We thank Dr Charles for his important comments.1

He raises the important question of whether CRB-65 is a useful tool to advise treatment limitations. If only 29% of those who finally died were at high risk of death at initial presentation (CRB-65 risk class 3), such a tool may be of limited value in this regard. In fact, we agree that the CRB-65 score (like any other such as the PSI) is not helpful for the decision to apply treatment restrictions. Such restrictions up to fully palliative treatment cannot be based primarily on considerations about the current risk of death but should be the result of a careful evaluation of the clinical state and overall prognosis of the patient, both initially and during follow-up, and such decisions should be decided with the patient or his legal social worker.

In this context, the CRB-65 severity score remains important as part of the initial clinical evaluation of all patients. Treatment restrictions must not follow a hidden agenda but must be openly discussed, communicated and documented. A predicted moderate to high risk of death from community-acquired pneumonia is a highly relevant piece of information required to mount an ethically valid treatment recommendation and decision, particularly in those patients with pneumonia regarded to be a terminal event. Nevertheless, we recalculated the predictions of the CRB-65 score excluding all those who died without having received any ventilator support during hospitalisation. The results are: overall death rate 8618, 2.5%, CRB-65 risk class 1: 0.5%, risk class 2: 1.7% and risk class 3: 12.2%. These numbers support the following conclusions: (1) the CRB-65 score remains useful in predicting deaths in a three class pattern; (2) obviously, virtually no previous study on community-acquired pneumonia truly excluded all patients with treatment limitations.