met the study’s criteria for good internal consistency.

For exploratory factor analysis, the researchers randomly divided the dataset into two halves. After listwise deletion, 285 responses formed the EFA sample and 286 formed the holdout sample for confirmatory factor analysis. EFA used a polychoric correlation matrix, principal-axis factoring, and oblique rotation to accommodate ordinal responses and correlated factors. Data suitability was supported by Kaiser–Meyer–Olkin values of 0.797 and 0.802 in the two samples and significant Bartlett tests of sphericity. Scree-plot inspection and parallel analysis based on 1,000 simulated datasets both indicated four factors, which explained 57.76% of the observed variance. The first four observed eigenvalues were 4.424, 1.051, 0.558, and 0.271, exceeding their corresponding simulated means; the fifth observed eigenvalue was 0.071, below its simulated mean of 0.116.

The exploratory solution grouped the four family items together: family members feeling upset or guilty, taking time off work, and experiencing financial consequences. Two items formed Child Self-Image, covering avoided smiling or laughing and avoided talking with other children. Trouble sleeping and irritability or frustration formed Child Psychology. Pain, difficulty drinking beverages, and difficulty eating certain foods formed Child Function and Symptoms, with pain also cross-loading on Family Impact but having a larger absolute loading on the functional factor. Pronunciation difficulty and missed daycare or preschool did not reach the study’s prespecified loading threshold of 0.40 on any factor. The avoided-talking item produced an unstable exploratory loading above 1, although it performed acceptably in the independent confirmatory analysis. The study did not directly compare the four-factor model with a six-domain confirmatory model, so the evidence concerns the tested one-factor and two-factor alternatives rather than a formal comparison with all six original domains.

Confirmatory factor analysis used weighted least squares mean- and variance-adjusted estimation, which is designed for ordinal items. The four-factor model showed a comparative fit index of 0.979, Tucker–Lewis index of 0.969, root mean square error of approximation of 0.070 with a 90% confidence interval of 0.052–0.088, and standardized root mean square residual of 0.049. It outperformed both a one-factor model, with a comparative fit index of 0.829 and RMSEA of 0.184, and a two-factor model corresponding to the original child-and-family structure, with values of 0.881 and 0.155. All standardized loadings in the four-factor model were statistically significant at p < 0.001 and ranged from 0.681 to 0.967. A modified model that allowed correlated errors between work absence and financial impact improved fit modestly, but the researchers retained the more parsimonious original model.

Average variance extracted was 0.625 for Family Impact, 0.850 for Child Self-Image, 0.551 for Child Psychology, and 0.642 for Child Function and Symptoms. Composite reliability values were 0.869, 0.919, 0.709, and 0.843, respectively. All average variance extracted estimates exceeded the study’s 0.50 criterion, while composite reliability met its 0.70 threshold. Discriminant-validity checks favored separation among the factors: all heterotrait–monotrait ratios ranged from 0.347 to 0.677, below the study’s 0.85 criterion, and each factor’s square root of average variance extracted exceeded its correlations with the other factors. The strongest factor correlation was 0.674 between Family Impact and Child Function and Symptoms, indicating that the two constructs were related but not interchangeable within this sample.

External convergent validity was supported by a moderate association between the total A-ECOHIS score and mothers’ global rating of their child’s oral health among 553 participants with complete data. The Spearman correlation was 0.443, with p < 0.001, linking higher impact scores to poorer perceived oral health. At the item level, missed daycare or preschool was reported at least sometimes by 43.5% of respondents, irritability by 42.7%, and pain by 37.5%. Conversely, 63.5% never reported avoided smiling or laughing, 61.0% never reported avoided talking, and 58.0% never reported a family member taking time off work. Those three items had high floor effects, yet their factor loadings remained substantial, whereas the two non-loading items did not have high floor effects. These are caregiver-reported frequencies, not clinical outcomes.

The study has important boundaries. Its cross-sectional, online design measured caregiver perceptions rather than changes over time, and convenience plus snowball recruitment may have introduced selection bias. Split-sample validation is internal cross-validation, not external replication, and the Arabic-language questionnaire was tested in a Saudi maternal population whose children ranged from 24 to 72 months. The study did not repeat test–retest reliability, and clinical caries information was available for only 150 participants and was not incorporated into the psychometric analysis, so criterion-related validity against objective measures such as decayed, missing and filled tooth scores was not assessed. The 11-item, four-factor structure should therefore be treated as an initial refinement requiring further item revision, particularly in Child Psychology, and confirmation in newly recruited samples from different settings before it can be considered a validated alternative to the 13-item A-ECOHIS.

Subject of Research: Psychometric evaluation of the Arabic Early Childhood Oral Health Impact Scale among Saudi mothers

Article Title: Factorial validation and psychometric evaluation of the Arabic early childhood oral health impact scale among Saudi mothers

Article References: Alissa, N. I., & AlZamil, H. Z. (2026). Factorial validation and psychometric evaluation of the Arabic early childhood oral health impact scale among Saudi mothers. Frontiers in Oral Health, 7, Article 1923290. https://doi.org/10.3389/froh.2026.1923290

Image Credits: AI Generated

DOI: 10.3389/froh.2026.1923290

Keywords: Arabic ECOHIS, Saudi Arabia, preschool children, oral health-related quality of life, factorial validity, factor analysis, psychometrics, caregiver reports, early childhood caries, scale validation, Factorial, validation