The government believes that a new study which attempts to calculate the number of accidents and errors in English hospitals each year might understate the problem.

Ministers and health officials say the figures from a trial collating information in 28 hospital trusts "are unreliable and not sufficiently robust". The evidence would put the level of harm done to patients lower than that suggested by previous research which has indicated that one in 10 patients - about 1 million a year - are harmed by so-called adverse incidents.

One study has indicated that a third of such incidents, often caused by wrong or late diagnosis, treatment errors or failures in medical devices, might lead to disability or death.

Now a pilot reporting scheme for the new national patient safety agency, set up by the Department of Health, has recorded more than 20,000 adverse incidents in six months in 28 NHS trusts. Roughly extrapolating that to all 355 trusts in England would take the overall figure over 500,000, half some earlier estimates. That has led the government to question the new figures.

The chief medical officer for England, Professor Sir Liam Donaldson, has been trying to introduce a "no blame" culture in which more mistakes are reported so NHS staff can learn from them. He and expert advisers have reported that about 400 people a year die or are seriously injured from incidents involving faulty medical devices, 10,000 patients suffer severe reactions to drugs, and 1,150 who have had contact with mental health services commit suicide.

At the same time, 28,000 people make written complaints about aspects of their clinical treatment in hospitals and the bill for negligence claims is about £400m a year.

The Department of Health last night denied covering up the extent of blunders. The raw figures from the pilot study were not sufficiently robust for official publication, it said. "There is nothing to hide and we are committed to publishing audited figures when they are available."

Sir Liam would give the preliminary figures at a conference in London today "with a 'health warning' about what conclusions can be drawn".

A spokesman added: "One of the major question marks about the quality of the data is that because the pilot scheme is in its early stages the number of reported incidents is less than international studies would suggest for comparable health care systems."

Error reporting is far more advanced in Australia and the US, and UK officials concede present systems are patchy.

Mike Stone, of the Patients' Association, said information about mistakes must not be suppressed, whatever the true figures. "Just because information does not put the department or a hospital in a good light, you cannot cover these things up."

The British Medical Association supported a move for more openness. "There is scope for important public debate but the figures must be robust so you can draw the right conclusions."