The good the bad and the ugly – what we really do when we identify the best and the worst organisations?

Performance and quality indicators are in widespread use in the health arena and beyond. They are used in such diverse applications as pay for performance for general practices, safety monitoring of hospitals and individual surgeons, and for examining geographic inequalities in public health issues. In the vast majority of cases these indicators are constructed by aggregating individual patient data and as such are based on finite samples. These finite samples result in chance influencing individual indicator values. Sometimes this chance is recognised and sometimes it is not. It is often assumed that using funnel plots, or calculating z-scores (with or without accounting for overdispersion) allows chance variation to be accounted for and are thus robust methods for identification of good and poor performance. I will examine three commonly used methods for identification of the best and worst performing organisations using simulation. These simulations show that, if a funnel plot is needed to represent the data, the data are not generally suitable for use as a quality indicator and will result in a high degree of misclassification whichever method is used.

If you would like to attend, please register at https://www.eventbrite.co.uk/e/what-we-really-do-when-we-identify-the-best-and-the-worst-organisations-tickets-42906460378