Social Sciences and Humanities Open (Jan 2024)
Using I-MAIHDA to extend understanding of engagement in early years interventions: an example using the Born in Bradford's Better Start (BiBBS) birth cohort data
Abstract
Background: Intervention in the early years is essential for reducing health and social inequalities across the lifespan. The success of pregnancy and early years programmes depends on engagement from target parents and families, yet there is no consensus on which factors predict engagement. This could be due to differences between interventions at the local level, or due to unexplored interactions between factors in their relationships with engagement. This study highlights the value of the intersectional application of the multilevel analysis of individual heterogeneity and discriminatory accuracy (I-MAIHDA) approach for adding nuance to our understanding of inequalities in engagement in interventions. Method: A context-driven, programme-specific analysis exploring factors relating to participation across the Better Start Bradford early years interventions acted as an applied example of the utility of I-MAIHDA. Two analyses were performed using data from the Born in Bradford's Better Start (BiBBS) birth cohort dataset to explore the effects of combinations of ethnicity, migrant status, spoken English ability and social support on participation in the interventions. Predicted prevalence of participation was obtained from the models. Results: Combinations of English language ability, migrant status, social support, and ethnicity were found to be related to differential prevalence of participation in the interventions, with inequalities in participation between strata. Discriminatory accuracy of the models was low but not negligible (∼5%), suggesting some of the variation in outcomes was due to the combined effect at the strata level. Conclusions: The I-MAIHDA approach showed promise for extending knowledge of engagement in interventions through context-driven analyses which incorporates complex relationships between multiple covariates. This approach will be of interest to anyone working to increase participation in interventions, especially in under-represented or marginalised groups.
