The centralisation of hyper acute stroke care will save lives and money, won’t it?

how many lives will be saved by the centralisation of hyper acute stroke units. How much money will this save.

Warning – this is long and pretty tricky.

The short answer is not many & not much. There may be higher value things to do.

I’m often asked this
We are centralising hyper acute stroke care. This is a national recommendation. Everyone is of the opinion that there will be many lives saved by this. Nobody really knows how much money will be saved (or indeed what the net cost would be).


I don’t see it like this, from a population perspective the view is different. I’m not of view there are millions of lives to be saved here (the Morris study in London v Manchester) and there are other things that might be better to focus energy into ( preventing stuff for example)


I’ve challenged on the evidence for my view a few times, so I had to do some homework.


Apologies this is a long and complex blog (it’s a complex issue). 

The issues often pertain to the interpretation of evidence – individual level vs population perspective – as much as the evidence itself.


Most of the evidence re centralisation / hyper acute relates to LOS and mortality. The research around centralisation leading to lower downstream morbidity & institutional care with hyper acute (compared to non centralised) has not been done, to my knowledge.


The central hypothesis

The establishment of hyper acute stroke unit is a key strategy to improve outcomes in ischemic stroke patients presenting within 6 hours.

The aim is to improve an organized response from multiple disciplines and clinical areas, a principal objective within this is to increase the proportion of patients receiving thrombolysis within 3 hours of stroke onset (nb onset to 3 hr, not door to needle time, but onset to needle time).

Achieving this in practice requires a significant reorganisation the rationale of which is sound. There are many other drivers including safety, workforce numbers and ability to staff rota 24/7

There are multiple approaches to reconfiguring stroke care, see postscript 2  for a more detailed commentary.



To the evidence…………………………..



Organised stroke care

Bray (BMJ 2013;346:f2827) and many others have demonstrated that ORGANISED stroke care is a good thing

Though didn’t consider thrombolysis in this.

This is fairly uncontested.



Centralisation and hyper acute

There are 2 key studies to my knowledge – both related to the same body of research.

The two most commonly referenced papers are Morris BMJ and Ramsay stroke

Looks like the full mashings on the NIHR website isn’t yet published.



Morris is the critical study

compared London to Manchester

selection of hospitals to  become sites for specialist stroke units – hub&spoke networks  during the first 72 hours after stroke


Description of reconfiguration

“Hub and spoke” model for acute stroke care. In London hyperacute care was provided to all patients with stroke. In Greater Manchester hyperacute care was provided to patients presenting within four hours of developing symptoms of stroke.


Slightly different approaches to reconfiguration – hyper acute available to all vs those presenting within 4hr



24 units designated to provide acute rehabilitation services, and eight of these were attached to a hyperacute stroke unit; five hospitals were no longer to provide acute stroke services.

Hospital selection was guided  by modelling exercise. Sites identified based on determination of need, including the travel times involved, with the intention that no Londoner would be more than a 30 minute ambulance journey away from the nearest hyperacute stroke unit.


G Manc

original intention was also to treat all pt in hyper acute unit – one24/7comprehensive stroke centre and two primary stroke centres running 7 am-7 pm,mon to fri).

Concerns about no of patients being transported greater distances, repatriation issues, and a view that access to specialist stroke centres was purely for thrombolysis led to situation where only pt presenting  within four hours of developing stroke symptoms were taken directly to a comprehensive stroke centre or primary stroke centre; all other patients were taken to one of 10 district stroke centres, which were designated to provide all aspects of post-thrombolysis care

No hospital stoped providing stroke services entirely as a result of the centralisation process in Greater Manchester




hyperacute stroke care was provided to all patients,

reduction in mortality and length of hospital stay

From memory (and not in this paper) £28m spend on the reconfig

1.4d reduction in LOS

1.1% reduction in 90d mortality (NNT= 100)


I’m not 100% sure whether this is judged as statistically significant.

Obviously it’s important, but whether stat sig……



Greater Manchester, 

hyperacute stroke care was provided to patients presenting within 4h

no impact on mortality but length of stay in hospital fell

2d reduction in LOS


The discrepancies re greater mortality reduction in London but greater LOS reduction (indicator of disability) in Manchester require some further insight by people cleverer than me…….


Processes of care

ambulance data – in 2011-12,

98.7% of patients with stroke were transported to the appropriate service:

95.7% were taken appropriately to a hyperacute stroke unit;

3% were taken appropriately to an emergency department;

1.3% were taken to an emergency department when they should have been taken to a hyperacute stroke unit.



G Mancs

A review of the first year

Of those who present with stroke within four hours, 36% were not taken to a comprehensive primary stroke centre. Thus – higher proportion of patients than planned were admitted to district hospitals, where access to specialist expertise and care was more limited.

Higher achievement of all care processes indicated in the National Audit comparing London to Manchester


It’s shame that some measure of disability (say discharge to residential care) isn’t included in the data.


What is also interesting is looking at fig2 at the very back of the Morris paper

The blue line is the trend in England.

Manchester doesn’t look to be hugely different from England

London does – extent to which it’s the hyper actute being available to ALL, vs the 4hr. Would be interesting to do the stats!


As always the rapid responses are worth a read


Elton – nuances in interpretation

“It is dangerous to be categorical that the there has been a much better outcome in London than in Greater Manchester based on the adjusted mortality following hospitalisation when some of the unadjusted mortality shows no significant difference, the length of stay (which may reflect disability) shows a greater decline in Greater Manchester as does the population stroke mortality”

Thus……….maybe hard to generalise from one place to another


Hill – cost to implement 

“The study authors fail to discuss the impact of significant financial investment in services in the results achieved. The London Stroke Strategy was supported by an additional total £23 million in capital funding as reported by Healthcare for London (1). The cost improvement programme highlighted the need for an additional £13 million investment in HASUs to provide the imaging and nursing needs to improve the service, and a £7 million staffing investment in SUs. My understanding is that similar investment in services was not made for Manchester, nor is there intent for similar investment elsewhere in the United Kingdom despite heavy promotion of the perceived merits of greater centralisation of stroke services.”

Show me the health econ analysis please!


Some bloke called Fell

NNT to prevent a death is c90

So got to treat 90 incident strokes (c1% ARR) to prevent a death


with all the inherent cost of the reconfiguration (one may never actually realise the benefit of this in cold hard cash unless you ACTUALLY take the capacity OUT of the system etc etc) and politics of closing units that don’t win the HASU beauty contest

the vast majority DONT get thrombolysed rapidly enough.


If you want to make most difference to pop outcomes the IOW study says thrombolysis makes a pretty marginal difference. Prevn of stroke far better bet.



Ramsay et al 2015

Stroke. 2015;46:2244-2251


Considered performance against core national audit indicators in more depth

Further cements the notion that it’s the fully centralised model that makes the difference

“Postcentralization, likelihood of receiving interventions increased in all areas.

London patients were overall significantly more likely to receive interventions, for example, brain scan within 3 hours:

Greater Manchester=65.2% (95% confidence interval=64.3–66.2); London=72.1% (71.4–72.8); comparator=55.5% (54.8–56.3).

Hyperacute units were significantly more likely to provide interventions, but fewer Greater Manchester patients were admitted to these (Greater Manchester=39%; London=93%).

Differences resulted from contrasting hyperacute unit referral criteria and how reliably they were followed.”


However – eyeballing the data

The unadjusted (table 1) & risk adjusted (table 2&3) before and after differences

It seems like it’s the brain scan at 3h and stroke unit at 4h that are the difference between London v comparator and Manchester v comparator

Improvement in other indicators looks to be across the board



The third chunk of evidence is on thrombolysis per se. Does thrombolysis make a difference

Of course all HASU is based on the premise that rapid access to thrombolysis is the critical issue that makes the difference


a) How effective is it

Cochrane 2014 review is the current definitive


You can read the detail at your leisure, key points of abstract are here for your convenience:-


  • Thrombolytic therapy, mostly administered up to six hours after ischaemic stroke, significantly reduced the proportion of participants who were dead or dependent at 3-6m by 15%
  • Treatment within 3 hours reduces risk of death or dependency by 35%
  • Thrombolytic therapy nearly quadruples the risk of symptomatic intracranial haemorrhage and increases risk of poor outcomes from other bleed related events.
  • Participants aged over 80 years benefited equally to those aged under 80 years, particularly if treated within three hours of stroke.


Relative vs absolute benefits (and risks)

the Cochrane review reports all the outcomes in relative terms (this is actually usual for a Cochrane review, perhaps reflecting the complexity of the review itself and the heterogeneity of the evidence).

The 2013 SSNAP report had an expression of the absolute benefit in absolute terms

the outcomes for 100 patients treated with clot busting drugs within 3 hours of stroke:

 fig 10


The review

A prior 2013 review by the NNT was rather more hardline and rather less complementary. It concluded that there was no benefit from thrombolysis.


Morgenstern -This review is also worth a read. A commentary on the available evidence

And from the conclusion

 b) Coverage of thrombolysis is a critical consideration in assessing population impact of something

It’s not great.


It’s thought that 20% eligible for thrombolysis but c12% get

(Fig 5)


closer to home c5% – 7.5% of Sheffield patients get thrombolysis



  1. c) Speed with which thrombolysis is delivered obviously matters

Cochrane tells us that thrombolysis is most effective within 3 hours of onset

I don’t know the number or proportion of patients with incident stroke who get thrombolysis within 3hrs


Door to needle time offers some clues (fig 11) – over half of patients receiving thrombolysis are administered the drug less than an hour after arriving. This is obviously good news.


This doesn’t, however, tell us about the time before the door – ie onset to door

onset to door

The 2013 audit gives some clues on onset to door time:

So given all this it doesn’t seem likely that a large proportion of patients receiving thrombolysis receive the drug within three hour window


How much we may improve may depend on the starting point>

as with all these things it depends on what the starting point is.

Sheffield (STH) thrombolysis rates are low (5-7%), our baseline mortality is not terrible. SSNAP has just published the first tranche of mortality data – Sheffield CCG is the red dot in the funnel plot.



Until somebody is able to look at our hospital configuration and baseline position relative to other areas we won’t be able to judge the mortality point (i.e. was that just solving a London problem).


So…… summary………

We are on this journey and not coming off it.

To be clear, I’m not necessarily arguing against it, just offering a population view and observing there may be more valuable things to to to save more lives, prevent more disability and save more money than this one (See the postcript for some quantifiable options).


  • Organised stroke care is v important – and probably the thing that makes the difference rather than thrombolysis per se.
  • On thrombolysis we’ve got low coverage of an intervention, with uncertain – but probably not high proportions of patients getting this intervention in the golden window, that doesn’t have amazing absolute benefit (expressed by NNT).
  • And we are not quite sure how much reorganisation of stroke care to improve this will cost or save, nor that if we do that it will save lots of lives
  • It’s important not to forget the whole stroke pathway not just HASU – for eg rehabilitation is important.
  • We just don’t know re downstream disability and entry to institutional care compared to either current or more organised model
  • We plain don’t know the economics of HASU establishment….These studies are starting to come through now –
  • Its obviously important we don’t loose sight of prevention. We are in danger of under focusing on two things that make a bigger difference:- reducing stroke INCIDENCE (back to our old friends prevention) and organised stroke care (mainly a downstream disability not a mortality point)/
  • There’s considerably more to gain pop outcomes and money wise through prevention of stroke – salt, BP management, anticoagulation etc
  • mortality after a stroke is c 10-15% at 1 year the Morris study on reconfiguration gives us some clues as to by how much this can be reduced through reconfiguration about an absolute risk reduction of 1%. Compare this the the reductiuon in stroke mortality risk of strokes prevented – ie no risk.
  • Moral = don’t just listen to those that come with shiny “we will save lots of lives” cases

This isn’t the first time I’ve said this, and I’m sure won’t be the last.


Postscript – but an important one.

Prevention versus acute treatment – which has biggest impact.

Estimating the health gains and cost impact of selected interventions to reduce stroke mortality and morbidity in England. LSE paper:


Again the key points from this paper are set out here for your convenience:-


Considered the population impact of six interventions to reduce the BoD from stroke:

(i) admitting all patients that presents to hospital to a stroke unit;

  1. ii) providing thrombolysis to 9% of stroke patients;

iii) prescribing a first line antihypertensive to all 55-year old or older people (unless they develop side effects);

  1. iv) prescribing a first line antihypertensive to all people with blood pressure above 140/90 mmHg (unless they develop side effects);
  2. v) reducing average daily sodium intake through processed food by 30% (by 3gr salt);
  3. vi) reducing average daily sodium intake by reducing salt content in bread (by 1.5gr salt per day).


NB…………….Its fair to say that Richard Lehman no less has issued me with a warning about the evidence base here:


he’s right…. act with some caution.




The current BoD from strokes occurring in one year is 663,000 QALY lost (474,000 discounting future outcomes).

Extending stroke units is estimated to have only a limited impact in reducing the BoD at the population level: by about 6%.

The limited impact from achieving universal admission of stroke patients to specialised stroke units dwarfs the minimal impact attributable to Thrombolysis: a reduction of the BoD of less than 0.5%.


Preventive interventions have a greater scope to reduce this burden than acute care interventions.

Assuming that all people above the age of 55 were to take a first line anti- hypertensive, the burden can be reduced by about 14%.

If the new prescribing were limited to hypertensive people 7% of the burden would be avoided.

A public health intervention to reduce the daily intake of salt (sodium) by 30% as proposed by the target set by the Food Standard Agency in England would reduce the burden by 18%.

In fact any reduction of average blood pressure in the population will have a substantial impact on reducing the BoD: a reduction by 2 mmHg by reducing salt content in bread is estimated to reduce the BoD by about 7%.

The reduction in the BoD achievable with preventive interventions are underestimate of the overall benefit because these are reductions in the burden caused by stroke only and do not take into account the reduction in mortality and morbidity associated with other coronary heart diseases.


Although both stroke units and thrombolysis are worth doing on a cost-effectiveness basis, they are insufficient to reduce substantially the burden caused by stroke.

We are not arguing against the promotion of stroke units and thrombolysis. In fact, thrombolysis increases health and saves money if we take into account future savings from better outcomes. The point we emphasise is that it would be mistaken to assume that these interventions offer potential materially to reduce the BoD, because, even if patients were receiving the best possible care, having a stroke would still be a major cause of mortality and disability.

Postscript 2

Permutations on stroke care reconfiguration and other issues
Advice and view I received from an expert in this area:-

Firstly, in terms of stroke service reconfigurations, there are (to simplify) three main types of stroke unit configuration. Firstly, the traditional stroke unit model where patients with stroke are cared for in a generic stroke unit, with there being very little networking between hospitals. Secondly, the Comprehensive Stroke Unit Model, that uses a hub and spoke model to provide a higher level of care for the subset of patients requiring particular, resource intensive interventions or care but with the remainder/majority being managed in a generic stroke unit. This is the dominant model in the USA and in parts of Europe, and was the model used in the GM reconfiguration. Thirdly, the London HASU model that uses a hub and spoke model but aims to admit all patients to the hub initially, so that all patients receive higher level care in the firsrt 72 hours. Patients then are transferred back out to downstream stroke units for ongoing medical care and rehab.


The key difference between the Comprehensive Stroke Unit model versus the HASU model is the patient selection: the former provides acute higher level care only to a subset (e.g. potential thrombolysis candidates) whilst the latter provides this to all. This “higher level care” is not particularly well defined internationally, but in London meant essentially a High Dependency Unit level of care, with more intensive physiological and neurological monitoring, higher nurse staffing levels, more consultant input and higher therapy staffing levels. The main driver here is not thrombolysis per se but is probably better thought of as providing a bigger “dose” in the very acute phase of the complex intervention that is stroke unit care. 


Both types of centralisation are logical ways to increase thrombolysis rates and, probably, increase door to needle times (we have published data on the latter), but as you say, thrombolysis provides relatively less improvements in patient outcomes than stroke unit care. It’s also worth emphasising that thrombolysis does not improve survival (indeed it increases early mortality, with equivalence by six months) and so changes in thombolysis rates can’t explain the mortality improvements with HASU care. Instead, the main mechanism of action is probably in preventing the complications of stroke : stroke associated pneumonia, VTE, pressure sores, sepsis. Preventing these is probably why stroke units reduce mortality and why HASUs confer additional benefit beyond the basic stroke unit model. And if we unpick this further, we have some (observational evidence) that higher nurse staffing ratios reduce mortality, and that earlier access to specialist dysphagia assessment reduces pneumonia incidence post stroke.


Trying to work out what is going on with the differences in LOS is very difficult and I think the evidence here is weak. The reason is that LOS is highly determined by the quality and availability of community and home based rehabilitation services. I suspect that the improvements in LOS in both GM and London were more about improving access to Early Supported Discharge services than the acute front door care.


Looking to the future, acute stroke is now definitely heading the way of cardiology of c15 years ago. For ischaemic stroke, the evidence for thrombectomy (interventional radiology) is now very strong (with an NNT of 5 to prevent disability + death endpoint) – the caveat being that only about 10% of patients are suitable (generally patients with the largest strokes). Similarly for intracerebral haemorrhage, stroke care is increasingly moving to early and aggressive BP management with iv antihypertensives. So the theme is is of increasingly intensive medical input in the first few hours after stroke, and stroke units looking more like coronary care units in terms of staffing and equipment. The health economics of all this is still very unclear. Early models of thrombectomy are very favourable, because of the huge impact on long term disability. Agressive medical management of ICH may pose harder health economic questions, since it is likely to be shifting individuals from death to severe disability. Both these interventions can only realistically be carried out in higher volume centres.


Beyond clinical effectiveness and health economics, there are also a few other points its worth considering. Firstly, inequalities in stroke care have reduced a lot over the past decade (with there being wide age based disparities in care previously) but, especially if you look internationally, you can see that the Comprehensive Stroke Unit model can lead to cherry picking patient groups. In some places of the world, you’d need very good luck to get into the well staffed CSU if you were older, more complicated, had learning disabilities etc. We have preliminary data from the UK that selecting out patients based on thrombolysis eligibility leads to non-thrombolysis patients getting less good therapy etc. Secondly, the staffing issues you mentioned are unfortunatly very pressing, with a big lack of consultants and nurses to staff stroke services adeqately.


Finally, in terms of prevention – most definitely! AF is probably the biggest and most tractable one here that could benefit from specific investment. Not sure that throwing more money at blood pressure management is really the issue to be honest – we need to implement better models of care (eg home based monitoring and self- management, which we know works well!) and prevention (salt, physical activity) rather than giving more money to GPs for this.


Machester has now switched to a HASU model and so there will be a nice natural experiment to see if the change in service configuration leads to different outcomes


 new health economic data coming out of SSNAP imminently – end of project report is about to be signed off with NHS England and the data will be going live this summer. Of note this will enable us to get much better estimates of the health and social care cost of avoidable stroke and so make the potential ROI on prevention much more visible (and quantifiable). It’s a patient level, continuous time simulation model stratified by age, sex, stroke, type and severity and so we can estimate for example, health and/or social care £ of stroke in all patients in each CCG who had AF but no OAC pre stroke. Helpful use cases for public health teams would be great so when it’s out let us know what would be useful and we can build the data into SSNAP outputs


London had a very low starting point, Manchester didn’t. Perhaps  explains some of the discrepancies in the mortality finding between the two places in the Morris study.


 mortality is a pretty poor proxy for outcomes of clincial care (as per cancer mortality represents incidence and quality of care, survival better represents quality of care – thus in this context stroke case fatality at XXXX months, institutional care and return to work are the best indicators of how ‘good’ things are. 


Stroke system mostly only cares what happens after stroke and if we’re lucky secondary prevention gets a cursory look in. Primary prevention (ie where most of the financial and clincial gain comes from) is always ‘someone else’s job’. This is fine and I accept the challenge, but if this always remains the case we will always get the outcomes we always get etc.

Some issues to consider in evaluation of this reconfig?



  1. Hello Greg. My name is Carly and I’m from SONIK (Save Our NHS in Kent). We are currently fighting a proposal for 3 HASUs in Kent to replace the care in numerous hospitals. Thanet area will be over an hour away by road from a HASU. We’ve been researching the issue and came across your piece. Would you be willing to come to one of our meetings?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s