As a medical director of a large acute hospital trust, how people die in our care occupies quite a lot of my waking thoughts (and quite a few of my sleeping ones too…). I thought I knew quite a lot about this subject – coming into the role from a background of critical care (where death is common). I have found though that my experience, whilst a useful primer in the topic, has demanded significant additional learning on my part – this post is my attempt to share some of that learning.

People Die in Hospital – That’s what they do

There is a curious symmetry in the social changes that have occurred at both the beginning and the end of life. The realisation over the last half century or so that many of what were believed to be inevitable tragedies at the beginning and end of life were in fact amenable to technological intervention and thus avoidable – and the change in the medical profession from overseer and commentator of natural tragedy to intervener and preventer of such – has driven these life events from a home based setting managed by community based services into a hospital setting managed by doctors, nurses and midwives.

This is largely a good thing – more mothers and babies survive the traumatic vagaries of child birth than ever before and people live longer healthier lives than ever before.

There is one rather stark asymmetry though – whilst death and tragedy in child birth is now a very rare event indeed – death at the end of life is still ultimately inevitable. As a society, even though we know we have to die eventually, we still haven’t worked out how we should die. This means that the majority of people end up dying in hospital (over 50% of people – thanks to Shaun Lintern and Craig Stenhouse for correcting this for me!) with doctors and nurses trying to stop them from doing so even when death is inevitable.

This is not a good thing because dying with someone trying to stop you with all the technological weapons of modern medicine (however well meaning) is not a good way to die.

The alternatives – provided by the heroes of the hospice movement – remains a cinderella service, patchily provided, reliant on charity and sparsley funded by public money.

Avoiding the Unexpected

So, people die in hospital – in our two hospitals that is about 1600 people a year (about 1 in every 75 admissions or put another way 4 or 5 admissions every day don’t make it home alive).

When these deaths are studied they can broadly be categorised into three groups according to whether they are expected or unexpected deaths and avoidable or unavoidable deaths.

Category 1Unavoidable expected deaths. This is overwhelmingly the largest category of deaths in hospital. People reaching the end of their natural lives, where their frail bodily systems have reached a point where they provide no defence against otherwise modest insults such as a fall, a urinary tract infection or the flu. Dying in hospital, as I alluded to above, has become the natural order of things in modern Britain (and many other developed health care systems in the world). Also included in this category are the deaths that we would recognise as ‘palliative care’ where even 21st century medicine genuinely has no response – terminal cancer or devastating neurological disease.

Category 2Unavoidable unexpected deaths. These are the genuine tragedies that punctuate life in a hospital. The relatively rare, but affecting cases that litter the landscape of a professional career. These are the maternal deaths from catastrophic amniotic fluid embolus, the brain haemorrhage from an unsuspected cerebral vascular anomaly, the tragically successful suicide or the disastrous major trauma from a road traffic accident (sadly I could go on). These deaths are often met with heroic efforts on the part of hospital staff to prevent them and are followed by dismay and soul searching when unsuccessful. When deconstructed at a later date there is often some learning (we can always do better) but just as often there is little to be learnt – bad things just happen.

Category 3Avoidable unexpected deaths. These are the ones where we got it wrong. Where lapses, incompetence, inexperience, neglect and system failure succeed in lining up in a single case to either actually cause death or prevent avoidable death. These are the cases that stay with you (believe me). This is the missed diagnosis because of a failure to follow up a test result, the failure to act on the deteriorating patient, the unacceptable delay in life saving intervention or the failure to take responsibility for a situation. They are thankfully rare (and overwhelmingly the smallest category of death) – not because the lapses and system failures are rare – they are not – but because (other than in exceptional circumstances) it takes several in a row to have such a devastating outcome, the good old Swiss cheese effect.

I suppose in theory there is a fourth category of avoidable expected death – that feels like a contradiction to me and I can’t conceive what it might be.

The ‘management’ task here is to attempt to assign every death to one of the three categories. That is not a trivial task – and it is not one for which you can take a statistical short cut, however compelling a notion that might be. Not least because in every category there will be deaths where on review there is identified a lapse of care – yet how many lapses does it take to move a death from either of the first two categories to the third?

Lies, Damn Lies and Mortality Statistics

Regular readers of my blog will know that I have been here before in my earlier post: Thinking about Mortality – Fast and Slow. I am not going to rehearse the arguments of the weaknesses of case mix adjusted, risk modified mortality statistics. Suffice to say that the reported ‘dramatic improvement’ in HSMR of the 11 Keogh Review hospitals will have had as much to do with scrutiny of their coding practices as the quality of care they provide.

Professor Nick Black through the PRISM (and soon to be published PRISM 2) study has shown little correlation between quality of care and HSMR (or SHMI or RAMI or any other way of measuring institutional mortality rates). I am trying not to be cynical about these – because they are in fact very useful tools, and cannot be ignored. However one has to be very careful in interpreting them because they tell you as much about what type of hospital you are and what type of patients you treat as they do about the quality of care you provide.

I can tell you without looking that:

Small provincial district general hospitals with proportionately large elderly care, stroke and fractured neck of femur services will have high mortality ratios (however well risk adjusted).

Large urban inner city general hospitals with relatively young transient populations will have low mortality ratios.

Tertiary Centres that carry out high risk care on relatively young patients (like liver transplants or major cancer surgery) will have high mortality ratios.

A Recipe for Managing Mortality

So where does this leave us? We can’t trust the stats and identifying the deaths where we got it wrong, sufficiently to have at least been a causative factor in the death, requires painstaking review of very large numbers of deaths and the judgement of Solomon.

Here are my tips:

1) Don’t ignore the stats – have a committee that looks at your death rates in all specialties and diagnostic groups – identifies the variances, the outliers and the alerts and investigates them even if CQC or Dr Foster don’t notice. You will learn a huge amount about how care is delivered in your hospital and how patients are moved about within it. You will also learn quite a lot about coding.

2) Have a system in place that reviews every death in the hospital. Some have chosen the Medical Examiner solution to this – paying a recently retired doctor to review the notes of every death and extract learning and identify those where more scrutiny is required. We have gone down a variant of the trigger tool methodology – every case is assessed against a set of criteria by a consultant not involved in the care of that patient, if enough triggers are present in one case it gets a table top review.

3) Have a method of sharing the learning – particularly with the consultants doing the screening (in our case that’s all of them), it is laborious work and it has to feel worthwhile. We are not so good at this yet – we have a news letter that goes out, we could and should do more.

4) Don’t trust Dr Foster – they are a commercial organisation that profits from scare stories. Don’t ignore them either.

5) You will find that 70% of patients that fall into category 3 are due to sepsis – invest in physiological track and trigger, sepsis screening tools and implement the sepsis 6 bundle and critical care outreach. Your avoidable mortality will fall.

6) You will also find that acute hospitals provide dreadful care to the majority of patients for whom death is unavoidable and expected (Category 1). The fortunate minority that find themselves under the care of palliative care doctors will get excellent care, the majority that don’t, won’t. Invest in end of life services – this has to be the greatest single priority for the NHS. We are about to publish our ‘Last Year of Life Audit’ – an investigation into the care provided in that critical period of a persons life. It will show that we admit to hospital on average three times in the last six months and start ‘supportive care’ on average 24 hours prior to death.

7) Finally – keep the politicians out, they don’t get it and never will and will do a great deal of harm meddling with stuff they don’t understand. You wouldn’t let them meddle with airport safety would you? why would you let them meddle with hospital mortality…?


  1. Thanks for a great summary of the risks of simplifying the complex.

  2. Great blog Roger. I agree with everything you say.
    One issue is that the press use these stats out of context and can cause real harm to the reputation of a hospital. 1000’s of excess deaths make great headlines but are statical nonsense. We all know this as must the press by now but they still do it.
    Don’t think we can get the genie back in the bottle but there ought to be a prominent health warning every time these statistics are released.

    • Thanks for comment Steve – you can cut the stats any way you want and its down to the reporter as to whether they use them to put down the NHS or big it up! – any chance of a tweet from BUTNHS 😉

  3. Thank you for your thoughts, as a hospital chaplain much of what you say makes perfect sense but as we are ‘not in the clinical loop’ we can only make guesstimations about the figures/causes etc. I think my trust does EoLC well but we suffer from those discharged home to die being returned by carers, family etc because they are ill! These of course affect the stats, but more importantly their death happens where they didn’t want it to!

  4. Kerrie (@kkstuart)

    Excellent blog and much needed to educate people on hospital deaths and HMSRs. As a Scot I have followed what has been happening in England for years now and can, as an outsider, see how the subject matter has been exploited by the press and politicians. The Scottish press also report HMSRs but always with a caveat that these stats are only an indication that something needs looked at rather than actual deaths. However when the same story is reported in England the caveat is missing.
    Steve is right to raise the point of reputational damage to a hospital, one only needs to look at Stafford to see its effects but there will also be an impact on patients as they become fearful of going into hospital as a result of all the hype.
    As far as HMSR goes, Scotland has been sceptical about this for some time and although still used in collaboration with case reviews, they view is to move away from them early next year.
    I also know a thing or two about stats and they can never be used on their own but see from England’s statistical bodies that the majority of NHS stats are used in that way, which is not only nonsensical but also dangerous, it means no one has a true picture of what is going on in the NHS.
    Finally the support needed at end of life needs to be looked at and structured/funded well as I can see the repercussions if it is not. Its not only what we owe to the patient at the end of their lives but also to their families. Twitter is awash with angry relatives that have been unable to go through the grieving process due to real/perceived failings in end of life care and therefore unable to move on with their own lives.

  5. Kerrie,
    I wholheartedly agree that statistics are so often misused, the ‘lies, damn lies and statistics’ saying is so often completely true!
    I also agree that we all need to manage to provide better and more sympathetic support for those who are reaching the ends of the lives, as well as supporting their families. The discredited LCP did make a start on doing this better, but it is my belief that until nationally we can accept that dying is (generally) a normal and expected thing, and the provision of services to those who are approaching death can be provided seamlessly by NHS and social services, we really are always going to struggle.

  6. Kerrie (@kkstuart)

    I agree with you Simon, I am old enough to remember the days when it was the norm for people to die in their own beds surrounded by their loved ones as no one would have ever thought of taking them to hospital to die. The wishes of the deceased would have been discussed with the family beforehand and the subsequent grief process was supported by the community, but then these were people who had lived through one/two world wars and were used to death. Perhaps we need a TV campaign similar to the organ donation campaign to ask people to discuss these matters with family members so that as a nation we can come to accept the inevitable will happen to all of us and that the grieving process is normal, in fact a healthy reaction to loss. I would also like to see more bereavement experts in the NHS/social care that relatives could call upon for support with palliative care leading up to and beyond the death of a loved one. Perhaps the funding that has been earmarked for care in the community could also include end of life care in the community.

    • Kerrie,
      I too am of that generation, and yes I agree that grieving can and usually should be a normal and healthy response to loss. I certainly like the idea of a TV campaign to educate people to this. My worry is always that by concentrating too much one aspect of dying (here mainly with the elderly) we would risk missing out those who suffer the loss of a loved one at the ‘wrong’ time. My work involves support of anyone and everyone in the hospitals I serve, sadly a lot of that work involves caring for young women whose baby is stillborn or miscarried , so the subject of death needs careful and sensitive handling. But we do need to break the taboo that surrounds the subject, and perhaps we could start by educating our colleagues in hospitals and the caring professions.

  7. Stroke is a great example of where any mortality assessment can never stand up to any scrutiny.

    Patient A and Patient B are the same age and have the same coded morbidities

    Patient A has a tiny lacunar infarct and can be discharged home after one day to do well. Mortality in hospital is tiny in the first 24 hours.

    Patient B has a massive ischaemic stroke taking out most of a cerebral hemisphere with oedema and is cheyne-stoking in ED. Mortality in hospital will be nearly 100%.

    To mortality assessments these two patients are identical as they both have had an ischaemic stroke and have the same expected mortality which is clearly not so. The major determinant as to survival from stroke is the size of your stroke. This is not measured or coded. Your mortality rate depends on how many patient A types and patient B types that you identify. If you choose to discharge from the ED all your patient A types your mortality will go up as the patient is not admitted and so does not register so you are left in a bind feeling what do I do here. BY failing to admit I am raising the mortality rate of my unit which will then raise questions over care when I am simply doing good medicine.

Leave a Reply to SimonE Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.