Comment on 'empirical assessment of methods for risk identification in healthcare data: results from the experiments of the Observational Medical Outcomes Partnership'