SETTING: Primary health centres in urban Lusaka, Zambia. OBJECTIVES: To describe the distribution and risk factors for delay among patients presenting with a cough to the urban health centres. DESIGN: A health systems research methodology was used. A participatory workshop analysed the problem and designed a cross-sectional survey of patients attending two urban health centres. Initial data analyses were performed in a second workshop, with results discussed with a broad range of policy-makers, health care staff and community members interested in tuberculosis. RESULTS: A total of 427 patients were interviewed; 35% had delayed for more than one month. Delay was associated with older age, severe underlying illness, poor perception of the health services, distance from the clinic and prior attendance at a private clinic. There was no relationship between delay and knowledge about tuberculosis, nor with education, socio-economic level or gender. Tuberculosis and HIV were felt to be closely linked and highly stigmatised, but stigmatising attitudes were not associated with longer delays. CONCLUSIONS: The health systems research methodology was an effective way to engage the staff of the district health services in action-oriented research. Investing in improvements in the health system and ensuring accessibility for older and more disabled patients is likely to reduce delays in diagnosis and help to improve tuberculosis control in Lusaka.