Data from the Option B+ prevention of mother-to-child transmission (PMTCT) program in Malawi show considerable variation between health facilities in retention on antiretroviral therapy (ART). In a programmatic setting, we studied whether the "model of care," based on the degree of integration of antenatal care (ANC), HIV testing and counselling (HTC) and ART service provision-influenced uptake of and retention on ART.We conducted a retrospective cohort study of pregnant women seeking ANC at rural primary health facilities in Zomba District, Malawi. Data were extracted from standardized national ANC registers, ART registers and ART master cards. The "model of care" of Option B+ service delivery was determined at each health facility, based on the degree of integration of ANC, HTC and ART. Full integration (Model 1) of HTC and ART initiation at ANC was compared with integration of HTC only into ANC services (Model 2) with subsequent referral to an existing ART clinic for treatment initiation.A total of 10,528 women were newly registered at ANC between October 2011 and March 2012 in 23 rural health facilities (12 were Model 1 and 11 Model 2). HIV status was ascertained in 8,572 (81%) women. Among 914/8,572 (9%) HIV-positive women enrolling at ANC, 101/914 (11%) were already on ART; of those not on treatment, 456/813 (56%) were started on ART. There was significantly higher ART uptake in Model 1 compared with Model 2 sites (63% vs. 51%; p=0.001), but significantly lower ART retention in Model 1 compared with Model 2 sites (79% vs. 87%; p=0.02). Multivariable analysis showed that initiation of ART on the same day as HIV diagnosis, but not model of care, was independently associated with reduced retention in the first six months (adjusted odds ratio 2.27; 95% CI: 1.34-3.85; p=0.002).HIV diagnosis and treatment on the same day was associated with reduced retention on ART, independent of the level of PMTCT service integration at ANC.