Fig. 1

Association between maternal HIV status and infant feeding practices. Graphs depict outputs from unadjusted and adjusted multinomial regression models comparing (i) not breastfeeding vs exclusive breastfeeding, and (ii) mixed feeding vs exclusive breastfeeding. The unadjusted (basic) model contains a priori covariates for maternal age, education and knowledge. The adjusted model contains additional covariates for parity, employment, relationship status, household assets, CQI exposure, and time. Both unadjusted and adjusted models account for clinic-level clustering of outcomes. CQI, continuous quality improvement; WLHIV, women living with HIV; WNLHIV, women not living with HIV