INS NYC 2024 Program

Paper	Paper Session 18 Program Schedule 02/17/2024 09:00 am - 10:30 am Room: West Side Ballroom - Salon 3 Paper Session 18: Neuromedical Disorders Final Abstract #6 Testing Measurement Invariance Across HIV Groups in a Low Resource Setting Using the Computerized Battery for Neuropsychological Evaluation of Children (BENCI) Rachel Maina, Tilburg University, Tilburg, Netherlands Jia He, Tilburg University, Tilburg, Netherlands Amina Abubakar, Aga Khan University, Nairobi, Kenya Miguel Perez-Garcia, University of Granada, Granada, Spain Karen Blackmon, Aga Khan University, Nairobi, Kenya Manasi Kumar, University of Nairobi, Nairobi, Kenya Jelte Wicherts, Tilburg University, Tilburg, Netherlands Category: Cross Cultural Neuropsychology/ Clinical Cultural Neuroscience Keyword 1: psychometrics Keyword 2: cross-cultural issues Keyword 3: computerized neuropsychological testing Objective: Developing culture-fair tests that measure constructs equivalently across different ethno-lingual groups is challenging, given the diverse cultural variations that impact neurocognitive measurement. Multi-level measurement invariance must be established before interpreting scores similarly across groups, both within and between cultures for meaningful comparisons. There is a dearth of studies evaluating measurement invariance of neurocognitive measures in Sub-Saharan Africa. Here, we considered whether The Computerized Battery for Neuropsychological Evaluation of Children (BENCI) exhibits measurement invariance between children living with and without HIV in Kenya. Participants and Methods: We evaluated performance on the BENCI in a sample of children living with (N = 274) and without HIV (N =330). Mean age was 9.48 years (SD = 1.31 years). Measurement invariance was evaluated with multi-group confirmatory factor analysis. We used a four-factor model that had an excellent fit in a Kenyan pooled sample when modified to have four correlated first-order factors (Fluency, Reasoning, Memory, and Flexibility). We first established configural invariance where all factor loadings, item intercepts and residual parameters were freely estimated across the two groups. We then specified a model for metric invariance where factor loadings were restrained and other parameters were freely estimated. We then specified a scalar invariance model where item intercepts and factor loadings were restrained and latent means were freely estimated to check latent mean differences in flexibility, fluency, verbal memory and reasoning across the two groups. We then specified a partial scalar invariance model using modification indices where we constrained one intercept for each indicator at a time and tested whether this restraint resulted in a significant chi-square difference. We compared each model to metric invariance model and checked for significant chi-square differences. Results: The four-factor model showed good model fit (RMSEA = 0.046, CFI = 0.960, TLI = 0.933), with metric (RMSEA = 0.040, CFI = 0.930, TLI = 0.893; Δχ2 = 23.26, DF = 8, p = 0.003) and partial scalar measurement invariance (RMSEA = .041, CFI = .920, TLI = .884; Δ χ²= 20.03, DF = 6, p >.001, ΔCFI 0.010) between the HIV positive and negative groups. In the metric invariance model, the indicator items had similar associations with the latent constructs across groups. In the scalar invariance model, two subtests resulted in a poorer fit compared to the metric invariance model (RMSEA = .045, CFI = .901, TLI = .860; Δχ² =46.77, DF= 8, p <.001, ΔCFI = 0.029). The Verbal Comprehension Figures and Visual Memory Delayed subtests did not fit the model but when freely estimated in the partial scalar model, they did show a good fit. Conclusions: Overall, results suggest that the four constructs of Fluency, Reasoning, Memory, and Flexibility from the BENCI can be compared across children with and without HIV in Kenya, as can mean differences between most of the subtests, except Verbal Comprehension Figures and Visual Memory Delayed. Multi-level measurement invariance should be customarily evaluated prior to interpretation of group differences in neuropsychological performance, especially in cultural settings that are different from those in which measures were originally developed and validated. Establishing measurement invariance is also critical for cross-cultural data harmonization in multi-national research programs.

Paper

02/17/2024
09:00 am - 10:30 am
Room: West Side Ballroom - Salon 3

Paper Session 18: Neuromedical Disorders

Final Abstract #6

Testing Measurement Invariance Across HIV Groups in a Low Resource Setting Using the Computerized Battery for Neuropsychological Evaluation of Children (BENCI)

Rachel Maina, Tilburg University, Tilburg, Netherlands
Jia He, Tilburg University, Tilburg, Netherlands
Amina Abubakar, Aga Khan University, Nairobi, Kenya
Miguel Perez-Garcia, University of Granada, Granada, Spain
Karen Blackmon, Aga Khan University, Nairobi, Kenya
Manasi Kumar, University of Nairobi, Nairobi, Kenya
Jelte Wicherts, Tilburg University, Tilburg, Netherlands

Category: Cross Cultural Neuropsychology/ Clinical Cultural Neuroscience

Keyword 1: psychometrics
Keyword 2: cross-cultural issues
Keyword 3: computerized neuropsychological testing

Objective:

Developing culture-fair tests that measure constructs equivalently across different ethno-lingual groups is challenging, given the diverse cultural variations that impact neurocognitive measurement. Multi-level measurement invariance must be established before interpreting scores similarly across groups, both within and between cultures for meaningful comparisons. There is a dearth of studies evaluating measurement invariance of neurocognitive measures in Sub-Saharan Africa. Here, we considered whether The Computerized Battery for Neuropsychological Evaluation of Children (BENCI) exhibits measurement invariance between children living with and without HIV in Kenya.

Participants and Methods:

We evaluated performance on the BENCI in a sample of children living with (N = 274) and without HIV (N =330). Mean age was 9.48 years (SD = 1.31 years). Measurement invariance was evaluated with multi-group confirmatory factor analysis. We used a four-factor model that had an excellent fit in a Kenyan pooled sample when modified to have four correlated first-order factors (Fluency, Reasoning, Memory, and Flexibility). We first established configural invariance where all factor loadings, item intercepts and residual parameters were freely estimated across the two groups. We then specified a model for metric invariance where factor loadings were restrained and other parameters were freely estimated. We then specified a scalar invariance model where item intercepts and factor loadings were restrained and latent means were freely estimated to check latent mean differences in flexibility, fluency, verbal memory and reasoning across the two groups. We then specified a partial scalar invariance model using modification indices where we constrained one intercept for each indicator at a time and tested whether this restraint resulted in a significant chi-square difference. We compared each model to metric invariance model and checked for significant chi-square differences.

Results:

The four-factor model showed good model fit (RMSEA = 0.046, CFI = 0.960, TLI = 0.933), with metric (RMSEA = 0.040, CFI = 0.930, TLI = 0.893; Δχ2 = 23.26, DF = 8, p = 0.003) and partial scalar measurement invariance (RMSEA = .041, CFI = .920, TLI = .884; Δ χ²= 20.03, DF = 6, p >.001, ΔCFI 0.010) between the HIV positive and negative groups. In the metric invariance model, the indicator items had similar associations with the latent constructs across groups. In the scalar invariance model, two subtests resulted in a poorer fit compared to the metric invariance model (RMSEA = .045, CFI = .901, TLI = .860; Δχ² =46.77, DF= 8, p <.001, ΔCFI = 0.029). The Verbal Comprehension Figures and Visual Memory Delayed subtests did not fit the model but when freely estimated in the partial scalar model, they did show a good fit.

Conclusions:

Overall, results suggest that the four constructs of Fluency, Reasoning, Memory, and Flexibility from the BENCI can be compared across children with and without HIV in Kenya, as can mean differences between most of the subtests, except Verbal Comprehension Figures and Visual Memory Delayed. Multi-level measurement invariance should be customarily evaluated prior to interpretation of group differences in neuropsychological performance, especially in cultural settings that are different from those in which measures were originally developed and validated. Establishing measurement invariance is also critical for cross-cultural data harmonization in multi-national research programs.