OBJECTIVES High early mortality after antiretroviral therapy (ART) initiation in resource-limited settings is associated with low baseline CD4 cell counts and a high burden of opportunistic infections. Our large urban HIV clinic in Uganda has made concerted efforts to initiate ART at higher CD4 cell counts and to improve diagnosis and care of patients coinfected with tuberculosis (TB). We sought to determine associated treatment outcomes. METHODS Routinely collected data for all patients who initiated ART from 2005 to 2009 were analysed. Median baseline CD4 cell counts by year of ART initiation were compared using the Cuzick test for trend. Mortality and TB incidence rates in the first year of ART were computed. Hazard ratios (HRs) were calculated using multivariable Cox proportional hazards models. RESULTS First-line ART was initiated in 7659 patients; 64% were women, and the mean age was 37 years (standard deviation 9 years). Median baseline CD4 counts increased from 2005 to 2009 [82 cells/μL (interquartile range (IQR) 24, 153) to 148 cells/μL (IQR 61, 197), respectively; Ptextless0.001]. The mortality rate fell from 6.5⁄100 person-years at risk (PYAR) [95% confidence interval (CI) 5.5-7.6 PYAR] to 3.6⁄100 PYAR (95% CI 2.2-5.8 PYAR). TB incidence rates increased from 8.2⁄100 PYAR (95% CI 7.1-9.5 PYAR) to 15.6⁄100 PYAR (95% CI 12.4-19.7 PYAR). A later year of ART initiation was independently associated with decreased mortality (HR 0.91; 95% CI 0.83-1.00; P=0.04). CONCLUSIONS Baseline CD4 cell counts have increased over time and are associated with decreased mortality. Additional reductions in mortality might be a result of a better standard of care and increased TB case finding. Further efforts to initiate ART earlier should be prioritized even in a setting of capped or reduced funding for ART programmes.