Around one in 10 patients admitted to hospital - nearly one million people a year - will suffer some accidental harm, from a minor fall to serious injury and death, it was acknowledged by government yesterday.

The admission came from the chief medical officer, Liam Donaldson, at a conference to reveal the results of a pilot study of an anonymous reporting system for doctors and other medical staff.

The national patient safety agency will be rolling out the scheme to all hospitals, GP practices and other NHS trusts within the year.

In the nine months since the pilot began in 28 trusts there were 27,110 incidents reported. A total of 135 patients died -around 80 from natural causes and 35 from suicides, some of which could probably have been prevented. The remaining 20 were victims of a variety of medical blunders in gynaecology, surgery and other disciplines. Some 2% of the incidents (542) were considered catastrophic or very serious.

There has been confusion about the figures which the Department of Health has attempted to say are not reliable. They suggest that if the scheme had been operating in all 700 trusts in England, there would have been around 680,000 incidents and 3,500 deaths, but that underplays the scale of the problem.

Most of the adverse events reported were in hospitals, which have a long track record of collecting data when something goes wrong. Asked how great the under-reporting of the GP practices was, Susan Williams, the agency's joint chief executive, said: "I think it's absolutely massive. I think they are not used to the culture of reporting {what goes wrong}."

Professor Donaldson acknowledged that only the hospital figures, showing that 10% of patients suffer an adverse event, could be taken at face value. "The acute trusts are closer to the ballpark figure of 10%," he said. "The primary care trusts are not used to this sort of reporting yet."

If all GP, mental health and ambulance figures were all put into the equation, the figure came down to less than 1%.

The pilot study hit big problems over the inability of various trust computers to talk to the national patient safety agency. The severity of almost 60% of the reported incidents could not be classified, because the data was missing. But the pilot gave the agency valuable information that will prevent future disasters.

Ms Williams told the conference the agency was issuing its first alert, over the storage of potassium chloride on general wards. They had received 31 reports of accidental harm through the use of the intravenously administered drug, in which three people died because they had been given it in neat form.

Potassium chloride is essential for some patients in maintaining a regular heartbeat but it can be deadly - it is used in lethal injections in the US. The agency said it should be kept in dilution on general wards.

There were 500 incidents involving childbirth. Ms Williams said the main reasons for problems were delays in obstetric staff arriving, the lack of available of senior staff and the results of tests not being available.

"We want the bad experience of a patient in Truro to be the saviour of the life of a patient in Darlington the next week, the next month, the next year," Prof Donaldson said.

The NHS would only learn from mistakes if staff felt confident that they could report them anonymously, she said. The agency would not use any information that came its way to pursue disciplinary action against any doctor. It was up to hospitals to investigate deaths and serious incidents.

Trainee doctors are being poorly prepared to diagnose and treat critically ill patients, according to a study which found worrying gaps in their knowledge of basic acute care, writes James Meikle .

The study's authors recommended that medical schools urgently overhauled their training and filled a gap that might be putting lives at risk and contributing towards high levels of hospital errors.

Their survey of 185 junior doctors at six hospitals, published in the Postgraduate Medical Journal today, concentrated on their ability to spot potential abnormalities that might indicate a deterioration in patients' conditions, critically important in a hospital system where such trainees were expected to make most of initial assessments and treatment decisions on acute wards.

Few identified all the signs of airway obstruction, and many could not describe fully how to use an oxygen mask.