Using path analysis to model the process of change in HbA1c among African Americans and Latinos in a community health worker diabetes intervention.

The study did not randomly assign people to group or one-on-one classes, so people who could attend group classes may have differed from those who needed one-on-one support. The intervention included several components together, including classes, home visits and accompanied clinic visits, and the home visit component was excluded from the model, so the contribution of any single component cannot be separated. Cohort 1 used 12-month knowledge and self-efficacy scores as a proxy for 6-month scores because those measures were not collected at 6 months. The analysis combined two cohorts from different time periods, one nonrandomized and one randomized with a delayed intervention group, which may introduce historical differences. Many outcomes were self-reported, and the self-management score combined questions that did not move together well, with medication adherence already near the top of the scale and unchanged. Participants were mostly women and were urban African American or Latina/o adults with type 2 diabetes, so results may not apply to rural settings, men, or other groups.

The authors declared no competing interests. The article is listed as supported by NIH and other government research funding.

Do not read the group-class findings as proof that group classes caused lower HbA1c. Participants were not randomly assigned to group or one-on-one format, and the intervention also included home visits and accompanied clinic visits, so the contribution of any single component cannot be separated.