IADR Abstract Archives
Measurement Invariance of the SOHO-5: School-Based and Clinical-Based Comparisons
Objectives: The present study aimed to evaluate the structural validity of the child’s self-reported version of the Scale of Oral Health Outcomes for 5-Years-Old Children (SOHO-5) and test its multi-group invariance.
Methods: Three datasets comprised the study, two from Brazil and one from the United Kingdom (UK). One of the Brazilian datasets was derived from clinical data collection (nbr-cli. = 193), while the others were from school-based epidemiological studies (nbr-sch. = 768, nuk-sch. = 296). The dimensionality was explored through parallel analysis and corroborated by unidimensional indexes. Internal consistency was assessed using Cronbach's alpha and McDonald’s omega. Prior to the measurement invariance, the model fit was tested through Confirmatory Factor Analysis (CFA) using the following thresholds to adjudge an acceptable for each dataset: χ2/df < 2.0, Comparative Fit Index (CFI) > .95, Root Mean Square Error of Approximation (RMSEA) < .06, and Standardised Root Mean Square Residual (SRMR) < .08. Measurement invariance across datasets was tested via multi-group CFA, contrasting the unidimensional first-order constructs in three steps, configural, metric and scalar invariance. Scalar invariance is supported when changes were within -0.010 units for the ΔCFI, 0.015 units for the ΔRMSEA, and 0.010 units for the ΔSRMR.
Results: The unidimensionality was empirically confirmed for the three datasets. The CFA models showed acceptable model fit indices and factor loadings between 0.48 and 0.894. Alpha and omega coefficients higher than 0.80 suggested adequate internal consistencies. The multigroup CFA tests reached the scalar invariance threshold between the Brazilian and UK school-based datasets (ΔCFI = -0.002, ΔRMSEA = 0.000, ΔSRMR = 0.001). However, there was no equivalence when comparing school-based with clinical datasets.
Conclusions: We conclude that child’s self-reported version of the SOHO-5 is psychometrically sound, and caution is needed when embarking on Oral health-related quality-of-life comparisons between clinic-based and school-based populations.
Methods: Three datasets comprised the study, two from Brazil and one from the United Kingdom (UK). One of the Brazilian datasets was derived from clinical data collection (nbr-cli. = 193), while the others were from school-based epidemiological studies (nbr-sch. = 768, nuk-sch. = 296). The dimensionality was explored through parallel analysis and corroborated by unidimensional indexes. Internal consistency was assessed using Cronbach's alpha and McDonald’s omega. Prior to the measurement invariance, the model fit was tested through Confirmatory Factor Analysis (CFA) using the following thresholds to adjudge an acceptable for each dataset: χ2/df < 2.0, Comparative Fit Index (CFI) > .95, Root Mean Square Error of Approximation (RMSEA) < .06, and Standardised Root Mean Square Residual (SRMR) < .08. Measurement invariance across datasets was tested via multi-group CFA, contrasting the unidimensional first-order constructs in three steps, configural, metric and scalar invariance. Scalar invariance is supported when changes were within -0.010 units for the ΔCFI, 0.015 units for the ΔRMSEA, and 0.010 units for the ΔSRMR.
Results: The unidimensionality was empirically confirmed for the three datasets. The CFA models showed acceptable model fit indices and factor loadings between 0.48 and 0.894. Alpha and omega coefficients higher than 0.80 suggested adequate internal consistencies. The multigroup CFA tests reached the scalar invariance threshold between the Brazilian and UK school-based datasets (ΔCFI = -0.002, ΔRMSEA = 0.000, ΔSRMR = 0.001). However, there was no equivalence when comparing school-based with clinical datasets.
Conclusions: We conclude that child’s self-reported version of the SOHO-5 is psychometrically sound, and caution is needed when embarking on Oral health-related quality-of-life comparisons between clinic-based and school-based populations.