The mysterious Dr Foster

Who is this Dr Foster, the source of the weekend's headlines about the NHS? He seems to be saying that the elaborate system for inspecting hospitals that was set up – and recently reformed – by the Government is failing to spot some pretty obvious examples of bad medicine.

Thousands of patients dying, hospitals failing to respond to safety alerts or unexpected deaths, swabs and drill bits left inside patients after operations... Isn't this the kind of thing any respectable Government inspector might be expected to pick up on? What on earth is going on?

Dr Foster is not, of course, a doctor at all, but a witty name chosen by two journalists when they set up a healthcare analysis company eight years ago, planning to exploit the mass of data churned out by the NHS (..."Dr Foster went to Gloucester, in a shower of rain")

Tim Kelsey from The Sunday Times and Roger Taylor from the Financial Times saw an unexploited commercial opportunity in the work of Professor Sir Brian Jarman of Imperial College. Professor Jarman had devised a way of measuring how well hospitals perform by using the Hospital Episode Statistics (HES) produced by the NHS.

Kelsey and Taylor's timing was good. The Department of Health under Alan Milburn needed evidence to support and monitor progress in implementing the NHS Plan and, after some initial hesitation, embraced Dr Foster warmly. The first Good Hospital Guide under the Dr Foster imprint appeared in 2001. Hospitals were under notice to collaborate with Dr Foster, however little they liked doing it.

So keen was Milburn's successor, Patricia Hewitt, on the product that she bought the company. Under a deal negotiated in 2006, £12m of public money was invested in a joint venture between Dr Foster and the NHS body, the Information Centre for Health and Social Care, to create the Dr Foster Intelligence consultancy.

The fact that this deal went through without a competitive tender raised some eyebrows, and was heavily criticised by the National Audit Office (NAO). The department had paid too much, and had failed to give others a chance to bid, the NAO ruled.

Professor Denise Lievesley, a statistician who had become chief executive of the Information Centre, protested to her bosses in the NHS, and again in 2007 when a contract to provide material to the NHS Choices online information service was awarded to Dr Foster without (in her view) proper procurement procedures. She was eased out of her job, with a gagging clause preventing her from telling her side of the story.

So the irony of the weekend's stories is that a company granted at least one sweetheart deal – and possibly two – by the Department of Health has been the bringer of exactly the sort of bad news the department would rather not have heard.

There have been hints that the department has fallen out of love with Dr Foster. The consultancy lost the contract for NHS Choices in July 2008, and set up a rival website for its Good Hospital Guide. There are suggestions that the Information Centre plans to sell its stake in the joint venture.

That said, is what Dr Foster says about the NHS to be relied on?

Professor Jarman's method of comparing hospitals is called the Hospital Standardised Mortality Ratio, or HSMR. From the HES data, the number of patients dying after a range of different procedures in hospitals is available. All things being equal, good hospitals will kill fewer patients than bad ones. The difficulty is that all things are not equal.

Hospitals vary in many ways – the area they serve, the age of their patients, the efficiency with which they record the data. The HSMR figures incorporate corrections for these variables, so that at the end of the process one hospital can be compared directly with another. Professor Jarman is a leading world authority on this subject, so nobody doubts the probity of the process.

But many do question the results. One striking feature of this year's results is a sharp fall of 7 per cent in HSMRs across England. This is good news – if slightly too good to believe without careful scrutiny – because it implies that many fewer patients are dying in England's hospitals than in previous years. But the fact that HSMRs across the board have fallen so fast means that the target has moved. The results have been rebased so that a hospital that exactly matched the national average would score 100 – those with more deaths above 100, those with fewer below 100. This means that some hospitals who know they have improved appear to have done worse, because they haven't improved quite as much as the average. Others say the Dr Foster method misrepresents how good they are.

Alternative methods for calculating HSMRs exist, and in some cases paint a much more flattering picture.

One reason for this may be that hospitals do not invariably "code" their patients fully. If a patient has four or five co-morbidities (conditions that make survival less likely) then this should be taken into account. If they are not recorded properly, then a hospital will have a higher HSMR than it should. (There may even – perish the thought – be the suggestion that clever hospitals are exaggerating the co-morbidities to improve their HSMRs. It wouldn't be the first time a target has been "gamed" in the NHS.) Another possible source of error is that Dr Foster counts only deaths that occur in hospital. If patients are discharged and die the next day, at home or in a hospice, their deaths go unrecorded.

So the Good Hospital Guide should be seen for what it is: a sincere attempt to measure something very elusive but very important. But is it more reliable than the Care Quality Commission (CQC) verdicts? Dr Foster can be said to provide an important corrective. Both approaches have their virtues, but careful scrutiny and a degree of scepticism should accompany the reading of either.