Effective health care models to scale up combination antiretroviral therapy (ART) are needed in rural southwestern China. We aimed to evaluate the ART treatment outcomes and their associations with patients' demographic characteristics and pre-treatment clinical features in a scaled-up provincial ART program serving eight heavily HIV-affected prefectures in Yunnan Province. We abstracted information from a computerized database for adults initiating ART between July 2007 and September 2008. Survival functions of mortality and treatment failure were calculated by age group, gender, transmission mode, and baseline CD4 count. Multivariable Cox regression analyses were conducted to find independent associations of various demographic and baseline clinical features with outcome variables. Of the 1967 patients in the mortality analysis, there were 110 deaths, of which 16 were coded as accidents or suicides. Adjusted hazard ratios (AHR) associated with mortality were greater for patients with baseline CD4 counts <100 cells/µl vs. patients with CD4 counts ≥200 cells/µl, for male vs. female, for single vs. married, and for those acquired HIV through injection drug use (IDU) vs. other modes of transmission. Successful treatment was 81.3% at six months after treatment started. Immunologic treatment failure was associated with baseline CD4 counts but not with demographic characteristics. Overall loss to follow-up rate was 2.1%. Collaboration between clinics and community networks are distinctive features of Yunnan's model for scaling up ART across a diverse, poor, and rural population. This study finds that the strategy can succeed even if 40% of the patients have a history of IDU.