Some of the most frail elderly patients are suffering "distressing" deaths because hospitals wrongly try to resuscitate them, a watchdog says.
The National Confidential Enquiry into Patient Outcome and Death reviewed the care given to 585 acutely-ill patients who ended up having a cardiac arrest.
The watchdog concluded that cardiopulmonary resuscitation (CPR) had wrongly become the default setting.
And it said a third of the cardiac arrests could have been prevented.
The report concluded assessing if resuscitation was necessary should become standard.
The review, which looked at patients with an average age of 77, also looked at the standards of care given to these patients.
It found that staff were not properly assessing their condition and were failing to spot the warning signs of an impending cardiac arrest.
Details of whether or not to give CPR was recorded in the notes of only 122 patients in the study of hospitals in England, Wales and Northern Ireland.
Of these, there were 52 cases where doctors had performed resuscitation on patients who had explicitly said they did not want it.
The experts said performing CPR in inappropriate cases could result in a distressing and undignified death.
They gave the example of an elderly patient with severe dementia who had CPR performed on them for 10 minutes until a senior doctor stopped the team. The report said resuscitation should not have taken place.
NCEPOD chairman Bertie Leigh said there needed to be a rethink by the NHS about what was possible for these patients.
"We are at a crossroads. All of us need to recognise and accept the limits of what can be achieved in medicine to the benefit of the patient."
Katherine Murphy, of the Patients Association, said there was a "huge degree of confusion" about the issue.
"Patients and relatives deserve to have all of their options communicated to them in full and then to take the decision that they feel is best for them.
"Once that decision has been taken, they should be able to trust clinicians to implement it."
Dr Mark Temple, an acute medicine fellow at the Royal College of Physicians, said the report did not make it clear why the patients had resuscitation, and whether this was because of poor documentation or staff handover.
He questioned whether it was an issue about having a correct "not for resuscitation" order made initially.
In an emergency, when a patient's status may change, a "snap decision" could be made by staff to proceed with CPR, he said.