Background: There is little agreement on the risks that concerns about the performance of general practitioners present to patients in primary care or how they should be investigated. This may result in a lack of consistent decisions about the management of serious concerns, and has implications for patient safety. Aim: To develop a set of criteria for assessment of risk to patients and for investigating performance concerns in general practice. Design:: Two-round Delphi questionnaire. Methods: Panellists were medical and non-medical people with extensive experience of assessing, investigating and managing performance concerns in primary care. Panellists were presented with scenarios about performance concerns, together with one of five possible investigation options:a medical record review, prescribing system review, practice-management assessment, GP suspension hearing or a death review. They then considered 95 scenarios, rating 69 according to risk and all 95 according to investigation options. In the second round, ratings were repeated after panellists had reviewed their own and group first-round responses. Consensus was defined in advance as 80% of responses in the upper third on a nine-point rating scale. Results: Consensus on high risk was achieved for 36 of the 69 (52%) risk scenarios. Consensus on the proposed investigation was achieved in 33 of the 95 (35%) investigation scenarios. Conclusions: This is the first study to report the development of consensus on the nature of performance concerns that pose a high risk to patients, and about their appropriate investigation in primary care. We have identified a series of high-risk performance concerns and linked these to appropriate methods of investigation. The management of performance concerns should be guided by explicit consensus criteria to improve the quality of decision making in managing poor performance in primary care. Patient safety may be compromised by inconsistent management of performance concerns.