So how many times can you die?

No this is not a morbid anticipation of my admission tomorrow for a hip revision operation but both a reflection and memory prompted by a news item on the BBC web site this morning : http://news.bbc.co.uk/1/hi/england/staffordshire/7301688.stm about a hospital which has a level of deaths outside the normal range and is about to be investigated. Already it appears that the appearance might not reflect reality and may be down to the way data has been recorded.

Despite the number of times that reviews of performance end up generating lots of heat but very little illumination it is still very necessary.

Lest we forget:

Beverley Allot,(http://www.guardian.co.uk/uk/2007/dec/06/ukcrime.health)
Bristol Children's Cardiac Deaths, (http://www.bristol-inquiry.org.uk/final_report/report/Summary3.htm)
Harold Shipman (http://news.bbc.co.uk/1/hi/uk/3391871.stm) and more recently
Colin Norris (http://news.bbc.co.uk/1/hi/england/west_yorkshire/7276700.stm)

should all remind us of the need for continuing vigilance.

After 2 years training as an NHS Administrator my first substantive post was a lowly administrator in charge of the general office of Pontefract General Infirmary in 1973. It was not at all what I had expected. There was an element of "man" management but to my mind much too much time spent on financial matters such as collecting all the cash from tills and petty cash floats throughout the hospital and explaining each week to disgruntled staff why the pay they received was less than they expected. The weekly petty cash accounting ended up with banking a vast amount of coins and cash and producing the necessary paperwork that showed everything balanced and could be reconciled with receipts issued by tills or various departments who either disbursed or received cash to/from patients and sometimes staff.

Just across the corridor from my office was a medical records outpost where the junior doctors would turn up and sign cremation certificates. As this was outside the normal range of their hospital duties they received payment on a per item basis. This was jokingly referred to as "Ash Cash". Depending on how much had been disbursed I would need to top up the float kept in the department and get a receipt signed.

Being somewhat sceptical I asked if "ash cash" might prove to be an incentive to provide less than the best care for some patients and although my suspicions were dismissed there was something in the way that they were dismissed that made me ask a supplementary question about whether any junior doctor seemed to sign a disproportionately high number of such certificates. There was also something in the way they responded that made me ponder this more - I think they had suspicions but could not bring themselves to speak them out loud - and having a penchant for data analysis I suggested to the Chief administrator that I might be allowed to record the times of deaths and locations to see if there was anything out of the ordinary and if there was whether this could be correlated with staff rotas. I was told that this was not something that hospital administrators needed to be concerned with and to get back to counting up the petty cash.

I was reminded of this a year or two back during an email exchange with a good friend who is the Chief Executive of a hospital in Melbourne. It caused a shiver when I realised that had I been allowed to pursue this weird request I might well have turned up evidence of Dr Harold Shipman's early murders of hospital patients, as he was one of the junior doctors working at Pontefract General Infirmary at the time and had already started on his destructive ways .

It always seemed to me important to retain a healthy scepticism about the purity of motives and abilities of hospital staff. After all they would share all the human frailties as the rest of us and "statistically" it was likely there would be some who were dangerous and needed to be identified early for remedial action to be taken.

Many years later I was approached by Jeremy Laurance, then of the Times to assist in the interpretation of hospital mortality statistics which were about to be published for the first time by the Department of Health. I was at the time an academic in charge of what I liked to call the "Informed Management Unit" of the Nuffield Instiute for Health Studies at Leeds University. ("informed management" as distinct from the "uninformed mismanagement" that I experienced around me in the NHS at the time).

I took the raw data which was standardised for age, sex and casemix and sorted it into a descending order and then plotted quartile scores onto a map of the UK. Of course I expected to see a picture resembling a well mixed fruit cake with the top quartile Standardised Mortality Ratios scattered throughout the UK. Instead I saw a significant clustering of the highest levels of hospital deaths along the M62 and down the northern section of the M1. In addition although somewhat less distinct was a concentration around the North East London and into Essex.

The article was held up from publication by one day because of the IRA bombing of the Baltic Exchange in the City of London on 24 April 1993, but the next day the map, sorted table and some comments from me appeared in the front page leading story. I had tried to tell Jeremy that from previous experience the explanations might lie in poor data, poor coding and classification or could be telling us something we did not know before.

When asked to speculate further I did say that possible reasons for high hospital mortality might reflect the underlying poorer health of the populations (quite likely given the industrial nature of the areas concerned), fewer resources, a greater willingness to undertake high risk cases or might reflect the poorer skills of the doctors attracted to work in the areas concerned. This analysis sounded more alarming when the relative position of the 4 possibilities was reversed and as a result it appeared that I was pointing the finger at the skills of doctors first and foremost!

A few days of senior NHS staff commenting on the figures particularly in the "worst" areas and hospitals followed. One leading Regional Medical Officer was quoted as saying that they were not worried about studying hospital mortality and the chief executive of one of the hospitals concerned said she had never seen these figures and did not routinely question them. Within a week the whole issue had "blown over".

2 years later when I was working in Riyadh I received a call from Jeremy who wanted to repeat the analysis on the "latest" figures to be issued by the Department of Health. I was flattered that he had tracked me down and wanted my assistance again. I arranged for the "latest" figures to be posted out to me and began my analysis only to discover they were exactly the same figures provided earlier. When I returned to the UK I visited the Department of Health Web site to obtain the next years figures and found they were no longer produced.

When I enquired why they were going backwards and no longer publishing hospital mortality figures I was told that they caused so much trouble when they did that they decided not to again. Now as it happens they were right to do this because further enquiry elicited that the reason for the clustering of high standardised mortality rates was related to the IT systems in use in the districts concerned.

The NHS had a few years earlier undergone a necessary upheaval in the way it recorded hospital activity. In the past, output (not outcome note!) had been recorded as discharges AND deaths - it did not matter in what state a patient left their bed as long as they did leave it. Some patients would in the course of a year be admitted many times to the same or different consultants for the same condition. As a result new data definitions and categories were designed to provide more meaningful information. So in future we could differentiate between consultant episodes and patient episodes. One patient episode might involve one or many consultant episodes.

What had happened in the early years of this reform was that some IT systems allowed coding staff to record the same patient death in hospital against all the consultants they might have seen as an in patient. In other words if a patient had been seen by 3 consultants ( say orthopaedic on admission, cardiac after a heart flutter and a general physician because of some other problem that required medication) and then died before being discharged, this counted as 3 deaths not one. Other hospital coding staff were recording the death either only against one consultant or ensured that the summarised reports of hospital deaths used patient episode data not consultant episode data.

When the Department corrected for this the highest mortality rates were more randomly scattered throughout the UK.

However despite these coding and classification problems (and they will always be there because its human nature to be well meaning but confused) the need for continued vigilance remains. If this recollection is read in years to come (as I hope it will) no doubt you will be able to extend the examples I provided earlier of why this is important to do and to get right - if for no other reason then to make sure that you really do only die once!