Knee joint replacement is an effective and cost-effective intervention for severe symptomatic osteoarthritis of the knee joint. However, utilisation rates vary hugely, there are no indications, it is difficult to know when (in the course of arthritis) it is best to operate, and some 10–20% of people who have this surgery are unhappy with the outcome, and have persistent pain. In this article we briefly discuss the variations in utilization of knee joint replacement, and then outline four different approaches to the selection and prioritisation of patients for this procedure. Consensus criteria, including appropriateness criteria are available, but if produced by professionals alone, they may conflict with the views of patients and the public. Databases and cohort studies can be used to attempt relating outcomes to baseline characteristics, but at present we can only account for a small percentage of the variance with this technique. Finally, we propose use of the ‘capacity to benefit framework’ to attempt providing guidance to both patients and healthcare professionals.