Abstract Problem: Need to improve the detection and management of depression in primary care.

Design: Prospective, before and after study of changes in detection and management following attempts to introduce a chronic disease management approach.

Background and setting: Two representative general practices in the north east of England that differed markedly in resources available and populations served.

Key measures of improvement: Number of cases on a depression register, number of cases accurately diagnosed, adherence to own clinical management guidelines.

Strategies for change: Multifaceted intervention to meet the needs of each practice modified by in-house steering group, including resources to develop a case register, an education and training programme on detection and management agreed by consensus, facilitation of meetings with secondary care staff, and support in developing a practice guideline.

Effects of change: Practice A (with six partners and serving a predominantly affluent white British population) improved case detection rate by 23%, reduced prescribing of sub-therapeutic doses of antidepressants by 36%, and adhered to the preferred treatment regimens. At Practice B (with three partners and two surgeries located in deprived urban inner city areas with high levels of unemployment and large ethnic minority populations) improvement in the sensitivity of case detection was accompanied by a reduction in specificity. The practice did not reach consensus on its own guideline and was unable to sustain the model.

Lessons learnt: A simple practice based approach improved the detection and management of depression in a team familiar with the philosophy of chronic disease management, with the capacity to commit to the programme, and with a critical mass of team members being open to change. This model failed to affect depression management when staff engagement with the project was passive rather than active and the practice was less well resourced and served an economically deprived and ethnically diverse population.