Cardiopulmonary resuscitation and mechanical ventilation have enabled us to prolong the life of patients, although this may not always be appropriate. To avoid patients experiencing a poor functional outcome or a long process of dying, Do-Not-Resuscitate (DNR) orders have been instituted. However, several studies suggest that DNR orders independently increase mortality. DNR orders are based on the patient's own preferences or because resuscitation is judged to be futile by the medical team. This latter assumes that the prognosis of the patient concerned is known. However, an accurate prognosis is extremely difficult, and prognostic models only show good agreement for patients in aggregate, not for individuals. These difficulties in prognostication could partly explain a potential increase in mortality due to DNR orders. Another cause may be unintentional but unjustified broadening of DNR orders by health care workers in patients with DNR orders, with the withholding of other treatments besides CPR and suboptimal care. Physicians should be aware of the unintentional impact of DNR orders. We recommend several precautions that can be taken to reduce the risk of increasing mortality due to DNR orders. Hospitals should develop clear guidelines, which describe indications for DNR orders and which emphasize that these orders have to be viewed in isolation from other treatment decisions. DNR orders should be made, if possible, at an early time, by combining prognostic models with the physician's own estimate of prognosis, jointly with other colleagues and in concordance with the patient's preferences. Because of the large differences between physicians and hospitals regarding the institution of DNR orders, together with the uncertainty of prognosis, and the self-fulfilling prophecy of treatment limitations, we need to be very cautious in recommending limitations of care. Copyright © 2012, Nederlandse Vereniging voor Intensive Care. All Rights Reserved.