SETTING: Lilongwe, the capital of Malawi, one of the countries in the world badly affected by the human immunodeficiency virus/acquired immune-deficiency syndrome (HIV/AIDS) epidemic. OBJECTIVE: In the face of a rising burden of tuberculosis (TB) fuelled by HIV, to evaluate the impact on the Lilongwe district tuberculosis programme performance of decentralisation of TB services, including extending the range of options for supervision of directly observed treatment (DOT) during the initial phase of treatment, and using a fully oral, intermittent regimen. DESIGN: Prospective assessment under programme conditions of 1) duration of hospital stay, 2) bed occupancy and 3) 8-month treatment outcomes in a cohort of patients registered before (1997) and after (1998) the introduction of decentralisation of TB services. RESULTS: The number of new patients (all forms) registered in Lilongwe district was 3144 in 1997 and 3761 in 1998. There were significant differences (P < 0.05) between all outcomes that were compared. In 1998, bed occupancy dropped by 38%; among smear-positive patients, the average length of hospital stay fell from 58 days in 1997 to 16, the cure rate was higher (64% vs. 56%), default rate was lower (5% vs. 19%), and treatment completion rate was lower (2% vs. 4%); among smear-negative patients, the treatment completion rate was higher (50% vs. 33%), default rate was lower (23% vs. 55%), and death rate was higher (17% vs. 4%). This death rate is attributable to improved follow-up and reporting of outcomes, rather than to increased deaths. CONCLUSION: Programme implementation of decentralised TB services in Lilongwe, including an extended range of supervision options for DOT and the use of an ambulatory treatment regimen, achieved reduced hospital stay and bed occupancy and good treatment outcomes.