Using Latent Class Analysis to Explore Social Determinants of Health Among Pregnant Women Receiving Home Visiting: Implication for Program Retention and Precision in Model Delivery

Background and Purpose: Home visiting is an evidence-based strategy for delivering services to low-income families to improve a range of outcomes. The MomsFirst program is a home visiting intervention designed to reduce racial disparities in birth outcomes. In an era of increasing accountability, high need for service, and budgetary reductions, researchers and program staff are called upon to demonstrate what works, for whom, and under what circumstances. Known as precision home visiting, program evaluation activities explore research questions beyond population-level impact to tailor services to needs. Social determinants of health, or the conditions in which people are born, develop, and live, contribute to health inequalities in birth outcomes and should be considered when providing prenatal services to vulnerable women. The purpose of the study is to (1) identify the unobserved latent classes of social determinants of health among participants, and (2) examine associations between social determinants of health classes and program retention.

Methods: Using a sample of 2,053 pregnant women who received their first home visit between 2014 and 2016, we conducted a Latent Class Analysis (LCA) to identify the latent structure of social determinants of health for the sample. We used 11 observed variables reflecting key areas of social determinants of health developed by Healthy People 2020 including income, employment, level of education, housing instability, residential mobility, homelessness, access to transportation, and partner support. Then we conducted an OLS regression to examine relationships between social determinants of health classes and program retention, controlling for individual demographic characteristics.

Results: Based on the evaluation of fit indices and substantive criteria, the LCA identified four heterogeneous classes. The first class, representing 34% of the sample, was characterized by participants with low socioeconomic status as demonstrated in low levels of education, high unemployment and no income (low socioeconomic status group). The second class represented women with housing stressors as indicated in high proportions of residential instability, unstable housing, homelessness, utility interruptions and transportation barriers (14% of sample, housing instability group). The third class comprised high proportions of women with no financial or emotional partner support. This group, named low partner support, represented 26% of the sample. The fourth class appeared more stable in relation to the other three. Women in this class (socioeconomic stability group) reported fewer socioeconomic disadvantages, lower levels of residential instability, and higher rates of partner support (26% of the sample). OLS regression analyses showed that the low socioeconomic status (β=1.82 p <.04) and low partner support groups (β= 6.43, p <.001) were more likely to be retained longer in the program than the socioeconomic stability group, perhaps indicating key features of the MomsFirst program that could be used for targeted recruitment.

Discussion and Implications: This study contributes to a better understanding of program operations by examining subgroups of participants using their social determinants of health factors. These results have important implications for MomsFirst effectiveness and efficiency. Our findings highlight the importance of precision home visiting for developing tailored services to meet families’ differing needs in order to retain them.