Monitoring the performance of general practices

Abstract

Performance indicators for general practice which reduce complex processes to simple counts can have little validity. Additionally, performance indicators are often statistically unreliable in small populations like general practices. Instead, it is possible to combine these measures of performance by using multiple regression to predict the outcome from a set of processes. This allows one to adjust the outcome for differences in the practice populations. It also improves the statistical reliability, because data from all practices are used to predict the outcome. This approach has statistical problems, because it is an ecological analysis, and does not pick out the poor performers ('bad apples'). The regression approach is similar to the concepts of continuous quality improvement (CQI). It is arguable that using CQI to improve quality is more likely to lead to cooperation from general practices than trying to pick out the poor performers.