Modelling stroke care systems: Evidence of the benefits

Over the past four and a half years I have spent a considerable amount of time trying to persuade NHS staff to use more modelling in stroke care. The response I’ve become accustomed to hearing is generally: ‘This looks impressive, but what evidence do you have that modelling can actually drive improvement?’

Luckily we can now point to concrete evidence in the shape of our work as part of the  NIHR CLAHRC South West Peninsula to improve treatement rates for patients with acute ischemic stroke.  In such cases the blood flow to the brain is cut-off by a blood clot. In England there are over 150,000 strokes each year with around 200,000 patients living with post-stroke disability burdens. In 2008 the total cost of strokes in the United Kingdom was estimated at £9 billion.

Time is brain

Ischemic strokes represent around 85% of all strokes and a select group of them can be treated with recombinant tissue plasminogen activator (rtPA) which dissolves the clot and restores blood flow.  However, this process, called thrombolysis, needs to be administered within 4½ hours of onset. Its effectiveness is extraordinarily time dependent.  Quite literally time is brain

Delays to treatment

So why is it difficult to treat strokes this fast?  The NHS’ FAST campaign (face, arms, speech, time): explains that someone has to recognise the symptoms and get the patient to hospital.  On arrival, the patient needs a CT (computerised tomography) scan to rule out a haemorrhagic stroke (which would worsen under thrombolysis) and a physician needs to check for contraindications to clot-busting drug.  If an A&E department is the way in there may be delay ahead of the scan and assessment.  And if the stroke occurs at night, it takes longer still.

Evidence that modelling works

We used simulation to test-drive some options for managing suspected stroke patients in order to minimise time to treat and maximise the benefit to patients.

As a result of this,  the Royal Devon and Exeter Foundation Trust now treats four times as many stroke patients in half the time.  This was achieved by implementing two process changes following the modelling.  First, ambulance paramedics now by-pass A&E with all suspected stroke patients.  The acute stroke team are instead alerted to pending arrivals as patients are transported to hospital.  Second, senior A&E nurses alert the acute stroke team of any suspected strokes that have self-presented as they are triaged.  This by-passes any lengthy wait for physicians in A&E.

So isn’t this obvious?

The question is why did we need to model it?   The trouble is that many ‘obvious’ improvements are simply not implemented successfully or sustainably.  You need to convince a lot of people to change their practice and the model helped to do just that.  In this case modelling translated the evidence of the clinical effectiveness of rtPA into a local context.  The magnitude of the improvement predicted by the model both in terms of treatment rates and post-stroke disability made it more real for clinicians in the hospital and convinced them to implement the changes.

The future

Modelling holds great potential in other areas of stroke care, particularly given NHS England’s push to centralise acute stroke services.  This creates a tension between increased pre-hospital travel times and reduced in-hospital processes that can only be understood by modelling.