In this paper we study demand effects of user charges in a district health care system using cross-sectional data from household and facility surveys. The effects are examined in public as well as in private health facilities. We also look briefly at the impact of fees on revenue and service quality in government facilities. During the period of cost-sharing in public clinics, attendance dropped by about 50%. This drop prompted the government to suspend the fees for approximately 20 months. Over the 7 months after suspension of fees, attendance at government health centres increased by 41%. The suspension further caused a notable movement of patients from the private sector to government health facilities. The revenue generated by user fees covered 2.4% of the recurrent health budget. Some 40% of the facilities did not spend the fee revenue they collected, mainly due to cumbersome procedures of expenditure approvals. The paper concludes with lessons from Kenya's experience with user charges.