OBJECTIVES: Self-reported adherence assessment in HIV-infected patients on antiretroviral therapy (ART) is challenging and may overestimate adherence. The aim of this study was to improve the ability of health care providers to elicit patients' reports of nonadherence using a "patient-centred" approach in a rural sub-Saharan African setting. METHODS: A prospective interventional cohort study of HIV-infected patients on ART for >/= 6 months attending an HIV clinic in rural Tanzania was carried out. The intervention consisted of a 2-day workshop for health care providers on patient-centred communication and the provision of an adherence assessment checklist for use in the consultations. Patients' self-reports of nonadherence (>/= 1 missed ART dose/4 weeks), subtherapeutic plasma ART concentrations (< 2.5th percentile of published population-based pharmacokinetic models), and virological and immunological failure according to the World Health Organization definition were assessed before and after (1-3 and 6-9 months after) the intervention. RESULTS: Before the intervention, only 3.3% of 299 patients included in the study reported nonadherence. Subtherapeutic plasma ART drug concentrations and virological and immunological failure were recorded in 6.5%, 7.7% and 14.5% of the patients, respectively. Two months after the intervention, health care providers detected significantly more patients reporting nonadherence compared with baseline (10.7 vs. 3.3%, respectively; P < 0.001), decreasing to 5.7% after 6-9 months. A time trend towards higher drug concentrations was observed for efavirenz but not for other drugs. The virological failure rate remained unchanged whereas the immunological failure rate decreased from 14.4 to 8.7% at the last visit (P = 0.002). CONCLUSIONS: Patient-centred communication can successfully be implemented with a simple intervention in rural Africa. It increases the likelihood of HIV-infected patients reporting problems with adherence to ART; however, sustainability remains a challenge.