Clinical prediction models to predict the risk of multiple binary outcomes: a comparison of approaches