BackHome
Refresh: OffView Personal Schedule
Individual Submission Summary
Share...

Direct link:

Measuring Parenting in Sub-Saharan Africa: HOME Inventory dimensionality assessment, IRT calibration, and scale validation

Wed, April 7, 1:10 to 2:40pm EDT (1:10 to 2:40pm EDT), Virtual

Abstract

Introduction. Several tools have been developed to try to capture the quality of a child’s home environments, being the most popular the Home Observation for the Measurement of the Environment (HOME) (Elardo, Bradley, and Caldwell, 1975). Multiple versions have been developed over time in order to target different age groups, physical settings, to address changes cultural norms for childcare, and to surpass cross-cultural and socio-economic barriers in low- and middle-income countries (LMIC).
Problem. The HOME has been used in observational and experimental studies of early child development as predictor, outcome, and a mediator between children’s developmental outcomes and context characteristics. However, with the exception of Jones et. al. (2017), who evaluated measurement invariance, little to no attention has been paid to the psychometric properties scale when used in LMIC. The current paper evaluates psychometric properties and validity of an adapted version of HOME (Singla et al., 2015) used in Rwanda.
Study population. The data were collected as part of a Cluster Randomized Trial of the Sugira Muryango family home-visiting program developed for use among the most vulnerable families in rural Rwanda (redacted, 2018, 2020), and the analysis sample is comprised of 1074 children aged 6–36 months living in low-income households.
Methods. Statistical analysis comprised reliability and dimensionality assessment, item calibration, and validity testing. Reliability considered the inspection of point-biserial correlations, and the KR-20 statistic. Dimensionality was evaluated using factor analytic techniques (Parallel Analysis and EFA). Calibration was carried using A 2-PL Item Response Theory (IRT) and Differential Item Functioning was tested using Logistic Regression. Convergent validity was evaluated using Pearson correlations between HOME IRT scores and the OMCI and known-group validity was evaluated using T-test comparing HOME Scores between developmentally on-track and delayed children according to ASQ cut-off.
Results. Item-level analyses flagged 7 items with low or negative point-biserial correlations. After removing them, dimensionality assessment yielded two factors. Upon inspection of the items, this differentiation seemed to reflect assessment modes (Mother-Reported vs. Observation-only factors), more than different conceptual dimensions. Two items were removed due to low factor loadings. The Unidimensional 2-PL IRT model yielded appropriate item discrimination and difficulty estimates, and a good Total Information Curve. The model also yielded very good model- and item-level fit statistics with neither significant Chi-2 values, nor large INFIT and OUTFIT statistics, and DIF analysis didn’t flag any item according to traditional Effect Sizes Lastly, IRT scores showed significant correlations (p<0.001) with the OMCI, and significant differences were found when comparing HOME scores between developmentally on-track vs delayed children according to ASQ traditional cut-off scores.
Conclusion: Overall, this reduced version was found to be valid for assessment of home environments in rural sub-Saharan settings. After removing problematic items, results point toward the validation of a reduced 31-items version of the HOME inventory. There was little support for the original subdimensions structure. The reduced version covers all the original domains while keeping in the scale the most informative items both from a statistical and theoretical perspective. Further research would help to evaluate cross-cultural invariance and to delve deeper into cultural fit.

Authors