With dramatic increases in antiretroviral therapy (ART) provision, many clinics in sub-Saharan Africa are congested, but little attention has focused on the efficiency of clinics. Between April and June 2008, we conducted a time-and-motion study to assess patient flow at three HIV clinics in Uganda. Mulago HIV Clinic had 6,700 active patients, compared with 2,700 at Mbarara Municipal Council Clinic (MMC) and 2,800 at Reachout Mbuya (ROM). Mulago had six doctors and eight nurses; MMC had two doctors and two nurses, and ROM had two doctors and 12 nurses. Mulago and MMC used a doctor-led model, whereas ROM used a nurse-led model. Randomly selected patients were tracked, with data collected on time waiting and time spent with providers. Patients were categorized as new, preparing for ART, early ART, stable ART, or non-ART. Doctors indicated whether the patients they saw warranted their consultation. Data were collected on 689 patients (230 at Mulago, 229 at MMC, and 230 at ROM). Overall waiting time was longest at ROM (274 min; 209-346) and Mulago ISS (270 min; 230-336) compared with MMC (183 min; 148-233). Nurse-clinicians at ROM spent twice the time with patients compared with the doctors at Mulago. At Mulago, doctors indicated that 27% of the patients they reviewed did not need to see a doctor, compared with 45% at MMC. Task-shifting may not be efficient in terms of time. More-effective triage and longer visit intervals could improve patient flow and capacity for cost-effective scale-up.