Q-squared as a mapping model performance metric: comparison of models mapping to EQ-5D in gastrointestinal disease

Abstract Background: Model performance in mapping studies between quality of life measures is often assessed using metrics such as root mean square error (RMSE), mean squared error (MSE), and R-squared. The Q-squared (Q²) statistic has recently been used as a measure of prognostic accuracy, but has not previously been applied to assess model performance in quality of life mapping studies. We aimed to estimate Q-squared for published mapping algorithms in gastrointestinal (GI) disease and evaluate its usefulness as a metric of prediction accuracy and model performance.