Second Wave

The summer holidays are over and schools are open again – sort of.

Restaurants, pubs and nightclubs are open again – sort of.

Gyms and leisure facilities are open again – sort of.

And after two months of gradual easing of social restrictions and massive expansion of test-and-trace we now have the spectre of a Second Wave looming.  It has happened in Australia, Italy, Spain and France so it can happen here.

As usual, the UK media are hyping up the general hysteria and we now also have rioting disbelievers claiming it is all a conspiracy and that re-applying local restrictions is an infringement of their liberty.

So, what is all the fuss about?

We need to side-step the gossip and get some hard data from a reliable source (i.e. not a newspaper). Here is what worldometer is sharing …

OMG!  It looks like The Second Wave is here already!  There are already as many cases now as in March and we still have the mantra “Stay At Home – Protect the NHS – Save Lives” ringing in our ears.  But something is not quite right.  No one is shouting that hospitals are bursting at the seams.  No one is reporting that the mortuaries are filling up.  Something is different.  What is going on?  We need more data.That is odd!  We can clearly see that cases and deaths went hand-in-hand in the First Wave with about 1:5 cases not making it.  But this time the deaths are not rising with the cases.

Ah ha!  Maybe that is because the virus has mutated into something much more benign and because we have got much better at diagnosing and treating this illness – the ventilators and steroids saved the day.  Hurrah!  It’s all a big fuss about nothing … we should still be able to have friends round for parties and go on pub crawls again!

But … what if there was a different explanation for the patterns on the charts above?

It is said that “data without context is meaningless” … and I’d go further than that … data without context is dangerous because if it leads to invalid conclusions and inappropriate decisions we can get well-intended actions that cause unintended harm.  Death.

So, we need to check the context of the data.

In the First Wave the availability of the antigen (swab) test was limited so it was only available to hospitals and the “daily new cases” were in patients admitted to hospital – the ones with severe enough symptoms to get through the NHS 111 telephone triage.  Most people with symptoms, even really bad ones, stayed at home to protect the NHS.  They didn’t appear in the statistics.

But did the collective sacrifice of our social lives save actual lives?

The original estimates of the plausible death toll in the UK ranged up to 500,000 from coronavirus alone (and no one knows how many more from the collateral effects of an overwhelmed NHS).  The COVID-19 body count to date is just under 50000, so putting a positive spin on that tragic statistic, 90% of the potential deaths were prevented.  The lock-down worked.  The NHS did not collapse.  The Nightingales stood ready and idle – an expensive insurance policy.  Lives were actually saved.

Why isn’t that being talked about?

And the context changed in another important way.  The antigen testing capacity was scaled up despite being mired in confusing jargon.  Who thought up the idea of calling them “pillars”?

But, if we dig about on the GOV.UK website long enough there is a definition:

So, Pillar 1 = NHS testing capacity Pillar 2 = commercial testing capacity and we don’t actually know how much was in-hospital testing and how much was in-community testing because the definitions seem to reflect budgets rather than patients.  Ever has it been thus in the NHS!

However, we can see from the chart below that testing activity (blue bars) has increased many-fold but the two testing streams (in hospital and outside hospital) are combined in one chart.  Well, it is one big pot of tax-payers cash after all and it is the same test.

To unravel this a bit we have to dig into the website, download the raw data, and plot it ourselves.  Looking at Pillar 2 (commercial) we can see they had a late start, caught the tail of the First Wave, and then ramped up activity as the population testing caught up with the available capacity (because hospital activity has been falling since late April).

Now we can see that the increased number of positive tests could be explained by the fact that we are now testing anyone with possible COVID-19 symptoms who steps up – mainly in the community.  And we were unable to do this before because the testing capacity did not exist.

The important message is that in the First Wave we were not measuring what was happening in the community – it was happening though – it must have been.  We measured the knock on effects: hospital admissions with positive tests and deaths after positive tests.

So, to present the daily positive tests as one time-series chart that conflates both ‘pillars’ is both meaningless and dangerous and it is no surprise that people are confused.


This raises a question: Can we estimate how many people there would have been in the community in the First Wave so that we can get a sense of what the rising positive test rate means now?

The way that epidemiologists do this is to build a generic simulation of the system dynamics of an epidemic (a SEIR multi-compartment model) and then use the measured data to calibrate the this model so that it can then be used for specific prediction and planning.

Here is an example of the output of a calibrated multi-compartment system dynamics model of the UK COVID-19 epidemic for a nominal 1.3 million population.  The compartments that are included are Susceptible, Exposed, Infectious, and Recovered (i.e. not infectious) and this model also simulates the severity of the illness i.e. Severe (in hospital), Critical (in ITU) and Died.

The difference in size of the various compartments is so great that the graph below requires two scales – the solid line (Infectious) is plotted on the left hand scale and the others are plotted on the right hand scale which is 10 times smaller.  The green line is today and the reported data up to that point has been used to calibrate the model and to estimate the historical metrics that we did not measure – such as how many people in the community were infectious (and would have tested positive).

At the peak of the First Wave, for this population of 1.3 million, the model estimates there were about 800 patients in hospital (which there were) and 24,000 patients in the community who would have tested positive if we had been able to test them.  24,000/800 = 30 which means the peak of the grey line is 30 x higher than the peak of the orange line – hence the need for the two Y-axes with a 10-fold difference in scale.

Note the very rapid rise in the number of infectious people from the beginning of March when the first UK death was announced, before the global pandemic was declared and before the UK lock-down was enacted in law and implemented.  Coronavirus was already spreading very rapidly.

Note how this rapid rise in the number of infectious people came to an abrupt halt when the UK lock-down was put into place in the third week of March 2020.  Social distancing breaks the chain of transmission from one infectious person to many other susceptible ones.

Note how the peaks of hospital admissions, critical care admissions and deaths lag after the rise in infectious people (because it takes time for the coronavirus to do its damage) and how each peak is smaller (because only about 1:30 get sick enough to need admission, and only 1:5 of hospital admissions do not survive.

Note how the fall in the infectious group was more gradual than the rise (because the lock-down was partial,  because not everyone could stay at home (essential services like the NHS had to continue), and because there was already a big pool of infectious people in the community.


So, by early July 2020 it was possible to start a gradual relaxation of the lock down and from then we can see a gradual rise in infectious people again.  But now we were measuring them because of the growing capacity to perform antigen tests in the community.  The relatively low level and the relatively slow rise are much less dramatic than what was happening in March (because of the higher awareness and the continued social distancing and use of face coverings).  But it is all too easy to become impatient and complacent.

But by early September 2020 it was clear that the number on infectious people was growing faster in the community – and then we saw hospital admissions reach a minimum and start to rise again.  And then the number if deaths reach a minimum and start to rise again.  And this evidence proves that the current level of social distancing is not enough to keep a lid on this disease.  We are in the foothills of a Second Wave.


So what do we do next?

First, we must estimate the effect that the current social distancing policies are having and one way to do that would be to stop doing them and see what happens.  Clearly that is not an ethical experiment to perform given what we already know.  But, we can simulate that experiment using our calibrated SEIR model.  Here is what is predicted to happen if we went back to the pre-lockdown behaviours: There would be a very rapid spread of the virus followed by a Second Wave that would be many times bigger than the first!!  Then it would burn itself out and those who had survived could go back to some semblance of normality.  The human sacrifice would be considerable though.

So, despite the problems that the current social distancing is causing, they pale into insignificance compared to what could happen if they were dropped.

The previous model shows what is predicted would happen if we continue as we are with no further easing of restrictions and assuming people stick to them.  In short, we will have COVID-for-Christmas and it could be a very nasty business indeed as it would come at the same time as other winter-associated infectious diseases such as influenza and norovirus.

The next chart shows what could happen if we squeeze the social distancing brake a bit harder by focusing only on the behaviours that the track-and-trace-and-test system is highlighting as the key drivers of the growth infections, admissions and deaths.

What we see is an arrest of the rise of the number of infectious people (as we saw before), a small and not sustained increase in hospital admissions, then a slow decline back to the levels that were achieved in early July – and at which point it would be reasonable to have a more normal Christmas.

And another potential benefit of a bit more social distancing might be a much less problematic annual flu epidemic because that virus would also find it harder to spread – plus we have a flu vaccination which we can use to reduce that risk further.


It is not going to be easy.  We will have to sacrifice a bit of face-to-face social life for a bit longer.  We will have to measure, monitor, model and tweak the plan as we go.

And one thing we can do immediately is to share the available information in a more informative and less histrionic way than we are seeing at the moment.


Update: Sunday 1st November 2020

Yesterday the Government had to concede that the policy of regional restrictions had failed and bluffing it out and ignoring the scientific advice was, with the clarity of hindsight, an unwise strategy.

In the face of the hard evidence of rapidly rising COVID+ve hospital admissions and deaths, the decision to re-impose a national 4-week lock-down was announced.  This is the only realistic option to prevent overwhelming the NHS at a time of year that it struggles with seasonal influenza causing a peak of admissions and deaths.

Paradoxically, this year the effect of influenza may be less because social distancing will reduce the spread of that as well and also because there is a vaccination for influenza.  Many will have had their flu jab early … I certainly did.

So, what is the predicted effect of a 4 week lock down?  Well, the calibrated model (also used to generate the charts above) estimates that it could indeed suppress the Second Wave and mitigate a nasty COVID-4-Christmas scenario.  But even with it the hospital admissions and associated mortality will continue to increase until the effect kicks in.

Brace yourselves.

Restoring Pride-in-Work

In 1986, Dr Don Berwick from Boston attended a 4-day seminar run by Dr W. Edwards Deming in Washington.  Dr Berwick was a 40 year old paediatrician who was also interested in health care management and improving quality and productivity.  Dr Deming was an 86 year old engineer and statistician who, when he was in his 40’s, helped the US to improve the quality and productivity of the industrial processes supporting the US and Allies in WWII.

Don Berwick describes attending the seminar as an emotionally challenging life-changing experience when he realised that his well-intended attempts to improve quality by inspection-and-correction was a counterproductive, abusive approach that led to fear, demotivation and erosion of pride-in-work.  His blinding new clarity of insight led directly to the Institute of Healthcare Improvement in the USA in the early 1990’s.

One of the tenets of Dr Deming’s theories is that the ingrained beliefs and behaviours that erode pride-in-work also lead to the very outcomes that management do not want – namely conflict between managers and workers and economic failure.

So, an explicit focus on improving pride-in-work as an early objective in any improvement exercise makes very good economic sense, and is a sign of wise leadership and competent management.


Last week a case study was published that illustrates exactly that principle in action.  The important message in the title is “restore the calm”.

One of the most demotivating aspects of health care that many complain about is the stress caused a chaotic environment, chronic crisis and perpetual firefighting.  So, anything that can restore calm will, in principle, improve motivation – and that is good for staff, patients and organisations.

The case study describes, in detail, how calm was restored in a chronically chaotic chemotherapy day unit … on Weds, June 19th 2019 … in one day and at no cost!

To say that the chemotherapy nurses were surprised and delighted is an understatement.  They were amazed to see that they could treat the same number of patients, with the same number of staff, in the same space and without the stress and chaos.  And they had time to keep up with the paperwork; and they had time for lunch; and they finished work 2 hours earlier than previously!

Such a thing was not possible surely? But here they were experiencing it.  And their patients noticed the flip from chaos-to-strangely-calm too.

The impact of the one-day-test was so profound that the nurses voted to adopt the design change the following week.  And they did.  And the restored calm has been sustained.


What happened next?

The chemotherapy nurses were able to catch up with their time-owing that had accumulated from the historical late finishes.  And the problem of high staff turnover and difficultly in recruitment evaporated.  Highly-trained chemotherapy nurses who had left because of the stressful chaos now want to come back.  Pride-in-work has been re-established.  There are no losers.  It is a win-win-win result for staff, patients and organisations.


So, how was this “miracle” achieved?

Well, first of all it was not a miracle.  The flip from chaos-to-calm was predicted to happen.  In fact, that was the primary objective of the design change.

So, how what this design change achieved?

By establishing the diagnosis first – the primary cause of the chaos – and it was not what the team believed it was.  And that is the reason they did not believe the design change would work; and that is the reason they were so surprised when it did.

So, how was the diagnosis achieved?

By using an advanced systems engineering technique called Complex Physical System (CPS) modelling.  That was the game changer!  All the basic quality improvement techniques had been tried and had not worked – process mapping, direct observation, control charts, respectful conversations, brainstorming, and so on.  The system structure was too complicated. The system behaviour was too complex (i.e. chaotic).

What CPS revealed was that the primary cause of the chaotic behaviour was the work scheduling policy.  And with that clarity of focus, the team were able to re-design the policy themselves using a simple paper-and-pen technique.  That is why it cost nothing to change.

So, why hadn’t they been able to do this before?

Because systems engineering is not a taught component of the traditional quality improvement offerings.  Healthcare is rather different to manufacturing! As the complexity of the health care system increases we need to learn the more advanced tools that are designed for this purpose.

What is the same is the principle of restoring pride-in-work and that is what Dr Berwick learned from Dr Deming in 1986, and what we saw happen on June 19th, 2019.

To read the story of how it was done click here.

Carveoutosis Multiforme Fulminans

This is the name given to an endemic, chronic, systemic, design disease that afflicts the whole NHS that very few have heard of, and even fewer understand.

This week marked two milestones in the public exposure of this elusive but eminently treatable health care system design illness that causes queues, delays, overwork, chaos, stress and risk for staff and patients alike.

The first was breaking news from the team in Swansea led by Chris Jones.

They had been grappling with the wicked problem of chronic queues, delays, chaos, stress, high staff turnover, and escalating costs in their Chemotherapy Day Unit (CDU) at the Singleton Hospital.

The breakthrough came earlier in the year when we used the innovative eleGANTT® system to measure and visualise the CDU chaos in real-time.

This rich set of data enabled us, for the first time, to apply a powerful systems engineering  technique called counterfactual analysis which revealed the primary cause of the chaos – the elusive and counter-intuitive design disease carvoutosis multiforme fulminans.

And this diagnosis implied that the chaos could be calmed quickly and at no cost.

But that news fell on slightly deaf ears because, not surprisingly, the CDU team were highly sceptical that such a thing was possible.

So, to convince them we needed to demonstrate the adverse effect of carveoutosis in a way that was easy to see.  And to do that we used some advanced technology: dice and tiddly winks.

The reaction of the CDU nurses was amazing.  As soon as they ‘saw’ it they clicked and immediately grasped how to apply it in their world.  They designed the change they needed to make in a matter of minutes.


But the proof-of-the-pudding-is-in-the eating and we arranged a one-day-test-of-change of their anti-carveout design.

The appointed day arrived, Wednesday 19th June.  The CDU nurses implemented their new design (which cost nothing to do).  Within an hour of the day starting they reported that the CDU was strangely calm.   And at the end of the day they reported that it had remained strangely calm all day; and that they had time for lunch; and that they had time to do all their admin as they went; and that they finished on time; and that the patients did not wait for their chemotherapy; and that the patients noticed the chaos-to-calm transformation too.

They treated just the same number of patients as usual with the same staff, in the same space and with the same equipment.  It cost nothing to make the change.

To say they they were surprised is an understatement!  They were so surprised and so delighted that they did not want to go back to the old design – but they had to because it was only a one-day-test-of-change.

So, on Thursday and Friday they reverted back to the carveoutosis design.  And the chaos returned.  That nailed it!  There was a riot!!  The CDU nurses refused to wait until later in the year to implement their new design and they voted unanimously to implement it from the following Monday.  And they did.  And calm was restored.


The second milestone happened on Thursday 11th July when we ran a Health Care Systems Engineering (HCSE) Masterclass on the very same topic … chronic systemic carveoutosis multiforme fulminans.

This time we used the dice and tiddly winks to demonstrate the symptoms, signs and the impact of treatment.  Then we explored the known pathophysiology of this elusive and endemic design disease in much more depth.

This is health care systems engineering in action.

It seems to work.

Warts-and-All

This week saw the publication of a landmark paper – one that will bring hope to many.  A paper that describes the first step of a path forward out of the mess that healthcare seems to be in.  A rational, sensible, practical, learnable and enjoyable path.


This week I also came across an idea that triggered an “ah ha” for me.  The idea is that the most rapid learning happens when we are making mistakes about half of the time.

And when I say ‘making a mistake’ I mean not achieving what we predicted we would achieve because that implies that our understanding of the world is incomplete.  In other words, when the world does not behave as we expect, we have an opportunity to learn and to improve our ability to make more reliable predictions.

And that ability is called wisdom.


When we get what we expect about half the time, and do not get what we expect about the other half of the time, then we have the maximum amount of information that we can use to compare and find the differences.

Was it what we did? Was it what we did not do? What are the acts and errors of commission and omission? What can we learn from those? What might we do differently next time? What would we expect to happen if we do?


And to explore this terrain we need to see the world as it is … warts and all … and that is the subject of the landmark paper that was published this week.


The context of the paper is improvement of cancer service delivery, and specifically of reducing waiting time from referral to first appointment.  This waiting is a time of extreme anxiety for patients who have suspected cancer.

It is important to remember that most people with suspected cancer do not have it, so most of the work of an urgent suspected cancer (USC) clinic is to reassure and to relieve the fear that the spectre of cancer creates.

So, the sooner that reassurance can happen the better, and for the unlucky minority who are diagnosed with cancer, the sooner they can move on to treatment the better.

The more important paragraph in the abstract is the second one … which states that seeing the system behaviour as it is, warts-and-all,  in near-real-time, allows us to learn to make better decisions of what to do to achieve our intended outcomes. Wiser decisions.

And the reason this is the more important paragraph is because if we can do that for an urgent suspected cancer pathway then we can do that for any pathway.


The paper re-tells the first chapter of an emerging story of hope.  A story of how an innovative and forward-thinking organisation is investing in building embedded capability in health care systems engineering (HCSE), and is now delivering a growing dividend.  Much bigger than the investment on every dimension … better safety, faster delivery, higher quality and more affordability. Win-win-win-win.

The only losers are the “warts” – the naysayers and the cynics who claim it is impossible, or too “wicked”, or too difficult, or too expensive.

Innovative reality trumps cynical rhetoric … and the full abstract and paper can be accessed here.

So, well done to Chris Jones and the whole team in ABMU.

And thank you for keeping the candle of hope alight in these dark, stormy and uncertain times for the NHS.

Congratulations Kate!

This week, it was my great pleasure to award the first Health Care Systems Engineering (HCSE) Level 2 Medal to Dr Kate Silvester, MBA, FRCOphth.

Kate is internationally recognised as an expert in health care improvement and over more than two decades has championed the adoption of improvement methods such as Lean and Quality Improvement in her national roles in the Modernisation Agency and then the NHS Institute for Innovation and Improvement.

Kate originally trained as a doctor and then left the NHS to learn manufacturing systems engineering with Lucas and Airbus.  Kate then brought these very valuable skills back with her into the NHS when she joined the Cancer Services Collaborative.

Kate is co-founder of the Journal of Improvement Science and over the last five years has been highly influential in the development of the Health Care Systems Engineering Programme – the first of its kind in the world that is designed by clinicians for clinicians.

The HCSE Programme is built on the pragmatic See One-Do Some-Teach Many principle of developing competence and confidence through being trained and coached by a more experienced practitioner while doing projects of increasing complexity and training and coaching others who are less experienced.

Competence is based on evidence-of-effectiveness, and Kate has achieved HCSE Level 2 by demonstrating that she can do HCSE and that she can teach and coach others how to do HCSE as well.

To illustrate, here is a recent FHJ paper that Kate has authored which illustrates the HCSE principles applied in practice in a real hospital.  This work was done as part of the Health Foundation’s Flow, Cost and Quality project that Kate led and recent evidence proves that the improvements have sustained and spread.  South Warwickshire NHS Foundation Trust is now one of the top-performing Trusts in the NHS.

More recently, Kate has trained and coached new practitioners in Exeter and North Devon who have delivered improvements and earned their HCSE 1 wings.

Congratulations Kate!

Filter-Pull versus Push-Carveout

It is November 2018, the clocks have changed back to GMT, the trick-and-treats are done, the fireworks light the night skies and spook the hounds, and the seasonal aisles in the dwindling number of high street stores are already stocked for Christmas.

I have been a bit quiet on the blog front this year but that is because there has been a lot happening behind the scenes and I have had to focus.

One output of is the recent publication of an article in Future Healthcare Journal on the topic of health care systems engineering (HCSE).  Click here to read the article and the rest of this excellent edition of FHJ that is dedicated to “systems”.

So, as we are back to the winter phase of the annual NHS performance cycle it is a good time to glance at the A&E Performance Radar and see who is doing well, and not-so-well.

Based on past experience, I was expecting Luton to be Top-of-the-Pops and so I was surprised (and delighted) to see that Barnsley have taken the lead.  And the chart shows that Barnsley has turned around a reasonable but sagging performance this year.

So I would be asking “What has happened at Barnsley that we can all learn from? What did you change and how did you know what and how to do that?

To be sure, Luton is still in the top three and it is interesting to explore who else is up there and what their A&E performance charts look like.

The data is all available for anyone with a web-browser to view – here.

For completeness, this is the chart for Luton, and we can see that, although the last point is lower than Barnsley, the performance-over-time is more consistent and less variable. So who is better?

NB. This is a meaningless question and illustrates the unhelpful tactic of two-point comparisons with others, and with oneself. The better question is “Is my design fit-for-purpose?”

The question I have for Luton is different. “How do you achieve this low variation and how do you maintain it? What can we all learn from you?”

And I have some ideas how they do that because in a recent HSJ interview they said “It is all about the filters“.


What do they mean by filters?

A filter is an essential component of any flow design if we want to deliver high safety, high efficiency, high effectiveness, and high productivity.  In other words, a high quality, fit-4-purpose design.

And the most important flow filters are the “upstream” ones.

The design of our upstream flow filters is critical to how the rest of the system works.  Get it wrong and we can get a spiralling decline in system performance because we can unintentionally trigger a positive feedback loop.

Queues cause delays and chaos that consume our limited resources.  So, when we are chasing cost improvement programme (CIP) targets using the “salami slicer” approach, and combine that with poor filter design … we can unintentionally trigger the perfect storm and push ourselves over the catastrophe cliff into perpetual, dangerous and expensive chaos.

If we look at the other end of the NHS A&E league table we can see typical examples that illustrate this pattern.  I have used this one only because it happens to be bottom this month.  It is not unique.

All other NHS trusts fall somewhere between these two extremes … stable, calm and acceptable and unstable, chaotic and unacceptable.

Most display the stable and chaotic combination – the “Zone of Perpetual Performance Pain”.

So what is the fundamental difference between the outliers that we can all learn from? The positive deviants like Barnsley and Luton, and the negative deviants like Blackpool.  I ask this because comparing the extremes is more useful than laboriously exploring the messy, mass-mediocrity in the middle.

An effective upstream flow filter design is a necessary component, but it is not sufficient. Triage (= French for sorting) is OK but it is not enough.  The other necessary component is called “downstream pull” and omitting that element of the design appears to be the primary cause of the chronic chaos that drags trusts and their staff down.

It is not just an error of omission though, the current design is an actually an error of commission. It is anti-pull; otherwise known as “push”.


This year I have been busy on two complicated HCSE projects … one in secondary care and the other in primary care.  In both cases the root cause of the chronic chaos is the same.  They are different systems but have the same diagnosis.  What we have revealed together is a “push-carveout” design which is the exact opposite of the “upstream-filter-plus-downstream-pull” design we need.

And if an engineer wanted to design a system to be chronically chaotic then it is very easy to do. Here is the recipe:

a) Set high average utilisation target of all resources as a proxy for efficiency to ensure everything is heavily loaded. Something between 80% and 100% usually does the trick.

b) Set a one-size-fits-all delivery performance target that is not currently being achieved and enforce it punitively.  Something like “>95% of patients seen and discharged or admitted in less than 4 hours, or else …”.

c) Divvy up the available resources (skills, time, space, cash, etc) into ring-fenced pots.

Chronic chaos is guaranteed.  The Laws of Physics decree it.


Unfortunately, the explanation of why this is the case is counter-intuitive, so it is actually better to experience it first, and then seek the explanation.  Reality first, reasoning second.

And, it is a bittersweet experience, so it needs to be done with care and compassion.

And that’s what I’ve been busy doing this year. Creating the experiences and then providing the explanations.  And if done gradually what then happens is remarkable and rewarding.

The FHJ article outlines one validated path to developing individual and organisational capability in health care systems engineering.

Seeing The Voice of the System

It is always a huge compliment to see an idea improved and implemented by inspired innovators.

Health care systems engineering (HCSE) brings together concepts from the separate domains of systems engineering and health care.  And one idea that emerged from this union is to regard the health care system as a living, evolving, adapting entity.

In medicine we have the concept of ‘vital signs’ … a small number of objective metrics that we can measure easily and quickly.  With these we can quickly assess the physical health of a patient and decide if we need to act, and when.

With a series of such measurements over time we can see the state of a patient changing … for better or worse … and we can use this to monitor the effect of our actions and to maintain the improvements we achieve.

For a patient, the five vital signs are conscious level, respiratory rate, pulse, blood pressure and temperature. To sustain life we must maintain many flows within healthy ranges and the most critically important is the flow of oxygen to every cell in the body.  Oxygen is carried by blood, so blood flow is critical.

So, what are the vital signs for a health care system where the flows are not oxygen and blood?  They are patients, staff, consumables, equipment, estate, data and cash.

The photograph shows a demonstration of a Vitals Dashboard for a part of the cancer care system in the ABMU health board in South Wales.  The inspirational innovators who created it are Imran Rao (left), Andy Jones (right) and Chris Jones (top left), and they are being supported by ABMU to do this as part of their HCSE training programme.

So well done guys … we cannot wait to hear how being better able to seeing the voice of your cancer system translates into improved care for patients, and improved working life for the dedicated NHS staff, and improved use of finite public resources.  Win-win-win.

The A.B.C.D.E. of Improvement

In medicine we use checklists as aide memoirs because they help us to avoid errors of omission, especially in an emergency when we are stressed and less able to think logically.

One that everyone learns if they do a First Aid course is A.B.C. and it stands for Airway, Breathing, Circulation.  It is designed to remind us what to do first because everything that follows depends on it, and then what to do next, and so on.  Avoiding the errors of omission improves outcomes.


In the world of improvement we are interested in change-for-the-better and there are many models of change that we can use to remind us not to omit necessary steps.

One of these is called the Six Steps model (or trans-theoretical model to use the academic title) and it is usually presented as a cycle starting with a state called pre-contemplation.

This change model arose from an empirical study of people who displayed addictive behaviours (e.g. smoking, drinking, drugs etc) and specifically, those who had overcome them without any professional assistance.

The researchers compared the stories from the successful self-healers with the accepted dogma for the management of addictions, and they found something very interesting.  The dogma advocated action, but the stories showed that there were some essential steps before action; steps that should not be omitted.  Specifically, the contemplation and determination steps.

If corrective actions were started too early then the success rate was low.  When the pre-action steps were added the success rate went up … a lot!


The first step is to raise awareness which facilitates a shift from pre-contemplation to contemplation.  The second step is to provide information that gradually increases the pros for change and at the same time gradually decreases the cons for change.

If those phases are managed skillfully then a tipping point is reached where the individual decides to make the change and moves themselves to the third step, the determination or planning phase.

Patience and persistence is required.  The contemplation phase can last a long time.  It is the phase of exploration, evidence and explanation. It is preparing the ground for change and can be summed up in one word: Study.

Often the trigger for determination (i.e. Plan) and then action (i.e. Do) is relatively small because when we are close to the tipping point it does not take much to nudge us to step across the line.


And there is an aide memoir we can use for this change cycle … one that is a bit easier to remember:

A = Awareness
B = Belief
C = Capability
D = Delivery
E = Excellence (+enjoyment, +evidence, +excitement, +engagement)

First we raise awareness of the issue.
Then we learn a solution is possible and that we can learn the know-how.
Then we plan the work.
Then we work the plan.
Then we celebrate what worked and learn from what did and what did not.

Experience shows that the process is not discrete and sequential and it cannot be project managed into defined time boxes.  Instead, it is a continuum and the phases overlap and blend from one to the next in a more fluid and adaptive way.


Raising awareness requires both empathy and courage because this issue is often treated as undiscussable, and even the idea of discussing it is undiscussable too. Taboo.

But for effective change we need to grasp the nettle, explore the current reality, and start the conversation.

The 85% Optimum Bed Occupancy Myth

A few years ago I had a rant about the dangers of the widely promoted mantra that 85% is the optimum average measured bed-occupancy target to aim for.

But ranting is annoying, ineffective and often counter-productive.

So, let us revisit this with some calm objectivity and disprove this Myth a step at a time.

The diagram shows the system of interest (SoI) where the blue box represents the beds, the coloured arrows are the patient flows, the white diamond is a decision and the dotted arrow is information about how full the hospital is (i.e. full/not full).

A new emergency arrives (red arrow) and needs to be admitted. If the hospital is not full the patient is moved to an empty bed (orange arrow), the medical magic happens, and some time later the patient is discharged (green arrow).  If there is no bed for the emergency request then we get “spillover” which is the grey arrow, i.e. the patient is diverted elsewhere (n.b. these are critically ill patients …. they cannot sit and wait).


This same diagram could represent patients trying to phone their GP practice for an appointment.  The blue box is the telephone exchange and if all the lines are busy then the call is dropped (grey arrow).  If there is a line free then the call is connected (orange arrow) and joins a queue (blue box) to be answered some time later (green arrow).

In 1917, a Danish mathematician/engineer called Agner Krarup Erlang was working for the Copenhagen Telephone Company and was grappling with this very problem: “How many telephone lines do we need to ensure that dropped calls are infrequent AND the switchboard operators are well utilised?

This is the perennial quality-versus-cost conundrum. The Value-4-Money challenge. Too few lines and the quality of the service falls; too many lines and the cost of the service rises.

Q: Is there a V4M ‘sweet spot” and if so, how do we find it? Trial and error?

The good news is that Erlang solved the problem … mathematically … and the not-so good news is that his equations are very scary to a non mathematician/engineer!  So this solution is not much help to anyone else.


Fortunately, we have a tool for turning scary-equations into easy-2-see-pictures; our trusty Excel spreadsheet. So, here is a picture called a heat-map, and it was generated from one of Erlang’s equations using Excel.

The Erlang equation is lurking in the background, safely out of sight.  It takes two inputs and gives one output.

The first input is the Capacity, which is shown across the top, and it represents the number of beds available each day (known as the space-capacity).

The second input is the Load (or offered load to use the precise term) which is down the left side, and is the number of bed-days required per day (e.g. if we have an average of 10 referrals per day each of whom would require an average 2-day stay then we have an average of 10 x 2 = 20 bed-days of offered load per day).

The output of the Erlang model is the probability that a new arrival finds all the beds are full and the request for a bed fails (i.e. like a dropped telephone call).  This average probability is displayed in the cell.  The colour varies between red (100% failure) and green (0% failure), with an infinite number of shades of red-yellow-green in between.

We can now use our visual heat-map in a number of ways.

a) We can use it to predict the average likelihood of rejection given any combination of bed-capacity and average offered load.

Suppose the average offered load is 20 bed-days per day and we have 20 beds then the heat-map says that we will reject 16% of requests … on average (bottom left cell).  But how can that be? Why do we reject any? We have enough beds on average! It is because of variation. Requests do not arrive in a constant stream equal to the average; there is random variation around that average.  Critically ill patients do not arrive at hospital in a constant stream; so our system needs some resilience and if it does not have it then failures are inevitable and mathematically predictable.

b) We can use it to predict how many beds we need to keep the average rejection rate below an arbitrary but acceptable threshold (i.e. the quality specification).

Suppose the average offered load is 20 bed-days per day, and we want to have a bed available more than 95% of the time (less than 5% failures) then we will need at least 25 beds (bottom right cell).

c) We can use it to estimate the maximum average offered load for a given bed-capacity and required minimum service quality.

Suppose we have 22 beds and we want a quality of >=95% (failure <5%) then we would need to keep the average offered load below 17 bed-days per day (i.e. by modifying the demand and the length of stay because average load = average demand * average length of stay).


There is a further complication we need to be mindful of though … the measured utilisation of the beds is related to the successful admissions (orange arrow in the first diagram) not to the demand (red arrow).  We can illustrate this with a complementary heat map generated in Excel.

For scenario (a) above we have an offered load of 20 bed-days per day, and we have 20 beds but we will reject 16% of requests so the accepted bed load is only 16.8 bed days per day  (i.e. (100%-16%) * 20) which is the reason that the average  utilisation is only 16.8/20 = 84% (bottom left cell).

For scenario (b) we have an offered load of 20 bed-days per day, and 25 beds and will only reject 5% of requests but the average measured utilisation is not 95%, it is only 76% because we have more beds (the accepted bed load is 95% * 20 = 19 bed-days per day and 19/25 = 76%).

For scenario (c) the average measured utilisation would be about 74%.


So, now we see the problem more clearly … if we blindly aim for an average, measured, bed-utilisation of 85% with the untested belief that it is always the optimum … this heat-map says it is impossible to achieve and at the same time offer an acceptable quality (>95%).

We are trading safety for money and that is not an acceptable solution in a health care system.


So where did this “magic” value of 85% come from?

From the same heat-map perhaps?

If we search for the combination of >95% success (<5% fail) and 85% average bed-utilisation then we find it at the point where the offered load reaches 50 bed-days per day and we have a bed-capacity of 56 beds.

And if we search for the combination of >99% success (<1% fail) and 85% average utilisation then we find it with an average offered load of just over 100 bed-days per day and a bed-capacity around 130 beds.

H’mm.  “Houston, we have a problem“.


So, even in this simplified scenario the hypothesis that an 85% average bed-occupancy is a global optimum is disproved.

The reality is that the average bed-occupancy associated with delivering the required quality for a given offered load with a specific number of beds is almost never 85%.  It can range anywhere between 50% and 100%.  Erlang knew that in 1917.


So, if a one-size-fits-all optimum measured average bed-occupancy assumption is not valid then how might we work out how many beds we need and predict what the expected average occupancy will be?

We would design the fit-4-purpose solution for each specific context …
… and to do that we need to learn the skills of complex adaptive system design …
… and that is part of the health care systems engineering (HCSE) skill-set.

 

Cognitive Traps for Hefalumps

One of the really, really cool things about the 1.3 kg of “ChimpWare” between our ears is the way it learns.

We have evolved the ability to predict the likely near-future based on just a small number of past experiences.

And we do that by creating stored mental models.

Not even the most powerful computers can do it as well as we do – and we do it without thinking. Literally. It is an unconscious process.

This ability to pro-gnose (=before-know) gave our ancestors a major survival advantage when we were wandering about on the savanna over 10 million years ago.  And we have used this amazing ability to build societies, mega-cities and spaceships.


But this capability is not perfect.  It has a flaw.  Our “ChimpOS” does not store a picture of reality like a digital camera; it stores a patchy and distorted perception of reality, and then fills in the gaps with guesses (i.e. gaffes).  And we do not notice – consciously.

The cognitive trap is set and sits waiting to be sprung.  And to trip us up.


Here is an example:

“Improvement implies change”

Yes. That is a valid statement because we can show that whenever improvement has been the effect, then some time before that a change happened.  And we can show that when there are no changes, the system continues to behave as it always has.  Status quo.

The cognitive trap is that our ChimpOS is very good at remembering temporal associations – for example an association between “improvement” and “change” because we remember in the present.  So, if two concepts are presented at the same time, and we spice-the-pie with a bit of strong emotion, then we are more likely to associate them. Which is OK.

The problem comes when we play back the memory … it can come back as …

“change implies improvement” which is not valid.  And we do not notice.

To prove it is not valid we just need to find one example where a change led to a deterioration; an unintended negative consequence, a surprising, confusing and disappointing failure to achieve our intended improvement.

An embarrassing gap between our intent and our impact.

And finding that evidence is not hard.  Failures and disappointments in the world of improvement are all too common.


And then we can fall into the same cognitive trap because we generalise from a single, bad experience and the lesson our ChimpOS stores for future reference is “change is bad”.

And forever afterwards we feel anxious whenever the idea of change is suggested.

It is a very effective survival tactic – for a hominid living on the African savanna 10 million years ago, and at risk of falling prey to sharp-fanged, hungry predators.  It is a less useful tactic in the modern world where the risk of being eaten-for-lunch is minimal, and where the pace of change is accelerating.  We must learn to innovate and improve to survive in the social jungle … and we are not well equipped!


Here is another common cognitive trap:

Excellence implies no failures.

Yes. If we are delivering a consistently excellent service then the absence of failures will be a noticeable feature.

No failures implies excellence.

This is not a valid inference.  If quality-of-service is measured on a continuum from Excrement-to-Excellent, then we can be delivering a consistently mediocre service, one that is barely adequate, and also have no failures.


The design flaw here is that our ChimpWare/ChimpOS memory system is lossy.

We do not remember all the information required to reconstruct an accurate memory of reality – because there is too much information.  So we distort, we delete and we generalise.  And we do that because when we evolved it was a good enough solution, and it enabled us to survive as a species, so the ChimpWare/ChimpOS genes were passed on.

We cannot reverse millions of years of evolution.  We cannot get a wetware or a software upgrade.  We need to learn to manage with the limitations of what we have between our ears.

And to avoid the cognitive traps we need to practice the discipline of bringing our unconscious assumptions up to conscious awareness … and we do that by asking carefully framed questions.

Here is another example:

A high-efficiency design implies high-utilisation of resources.

Yes, that is valid. Idle resources means wasted resources which means lower efficiency.

Q1: Is the converse also valid?
Q2: Is there any evidence that disproves the converse is valid?

If high-utilisation does not imply high-efficiency, what are the implications of falling into this cognitive trap?  What is the value of measuring utilisation? Does it have a value?

These are useful questions.

The Strangeness of LoS

It had been some time since Bob and Leslie had chatted so an email from the blue was a welcome distraction from a complex data analysis task.

<Bob> Hi Leslie, great to hear from you. I was beginning to think you had lost interest in health care improvement-by-design.

<Leslie> Hi Bob, not at all.  Rather the opposite.  I’ve been very busy using everything that I’ve learned so far.  It’s applications are endless, but I have hit a problem that I have been unable to solve, and it is driving me nuts!

<Bob> OK. That sounds encouraging and interesting.  Would you be able to outline this thorny problem and I will help if I can.

<Leslie> Thanks Bob.  It relates to a big issue that my organisation is stuck with – managing urgent admissions.  The problem is that very often there is no bed available, but there is no predictability to that.  It feels like a lottery; a quality and safety lottery.  The clinicians are clamoring for “more beds” but the commissioners are saying “there is no more money“.  So the focus has turned to reducing length of stay.

<Bob> OK.  A focus on length of stay sounds reasonable.  Reducing that can free up enough beds to provide the necessary space-capacity resilience to dramatically improve the service quality.  So long as you don’t then close all the “empty” beds to save money, or fall into the trap of believing that 85% average bed occupancy is the “optimum”.

<Leslie> Yes, I know.  We have explored all of these topics before.  That is not the problem.

<Bob> OK. What is the problem?

<Leslie> The problem is demonstrating objectively that the length-of-stay reduction experiments are having a beneficial impact.  The data seems to say they they are, and the senior managers are trumpeting the success, but the people on the ground say they are not. We have hit a stalemate.


<Bob> Ah ha!  That old chestnut.  So, can I first ask what happens to the patients who cannot get a bed urgently?

<Leslie> Good question.  We have mapped and measured that.  What happens is the most urgent admission failures spill over to commercial service providers, who charge a fee-per-case and we have no choice but to pay it.  The Director of Finance is going mental!  The less urgent admission failures just wait on queue-in-the-community until a bed becomes available.  They are the ones who are complaining the most, so the Director of Governance is also going mental.  The Director of Operations is caught in the cross-fire and the Chief Executive and Chair are doing their best to calm frayed tempers and to referee the increasingly toxic arguments.

<Bob> OK.  I can see why a “Reduce Length of Stay Initiative” would tick everyone’s Nice If box.  So, the data analysts are saying “the length of stay has come down since the Initiative was launched” but the teams on the ground are saying “it feels the same to us … the beds are still full and we still cannot admit patients“.

<Leslie> Yes, that is exactly it.  And everyone has come to the conclusion that demand must have increased so it is pointless to attempt to reduce length of stay because when we do that it just sucks in more work.  They are feeling increasingly helpless and hopeless.

<Bob> OK.  Well, the “chronic backlog of unmet need” issue is certainly possible, but your data will show if admissions have gone up.

<Leslie> I know, and as far as I can see they have not.

<Bob> OK.  So I’m guessing that the next explanation is that “the data is wonky“.

<Leslie> Yup.  Spot on.  So, to counter that the Information Department has embarked on a massive push on data collection and quality control and they are adamant that the data is complete and clean.

<Bob> OK.  So what is your diagnosis?

<Leslie> I don’t have one, that’s why I emailed you.  I’m stuck.


<Bob> OK.  We need a diagnosis, and that means we need to take a “history” and “examine” the process.  Can you tell me the outline of the RLoS Initiative.

<Leslie> We knew that we would need a baseline to measure from so we got the historical admission and discharge data and plotted a Diagnostic Vitals Chart®.  I have learned something from my HCSE training!  Then we planned the implementation of a visual feedback tool that would show ward staff which patients were delayed so that they could focus on “unblocking” the bottlenecks.  We then planned to measure the impact of the intervention for three months, and then we planned to compare the average length of stay before and after the RLoS Intervention with a big enough data set to give us an accurate estimate of the averages.  The data showed a very obvious improvement, a highly statistically significant one.

<Bob> OK.  It sounds like you have avoided the usual trap of just relying on subjective feedback, and now have a different problem because your objective and subjective feedback are in disagreement.

<Leslie> Yes.  And I have to say, getting stuck like this has rather dented my confidence.

<Bob> Fear not Leslie.  I said this is an “old chestnut” and I can say with 100% confidence that you already have what you need in your T4 kit bag?

<Leslie>Tee-Four?

<Bob> Sorry, a new abbreviation. It stands for “theory, techniques, tools and training“.

<Leslie> Phew!  That is very reassuring to hear, but it does not tell me what to do next.

<Bob> You are an engineer now Leslie, so you need to don the hard-hat of Improvement-by-Design.  Start with your Needs Analysis.


<Leslie> OK.  I need a trustworthy tool that will tell me if the planned intervention has has a significant impact on length of stay, for better or worse or not at all.  And I need it to tell me that quickly so I can decide what to do next.

<Bob> Good.  Now list all the things that you currently have that you feel you can trust.

<Leslie> I do actually trust that the Information team collect, store, verify and clean the raw data – they are really passionate about it.  And I do trust that the front line teams are giving accurate subjective feedback – I work with them and they are just as passionate.  And I do trust the systems engineering “T4” kit bag – it has proven itself again-and-again.

<Bob> Good, and I say that because you have everything you need to solve this, and it sounds like the data analysis part of the process is a good place to focus.

<Leslie> That was my conclusion too.  And I have looked at the process, and I can’t see a flaw. It is driving me nuts!

<Bob> OK.  Let us take a different tack.  Have you thought about designing the tool you need from scratch?

<Leslie> No. I’ve been using the ones I already have, and assume that I must be using them incorrectly, but I can’t see where I’m going wrong.

<Bob> Ah!  Then, I think it would be a good idea to run each of your tools through a verification test and check that they are fit-4-purpose in this specific context.

<Leslie> OK. That sounds like something I haven’t covered before.

<Bob> I know.  Designing verification test-rigs is part of the Level 2 training.  I think you have demonstrated that you are ready to take the next step up the HCSE learning curve.

<Leslie> Do you mean I can learn how to design and build my own tools?  Special tools for specific tasks?

<Bob> Yup.  All the techniques and tools that you are using now had to be specified, designed, built, verified, and validated. That is why you can trust them to be fit-4-purpose.

<Leslie> Wooohooo! I knew it was a good idea to give you a call.  Let’s get started.


[Postscript] And Leslie, together with the other stakeholders, went on to design the tool that they needed and to use the available data to dissolve the stalemate.  And once everyone was on the same page again they were able to work collaboratively to resolve the flow problems, and to improve the safety, flow, quality and affordability of their service.  Oh, and to know for sure that they had improved it.

The Turkeys Voting For Xmas Trap

One of the quickest and easiest ways to kill an improvement initiative stone dead is to label it as a “cost improvement program” or C.I.P.

Everyone knows that the biggest single contributor to cost is salaries.

So cost reduction means head count reduction which mean people lose their jobs and their livelihood.

Who is going to sign up to that?

It would be like turkeys voting for Xmas.

There must be a better approach?

Yes. There is.


Over the last few weeks, groups of curious skeptics have experienced the immediate impact of systems engineering theory, techniques and tools in a health care context.

They experienced queues, delays and chaos evaporate in front of their eyes … and it cost nothing to achieve. No extra resources. No extra capacity. No extra cash.

Their reaction was “surprise and delight”.

But … it also exposed a problem.  An undiscussable problem.


Queues and chaos require expensive resources to manage.

We call them triagers, progress-chasers, and fire-fighters.  And when the queues and chaos evaporate then their jobs do too.

The problem is that the very people who are needed to make the change happen are the ones who become surplus-to-requirement as a result of the change.

So change does not happen.

It would like turkeys voting for Xmas.


The way around this impasse is to anticipate the effect and to proactively plan to re-invest the resource that is released.  And to re-invest it doing a more interesting and more worthwhile jobs than queue-and-chaos management.

One opportunity for re-investment is called time-buffering which is an effective way to improve resilience to variation, especially in an unscheduled care context.

Another opportunity for re-investment is tail-gunning the chronic backlogs until they are down to a safe and sensible size.

And many complain that they do not have time to learn about improvement because they are too busy managing the current chaos.

So, another opportunity for re-investment is training – oneself first and then others.


R.I.P.    C.I.P.

The Disbelief to Belief Transition

The NHS appears to be descending in a frenzy of fear as the winter looms and everyone says it will be worse than last and the one before that.

And with that we-are-going-to-fail mindset, it almost certainly will.

Athletes do not start a race believing that they are doomed to fail … they hold a belief that they can win the race and that they will learn and improve even if they do not. It is a win-win mindset.

But to succeed in sport requires more than just a positive attitude.

It also requires skills, training, practice and experience.

The same is true in healthcare improvement.


That is not the barrier though … the barrier is disbelief.

And that comes from not having experienced what it is like to take a system that is failing and transform it into one that is succeeding.

Logically, rationally, enjoyably and surprisingly quickly.

And, the widespread disbelief that it is possible is paradoxical because there are plenty of examples where others have done exactly that.

The disbelief seems to be “I do not believe that will work in my world and in my hands!

And the only way to dismantle that barrier-of-disbelief is … by doing it.


How do we do that?

The emotionally safest way is in a context that is carefully designed to enable us to surface the unconscious assumptions that are the bricks in our individual Barriers of Disbelief.

And to discard the ones that do not pass a Reality Check, and keep the ones that are OK.

This Disbelief-Busting design has been proven to be effective, as evidenced by the growing number of individuals who are learning how to do it themselves, and how to inspire, teach and coach others to as well.


So, if you would like to flip disbelief-and-hopeless into belief-and-hope … then the door is here.

Ability minus Awareness equals Engagement

It is always rewarding when separate but related ideas come together and go “click”.

And this week I had one of those “ah ha” moments while attempting to explain how the process of engagement works.

Many years ago I was introduced to the conscious-competence model of learning which I found really insightful.  Sometime later I renamed it as the awareness-ability model because the term competence felt too judgmental.

The idea is that when we learn we all start from a position of being unaware of our inability.

A state called blissful ignorance.

And it is only when we try to do something that we become aware of what we cannot do; which can lead to temper tantrums!

As we concentrate and practice our ability improves and we enter the zone of know how.  We become able to demonstrate what we can do, and explain how we are doing it.

The final phase comes when it becomes so habitual that we forget how we learned our skill – it has become second nature.


Some years later I was introduced to the Nerve Curve which is the emotional roller-coaster ride that accompanies change.  Any form of change.

A five-step model was described in the context of bereavement by psychiatrist Elisabeth Kübler-Ross in her 1969 book “On Death & Dying: What the Dying Have to Teach Doctors, Nurses, Clergy and their Families.

More recently this has been extended and applied by authors such as William Bridges and John Fisher in the less emotionally traumatic contexts called transitions.

The characteristic sequence of emotions are triggered by external events are:

  • shock
  • denial
  • frustration
  • blame
  • guilt
  • depression
  • acceptance
  • engagement
  • excitement.

The important messages in both of these models is that we can get stuck along the path of transition, and we can disengage at several points, signalling to others that we have come off the track.  When we do that we exhibit behaviours such as denial, disillusionment and hostility.


More recently I was introduced to the work of the late Chris Argyris and specifically the concept of “defensive reasoning“.

The essence of the concept:  As we start to become aware of a gap between our intentions and our impact, then we feel threatened and our natural reaction is defensive.  This is the essence of the behaviour called “resistance to change”, and it is interesting to note that “smart” people are particularly adept at it.


These three concepts are clearly related in some way … but how?


As a systems engineer I am used to cyclical processes and the concepts of wavelength, amplitude, phase and offset, and I found myself looking at the Awareness-Ability cycle and asking:

“How could that cycle generate the characteristic shape of the transition curve?”

Then the Argyris idea of the gap between intent and impact popped up and triggered another question:

“What if we look at the gap between our ability and our awareness?”

So, I conducted a thought experiment and imagined myself going around the cycle – and charting my ability, awareness and emotional state along the way … and this sketch emerged. Ah ha!

When my awareness exceeded my ability I felt disheartened. That is the defensive reasoning that Chris Argyris talks about, the emotional barrier to self-improvement.


Ability – Awareness = Engagement


This suggested to me that the process of building self-engagement requires opening the ability-versus-awareness gap a little-bit-at-a-time, sensing the emotional discomfort, and then actively releasing the tension by learning a new concept, principle, technique or tool (and usually all four).

Eureka!

I wonder if the same strategy would work elsewhere?

Evidence-Based Co-Design

The first step in a design conversation is to understand the needs of the customer.

It does not matter if you are designing a new kitchen, bathroom, garden, house, widget, process, or system.  It is called a “needs analysis”.

Notice that it is not called a “wants analysis”.  They are not the same thing because there is often a gap between what we want (and do not want) and what we need (and do not need).

The same is true when we are looking to use a design-based approach to improve something that we already have.


This is especially true when we are improving services because the the needs and wants of a service tend to drift and shift continuously, and we are in a continual state of improvement.

For design to work the “customers” and the “suppliers” need work collaboratively to ensure that they both get what they need.

Frustration and fragmentation are the symptoms of a combative approach where a “win” for one is a “lose” for the other (NB. In absolute terms both will end up worse off than they started so both lose in the long term.)


And there is a tried and tested process to collaborative improvement-by-design.

One version is called “experience based co-design” (EBCD) and it was cooked up in a health care context about 20 years ago and shown to work in a few small pilot studies.

The “experience” that triggered the projects was almost always a negative one and was associated with feelings of frustration, anxiety and disappointment. So, the EBCD case studies were more focused on helping the protagonists to share their perspectives, in the belief that will be enough to solve the problem.  And it is indeed a big step forwards.

It has a limitation though.  It assumes that the staff and patients know how to design processes so that they are fit-4-purpose, and the evidence to support that assumption is scanty.

In one pilot in mental health, the initial improvement (a fall in patient and carer complaints) was not sustained.  The reason given was that the staff who were involved in the pilot inevitably moved on, and as they did the old attitudes, beliefs and behaviours returned.


So, an improved version of EBCD is needed.  One that is based on hard evidence of what works and what does not.  One that is also focused on moving towards a future-purpose rather than just moving away from past-problems.

Let us call this improved version “Evidence-Based Co-Design“.

And we already know that by a different name:

Health Care Systems Engineering (HCSE).

O.O.D.A.

OODA is something we all do thousands of times a day without noticing.

Observe – Orient – Decide – Act.

The term is attributed to Colonel John Boyd, a real world “Top Gun” who studied economics and engineering, then flew and designed fighter planes, then became a well-respected military strategist.

OODA is a continuous process of updating our mental model based on sensed evidence.

And it is a fast process because happens largely out of awareness.

This was Boyd’s point: In military terms, the protagonist that can make wiser and faster decisions are more likely to survive in combat.


And notice that it is not a simple linear sequence … it is a system … there are parallel paths and both feed-forward and feed-backward loops … there are multiple information flow paths.

And notice that the Implicit Guidance & Control links do not go through Decision – this means they operate out of awareness and are much faster.

And notice the Feed Forward links link the OODA steps – this is the conscious, sequential, future looking process that we know by another name:

Study-Adjust-Plan-Do.


We use the same process in medicine: first we study the patient and the problem they are presenting (history, examination, investigation), then we adjust our generic mental model of how the body works to the specific patient (diagnosis), then we plan and decide a course of action to achieve the intended outcome, and then we act, we do it (treatment).

But at any point we can jump back to an earlier step and we can jump forwards to a later one.  The observe, orient, decide, act modes are running in parallel.

And the more experience we have of similar problems the faster we can complete the OODA (or SAPD) work because we learn what is the most useful information to attend to, and we learn how to interpret it.

We learn the patterns and what to look for – and that speeds up the process – a lot!


This emergent learning is then re-inforced if the impact of our action matches our intent and prediction and our conscious learning is then internalised as unconscious “rules of thumb” called heuristics.


We start by thinking our way consciously and slowly … and … we finish by feeling our way unconsciously and quickly.


Until … we  encounter a novel problem that does not fit any of our learned pattern matching neural templates. When that happens, our unconscious, parallel processing, pattern-matching system alerts us with a feeling of confusion and bewilderment – and we freeze (often with fright!)

Now we have a choice: We can retreat to using familiar, learned, reactive, knee-jerk patterns of behaviour (presumably in the hope that they will work) or we can switch into a conscious learning loop and start experimenting with novel ideas.

If we start at Hypothesis then we have the Plan-Do-Study-Act cycle; where we generate novel hypotheses to explain the unexpected, and we then plan experiments to test our hypotheses; and we then study the outcome of the experiments and we then we act on our conclusions.

This mindful mode of thinking is well described in the book “Managing the Unexpected” by Weick and Sutcliffe and is the behaviour that underpins the success of HROs – High Reliability Organisations.

The image is of the latest (3rd edition) but the previous (2nd edition) is also worth reading.

So we have two interdependent problem solving modes – the parallel OODA system and the sequential SAPD process.

And we can switch between them depending on the context.


Which is an effective long-term survival strategy because the more we embrace the unexpected, the more opportunities we will have to switch into exploration mode and learn new patterns; and the more patterns we recognise the more efficient and effective our unconscious decision-making process will become.

This complex adaptive system behaviour has another name … Resilience.

Unknown-Knowns

This is the now-infamous statement that Donald Rumsfeld made at a Pentagon Press Conference which triggered some good-natured jesting from the assembled journalists.

But there is a problem with it.

There is a fourth combination that he does not mention: the Unknown-Knowns.

Which is a shame because they are actually the most important because they cause the most problems.  Avoidable problems.


Suppose there is a piece of knowledge that someone knows but that someone else does not; then we have an unknown-known.

None of us know everything and we do not need to, because knowledge that is of no value to us is irrelevant for us.

But what happens when the unknown-known is of value to us, and more than that; what happens when it would be reasonable for someone else to expect us to know it; because it is our job to know.


A surgeon would be not expected to know a lot about astronomy, but they would be expected to know a lot about anatomy.


So, what happens if we become aware that we are missing an important piece of knowledge that is actually already known?  What is our normal human reaction to that discovery?

Typically, our first reaction is fear-driven and we express defensive behaviour.  This is because we fear the potential loss-of-face from being exposed as inept.

From this sudden shock we then enter a characteristic emotional pattern which is called the Nerve Curve.

After the shock of discovery we quickly flip into denial and, if that does not work then to anger (i.e. blame).  We ignore the message and if that does not work we shoot the messenger.


And when in this emotionally charged state, our rationality tends to take a back seat.  So, if we want to benefit from the discovery of an unknown-known, then we have to learn to bite-our-lip, wait, let the red mist dissipate, and then re-examine the available evidence with a cool, curious, open mind.  A state of mind that is receptive and open to learning.


Recently, I was reminded of this.


The context is health care improvement, and I was using a systems engineering framework to conduct some diagnostic data analysis.

My first task was to run a data-completeness-verification-test … and the data I had been sent did not pass the test.  There was some missing.  It was an error of omission (EOO) and they are the hardest ones to spot.  Hence the need for the verification test.

The cause of the EOO was an unknown-known in the department that holds the keys to the data warehouse.  And I have come across this EOO before, so I was not surprised.

Hence the need for the verification test.

I was not annoyed either.  I just fed back the results of the test, explained what the issue was, explained the cause, and they listened and learned.


The implication of this specific EOO is quite profound though because it appears to be ubiquitous across the NHS.

To be specific it relates to the precise details of how raw data on demand, activity, length of stay and bed occupancy is extracted from the NHS data warehouses.

So it is rather relevant to just about everything the NHS does!

And the error-of-omission leads to confusion at best; and at worst … to the following sequence … incomplete data =>  invalid analysis => incorrect conclusion => poor decision => counter-productive action => unintended outcome.

Does that sound at all familiar?


So, if would you like to learn about this valuable unknown-known is then I recommend the narrative by Dr Kate Silvester, an internationally recognised expert in healthcare improvement.  In it, Kate re-tells the story of her emotional roller-coaster ride when she discovered she was making the same error.


Here is the link to the full abstract and where you can download and read the full text of Kate’s excellent essay, and help to make it a known-known.

That is what system-wide improvement requires – sharing the knowledge.

Catch-22

There is a Catch-22 in health care improvement and it goes a bit like this:

Most people are too busy fire-fighting the chronic chaos to have time to learn how to prevent the chaos, so they are stuck.

There is a deeper Catch-22 as well though:

The first step in preventing chaos is to diagnose the root cause and doing that requires experience, and we don’t have that experience available, and we are too busy fire-fighting to develop it.


Health care is improvement science in action – improving the physical and psychological health of those who seek our help. Patients.

And we have a tried-and-tested process for doing it.

First we study the problem to arrive at a diagnosis; then we design alternative plans to achieve our intended outcome and we decide which plan to go with; and then we deliver the plan.

Study ==> Plan ==> Do.

Diagnose  ==> Design & Decide ==> Deliver.

But here is the catch. The most difficult step is the first one, diagnosis, because there are many different illnesses and they often present with very similar patterns of symptoms and signs. It is not easy.

And if we make a poor diagnosis then all the action plans that follow will be flawed and may lead to disappointment and even harm.

Complaints and litigation follow in the wake of poor diagnostic ability.

So what do we do?

We defer reassuring our patients, we play safe, we request more tests and we refer for second opinions from specialists. Just to be on the safe side.

These understandable tactics take time, cost money and are not 100% reliable.  Diagnostic tests are usually precisely focused to answer specific questions but can have false positive and false negative results.

To request a broad batch of tests in the hope that the answer will appear like a rabbit out of a magician’s hat is … mediocre medicine.


This diagnostic dilemma arises everywhere: in primary care and in secondary care, and in non-urgent and urgent pathways.

And it generates extra demand, more work, bigger queues, longer delays, growing chaos, and mounting frustration, disappointment, anxiety and cost.

The solution is obvious but seemingly impossible: to ensure the most experienced diagnostician is available to be consulted at the start of the process.

But that must be impossible because if the consultants were seeing the patients first, what would everyone else do?  How would they learn to become more expert diagnosticians? And would we have enough consultants?


When I was a junior surgeon I had the great privilege to have the opportunity to learn from wise and experienced senior surgeons, who had seen it, and done it and could teach it.

Mike Thompson is one of these.  He is a general surgeon with a special interest in the diagnosis and treatment of bowel cancer.  And he has a particular passion for improving the speed and accuracy of the diagnosis step; because it can be a life-saver.

Mike is also a disruptive innovator and an early pioneer of the use of endoscopy in the outpatient clinic.  It is called point-of-care testing nowadays, but in the 1980’s it was a radically innovative thing to do.

He also pioneered collecting the symptoms and signs from every patient he saw, in a standard way using a multi-part printed proforma. And he invested many hours entering the raw data into a computer database.

He also did something that even now most clinicians do not do; when he knew the outcome for each patient he entered that into his database too – so that he could link first presentation with final diagnosis.


Mike knew that I had an interest in computer-aided diagnosis, which was a hot topic in the early 1980’s, and also that I did not warm to the Bayesian statistical models that underpinned it.  To me they made too many simplifying assumptions.

The human body is a complex adaptive system. It defies simplification.

Mike and I took a different approach.  We  just counted how many of each diagnostic group were associated with each pattern of presenting symptoms and signs.

The problem was that even his database of 8000+ patients was not big enough! This is why others had resorted to using statistical simplifications.

So we used the approach that an experienced diagnostician uses.  We used the information we had already gleaned from a patient to decide which question to ask next, and then the next one and so on.


And we always have three pieces of information at the start – the patient’s age, gender and presenting symptom.

What surprised and delighted us was how easy it was to use the database to help us do this for the new patients presenting to his clinic; the ones who were worried that they might have bowel cancer.

And what surprised us even more was how few questions we needed to ask arrive at a statistically robust decision to reassure-or-refer for further tests.

So one weekend, I wrote a little computer program that used the data from Mike’s database and our simple bean-counting algorithm to automate this process.  And the results were amazing.  Suddenly we had a simple and reliable way of using past experience to support our present decisions – without any statistical smoke-and-mirror simplifications getting in the way.

The computer program did not make the diagnosis, we were still responsible for that; all it did was provide us with reliable access to a clear and comprehensive digital memory of past experience.


What it then enabled us to do was to learn more quickly by exploring the complex patterns of symptoms, signs and outcomes and to develop our own diagnostic “rules of thumb”.

We learned in hours what it would take decades of experience to uncover. This was hot stuff, and when I presented our findings at the Royal Society of Medicine the audience was also surprised and delighted (and it was awarded the John of Arderne Medal).

So, we called it the Hot Learning System, and years later I updated it with Mike’s much bigger database (29,000+ records) and created a basic web-based version of the first step – age, gender and presenting symptom.  You can have a play if you like … just click HERE.


So what are the lessons here?

  1. We need to have the most experienced diagnosticians at the start of the improvement process.
  2. The first diagnostic assessment can be very quick so long as we have developed evidence-based heuristics.
  3. We can accelerate the training in diagnostic skills using simple information technology and basic analysis techniques.

And exactly the same is true in the health care system improvement.

We need to have an experienced health care improvement practitioner involved at the start, because if we skip this critical study step and move to plan without a correct diagnosis, then we will make errors, poor decisions, and counter-productive actions.  And then generate more work, more queues, more delays, more chaos, more distress and increased costs.

Exactly the opposite of what we want.

Q1: So, how do we develop experienced improvement practitioners more quickly?

Q2: Is there a hot learning system for improvement science?

A: Yes, there is. It can be found here.

The Marmite Effect

Have you heard the phrase “you either love it or you hate it“?  It is called the Marmite Effect.

Improvement science has Marmite-like effect on some people, or more specifically, the theory part does.

Both evidence and experience show that most people prefer to learn-by-doing first; and then consolidate their learning with the minimum, necessary amount of supporting theory.

But that is not how we usually share what we know with others.  We usually attempt to teach the theory first, perhaps in the belief that it will speed up the process of learning.

Sadly, it usually has the opposite effect. Too much theory too soon often creates a barrier to engagement. It actually slows learning down! Which was not the impact we were intending.


The implications of this is that teachers of the science of improvement need to provide a range of different ways to engage with the subject.  Complementary ways.  And leave the choice of which suits whom … to the learner.

And the way to tell if it is working is … the sound of laughter.

Why is that?


Laughing is a complex behaviour that leaves us feeling happier. Which is good.

Comedians make a living from being able to trigger this behaviour in their audiences, and we will gladly part with hard cash when we know something will make us feel better.

And laughing is one of the healthiest ways to feel better!

So why do we laugh when we are learning?

It is believed that one trigger for the laughter reaction is the sudden shift from one perspective to another.  More specifically, a mental shift that relieves a growing emotional tension.  The punch line of a really good joke for example.

And later-in-life learning is often more a process of unlearning.

When we challenge a learned assumption with evidence and if we disprove it … we are unlearning.  And doing that generates emotional tension. We are often very attached to our unconscious assumptions and will usually resist them being challenged.

The way to unlearn effectively is to use the evidence of our own eyes to raise doubts about our unconscious assumptions.  We need to actively generate a bit of confusion.

Then, we resolve the apparent paradox by creatively shifting perspective, often with a real example, a practical explanation or a hands-on demonstration.

And when we experience the “Ah ha! Now I see!” reaction, and we emerge from the fog of confusion, we will relieve the emotional tension and our involuntary reaction is to laugh.

But if our teacher unintentionally triggers a Marmite effect; a “Yeuk, I am NOT enjoying this!” feeling, then we need to respect that, and step back, and adopt a different tack.


Over the last few months I have been experimenting with different approaches to introducing the principles of improvement-by-design.

And the results are clear.

A minority prefer to start with the abstract theory, and then apply it in practice.

The majority have various degrees of Marmite reaction to the theory, and some are so put off that they actively disengage.  But when they have an opportunity to see the same principles demonstrated in a concrete, practical way; they learn and laugh.

Unlearning-by-doing seems to work better for the majority.

So, if you want to have fun and learn how to deliver significant and sustained improvements … then the evidence points to this as the starting point …

… the Flow Design Practical Skills One Day Workshop.

And if you also want to dip into a bit of the tried-and-tested theory that underpins improvement-by-design then you can do that as well, either before or later (when it becomes necessary), or both.


So, to have lots of fun and learn some valuable improvement-by-design practical skills at the same time …  click here.

The Pathology of Variation I

In medical training we have to learn about lots of things. That is one reason why it takes a long time to train a competent and confident clinician.

First, we learn the anatomy (structure) and the physiology (function) of the normal, healthy human.

Then we learn about how this amazingly complicated system can go wrong.  We learn about pathology.  And we do that so that we understand the relationship between the cause (disease) and the effect (symptoms and signs).

Then we learn about diagnostics – which is how to work backwards from the effects to the most likely cause(s).

And only then can we learn about therapeutics – the design and delivery of a treatment plan that we are confident will relieve the symptoms by curing the disease.

And we learn about prevention – how to avoid some illnesses (and delay others) by addressing the root causes earlier.  Much of the increase in life expectancy over the last 200 years has come from prevention, not from cure.


The NHS is an amazingly complicated system, and it too can go wrong.  It can exhibit a wide spectrum of symptoms and signs; medical errors, long delays, unhappy patients, burned-out staff, and overspent budgets.

But, there is no equivalent training in how to diagnose and treat a sick health care system.  And this is not acceptable, especially given that the knowledge of how to do this is already available.

It is called complex adaptive systems engineering (CASE).


Before the Renaissance, the understanding of how the body works was primitive and it was believed that illness was “God’s Will” so we had to just grin-and-bear (and pray).

The Scientific Revolution brought us new insights, profound theories, innovative techniques and capability-extending tools.  And the impact has been dramatic.  Those who do have access to this knowledge live better and longer than ever.  Those who do not … do not.

Our current understanding of how health care systems work is, to be blunt, medieval.  The current approaches amount to little more than rune reading, incantations and the prescription of purgatives and leeches.  And the impact is about as effective.

So we need to study the anatomy, physiology, pathology, diagnostics and therapeutics of complex adaptive systems like healthcare.  And most of all we need to understand how to prevent catastrophes happening in the first place.  We need the NHS to be immortal.


And this week a prototype complex adaptive pathology training system was tested … and it employed cutting-edge 21st Century technology: Pasta Twizzles.

The specific topic under scrutiny was variation.  A brain-bending concept that is usually relegated to the mystical smoke-and-mirrors world called “Sadistics”.

But no longer!

The Mists-of-Jargon and Fog-of-Formulae were blown away as we switched on the Fan-of-Facilitation and the Light-of-Simulation and went exploring.

Empirically. Pragmatically.


And what we discovered was jaw-dropping.

A disease called the “Flaw of Averages” and its malignant manifestation “Carveoutosis“.


And with our new knowledge we opened the door to a previously hidden world of opportunity and improvement.

Then we activated the Laser-of-Insight and evaporated the queues and chaos that, before our new understanding, we had accepted as inevitable and beyond our understanding or control.

They were neither. And never had been. We were deluding ourselves.

Welcome to the Resilient Design – Practical Skills – One Day Workshop.

Validation Test: Passed.

Diagnose-Design-Deliver

A story was shared this week.

A story of hope for the hard-pressed NHS, its patients, its staff and its managers and its leaders.

A story that says “We can learn how to fix the NHS ourselves“.

And the story comes with evidence; hard, objective, scientific, statistically significant evidence.


The story starts almost exactly three years ago when a Clinical Commissioning Group (CCG) in England made a bold strategic decision to invest in improvement, or as they termed it “Achieving Clinical Excellence” (ACE).

They invited proposals from their local practices with the “carrot” of enough funding to allow GPs to carve-out protected time to do the work.  And a handful of proposals were selected and financially supported.

This is the story of one of those proposals which came from three practices in Sutton who chose to work together on a common problem – the unplanned hospital admissions in their over 70’s.

Their objective was clear and measurable: “To reduce the cost of unplanned admissions in the 70+ age group by working with hospital to reduce length of stay.

Did they achieve their objective?

Yes, they did.  But there is more to this story than that.  Much more.


One innovative step they took was to invest in learning how to diagnose why the current ‘system’ was costing what it was; then learning how to design an improvement; and then learning how to deliver that improvement.

They invested in developing their own improvement science skills first.

They did not assume they already knew how to do this and they engaged an experienced health care systems engineer (HCSE) to show them how to do it (i.e. not to do it for them).

Another innovative step was to create a blog to make it easier to share what they were learning with their colleagues; and to invite feedback and suggestions; and to provide a journal that captured the story as it unfolded.

And they measured stuff before they made any changes and afterwards so they could measure the impact, and so that they could assess the evidence scientifically.

And that was actually quite easy because the CCG was already measuring what they needed to know: admissions, length of stay, cost, and outcomes.

All they needed to learn was how to present and interpret that data in a meaningful way.  And as part of their IS training,  they learned how to use system behaviour charts, or SBCs.


By Jan 2015 they had learned enough of the HCSE techniques and tools to establish the diagnosis and start to making changes to the parts of the system that they could influence.


Two years later they subjected their before-and-after data to robust statistical analysis and they had a surprise. A big one!

Reducing hospital mortality was not a stated objective of their ACE project, and they only checked the mortality data to be sure that it had not changed.

But it had, and the “p=0.014” part of the statement above means that the probability that this 20.0% reduction in hospital mortality was due to random chance … is less than 1.4%.  [This is well below the 5% threshold that we usually accept as “statistically significant” in a clinical trial.]

But …

This was not a randomised controlled trial.  This was an intervention in a complicated, ever-changing system; so they needed to check that the hospital mortality for comparable patients who were not their patients had not changed as well.

And the statistical analysis of the hospital mortality for the ‘other’ practices for the same patient group, and the same period of time confirmed that there had been no statistically significant change in their hospital mortality.

So, it appears that what the Sutton ACE Team did to reduce length of stay (and cost) had also, unintentionally, reduced hospital mortality. A lot!


And this unexpected outcome raises a whole raft of questions …


If you would like to read their full story then you can do so … here.

It is a story of hunger for improvement, of humility to learn, of hard work and of hope for the future.

Levels of Resistance

Improvement implies change, but change does not imply improvement.

We have all experienced the pain of disappointment when a change that promised much delivered no improvement, or even worse, a negative impact.

We have learned to become wary and skeptical about change.

We have learned a whole raft of tactics for deflection and diffusion of the enthusiasm of others.  And by doing so we don the black hat of the healthy skeptic and the tell tale mantra of “Yes, but …”.

So here is an onion diagram to use as a reference.  It comes from a recently published essay that compares and contrasts two schools of flow improvement.  Eli Goldratt’s “Theory of Constraints” and a translation of Systems Engineering called 6M Design®.


The first five layers can be described as “denial”, the second four as “grudging acceptance” … and the last one is the sound of the final barrier coming down and revealing the raw emotion underpinning our reluctance to change. Fear.


The good news is that this diagram helps us to shape and steer change in a way that improves its chances of success, because if we can learn to peel back these layers by sharing information that soothes the fear of the unknown, then we can align and engage.  And that is essential for emotional momentum to build.

So when we meet resistance do we push or not?

Ask yourself. How would prefer to be engaged? Pushed or not?

Hugh, Louise and Bob

Bob Jekyll was already sitting at a table, sipping a pint of Black Sheep and nibbling on a bowl of peanuts when Hugh and Louise arrived.

<Hugh> Hello, are you Bob?

<Bob> Yes, indeed! You must be Hugh and Louise. Can I get you a thirst quencher?

<Louise> Lime and soda for me please.

<Hugh> I’ll have the same as you, a Black Sheep.

<Bob> On the way.

<Hugh> Hello Louise, I’m Hugh Lewis.  I am the ops manager for acute medicine at St. Elsewhere’s Hospital. It is good to meet you at last. I have seen your name on emails and performance reports.

<Louise> Good to meet you too Hugh. I am senior data analyst for St. Elsewhere’s and I think we may have met before, but I’m not sure when.  Do you know what this is about? Your invitation was a bit mysterious.

<Hugh> Yes. Sorry about that. I was chatting to a friend of mine at the golf club last week, Dr Bill Hyde who is one of our local GPs.  As you might expect, we got to talking about the chronic pressure we are all under in both primary and secondary care.  He said he has recently crossed paths with an old chum of his from university days who he’d had a very interesting conversation with in this very pub, and he recommended I email him. So I did. And that led to a phone conversation with Bob Jekyll. I have to say he asked some very interesting questions that left me feeling a mixture of curiosity and discomfort. After we talked Bob suggested that we meet for a longer chat and that I invite my senior data analyst along. So here we are.

<Louise> I have to say my curiosity was pricked by your invitation, specifically the phrase ‘system behaviour charts’. That is a new one on me and I have been working in the NHS for some time now. It is too many years to mention since I started as junior data analyst, fresh from university!

<Hugh> That is the term Bob used, and I confess it was new to me too.

<Bob> Here we are, Black Sheep, lime soda and more peanuts.  Thank you both for coming, so shall we talk about the niggle that Hugh raised when we spoke on the phone?

<Hugh> Ah! Louise, please accept my apologies in advance. I think Bob might be referring to when I said that “90% of the performance reports don’t make any sense to me“.

<Louise> There is no need to apologise Hugh. I am actually reassured that you said that. They don’t make any sense to me either! We only produce them that way because that is what we are asked for.  My original degree was geography and I discovered that I loved data analysis! My grandfather was a doctor so I guess that’s how I ended up in doing health care data analysis. But I must confess, some days I do not feel like I am adding much value.

<Hugh> Really? I believe we are in heated agreement! Some days I feel the same way.  Is that why you invited us both Bob?

<Bob> Yes.  It was some of the things that Hugh said when we talked on the phone.  They rang some warning bells for me because, in my line of work, I have seen many people fall into a whole minefield of data analysis traps that leave them feeling confused and frustrated.

<Louise> What exactly is your line of work, Bob?

<Bob> I am a systems engineer.  I design, build, verify, integrate, implement and validate systems. Fit-for-purpose systems.

<Louise> In health care?

<Bob> Not until last week when I bumped into Bill Hyde, my old chum from university.  But so far the health care system looks just like all the other ones I have worked in, so I suspect some of the lessons from other systems are transferable.

<Hugh> That sounds interesting. Can you give us an example?

<Bob> OK.  Hugh, in our first conversation, you often used the words “demand”  and “capacity”. What do you mean by those terms?

<Hugh> Well, demand is what comes through the door, the flow of requests, the workload we are expected to manage.  And capacity is the resources that we have to deliver the work and to meet our performance targets.  Capacity is the staff, the skills, the equipment, the chairs, and the beds. The stuff that costs money to provide.  As a manager, I am required to stay in-budget and that consumes a big part of my day!

<Bob> OK. Speaking as an engineer I would like to know the units of measurement of “demand” and “capacity”?

<Hugh> Oh! Um. Let me think. Er. I have never been asked that question before. Help me out here Louise.  I told you Bob asks tricky questions!

<Louise> I think I see what Bob is getting at.  We use these terms frequently but rather loosely. On reflection they are not precisely defined, especially “capacity”. There are different sorts of capacity all of which will be measured in different ways so have different units. No wonder we spend so much time discussing and debating the question of if we have enough capacity to meet the demand.  We are probably all assuming different things.  Beds cannot be equated to staff, but too often we just seem to lump everything together when we talk about “capacity”.  So by doing that what we are really asking is “do we have enough cash in the budget to pay for the stuff we thing we need?”. And if we are failing one target or another we just assume that the answer is “No” and we shout for “more cash”.

<Bob> Exactly my point. And this was one of the warning bells.  Lack of clarity on these fundamental definitions opens up a minefield of other traps like the “Flaw of Averages” and “Time equals Money“.  And if we are making those errors then they will, unwittingly, become incorporated into our data analysis.

<Louise> But we use averages all the time! What is wrong with an average?

<Bob> I can sense you are feeling a bit defensive Louise.  There is no need to.  An average is perfectly OK and is very useful tool.  The “flaw” is when it is used inappropriately.  Have you heard of Little’s Law?

<Louise> No. What’s that?

<Bob> It is the mathematically proven relationship between flow, work-in-progress and lead time.  It is a fundamental law of flow physics and it uses averages. So averages are OK.

<Hugh> So what is the “Flaw of Averages”?

<Bob> It is easier to demonstrate it than to describe it.  Let us play a game.  I have some dice and we have a big bowl of peanuts.  Let us simulate a simple two step process.  Hugh you are Step One and Louise you are Step Two.  I will be the the source of demand.

I will throw a dice and count that many peanuts out of the bowl and pass them to Hugh.  Hugh, you then throw the dice and move that many peanuts from your heap to Louise, then Louise throws the dice and moves that many from her pile to the final heap which we will call activity.

<Hugh> Sounds easy enough.  If we all use the same dice then the average flow through each step will be the same so after say ten rounds we should have, um …

<Louise> … thirty five peanuts in the activity heap.  On average.

<Bob> OK.  That’s the theory, let’s see what happens in reality.  And no eating the nuts-in-progress please.


They play the game and after a few minutes they have completed the ten rounds.


<Hugh> That’s odd.  There are only 30 nuts in the activity heap and we expected 35.  Nobody nibbled any nuts so its just chance I suppose.  Lets play again. It should average out.

…..  …..

<Louise> Thirty four this time which is better, but is still below the predicted average.  That could still be a chance effect though.  Let us run the ‘nutty’ game this a few more times.

….. …..

<Hugh> We have run the same game six times with the same nuts and the same dice and we delivered activities of 30, 34, 30, 24, 23 and 31 and there are usually nuts stuck in the process at the end of each game, so it is not due to a lack of demand.  We are consistently under-performing compared with our theoretical prediction.  That is weird.  My head says we were just unlucky but I have a niggling doubt that there is more to it.

<Louise> Is this the Flaw of Averages?

<Bob> Yes, it is one of them. If we set our average future flow-capacity to the average historical demand and there is any variation anywhere in the process then we will see this effect.

<Hugh> H’mmm.  But we do this all the time because we assume that the variation will average out over time. Intuitively it must average out over time.  What would happen if we kept going for more cycles?

<Bob> That is a very good question.  And your intuition is correct.  It does average out eventually but there is a catch.

<Hugh> What is the catch?

<Bob>  The number of peanuts in the process and the time it takes for one peanut to get through is very variable.

<Louise> Is there any pattern to the variation? Is it predictable?

<Bob> Another excellent question.  Yes, there is a pattern.  It is called “chaos”.  Predictable chaos if you like.

<Hugh> So is that the reason you said on the phone that we should present our metrics as time-series charts?

<Bob> Yes, one of them.  The appearance of chaotic system behaviour is very characteristic on a time-series chart.

<Louise> And if we see the chaos pattern on our charts then we could conclude that we have made the Flaw of Averages error?

<Bob> That would be a reasonable hypothesis.

<Hugh> I think I understand the reason you invited us to a face-to-face demonstration.  It would not have worked if you had just described it.  You have to experience it because it feels so counter-intuitive.  And this is starting to feel horribly familiar; perpetual chaos about sums up my working week!

<Louise> You also mentioned something you referred to as the “time equals money” trap.  Is that somehow linked to this?

<Bob> Yes.  We often equate time and money but they do not behave the same way.  If have five pounds today and I only spend four pounds then I can save the remaining one pound for tomorrow and spend it then – so the Law of Averages works.  But if I have five minutes today and I only use four minutes then the other minute cannot be saved and used tomorrow, it is lost forever.  That is why the Law of Averages does not work for time.

<Hugh> But that means if we set our budgets based on the average demand and the cost of people’s time then not only will we have queues, delays and chaos, we will also consistently overspend the budget too.  This is sounding more and more familiar by the minute!  This is nuts, if you will excuse the pun.

<Louise> So what is the solution?  I hope you would not have invited us here if there was no solution.

<Bob> Part of the solution is to develop our knowledge of system behaviour and how we need to present it in a visual format. With that we develop a deeper understanding of what the system behaviour charts are saying to us.  With that we can develop our ability to make wiser decisions that will lead to effective actions which will eliminate the queues, delays, chaos and cost-pressures.

<Hugh> This is possible?

<Bob> Yes. It is called systems engineering. That’s what I do.

<Louise> When do we start?

<Bob> We have started.

Dr Hyde and Mr Jekyll

Dr Bill Hyde was already at the bar when Bob Jekyll arrived.

Bill and  Bob had first met at university and had become firm friends, but their careers had diverged and it was only by pure chance that their paths had crossed again recently.

They had arranged to meet up for a beer and to catch up on what had happened in the 25 years since they had enjoyed the “good old times” in the university bar.

<Dr Bill> Hi Bob, what can I get you? If I remember correctly it was anything resembling real ale. Will this “Black Sheep” do?

<Bob> Hi Bill, Perfect! I’ll get the nibbles. Plain nuts OK for you?

<Dr Bill> My favourite! So what are you up to now? What doors did your engineering degree open?

<Bob> Lots!  I’ve done all sorts – mechanical, electrical, software, hardware, process, all except civil engineering. And I love it. What I do now is a sort of synthesis of all of them.  And you? Where did your medical degree lead?

<Dr Bill> To my hearts desire, the wonderful Mrs Hyde, and of course to primary care. I am a GP. I always wanted to be a GP since I was knee-high to a grasshopper.

<Bob> Yes, you always had that “I’m going to save the world one patient at a time!” passion. That must be so rewarding! Helping people who are scared witless by the health horror stories that the media pump out.  I had a fright last year when I found a lump.  My GP was great, she confidently diagnosed a “hernia” and I was all sorted in a matter of weeks with a bit of nifty day case surgery. I was convinced my time had come. It just shows how damaging the fear of the unknown can be!

<Dr Bill> Being a GP is amazingly rewarding. I love my job. But …

<Bob> But what? Are you alright Bill? You suddenly look really depressed.

<Dr Bill> Sorry Bob. I don’t want to be a damp squib. It is good to see you again, and chat about the old days when we were teased about our names.  And it is great to hear that you are enjoying your work so much. I admit I am feeling low, and frankly I welcome the opportunity to talk to someone I know and trust who is not part of the health care system. If you know what I mean?

<Bob> I know exactly what you mean.  Well, I can certainly offer an ear, “a problem shared is a problem halved” as they say. I can’t promise to do any more than that, but feel free to tell me the story, from the beginning. No blood-and-guts gory details though please!

<Dr Bill> Ha! “Tell me the story from the beginning” is what I say to my patients. OK, here goes. I feel increasingly overwhelmed and I feel like I am drowning under a deluge of patients who are banging on the practice door for appointments to see me. My intuition tells me that the problem is not the people, it is the process, but I can’t seem to see through the fog of frustration and chaos to a clear way forward.

<Bob> OK. I confess I know nothing about how your system works, so can you give me a bit more context.

<Dr Bill> Sorry. Yes, of course. I am what is called a single-handed GP and I have a list of about 1500 registered patients and I am contracted to provide primary care for them. I don’t have to do that 24 x 7, the urgent stuff that happens in the evenings and weekends is diverted to services that are designed for that. I work Monday to Friday from 9 AM to 5 PM, and I am contracted to provide what is needed for my patients, and that means face-to-face appointments.

<Bob> OK. When you say “contracted” what does that mean exactly?

<Dr Bill> Basically, the St. Elsewhere’s® Practice is like a small business. It’s annual income is a fixed amount per year for each patient on the registration list, and I have to provide the primary care service for them from that pot of cash. And that includes all the costs, including my income, our practice nurse, and the amazing Mrs H. She is the practice receptionist, manager, administrator and all-round fixer-of-anything.

<Bob> Wow! What a great design. No need to spend money on marketing, research, new product development, or advertising! Just 100% pure service delivery of tried-and-tested medical know-how to a captive audience for a guaranteed income. I have commercial customers who would cut off their right arms for an offer like that!

<Dr Bill> Really? It doesn’t feel like that to me. It feels like the more I offer, the more the patients expect. The demand is a bottomless well of wants, but the income is capped and my time is finite!

<Bob> H’mm. Tell me more about the details of how the process works.

<Dr Bill> Basically, I am a problem-solving engine. Patients phone for an appointment, Mrs H books one, the patient comes at the appointed time, I see them, and I diagnose and treat the problem, or I refer on to a specialist if it’s more complicated. That’s basically it.

<Bob> OK. Sounds a lot simpler than 99% of the processes that I’m usually involved with. So what’s the problem?

<Dr Bill> I don’t have enough capacity! After all the appointments for the day are booked Mrs H has to say “Sorry, please try again tomorrow” to every patient who phones in after that.  The patients who can’t get an appointment are not very happy and some can get quite angry. They are anxious and frustrated and I fully understand how they feel. I feel the same.

<Bob> We will come back to what you mean by “capacity”. Can you outline for me exactly how a patient is expected to get an appointment?

<Dr Bill> We tell them to phone at 8 AM for an appointment, there is a fixed number of bookable appointments, and it is first-come-first-served.  That is the only way I can protect myself from being swamped and is the fairest solution for patients.  It wasn’t my idea; it is called Advanced Access. Each morning at 8 AM we switch on the phones and brace ourselves for the daily deluge.

<Bob> You must be pulling my leg! This design is a batch-and-queue phone-in appointment booking lottery!  I guess that is one definition of “fair”.  How many patients get an appointment on the first attempt?

<Dr Bill> Not many.  The appointments are usually all gone by 9 AM and a lot are to people who have been trying to get one for several days. When they do eventually get to see me they are usually grumpy and then spring the trump card “And while I’m here doctor I have a few other things that I’ve been saving up to ask you about“. I help if I can but more often than not I have to say, “I’m sorry, you’ll have to book another appointment!“.

<Bob> I’m not surprised you patients are grumpy. I would be too. And my recollection of seeing my GP with my scary lump wasn’t like that at all. I phoned at lunch time and got an appointment the same day. Maybe I was just lucky, or maybe my GP was as worried as me. But it all felt very calm. When I arrived there was only one other patient waiting, and I was in and out in less than ten minutes – and mightily reassured I can tell you! It felt like a high quality service that I could trust if-and-when I needed it, which fortunately is very infrequently.

<Dr Bill> I dream of being able to offer a service like that! I am prepared to bet you are registered with a group practice and you see whoever is available rather than your own GP. Single-handed GPs like me who offer the old fashioned personal service are a rarity, and I can see why. We must be suckers!

<Bob> OK, so I’m starting to get a sense of this now. Has it been like this for a long time?

<Dr Bill> Yes, it has. When I was younger I was more resilient and I did not mind going the extra mile.  But the pressure is relentless and maybe I’m just getting older and grumpier.  My real fear is I end up sounding like the burned-out cynics that I’ve heard at the local GP meetings; the ones who crow about how they are counting down the days to when they can retire and gloat.

<Bob> You’re the same age as me Bill so I don’t think either of us can use retirement as an exit route, and anyway, that’s not your style. You were never a quitter at university. Your motto was always “when the going gets tough the tough get going“.

<Dr Bill> Yeah I know. That’s why it feels so frustrating. I think I lost my mojo a long time back. Maybe I should just cave in and join up with the big group practice down the road, and accept the inevitable loss of the personal service. They said they would welcome me, and my list of 1500 patients, with open arms.

<Bob> OK. That would appear to be an option, or maybe a compromise, but I’m not sure we’ve exhausted all the other options yet.  Tell me, how do you decide how long a patient needs for you to solve their problem?

<Dr Bill> That’s easy. It is ten minutes. That is the time recommended in the Royal College Guidelines.

<Bob> Eh? All patients require exactly ten minutes?

<Dr Bill> No, of course not!  That is the average time that patients need.  The Royal College did a big survey and that was what most GPs said they needed.

<Bob> Please tell me if I have got this right.  You work 9-to-5, and you carve up your day into 10-minute time-slots called “appointments” and, assuming you are allowed time to have lunch and a pee, that would be six per hour for seven hours which is 42 appointments per day that can be booked?

<Dr Bill> No. That wouldn’t work because I have other stuff to do as well as see patients. There are only 25 bookable 10-minute appointments per day.

<Bob> OK, that makes more sense. So where does 25 come from?

<Dr Bill> Ah! That comes from a big national audit. For an average GP with and average  list of 1,500 patients, the average number of patients seeking an appointment per day was found to be 25, and our practice population is typical of the national average in terms of age and deprivation.  So I set the upper limit at 25. The workload is manageable but it seems to generate a lot of unhappy patients and I dare not increase the slots because I’d be overwhelmed with the extra workload and I’m barely coping now.  I feel stuck between a rock and a hard place!

<Bob> So you have set the maximum slot-capacity to the average demand?

<Dr Bill> Yes. That’s OK isn’t it? It will average out over time. That is what average means! But it doesn’t feel like that. The chaos and pressure never seems to go away.


There was a long pause while Bob mulls over what he had heard, sips his pint of Black Sheep and nibbles on the dwindling bowl of peanuts.  Eventually he speaks.


<Bob> Bill, I have some good news and some not-so-good news and then some more good news.

<Dr Bill> Oh dear, you sound just like me when I have to share the results of tests with one of my patients at their follow up appointment. You had better give me the “bad news sandwich”!

<Bob> OK. The first bit of good news is that this is a very common, and easily treatable flow problem.  The not-so-good news is that you will need to change some things.  The second bit of good news is that the changes will not cost anything and will work very quickly.

<Dr Bill> What! You cannot be serious!! Until ten minutes ago you said that you knew nothing about how my practice works and now you are telling me that there is a quick, easy, zero cost solution.  Forgive me for doubting your engineering know-how but I’ll need a bit more convincing than that!

<Bob> And I would too if I were in your position.  The clues to the diagnosis are in the story. You said the process problem was long-standing; you said that you set the maximum slot-capacity to the average demand; and you said that you have a fixed appointment time that was decided by a subjective consensus.  From an engineering perspective, this is a perfect recipe for generating chronic chaos, which is exactly the symptoms you are describing.

<Dr Bill> Is it? OMG. You said this is well understood and resolvable? So what do I do?

<Bob> Give me a minute.  You said the average demand is 25 per day. What sort of service would you like your patients to experience? Would “90% can expect a same day appointment on the first call” be good enough as a starter?

<Dr Bill> That would be game changing!  Mrs H would be over the moon to be able to say “Yes” that often. I would feel much less anxious too, because I know the current system is a potentially dangerous lottery. And my patients would be delighted and relieved to be able to see me that easily and quickly.

<Bob> OK. Let me work this out. Based on what you’ve said, some assumptions, and a bit of flow engineering know-how; you would need to offer up to 31 appointments per day.

<Dr Bill> What! That’s impossible!!! I told you it would be impossible! That would be another hour a day of face-to-face appointments. When would I do the other stuff? And how did you work that out anyway?

<Bob> I did not say they would have to all be 10-minute appointments, and I did not say you would expect to fill them all every day. I did however say you would have to change some things.  And I did say this is a well understood flow engineering problem.  It is called “resilience design“. That’s how I was able to work it out on the back of this Black Sheep beer mat.

<Dr Bill> H’mm. That is starting to sound a bit more reasonable. What things would I have to change? Specifically?

<Bob> I’m not sure what specifically yet.  I think in your language we would say “I have taken a history, and I have a differential diagnosis, so next I’ll need to examine the patient, and then maybe do some tests to establish the actual diagnosis and to design and decide the treatment plan“.

<Dr Bill> You are learning the medical lingo fast! What do I need to do first? Brace myself for the forensic rubber-gloved digital examination?

<Bob> Alas, not yet and certainly not here. Shall we start with the vital signs? Height, weight, pulse, blood pressure, and temperature? That’s what my GP did when I went with my scary lump.  The patient here is not you, it is your St. Elsewhere’s® Practice, and we will need to translate the medical-speak into engineering-speak.  So one thing you’ll need to learn is a bit of the lingua-franca of systems engineering.  By the way, that’s what I do now. I am a systems engineer, or maybe now a health care systems engineer?

<Dr Bill> Point me in the direction of the HCSE dictionary! The next round is on me. And the nuts!

<Bob> Excellent. I’ll have another Black Sheep and some of those chilli-coated ones. We have work to do.  Let me start by explaining what “capacity” actually means to an engineer. Buckle up. This ride might get a bit bumpy.


This story is fictional, but the subject matter is factual.

Bob’s diagnosis and recommendations are realistic and reasonable.

Chapter 1 of the HCSE dictionary can be found here.

And if you are a GP who recognises these “symptoms” then this may be of interest.

Miracle on Tavanagh Avenue

Sometimes change is dramatic. A big improvement appears very quickly. And when that happens we are caught by surprise (and delight).

Our emotional reaction is much faster than our logical response. “Wow! That’s a miracle!


Our logical Tortoise eventually catches up with our emotional Hare and says “Hare, we both know that there is no such thing as miracles and magic. There must be a rational explanation. What is it?

And Hare replies “I have no idea, Tortoise.  If I did then it would not have been such a delightful surprise. You are such a kill-joy! Can’t you just relish the relief without analyzing the life out of it?

Tortoise feels hurt. “But I just want to understand so that I can explain to others. So that they can do it and get the same improvement.  Not everyone has a ‘nothing-ventured-nothing-gained’ attitude like you! Most of us are too fearful of failing to risk trusting the wild claims of improvement evangelists. We have had our fingers burned too often.


The apparent miracle is real and recent … here is a snippet of the feedback:

Notice carefully the last sentence. It took a year of discussion to get an “OK” and a month of planning to prepare the “GO”.

That is not a miracle and some magic … that took a lot of hard work!

The evangelist is the customer. The supplier is an engineer.


The context is the chronic niggle of patients trying to get an appointment with their GP, and the chronic niggle of GPs feeling overwhelmed with work.

Here is the back story …

In the opening weeks of the 21st Century, the National Primary Care Development Team (NPDT) was formed.  Primary care was a high priority and the government had allocated £168m of investment in the NHS Plan, £48m of which was earmarked to improve GP access.

The approach the NPDT chose was:

harvest best practice +
use a panel of experts +
disseminate best practice.

Dr (later Sir) John Oldham was the innovator and figure-head.  The best practice was copied from Dr Mark Murray from Kaiser Permanente in the USA – the Advanced Access model.  The dissemination method was copied from from Dr Don Berwick’s Institute of Healthcare Improvement (IHI) in Boston – the Collaborative Model.

The principle of Advanced Access is “today’s-work-today” which means that all the requests for a GP appointment are handled the same day.  And the proponents of the model outlined the key elements to achieving this:

1. Measure daily demand.
2. Set capacity so that is sufficient to meet the daily demand.
3. Simple booking rule: “phone today for a decision today”.

But that is not what was rolled out. The design was modified somewhere between aspiration and implementation and in two important ways.

First, by adding a policy of “Phone at 08:00 for an appointment”, and second by adding a policy of “carving out” appointment slots into labelled pots such as ‘Dr X’ or ‘see in 2 weeks’ or ‘annual reviews’.

Subsequent studies suggest that the tweaking happened at the GP practice level and was driven by the fear that, by reducing the waiting time, they would attract more work.

In other words: an assumption that demand for health care is supply-led, and without some form of access barrier, the system would be overwhelmed and never be able to cope.


The result of this well-intended tampering with the Advanced Access design was to invalidate it. Oops!

To a systems engineer this is meddling was counter-productive.

The “today’s work today” specification is called a demand-led design and, if implemented competently, will lead to shorter waits for everyone, no need for urgent/routine prioritization and slot carve-out, and a simpler, safer, calmer, more efficient, higher quality, more productive system.

In this context it does not mean “see every patient today” it means “assess and decide a plan for every patient today”.

In reality, the actual demand for GP appointments is not known at the start; which is why the first step is to implement continuous measurement of the daily number and category of requests for appointments.

The second step is to feed back this daily demand information in a visual format called a time-series chart.

The third step is to use this visual tool for planning future flow-capacity, and for monitoring for ‘signals’, such as spikes, shifts, cycles and slopes.

That was not part of the modified design, so the reasonable fear expressed by GPs was (and still is) that by attempting to do today’s-work-today they would unleash a deluge of unmet need … and be swamped/drowned.

So a flood defense barrier was bolted on; the policy of “phone at 08:00 for an appointment today“, and then the policy of  channeling the over spill into pots of “embargoed slots“.

The combined effect of this error of omission (omitting the measured demand visual feedback loop) and these errors of commission (the 08:00 policy and appointment slot carve-out policy) effectively prevented the benefits of the Advanced Access design being achieved.  It was a predictable failure.

But no one seemed to realize that at the time.  Perhaps because of the political haste that was driving the process, and perhaps because there were no systems engineers on the panel-of-experts to point out the risks of diluting the design.

It is also interesting to note that the strategic aim of the NPCT was to develop a self-sustaining culture of quality improvement (QI) in primary care. That didn’t seem to have happened either.


The roll out of Advanced Access was not the success it was hoped. This is the conclusion from the 300+ page research report published in 2007.


The “Miracle on Tavanagh Avenue” that was experienced this week by both patients and staff was the expected effect of this tampering finally being corrected; and the true potential of the original demand-led design being released – for all to experience.

Remember the essential ingredients?

1. Measure daily demand and feed it back as a visual time-series chart.
2. Set capacity so that is sufficient to meet the daily demand.
3. Use a simple booking rule: “phone anytime for a decision today”.

But there is also an extra design ingredient that has been added in this case, one that was not part of the original Advanced Access specification, one that frees up GP time to provide the required “resilience” to sustain a same-day service.

And that “secret” ingredient is how the new design worked so quickly and feels like a miracle – safe, calm, enjoyable and productive.

This is health care systems engineering (HCSE) in action.


So congratulations to Harry Longman, the whole team at GP Access, and to Dr Philip Lusty and the team at Riverside Practice, Tavangh Avenue, Portadown, NI.

You have demonstrated what was always possible.

The fear of failure prevented it before, just as it prevented you doing this until you were so desperate you had no other choices.

To read the fuller story click here.

PS. Keep a close eye on the demand time-series chart and if it starts to rise then investigate the root cause … immediately.


The Power of Pictures

I am a big fan of pictures that tell a story … and this week I discovered someone who is creating great pictures … Hayley Lewis.

This is one of Hayley’s excellent sketch notes … the one that captures the essence of the Bruce Tuckman model of team development.

The reason that I share this particular sketch-note is because my experience of developing improvement-by-design teams is that it works just like this!

The tricky phase is the STORMING one because not all teams survive it!

About half sink in the storm – and that seems like an awful waste – and I believe it is avoidable.

This means that before starting the team development cycle, the leader needs to be aware of how to navigate themselves and the team through the storm phase … and that requires training, support and practice.

Which is the reason why coaching from a independent, experienced, capable practitioner is a critical element of the improvement process.

How Do We Know We Have Improved?

Phil and Pete are having a coffee and a chat.  They both work in the NHS and have been friends for years.

They have different jobs. Phil is a commissioner and an accountant by training, Pete is a consultant and a doctor by training.

They are discussing a challenge that affects them both on a daily basis: unscheduled care.

Both Phil and Pete want to see significant and sustained improvements and how to achieve them is often the focus of their coffee chats.


<Phil> We are agreed that we both want improvement, both from my perspective as a commissioner and from your perspective as a clinician. And we agree that what we want to see improvements in patient safety, waiting, outcomes, experience for both patients and staff, and use of our limited NHS resources.

<Pete> Yes. Our common purpose, the “what” and “why”, has never been an issue.  Where we seem to get stuck is the “how”.  We have both tried many things but, despite our good intentions, it feels like things are getting worse!

<Phil> I agree. It may be that what we have implemented has had a positive impact and we would have been even worse off if we had done nothing. But I do not know. We clearly have much to learn and, while I believe we are making progress, we do not appear to be learning fast enough.  And I think this knowledge gap exposes another “how” issue: After we have intervened, how do we know that we have (a) improved, (b) not changed or (c) worsened?

<Pete> That is a very good question.  And all that I have to offer as an answer is to share what we do in medicine when we ask a similar question: “How do I know that treatment A is better than treatment B?”  It is the essence of medical research; the quest to find better treatments that deliver better outcomes and at lower cost.  The similarities are strong.

<Phil> OK. How do you do that? How do you know that “Treatment A is better than Treatment B” in a way that anyone will trust the answer?

 <Pete> We use a science that is actually very recent on the scientific timeline; it was only firmly established in the first half of the 20th century. One reason for that is that it is rather a counter-intuitive science and for that reason it requires using tools that have been designed and demonstrated to work but which most of us do not really understand how they work. They are a bit like magic black boxes.

<Phil> H’mm. Please forgive me for sounding skeptical but that sounds like a big opportunity for making mistakes! If there are lots of these “magic black box” tools then how do you decide which one to use and how do you know you have used it correctly?

<Pete> Those are good questions! Very often we don’t know and in our collective confusion we generate a lot of unproductive discussion.  This is why we are often forced to accept the advice of experts but, I confess, very often we don’t understand what they are saying either! They seem like the medieval Magi.

<Phil> H’mm. So these experts are like ‘magicians’ – they claim to understand the inner workings of the black magic boxes but are unable, or unwilling, to explain in a language that a ‘muggle’ would understand?

<Pete> Very well put. That is just how it feels.

<Phil> So can you explain what you do understand about this magical process? That would be a start.


<Pete> OK, I will do my best.  The first thing we learn in medical research is that we need to be clear about what it is we are looking to improve, and we need to be able to measure it objectively and accurately.

<Phil> That  makes sense. Let us say we want to improve the patient’s subjective quality of the A&E experience and objectively we want to reduce the time they spend in A&E. We measure how long they wait. 

<Pete> The next thing is that we need to decide how much improvement we need. What would be worthwhile? So in the example you have offered we know that reducing the average time patients spend in A&E by just 30 minutes would have a significant effect on the quality of the patient and staff experience, and as a by-product it would also dramatically improve the 4-hour target performance.

<Phil> OK.  From the commissioning perspective there are lots of things we can do, such as commissioning alternative paths for specific groups of patients; in effect diverting some of the unscheduled demand away from A&E to a more appropriate service provider.  But these are the sorts of thing we have been experimenting with for years, and it brings us back to the question: How do we know that any change we implement has had the impact we intended? The system seems, well, complicated.

<Pete> In medical research we are very aware that the system we are changing is very complicated and that we do not have the power of omniscience.  We cannot know everything.  Realistically, all we can do is to focus on objective outcomes and collect small samples of the data ocean and use those in an attempt to draw conclusions can trust. We have to design our experiment with care!

<Phil> That makes sense. Surely we just need to measure the stuff that will tell us if our impact matches our intent. That sounds easy enough. What’s the problem?

<Pete> The problem we encounter is that when we measure “stuff” we observe patient-to-patient variation, and that is before we have made any changes.  Any impact that we may have is obscured by this “noise”.

<Phil> Ah, I see.  So if the our intervention generates a small impact then it will be more difficult to see amidst this background noise. Like trying to see fine detail in a fuzzy picture.

<Pete> Yes, exactly like that.  And it raises the issue of “errors”.  In medical research we talk about two different types of error; we make the first type of error when our actual impact is zero but we conclude from our data that we have made a difference; and we make the second type of error when we have made an impact but we conclude from our data that we have not.

<Phil> OK. So does that imply that the more “noise” we observe in our measure for-improvement before we make the change, the more likely we are to make one or other error?

<Pete> Precisely! So before we do the experiment we need to design it so that we reduce the probability of making both of these errors to an acceptably low level.  So that we can be assured that any conclusion we draw can be trusted.

<Phil> OK. So how exactly do you do that?

<Pete> We know that whenever there is “noise” and whenever we use samples then there will always be some risk of making one or other of the two types of error.  So we need to set a threshold for both. We have to state clearly how much confidence we need in our conclusion. For example, we often use the convention that we are willing to accept a 1 in 20 chance of making the Type I error.

<Phil> Let me check if I have heard you correctly. Suppose that, in reality, our change has no impact and we have set the risk threshold for a Type 1 error at 1 in 20, and suppose we repeat the same experiment 100 times – are you saying that we should expect about five of our experiments to show data that says our change has had the intended impact when in reality it has not?

<Pete> Yes. That is exactly it.

<Phil> OK.  But in practice we cannot repeat the experiment 100 times, so we just have to accept the 1 in 20 chance that we will make a Type 1 error, and we won’t know we have made it if we do. That feels a bit chancy. So why don’t we just set the threshold to 1 in 100 or 1 in 1000?

<Pete> We could, but doing that has a consequence.  If we reduce the risk of making a Type I error by setting our threshold lower, then we will increase the risk of making a Type II error.

<Phil> Ah! I see. The old swings-and-roundabouts problem. By the way, do these two errors have different names that would make it  easier to remember and to explain?

<Pete> Yes. The Type I error is called a False Positive. It is like concluding that a patient has a specific diagnosis when in reality they do not.

<Phil> And the Type II error is called a False Negative?

<Pete> Yes.  And we want to avoid both of them, and to do that we have to specify a separate risk threshold for each error.  The convention is to call the threshold for the false positive the alpha level, and the threshold for the false negative the beta level.

<Phil> OK. So now we have three things we need to be clear on before we can do our experiment: the size of the change that we need, the risk of the false positive that we are willing to accept, and the risk of a false negative that we are willing to accept.  Is that all we need?

<Pete> In medical research we learn that we need six pieces of the experimental design jigsaw before we can proceed. We only have three pieces so far.

<Phil> What are the other three pieces then?

<Pete> We need to know the average value of the metric we are intending to improve, because that is our baseline from which improvement is measured.  Improvements are often framed as a percentage improvement over the baseline.  And we need to know the spread of the data around that average, the “noise” that we referred to earlier.

<Phil> Ah, yes!  I forgot about the noise.  But that is only five pieces of the jigsaw. What is the last piece?

<Pete> The size of the sample.

<Phil> Eh?  Can’t we just go with whatever data we can realistically get?

<Pete> Sadly, no.  The size of the sample is how we control the risk of a false negative error.  The more data we have the lower the risk. This is referred to as the power of the experimental design.

<Phil> OK. That feels familiar. I know that the more experience I have of something the better my judgement gets. Is this the same thing?

<Pete> Yes. Exactly the same thing.

<Phil> OK. So let me see if I have got this. To know if the impact of the intervention matches our intention we need to design our experiment carefully. We need all six pieces of the experimental design jigsaw and they must all fall inside our circle of control. We can measure the baseline average and spread; we can specify the impact we will accept as useful; we can specify the risks we are prepared to accept of making the false positive and false negative errors; and we can collect the required amount of data after we have made the intervention so that we can trust our conclusion.

<Pete> Perfect! That is how we are taught to design research studies so that we can trust our results, and so that others can trust them too.

<Phil> So how do we decide how big the post-implementation data sample needs to be? I can see we need to collect enough data to avoid a false negative but we have to be pragmatic too. There would appear to be little value in collecting more data than we need. It would cost more and could delay knowing the answer to our question.

<Pete> That is precisely the trap than many inexperienced medical researchers fall into. They set their sample size according to what is achievable and affordable, and then they hope for the best!

<Phil> Well, we do the same. We analyse the data we have and we hope for the best.  In the magical metaphor we are asking our data analysts to pull a white rabbit out of the hat.  It sounds rather irrational and unpredictable when described like that! Have medical researchers learned a way to avoid this trap?

<Pete> Yes, it is a tool called a power calculator.

<Phil> Ooooo … a power tool … I like the sound of that … that would be a cool tool to have in our commissioning bag of tricks. It would be like a magic wand. Do you have such a thing?

<Pete> Yes.

<Phil> And do you understand how the power tool magic works well enough to explain to a “muggle”?

<Pete> Not really. To do that means learning some rather unfamiliar language and some rather counter-intuitive concepts.

<Phil> Is that the magical stuff I hear lurks between the covers of a medical statistics textbook?

<Pete> Yes. Scary looking mathematical symbols and unfathomable spells!

<Phil> Oh dear!  Is there another way for to gain a working understanding of this magic? Something a bit more pragmatic? A path that a ‘statistical muggle’ might be able to follow?

<Pete> Yes. It is called a simulator.

<Phil> You mean like a flight simulator that pilots use to learn how to control a jumbo jet before ever taking a real one out for a trip?

<Pete> Exactly like that.

<Phil> Do you have one?

<Pete> Yes. It was how I learned about this “stuff” … pragmatically.

<Phil> Can you show me?

<Pete> Of course.  But to do that we will need a bit more time, another coffee, and maybe a couple of those tasty looking Danish pastries.

<Phil> A wise investment I’d say.  I’ll get the the coffee and pastries, if you fire up the engines of the simulator.

The Lost Tribe

figures_lost_looking_at_map_anim_150_wht_15601

“Jingle Bells, Jingle Bells” announced Bob’s computer as he logged into the Webex meeting with Lesley.

<Bob> Hi Lesley, in case I forget later I’d like to wish you a Happy Christmas and hope that 2017 brings you new opportunity for learning and fun.

<Lesley> Thanks Bob, and I wish you the same. And I believe the blog last week pointed to some.

<Bob> Thank you and I agree;  every niggle is an opportunity for improvement and the “Houston we have a problem!” one is a biggie.

<Lesley> So how do we start on this one? It is massive!

<Bob> The same way we do on all niggles; we diagnose the root cause first. What do you feel they might be?

<Lesley> Well, following it backwards from your niggle, the board reports are created by the data analysts, and they will produce whatever they are asked to. It must be really irritating for them to have their work rubbished!

<Bob> Are you suggesting that they understand the flaws in what they are asked to do but keep quiet?

<Lesley> I am not sure they do, but there is clearly a gap between their intent and their impact. Where would they gain the insight? Do they have access to the sort of training I have am getting?

<Bob> That is a very good question, and until this week I would not have been able to answer, but an interesting report by the Health Foundation was recently published on that very topic. It is entitled “Understanding Analytical Capability In Health Care” and what it says is that there is a lost tribe of data analysts in the NHS.

<Lesley> How interesting! That certainly resonates with my experience.  All the data analysts I know seem to be hidden away behind their computers, caught in the cross-fire between between the boards and the wards, and very sensibly keeping their heads down and doing what they are asked to.

<Bob> That would certainly help to explain what we are seeing! And the good news is that Martin Bardsley, the author of the paper, has interviewed many people across the system, gathered their feedback, and offered some helpful recommendations.  Here is a snippet.

analysiscapability

<Lesley> I like these recommendations, especially the “in-work training programmes” and inclusion “in general management and leadership training“. But isn’t that one of the purposes of the CHIPs training?

<Bob> It is indeed, which is why it is good to see that Martin has specifically recommended it.

saasoftrecommended

<Lesley> Excellent! That means that my own investment in the CHIPs training has just gained in street value and that’s good for my CV. An unexpected early Xmas present. Thank you!

Pride and Joy

stick_figure_superhero_anim_150_wht_1857Have you heard the phrase “Pride comes before a fall“?

What does this mean? That the feeling of pride is the reason for the subsequent fall?

So by following that causal logic, if we do not allow ourselves to feel proud then we can avoid the fall?

And none of us like the feeling of falling and failing. We are fearful of that negative feeling, so with this simple trick we can avoid feeling bad. Yes?

But we all know the positive feeling of achievement – we feel pride when we have done good work, when our impact matches our intent.  Pride in our work.

Is that bad too?

Should we accept under-achievement and unexceptional mediocrity as the inevitable cost of avoiding the pain of possible failure?  Is that what we are being told to do here?


The phrase comes from the Bible, from the Book of Proverbs 16:18 to be precise.

proverb

And the problem here is that the phrase “pride comes before a fall” is not the whole proverb.

It has been simplified. Some bits have been omitted. And those omissions lead to ambiguity and the opportunity for obfuscation and re-interpretation.

pride_goes_before_a_fall
In the fuller New International Version we see a missing bit … the “haughty spirit” bit.  That is another way of saying “over-confident” or “arrogant”.


But even this “authorised” version is still ambiguous and more questions spring to mind:

Q1. What sort of pride are we referring to? Just the confidence version? What about the pride that follows achievement?

Q2. How would we know if our feeling of confidence is actually justified?

Q3. Does a feeling of confidence always precede a fall? Is that how we diagnose over-confidence? Retrospectively? Are there instances when we feel confident but we do not fail? Are there instances when we do not feel confident and then fail?

Q4. Does confidence cause the fall or it is just a temporal association? Is there something more fundamental that causes both high-confidence and low-competence?


There is a well known model called the Conscious-Competence model of learning which generates a sequence of four stages to achieving a new skill. Such as one we need to achieve our intended outcomes.

We all start in the “blissful ignorance” zone of unconscious incompetence.  Our unknowns are unknown to us.  They are blind spots.  So we feel unjustifiably confident.

hierarchy_of_competence

In this model the first barrier to progress is “wrong intuition” which means that we actually have unconscious assumptions that are distorting our perception of reality.

What we perceive makes sense to us. It is clear and obvious. We feel confident. We believe our own rhetoric.

But our unconscious assumptions can trick us into interpreting information incorrectly.  And if we derive decisions from unverified assumptions and invalid analysis then we may do the wrong thing and not achieve our intended outcome.  We may unintentionally cause ourselves to fail and not be aware of it.  But we are proud and confident.

Then the gap between our intent and our impact becomes visible to all and painful to us. So we are tempted to avoid the social pain of public failure by retreating behind the “Yes, But” smokescreen of defensive reasoning. The “doom loop” as it is sometimes called. The Victim Vortex. “Don’t name, shame and blame me, I was doing my best. I did not intent that to happen. To err is human”.


The good news is that this learning model also signposts a possible way out; a door in the black curtain of ignorance.  It suggests that we can learn how to correct our analysis by using feedback from reality to verify our rhetorical assumptions.  Those assumptions which pass the “reality check” we keep, those which fail the “reality check” we redesign and retest until they pass.  Bit by bit our inner rhetoric comes to more closely match reality and the wisdom of our decisions will improve.

And what we then see is improvement.  Our impact moves closer towards our intent. And we can justifiably feel proud of that achievement. We do not need to be best-compared-with-the-rest; just being better-than-we-were-before is OK. That is learning.

the_learning_curve

And this is how it feels … this is the Learning Curve … or the Nerve Curve as we call it.

What it says is that to be able to assess confidence we must also measure competence. Outcomes. Impact.

And to achieve excellence we have to be prepared to actively look for any gap between intent and impact.  And we have to be prepared to see it as an opportunity rather than as a threat. And we will need to be able to seek feedback and other people’s perspectives. And we need to be to open to asking for examples and explanations from those who have demonstrated competence.

It says that confidence is not a trustworthy surrogate for competence.

It says that we want the confidence that flows from competence because that is the foundation of trust.

Improvement flows at the speed of trust and seeing competence, confidence and trust growing is a joyous thing.

Pride and Joy are OK.

Arrogance and incompetence comes before a fall would be a better proverb.

Defensive Reasoning

monkey_on_back_anim_150_wht_11200

About 25 years ago a paper was published in the Harvard Business Review with the interesting title of “Teaching Smart People How To Learn

The uncomfortable message was that many people who are top of the intellectual rankings are actually very poor learners.

This sounds like a paradox.  How can people be high-achievers and yet be unable to learn?


Health care systems are stuffed full of super-smart, high-achieving professionals. The cream of educational crop. The top 2%. They are called “doctors”.

And we have a problem with improvement in health care … a big problem … the safety, delivery, quality and affordability of the NHS is getting worse. Not better.

Improvement implies change and change implies learning, so if smart people struggle to learn then could that explain why health care systems find self-improvement so difficult?

This paragraph from the 1991 HBR paper feels uncomfortably familiar:

defensive_reasoning_2

The author, Chris Argyris, refers to something called “single-loop learning” and if we translate this management-speak into the language of medicine it would come out as “treating the symptom and ignoring the disease“.  That is poor medicine.

Chris also suggests an antidote to this problem and gave it the label “double-loop learning” which if translated into medical speak becomes “diagnosis“.  And that is something that doctors can relate to because without a diagnosis, a justifiable treatment is difficult to formulate.


We need to diagnose the root cause(s) of the NHS disease.


The 1991 HBR paper refers back to an earlier 1977 HBR paper called Double Loop Learning in Organisations where we find the theory that underpins it.

The proposed hypothesis is that we all have cognitive models that we use to decide our actions (and in-actions), what I have referred to before as ChimpWare.  In it is a reference to a table published in a 1974 book and the message is that Single-Loop learning is a manifestation of a Model 1 theory-in-action.

defensive_reasoning_models


And if we consider the task that doctors are expected to do then we can empathize with their dominant Model 1 approach.  Health care is a dangerous business.  Doctors can cause a lot of unintentional harm – both physical and psychological.  Doctors are dealing with a very, very complex system – a human body – that they only partially understand.  No two patients are exactly the same and illness is a dynamic process.  Everyone’s expectations are high. We have come a long way since the days of blood-letting and leeches!  Failure is not tolerated.

Doctors are intelligent and competitive … they had to be to win the education race.

Doctors must make tough decisions and have to have tough conversations … many, many times … and yet not be consumed in the process.  They often have to suppress emotions to be effective.

Doctors feel the need to protect patients from harm – both physical and emotional.

And collectively they do a very good job.  Doctors are respected and trusted professionals.


But …  to quote Chris Argyris …

“Model I blinds people to their weaknesses. For instance, the six corporate presidents were unable to realize how incapable they were of questioning their assumptions and breaking through to fresh understanding. They were under the illusion that they could learn, when in reality they just kept running around the same track.”

This blindness is self-reinforcing because …

“All parties withheld information that was potentially threatening to themselves or to others, and the act of cover-up itself was closed to discussion.”


How many times have we seen this in the NHS?

The Mid-Staffordshire Hospital debacle that led to the Francis Report is all the evidence we need.


So what is the way out of this double-bind?

Chris gives us some hints with his Model II theory-in-use.

  1. Valid information – Study.
  2. Free and informed choice – Plan.
  3. Constant monitoring of the implementation – Do.

The skill required is to question assumptions and break through to fresh understanding and we can do that with design-led approach because that is what designers do.

They bring their unconscious assumptions up to awareness and ask “Is that valid?” and “What if” questions.

It is called Improvement-by-Design.

And the good news is that this Model II approach works in health care, and we know that because the evidence is accumulating.

 

Value, Verify and Validate

thinker_figure_unsolve_puzzle_150_wht_18309Many of the challenges that we face in delivering effective and affordable health care do not have well understood and generally accepted solutions.

If they did there would be no discussion or debate about what to do and the results would speak for themselves.

This lack of understanding is leading us to try to solve a complicated system design challenge in our heads.  Intuitively.

And trying to do it this way is fraught with frustration and risk because our intuition tricks us. It was this sort of challenge that led Professor Rubik to invent his famous 3D Magic Cube puzzle.

It is difficult enough to learn how to solve the Magic Cube puzzle by trial and error; it is even more difficult to attempt to do it inside our heads! Intuitively.


And we know the Rubik Cube puzzle is solvable, so all we need are some techniques, tools and training to improve our Rubik Cube solving capability.  We can all learn how to do it.


Returning to the challenge of safe and affordable health care, and to the specific problem of unscheduled care, A&E targets, delayed transfers of care (DTOC), finance, fragmentation and chronic frustration.

This is a systems engineering challenge so we need some systems engineering techniques, tools and training before attempting it.  Not after failing repeatedly.

se_vee_diagram

One technique that a systems engineer will use is called a Vee Diagram such as the one shown above.  It shows the sequence of steps in the generic problem solving process and it has the same sequence that we use in medicine for solving problems that patients present to us …

Diagnose, Design and Deliver

which is also known as …

Study, Plan, Do.


Notice that there are three words in the diagram that start with the letter V … value, verify and validate.  These are probably the three most important words in the vocabulary of a systems engineer.


One tool that a systems engineer always uses is a model of the system under consideration.

Models come in many forms from conceptual to physical and are used in two main ways:

  1. To assist the understanding of the past (diagnosis)
  2. To predict the behaviour in the future (prognosis)

And the process of creating a system model, the sequence of steps, is shown in the Vee Diagram.  The systems engineer’s objective is a validated model that can be trusted to make good-enough predictions; ones that support making wiser decisions of which design options to implement, and which not to.


So if a systems engineer presented us with a conceptual model that is intended to assist our understanding, then we will require some evidence that all stages of the Vee Diagram process have been completed.  Evidence that provides assurance that the model predictions can be trusted.  And the scope over which they can be trusted.


Last month a report was published by the Nuffield Trust that is entitled “Understanding patient flow in hospitals”  and it asserts that traffic flow on a motorway is a valid conceptual model of patient flow through a hospital.  Here is a direct quote from the second paragraph in the Executive Summary:

nuffield_report_01
Unfortunately, no evidence is provided in the report to support the validity of the statement and that omission should ring an alarm bell.

The observation that “the hospitals with the least free space struggle the most” is not a validation of the conceptual model.  Validation requires a concrete experiment.


To illustrate why observation is not validation let us consider a scenario where I have a headache and I take a paracetamol and my headache goes away.  I now have some evidence that shows a temporal association between what I did (take paracetamol) and what I got (a reduction in head pain).

But this is not a valid experiment because I have not considered the other seven possible combinations of headache before (Y/N), paracetamol (Y/N) and headache after (Y/N).

An association cannot be used to prove causation; not even a temporal association.

When I do not understand the cause, and I am without evidence from a well-designed experiment, then I might be tempted to intuitively jump to the (invalid) conclusion that “headaches are caused by lack of paracetamol!” and if untested this invalid judgement may persist and even become a belief.


Understanding causality requires an approach called counterfactual analysis; otherwise known as “What if?” And we can start that process with a thought experiment using our rhetorical model.  But we must remember that we must always validate the outcome with a real experiment. That is how good science works.

A famous thought experiment was conducted by Albert Einstein when he asked the question “If I were sitting on a light beam and moving at the speed of light what would I see?” This question led him to the Theory of Relativity which completely changed the way we now think about space and time.  Einstein’s model has been repeatedly validated by careful experiment, and has allowed engineers to design and deliver valuable tools such as the Global Positioning System which uses relativity theory to achieve high positional precision and accuracy.


So let us conduct a thought experiment to explore the ‘faster movement requires more space‘ statement in the case of patient flow in a hospital.

First, we need to define what we mean by the words we are using.

The phrase ‘faster movement’ is ambiguous.  Does it mean higher flow (more patients per day being admitted and discharged) or does it mean shorter length of stage (the interval between the admission and discharge events for individual patients)?

The phrase ‘more space’ is also ambiguous. In a hospital that implies physical space i.e. floor-space that may be occupied by corridors, chairs, cubicles, trolleys, and beds.  So are we actually referring to flow-space or storage-space?

What we have in this over-simplified statement is the conflation of two concepts: flow-capacity and space-capacity. They are different things. They have different units. And the result of conflating them is meaningless and confusing.


However, our stated goal is to improve understanding so let us consider one combination, and let us be careful to be more precise with our terminology, “higher flow always requires more beds“. Does it? Can we disprove this assertion with an example where higher flow required less beds (i.e. space-capacity)?

The relationship between flow and space-capacity is well understood.

The starting point is Little’s Law which was proven mathematically in 1961 by J.D.C. Little and it states:

Average work in progress = Average lead time  X  Average flow.

In the hospital context, work in progress is the number of occupied beds, lead time is the length of stay and flow is admissions or discharges per time interval (which must be the same on average over a long period of time).

(NB. Engineers are rather pedantic about units so let us check that this makes sense: the unit of WIP is ‘patients’, the unit of lead time is ‘days’, and the unit of flow is ‘patients per day’ so ‘patients’ = ‘days’ * ‘patients / day’. Correct. Verified. Tick.)

So, is there a situation where flow can increase and WIP can decrease? Yes. When lead time decreases. Little’s Law says that is possible. We have disproved the assertion.


Let us take the other interpretation of higher flow as shorter length of stay: i.e. shorter length of stay always requires more beds.  Is this correct? No. If flow remains the same then Little’s Law states that we will require fewer beds. This assertion is disproved as well.

And we need to remember that Little’s Law is proven to be valid for averages, does that shed any light on the source of our confusion? Could the assertion about flow and beds actually be about the variation in flow over time and not about the average flow?


And this is also well understood. The original work on it was done almost exactly 100 years ago by Agner Krarup Erlang and the problem he looked at was the quality of customer service of the early telephone exchanges. Specifically, how likely was the caller to get the “all lines are busy, please try later” response.

What Erlang showed was there there is a mathematical relationship between the number of calls being made (the demand), the probability of a call being connected first time (the service quality) and the number of telephone circuits and switchboard operators available (the service cost).


So it appears that we already have a validated mathematical model that links flow, quality and cost that we might use if we substitute ‘patients’ for ‘calls’, ‘beds’ for ‘telephone circuits’, and ‘being connected’ for ‘being admitted’.

And this topic of patient flow, A&E performance and Erlang queues has been explored already … here.

So a telephone exchange is a more valid model of a hospital than a motorway.

We are now making progress in deepening our understanding.


The use of an invalid, untested, conceptual model is sloppy systems engineering.

So if the engineering is sloppy we would be unwise to fully trust the conclusions.

And I share this feedback in the spirit of black box thinking because I believe that there are some valuable lessons to be learned here – by us all.


To vote for this topic please click here.
To subscribe to the blog newsletter please click here.
To email the author please click here.

Patient Traffic Engineering

motorway[Beep] Bob’s computer alerted him to Leslie signing on to the Webex session.

<Bob> Good afternoon Leslie, how are you? It seems a long time since we last chatted.

<Leslie> Hi Bob. I am well and it has been a long time. If you remember, I had to loop out of the Health Care Systems Engineering training because I changed job, and it has taken me a while to bring a lot of fresh skeptics around to the idea of improvement-by-design.

<Bob> Good to hear, and I assume you did that by demonstrating what was possible by doing it, delivering results, and describing the approach.

<Leslie> Yup. And as you know, even with objective evidence of improvement it can take a while because that exposes another gap, the one between intent and impact.  Many people get rather defensive at that point, so I have had to take it slowly. Some people get really fired up though.

 <Bob> Yes. Respect, challenge, patience and persistence are all needed. So, where shall we pick up?

<Leslie> The old chestnut of winter pressures and A&E targets.  Except that it is an all-year problem now and according to what I read in the news, everyone is predicting a ‘melt-down’.

<Bob> Did you see last week’s IS blog on that very topic?

<Leslie> Yes, I did!  And that is what prompted me to contact you and to re-start my CHIPs coaching.  It was a real eye opener.  I liked the black swan code-named “RC9” story, it makes it sound like a James Bond film!

<Bob> I wonder how many people dug deeper into how “RC9” achieved that rock-steady A&E performance despite a rising tide of arrivals and admissions?

<Leslie> I did, and I saw several examples of anti-carve-out design.  I have read though my notes and we have talked about carve out many times.

<Bob> Excellent. Being able to see the signs of competent design is just as important as the symptoms of inept design. So, what shall we talk about?

<Leslie> Well, by co-incidence I was sent a copy of of a report entitled “Understanding patient flow in hospitals” published by one of the leading Think Tanks and I confess it made no sense to me.  Can we talk about that?

<Bob> OK. Can you describe the essence of the report for me?

<Leslie> Well, in a nutshell it said that flow needs space so if we want hospitals to flow better we need more space, in other words more beds.

<Bob> And what evidence was presented to support that hypothesis?

<Leslie> The authors equated the flow of patients through a hospital to the flow of traffic on a motorway. They presented a table of numbers that made no sense to me, I think partly because there are no units stated for some of the numbers … I’ll email you a picture.

traffic_flow_dynamics

<Bob> I agree this is not a very informative table.  I am not sure what the definition of “capacity” is here and it may be that the authors may be equating “hospital bed” to “area of tarmac”.  Anyway, the assertion that hospital flow is equivalent to motorway flow is inaccurate.  There are some similarities and traffic engineering is an interesting subject, but they are not equivalent.  A hospital is more like a busy city with junctions, cross-roads, traffic lights, roundabouts, zebra crossings, pelican crossings and all manner of unpredictable factors such as cyclists and pedestrians. Motorways are intentionally designed without these “impediments”, for obvious reasons! A complex adaptive flow system like a hospital cannot be equated to a motorway. It is a dangerous over-simplification.

<Leslie> So, if the hospital-motorway analogy is invalid then the conclusions are also invalid?

<Bob> Sometimes, by accident, we get a valid conclusion from an invalid method. What were the conclusions?

<Leslie> That the solution to improving A&E performance is more space (i.e. hospital beds) but there is no more money to build them or people to staff them.  So the recommendations are to reduce volume, redesign rehabilitation and discharge processes, and improve IT systems.

<Bob> So just re-iterating the habitual exhortations and nothing about using well-understood systems engineering methods to accurately diagnose the actual root cause of the ‘symptoms’, which is likely to be the endemic carveoutosis multiforme, and then treat accordingly?

<Leslie> No. I could not find the term “carve out” anywhere in the document.

<Bob> Oh dear.  Based on that observation, I do not believe this latest Think Tank report is going to be any more effective than the previous ones.  Perhaps asking “RC9” to write an account of what they did and how they learned to do it would be more informative?  They did not reduce volume, and I doubt they opened more beds, and their annual report suggests they identified some space and flow carveoutosis and treated it. That is what a competent systems engineer would do.

<Leslie> Thanks Bob. Very helpful as always. What is my next step?

<Bob> Some ISP-2 brain-teasers, a juicy ISP-2 project, and some one day training workshops for your all-fired-up CHIPs.

<Leslie> Bring it on!


For more posts like this please vote here.
For more information please subscribe here.

DIKUW

This 100 second video of the late Russell Ackoff is solid gold!

In it he describes the DIKUW hierarchy – data, information, knowledge, understanding and wisdom – and how it is critical to put effectiveness before efficiency.

A wise objective is a purpose … the intended outcome … and a well designed system will be both effective and efficient.  That is the engineers definition of productivity.  Doing the right thing first, and doing it right second.

So how do we transform data into wisdom? What are needs to be added or taken away? What is the process?

Data is what we get from our senses.

To convert data into information we add context.

To convert information into knowledge we use memory.

To convert knowledge into understanding we need to learn-by-doing.

And the test of understanding is to be able to teach someone else what we know and to be able to support them developing an understanding through practice.

To convert understanding into wisdom requires years of experience of seeing, doing and teaching.

There are no short cuts.

So the sooner we start learning-by-doing the quicker we will develop the wisdom of purpose, and the understanding of process.


For more posts like this please vote here.
For more information please subscribe here.

Crash Test Dummy

CrashTestDummyThere are two complementary approaches to safety and quality improvement: desire and design.

In the improvement-by-desire world we use a suck-it-and-see approach to fix a problem.  It is called PDSA.

Sometimes this works and we pat ourselves on the back, and remember the learning for future use.

Sometimes it works for us but has a side effect: it creates a problem for someone else.  And we may not be aware of the unintended consequence unless someone shouts “Oi!” It may be too late by then of course.


The more parts in a system, and the more interconnected they are, the more likely it is that a well-intended suck-it-and-see change will create an unintended negative impact.

And in that situation our temptation is to … do nothing … and put up with the problems. It seems the safest option.


In the improvement-by-design world we choose to study first, and to find the causal roots of the system behaviour we are seeing.  Our first objective is a diagnosis.

With that we can propose rational design changes that we anticipate will deliver the improvement we seek without creating adverse effects.

But we have learned the hard way that our intuition can trick us … so we need a way to test our designs … a safe and controlled way.  We need a crash test dummy!


What they do is to deliberately experience our design in a controlled experiment, and what they generate for us is constructive feedback. What did work, and what did not.

A crash test dummy is tough and sensitive at the same time.  They do not break easily and yet they feel the pain and gain too.  They are resilient.


And with their feedback we can re-visit our design and improve it further, or we can use it to offer evidence-based assurance that our design is fit-for-purpose.

Safety and Quality Assurance is improvement-by-design. Diagnosis-and-treatment.

Safety and Quality Control is improvement-by-desire. Suck-and-see.

If you were a passenger or a patient … which option would you prefer?

Fragmentation Cost

figure_falling_with_arrow_17621The late Russell Ackoff used to tell a great story. It goes like this:

“A team set themselves the stretch goal of building the World’s Best Car.  So the put their heads together and came up with a plan.

First they talked to drivers and drew up a list of all the things that the World’s Best Car would need to have. Safety, speed, low fuel consumption, comfort, good looks, low emissions and so on.

Then they drew up a list of all the components that go into building a car. The engine, the wheels, the bodywork, the seats, and so on.

Then they set out on a quest … to search the world for the best components … and to bring the best one of each back.

Then they could build the World’s Best Car.

Or could they?

No.  All they built was a pile of incompatible parts. The WBC did not work. It was a futile exercise.


Then the penny dropped. The features in their wish-list were not associated with any of the separate parts. Their desired performance emerged from the way the parts worked together. The working relationships between the parts were as necessary as the parts themselves.

And a pile of average parts that work together will deliver a better performance than a pile of best parts that do not.

So the relationships were more important than the parts!


From this they learned that the quickest, easiest and cheapest way to degrade performance is to make working-well-together a bit more difficult.  Irrespective of the quality of the parts.


Q: So how do we reverse this degradation of performance?

A: Add more failure-avoidance targets of course!

But we just discovered that the performance is the effect of how the parts work well together?  Will another failure-metric-fueled performance target help? How will each part know what it needs to do differently – if anything?  How will each part know if the changes they have made are having the intended impact?

Fragmentation has a cost.  Fear, frustration, futility and ultimately financial failure.

So if performance is fading … the quality of the working relationships is a good place to look for opportunities for improvement.

Early Warning System

radar_screen_anim_300_clr_11649The most useful tool that a busy operational manager can have is a reliable and responsive early warning system (EWS).

One that alerts when something is changing and that, if missed or ignored, will cause a big headache in the future.

Rather like the radar system on an aircraft that beeps if something else is approaching … like another aircraft or the ground!


Operational managers are responsible for delivering stuff on time.  So they need a radar that tells them if they are going to deliver-on-time … or not.

And their on-time-delivery EWS needs to alert them soon enough that they have time to diagnose the ‘threat’, design effective plans to avoid it, decide which plan to use, and deliver it.

So what might an effective EWS for a busy operational manager look like?

  1. It needs to be reliable. No missed threats or false alarms.
  2. It needs to be visible. No tomes of text and tables of numbers.
  3. It needs to be simple. Easy to learn and quick to use.

And what is on offer at the moment?

The RAG Chart
This is a table that is coloured red, amber and green. Red means ‘failing’, green means ‘not failing’ and amber means ‘not sure’.  So this meets the specification of visible and simple, but it is reliable?

It appears not.  RAG charts do not appear to have helped to solve the problem.

A RAG chart is generated using historic data … so it tells us where we are now, not how we got here, where we are going or what else is heading our way.  It is a snapshot. One frame from the movie.  Better than complete blindness perhaps, but not much.

The SPC Chart
This is a statistical process control chart and is a more complicated beast.  It is a chart of how some measure of performance has changed over time in the past.  So like the RAG chart it is generated using historic data.  The advantage is that it is not just a snapshot of where were are now, it is a picture of story of how we got to where we are, so it offers the promise of pointing to where we may be heading.  It meets the specification of visible, and while more complicated than a RAG chart, it is relatively easy to learn and quick to use.

Luton_A&E_4Hr_YieldHere is an example. It is the SPC  chart of the monthly A&E 4-hour target yield performance of an acute NHS Trust.  The blue lines are the ‘required’ range (95% to 100%), the green line is the average and the red lines are a measure of variation over time.  What this charts says is: “This hospital’s A&E 4-hour target yield performance is currently acceptable, has been so since April 2012, and is improving over time.”

So that is much more helpful than a RAG chart (which in this case would have been green every month because the average was above the minimum acceptable level).


So why haven’t SPC charts replaced RAG charts in every NHS Trust Board Report?

Could there be a fly-in-the-ointment?

The answer is “Yes” … there is.

SPC charts are a quality audit tool.  They were designed nearly 100 years ago for monitoring the output quality of a process that is already delivering to specification (like the one above).  They are designed to alert the operator to early signals of deterioration, called ‘assignable cause signals’, and they prompt the operator to pay closer attention and to investigate plausible causes.

SPC charts are not designed for predicting if there is a flow problem looming over the horizon.  They are not designed for flow metrics that exhibit expected cyclical patterns.  They are not designed for monitoring metrics that have very skewed distributions (such as length of stay).  They are not designed for metrics where small shifts generate big cumulative effects.  They are not designed for metrics that change more slowly than the frequency of measurement.

And these are exactly the sorts of metrics that a busy operational manager needs to monitor, in reality, and in real-time.

Demand and activity both show strong cyclical patterns.

Lead-times (e.g. length of stay) are often very skewed by variation in case-mix and task-priority.

Waiting lists are like bank accounts … they show the cumulative sum of the difference between inflow and outflow.  That simple fact invalidates the use of the SPC chart.

Small shifts in demand, activity, income and expenditure can lead to big cumulative effects.

So if we abandon our RAG charts and we replace them with SPC charts … then we climb out of the RAG frying pan and fall into the SPC fire.

Oops!  No wonder the operational managers and financial controllers have not embraced SPC.


So is there an alternative that works better?  A more reliable EWS that busy operational managers and financial controllers can use?

Yes, there is, and here is a clue …

… but tread carefully …

… building one of these Flow-Productivity Early Warning Systems is not as obvious as it might first appear.  There are counter-intuitive traps for the unwary and the untrained.

You may need the assistance of a health care systems engineer (HCSE).

A Recipe for Chaos

growing_workload_anim_6858There is a very easy and quick-to-cook recipe for chaos.

All we have to do is to ensure that the maximum number of jobs that we can do in a given time is set equal to the average number of jobs that we are required to do in the same period of time.

Eh?

That does not make sense.  Our intuition says that looks like the perfect recipe for a hyper-efficient, zero-waste, zero idle-time design which is what we want.


I know it does, but it isn’t.  Our intuition is tricking us.

It is the recipe for chaos – and to prove it all we will have to do a real world experiment – because to prove it using maths is really difficult. So difficult in fact that the formula was not revealed until 1962 – by a mathematician called John Kingman while a postgraduate student at Pembroke College, Cambridge.

The empirical experiment is very easy to do – all we need is a single step process – and a stream of jobs to do.

And we could do it for real, or we can simulate it using an Excel spreadsheet – which is much quicker.


So we set up our spreadsheet to simulate a new job arriving every X minutes and each job taking X minutes to complete.

Our operator can only do one job at a time so if a job arrives and the operator is busy the job joins the back of a queue of jobs and waits.

When the operator finishes a job it takes the next one from the front of the queue, the one that has been waiting longest.

And if there is no queue the operator will wait until the next job arrives.

Simple.

And when we run simulation the we see that there is indeed no queue, no jobs waiting and the operator is always busy (i.e. 100% utilised). Perfection!

BUT ….

This is not a realistic scenario.  In reality there is always some random variation.  Not all jobs require the same length of time, and jobs do not arrive at precisely the right intervals.

No matter, our confident intuition tells us. It will average out.  Swings-and-roundabouts. Give-and-take.

It doesn’t.

And if you do not believe me just build the simple Excel model outlined above, verify that it works, then add some random variation to the time it takes to do each job … and observe what happens to the average waiting time.

What you will discover is that as soon as we add even a small amount of random variation we get a queue, and waiting and idle resources as well!

But not a steady, stable, predictable queue … Oh No! … We get an unsteady, unstable and unpredictable queue … we get chaos.

Try it.


So what? How does this abstract ‘queue theory’ apply to the real world?


Well, suppose we have a single black box system called ‘a hospital’ – patients arrive and we work hard to diagnose and treat them.  And so long as we have enough resource-time to do all the jobs we are OK. No unstable queues. No unpredictable waiting.

But time-costs-money and we have an annual cost improvement target (CIP) that we are required to meet so we need to ‘trim’ resource-time capacity to push up resource utilisation.  And we will call that an ‘efficiency improvement’ which is good … yes?

It isn’t actually.  I can just as easily push up my ‘utilisation’ by working slower, or doing stuff I do not need to, or by making mistakes that I have to check for and then correct.  I can easily make myself busier and delude myself I am working harder.

And we are also a victim of our own success … the better we do our job … the longer people live and the more workload they put on the health and social care system.

So we have the perfect storm … the perfect recipe for chaos … slowly rising demand … slowly shrinking budgets … and an inefficient ‘business’ design.

And that in a nutshell is the reason the NHS is descending into chaos.


So what is the solution?

Reduce demand? Stop people getting sick? Or make them sicker so they die quicker?

Increase budgets? Where will the money come from? Beg? Borrow? Steal? Economic growth?

Improve the design?  Now there’s a thought. But how? By using the same beliefs and behaviours that have created the current chaos?

Maybe we need to challenge some invalid beliefs and behaviours … and replace those that fail the Reality Test with some more effective ones.

System of Profound Knowledge

 

Don_Berwick_2016

This week I had the great pleasure of watching Dr Don Berwick sharing the story of his own ‘near religious experience‘ and his conversion to a belief that a Science of Improvement exists.  In 1986, Don attended one of W.Edwards Deming’s famous 4-day workshops.  It was an emotional roller coaster ride for Don! See here for a link to the whole video … it is worth watching all of it … the best bit is at the end.


Don outlines Deming’s System of Profound Knowledge (SoPK) and explores each part in turn. Here is a summary of SoPK from the Deming website.

Deming_SOPK

W.Edwards Deming was a physicist and statistician by training and his deep understanding of variation and appreciation for a system flows from that.  He was not trained as a biologist, psychologist or educationalist and those parts of the SoPK appear to have emerged later.

Here are the summaries of these parts – psychology first …

Deming_SOPK_Psychology

Neurobiologists and psychologists now know that we are the product of our experiences and our learning. What we think consciously is just the emergent tip of a much bigger cognitive iceberg. Most of what is happening is operating out of awareness. It is unconscious.  Our outward behaviour is just a visible manifestation of deeply ingrained values and beliefs that we have learned – and reinforced over and over again.  Our conscious thoughts are emergent effects.


So how do we learn?  How do we accumulate these values and beliefs?

This is the summary of Deming’s Theory of Knowledge …

Deming_SOPK_PDSA

But to a biologist, neuroanatomist, neurophysiologist, doctor, system designer and improvement coach … this does not feel correct.

At the most fundamental biological level we do not learn by starting with a theory; we start with a sensory.  The simplest element of the animal learning system – the nervous system – is called a reflex arc.

Sensor_Processor_EffectorFirst, we have some form of sensor to gather data from the outside world. Eyes, ears, smell, taste, touch, temperature, pain and so on.  Let us consider pain.

That signal is transmitted via a sensory nerve to the processor, the grey matter in this diagram, where it is filtered, modified, combined with other data, filtered again and a binary output generated. Act or Not.

If the decision is ‘Act’ then this signal is transmitted by a motor nerve to an effector, in this case a muscle, which results in an action.  The muscle twitches or contracts and that modifies the outside world – we pull away from the source of pain.  It is a harm avoidance design. Damage-limitation. Self-preservation.

Another example of this sensor-processor-effector design template is a knee-jerk reflex, so-named because if we tap the tendon just below the knee we can elicit a reflex contraction of the thigh muscle.  It is actually part of a very complicated, dynamic, musculoskeletal stability cybernetic control system that allows us to stand, walk and run … with almost no conscious effort … and no conscious awareness of how we are doing it.

But we are not born able to walk. As youngsters we do not start with a theory of how to walk from which we formulate a plan. We see others do it and we attempt to emulate them. And we fail repeatedly. Waaaaaaah! But we learn.


Human learning starts with study. We then process the sensory data using our internal mental model – our rhetoric; we then decide on an action based on our ‘current theory’; and then we act – on the external world; and then we observe the effect.  And if we sense a difference between our expectation and our experience then that triggers an ‘adjustment’ of our internal model – so next time we may do better because our rhetoric and the reality are more in sync.

The biological sequence is Study-Adjust-Plan-Do-Study-Adjust-Plan-Do and so on, until we have achieved our goal; or until we give up trying to learn.


So where does psychology come in?

Well, sometimes there is a bigger mismatch between our rhetoric and our reality. The world does not behave as we expect and predict. And if the mismatch is too great then we are left with feelings of confusion, disappointment, frustration and fear.  (PS. That is our unconscious mind telling us that there is a big rhetoric-reality mismatch).

We can see the projection of this inner conflict on the face of a child trying to learn to walk.  They screw up their faces in conscious effort, and they fall over, and they hurt themselves and they cry.  But they do not want us to do it for them … they want to learn to do it for themselves. Clumsily at first but better with practice. They get up and try again … and again … learning on each iteration.

Study-Adjust-Plan-Do over and over again.


There is another way to avoid the continual disappointment, frustration and anxiety of learning.  We can distort our sensation of external reality to better fit with our internal rhetoric.  When we do that the inner conflict goes away.

We learn how to tamper with our sensory filters until what we perceive is what we believe. Inner calm is restored (while outer chaos remains or increases). We learn the psychological defense tactics of denial and blame.  And we practice them until they are second-nature. Unconscious habitual reflexes. We build a reality-distortion-system (RDS) and it has a name – the Ladder of Inference.


And then one day, just by chance, somebody or something bypasses our RDS … and that is the experience that Don Berwick describes.

Don went to a 4-day workshop to hear the wisdom of W.Edwards Deming first hand … and he was forced by the reality he saw to adjust his inner model of the how the world works. His rhetoric.  It was a stormy transition!

The last part of his story is the most revealing.  It exposes that his unconscious mind got there first … and it was his conscious mind that needed to catch up.

Study-(Adjust)-Plan-Do … over-and-over again.


In Don’s presentation he suggests that Frederick W. Taylor is the architect of the failure of modern management. This is a commonly held belief, and everyone is equally entitled to an opinion, that is a definition of mutual respect.

But before forming an individual opinion on such a fundamental belief we should study the raw evidence. The words written by the person who wrote them not just the words written by those who filtered the reality through their own perceptual lenses.  Which we all do.

Culture – cause or effect?

The Harvard Business Review is worth reading because many of its articles challenge deeply held assumptions, and then back up the challenge with the pragmatic experience of those who have succeeded to overcome the limiting beliefs.

So the heading on the April 2016 copy that awaited me on my return from an Easter break caught my eye: YOU CAN’T FIX CULTURE.


 

HBR_April_2016

The successful leaders of major corporate transformations are agreed … the cultural change follows the technical change … and then the emergent culture sustains the improvement.

The examples presented include the Ford Motor Company, Delta Airlines, Novartis – so these are not corporate small fry!

The evidence suggests that the belief of “we cannot improve until the culture changes” is the mantra of failure of both leadership and management.


A health care system is characterised by a culture of risk avoidance. And for good reason. It is all too easy to harm while trying to heal!  Primum non nocere is a core tenet – first do no harm.

But, change and improvement implies taking risks – and those leaders of successful transformation know that the bigger risk by far is to become paralysed by fear and to do nothing.  Continual learning from many small successes and many small failures is preferable to crisis learning after a catastrophic failure!

The UK healthcare system is in a state of chronic chaos.  The evidence is there for anyone willing to look.  And waiting for the NHS culture to change, or pushing for culture change first appears to be a guaranteed recipe for further failure.

The HBR article suggests that it is better to stay focussed; to work within our circles of control and influence; to learn from others where knowledge is known, and where it is not – to use small, controlled experiments to explore new ground.


And I know this works because I have done it and I have seen it work.  Just by focussing on what is important to every member on the team; focussing on fixing what we could fix; not expecting or waiting for outside help; gathering and sharing the feedback from patients on a continuous basis; and maintaining patient and team safety while learning and experimenting … we have created a micro-culture of high safety, high efficiency, high trust and high productivity.  And we have shared the evidence via JOIS.

The micro-culture required to maintain the safety, flow, quality and productivity improvements emerged and evolved along with the improvements.

It was part of the effect, not the cause.


So the concept of ‘fix the system design flaws and the continual improvement culture will emerge’ seems to work at macro-system and at micro-system levels.

We just need to learn how to diagnose and treat healthcare system design flaws. And that is known knowledge.

So what is the next excuse?  Too busy?

Type II Error

figure_pointing_out_chart_data_150_clr_8005It was the time for Bob and Leslie’s regular Improvement Science coaching session.

<Leslie> Hi Bob, how are you today?

<Bob> I am getting over a winter cold but otherwise I am good.  And you?

<Leslie> I am OK and I need to talk something through with you because I suspect you will be able to help.

<Bob> OK. What is the context?

<Leslie> Well, one of the projects that I am involved with is looking at the elderly unplanned admission stream which accounts for less than half of our unplanned admissions but more than half of our bed days.

<Bob> OK. So what were you looking to improve?

<Leslie> We want to reduce the average length of stay so that we free up beds to provide resilient space-capacity to ease the 4-hour A&E admission delay niggle.

<Bob> That sounds like a very reasonable strategy.  So have you made any changes and measured any improvements?

<Leslie> We worked through the 6M Design® sequence. We studied the current system, diagnosed some time traps and bottlenecks, redesigned the ones we could influence, modified the system, and continued to measure to monitor the effect.

<Bob> And?

<Leslie> It feels better but the system behaviour charts do not show an improvement.

<Bob> Which charts, specifically?

<Leslie> The BaseLine XmR charts of average length of stay for each week of activity.

<Bob> And you locked the limits when you made the changes?

<Leslie> Yes. And there still were no red flags. So that means our changes have not had a significant effect. But it definitely feels better. Am I deluding myself?

<Bob> I do not believe so. Your subjective assessment is very likely to be accurate. Our Chimp OS 1.0 is very good at some things! I think the issue is with the tool you are using to measure the change.

<Leslie> The XmR chart?  But I thought that was THE tool to use?

<Bob> Like all tools it is designed for a specific purpose.  Are you familiar with the term Type II Error.

<Leslie> Doesn’t that come from research? I seem to remember that is the error we make when we have an under-powered study.  When our sample size is too small to confidently detect the change in the mean that we are looking for.

<Bob> A perfect definition!  The same error can happen when we are doing before and after studies too.  And when it does, we see the pattern you have just described: the process feels better but we do not see any red flags on our BaseLine© chart.

<Leslie> But if our changes only have a small effect how can it feel better?

<Bob> Because some changes have cumulative effects and we omit to measure them.

<Leslie> OMG!  That makes complete sense!  For example, if my bank balance is stable my average income and average expenses are balanced over time. So if I make a small-but-sustained improvement to my expenses, like using lower cost generic label products, then I will see a cumulative benefit over time to the balance, but not the monthly expenses; because the noise swamps the signal on that chart!

<Bob> An excellent analogy!

<Leslie> So the XmR chart is not the tool for this job. And if this is the only tool we have then we risk making a Type II error. Is that correct?

<Bob> Yes. We do still use an XmR chart first though, because if there is a big enough and fast enough shift then the XmR chart will reveal it.  If there is not then we do not give up just yet; we reach for our more sensitive shift detector tool.

<Leslie> Which is?

<Bob> I will leave you to ponder on that question.  You are a trained designer now so it is time to put your designer hat on and first consider the purpose of this new tool, and then create the outline a fit-for-purpose design.

<Leslie> OK, I am on the case!

Raising Awareness

SaveTheNHSGameThe first step in the process of improvement is raising awareness, and this has to be done carefully.

Most of us spend most of our time in a mental state called blissful ignorance.  We are happily unaware of the problems, and of their solutions.

Some of us spend some of our time in a different mental state called denial.

And we enter that from yet another mental state called painful awareness.

By raising awareness we are deliberately nudging ourselves, and others, out of our comfort zones.

But suddenly moving from blissful ignorance to painful awareness is not a comfortable transition. It feels like a shock. We feel confused. We feel vulnerable. We feel frightened. And we have a choice: freeze, flee or fight.

Freeze is shock. We feel paralysed by the mismatch between rhetoric and reality.

Flee is denial.  We run away from a new and uncomfortable reality.

Fight is anger. Directed first at others (blame) and then at ourselves (guilt).

It is this anger-passion that we must learn to channel and focus as determination to listen, learn and then lead.


The picture is of a recent awareness-raising event; it happened this week.

The audience is a group of NHS staff from across the depth and breadth of a health and social care system.

On the screen is the ‘Save the NHS Game’.  It is an interactive, dynamic flow simulation of a whole health care system; and its purpose is educational.  It is designed to illustrate the complex and counter-intuitive flow behaviour of a system of interdependent parts: primary care, an acute hospital, intermediate care, residential care, and so on.

We all became aware of a lot of unfamiliar concepts in a short space of time!

We all learned that a flow system can flip from calm to chaotic very quickly.

We all learned that a small change in one part of a system of interdependent parts can have a big effect in another part – either harmful or beneficial and often both.

We all learned that there is often a long time-lag between the change and the effect.

We all learned that we cannot reverse the effect just by reversing the change.

And we all learned that this high sensitivity to small changes is the result of the design of our system; i.e. our design.


Learning all that in one go was a bit of a shock!  Especially the part where we realised that we had, unintentionally, created near perfect conditions for chaos to emerge. Oh dear!

Denial felt like a very reasonable option; as did blame and guilt.

What emerged was a collective sense of determination.  “Let’s Do It!” captured the mood.


puzzle_lightbulb_build_PA_150_wht_4587The second step in the process of improvement is to show the door to the next phase of learning; the phase called ‘know how’.

This requires demonstrating that there is an another way out of the zone of painful awareness.  An alternative to denial.

This is where how-to-diagnose-and-correct-the-design-flaws needs to be illustrated. A step-at-a-time.

And when that happens it feels like a light bulb has been switched on.  What before was obscure and confusing suddenly becomes clear and understandable; and we say ‘Ah ha!’


So, if we deliberately raise awareness about a problem then, as leaders of change and improvement, we also have the responsibility to raise awareness about feasible solutions.


Because only then are we able to ask “Would we like to learn how to do this ourselves!”

And ‘Yes, please’ is what 68% of the people said after attending the awareness raising event.  Only 15% said ‘No, thank you’ and only 17% abstained.

Raising awareness is the first step to improvement.
Choosing the path out of the pain towards knowledge is the second.
And taking the first step on that path is the third.

Anti-Chaos

Hypothesis: Chaotic behaviour of healthcare systems is inevitable without more resources.

This appears to be a rather widely held belief, but what is the evidence?

Can we disprove this hypothesis?

Chaos is a predictable, emergent behaviour of many systems, both natural and man made, a discovery that was made rather recently, in the 1970’s.  Chaotic behaviour is not the same as random behaviour.  The fundamental difference is that random implies independence, while chaos requires the opposite: chaotic systems have interdependent parts.

Chaotic behaviour is complex and counter-intuitive, which may explain why it took so long for the penny to drop.


Chaos is a complex behaviour and it is tempting to assume that complicated structures always lead to complex behaviour.  But they do not.  A mechanical clock is a complicated structure but its behaviour is intentionally very stable and highly predictable – that is the purpose of a clock.  It is a fit-for-purpose design.

The healthcare system has many parts; it too is a complicated system; it has a complicated structure.  It is often seen to demonstrate chaotic behaviour.

So we might propose that a complicated system like healthcare could also be stable and predictable. If it were designed to be.


But there is another critical factor to take into account.

A mechanical clock only has inanimate cogs and springs that only obey the Laws of Physics – and they are neither adaptable nor negotiable.

A healthcare system is different. It is a living structure. It has patients, providers and purchasers as essential components. And the rules of how people work together are both negotiable and adaptable.

So when we are thinking about a healthcare system we are thinking about a complex adaptive system or CAS.

And that changes everything!


The good news is that adaptive behaviour can be a very effective anti-chaos strategy, if it is applied wisely.  The not-so-good news is that if it is not applied wisely then it can actually generate even more chaos.


Which brings us back to our hypothesis.

What if the chaos we are observing on out healthcare system is actually iatrogenic?

What if we are unintentionally and unconsciously generating it?

These questions require an answer because if we are unwittingly contributing to the chaos, with insight, understanding and wisdom we can intentionally calm it too.

These questions also challenge us to study our current way of thinking and working.  And in that challenge we will need to demonstrate a behaviour called humility. An ability to acknowledge that there are gaps in our knowledge and our understanding. A willingness to learn.


This all sounds rather too plausible in theory. What about an example?

Let us consider the highest flow process in healthcare: the outpatient clinic stream.

The typical design is a three-step process called the New-Test-Review design. This sequential design is simpler because the steps are largely independent of each other. And this simplicity is attractive because it is easier to schedule so is less likely to be chaotic. The downsides are the queues and delays between the steps and the risk of getting lost in the system. So if we are worried that a patient may have a serious illness that requires prompt diagnosis and treatment (e.g. cancer), then this simpler design is actually a potentially unsafe design.

A one-stop clinic is a better design because the New-Test-Review steps are completed in one visit, and that is better for everyone. But, a one-stop clinic is a more challenging scheduling problem because all the steps are now interdependent, and that is fertile soil for chaos to emerge.  And chaos is exactly what we often see.

Attending a chaotic one-stop clinic is frustrating experience for both patients and staff, and it is also less productive use of resources. So the chaos and cost appears to be price we are asked to pay for a quicker and safer design.

So is the one stop clinic chaos inevitable, or is it avoidable?

Simple observation of a one stop clinic shows that the chaos is associated with queues – which are visible as a waiting room full of patients and front-of-house staff working very hard to manage the queue and to signpost and soothe the disgruntled patients.

What if the one stop clinic queue and chaos is iatrogenic? What if it was avoidable without investing in more resources? Would the chaos evaporate? Would the quality improve?  Could we have a safer, calmer, higher quality and more productive design?

Last week I shared evidence that proved the one-stop clinic chaos was iatrogenic – by showing it was avoidable.

A team of healthcare staff were shown how to diagnose the cause of the queue and were then able to remove that cause, and to deliver the same outcome without the queue and the associated chaos.

And the most surprising lesson that the team learned was that they achieved this improvement using the same resources as before; and that those resources also felt the benefit of the chaos evaporating. Their work was easier, calmer and more predictable.

The impossible-without-more-resources hypothesis had been disproved.

So, where else in our complicated and complex healthcare system might we apply anti-chaos?

Everywhere?


And for more about complexity science see Santa Fe Institute

The Magic Black Box

stick_figure_magic_carpet_150_wht_5040It was the appointed time for Bob and Leslie’s regular coaching session as part of the improvement science practitioner programme.

<Leslie> Hi Bob, I am feeling rather despondent today so please excuse me in advance if you hear a lot of “Yes, but …” language.

<Bob> I am sorry to hear that Leslie. Do you want to talk about it?

<Leslie> Yes, please.  The trigger for my gloom was being sent on a mandatory training workshop.

<Bob> OK. Training to do what?

<Leslie> Outpatient demand and capacity planning!

<Bob> But you know how to do that already, so what is the reason you were “sent”?

<Leslie> Well, I am no longer sure I know how to it.  That is why I am feeling so blue.  I went more out of curiosity and I came away utterly confused and with my confidence shattered.

<Bob> Oh dear! We had better start at the beginning.  What was the purpose of the workshop?

<Leslie> To train everyone in how to use an Outpatient Demand and Capacity planning model, an Excel one that we were told to download along with the User Guide.  I think it is part of a national push to improve waiting times for outpatients.

<Bob> OK. On the surface that sounds reasonable. You have designed and built your own Excel flow-models already; so where did the trouble start?

<Leslie> I will attempt to explain.  This was a paragraph in the instructions. I felt OK with this because my Improvement Science training has given me a very good understanding of basic demand and capacity theory.

IST_DandC_Model_01<Bob> OK.  I am guessing that other delegates may have felt less comfortable with this. Was that the case?

<Leslie> The training workshops are targeted at Operational Managers and the ones I spoke to actually felt that they had a good grasp of the basics.

<Bob> OK. That is encouraging, but a warning bell is ringing for me. So where did the trouble start?

<Leslie> Well, before going to the workshop I decided to read the User Guide so that I had some idea of how this magic tool worked.  This is where I started to wobble – this paragraph specifically …

IST_DandC_Model_02

<Bob> H’mm. What did you make of that?

<Leslie> It was complete gibberish to me and I felt like an idiot for not understanding it.  I went to the workshop in a bit of a panic and hoped that all would become clear. It didn’t.

<Bob> Did the User Guide explain what ‘percentile’ means in this context, ideally with some visual charts to assist?

<Leslie> No and the use of ‘th’ and ‘%’ was really confusing too.  After that I sort of went into a mental fog and none of the workshop made much sense.  It was all about practising using the tool without any understanding of how it worked. Like a black magic box.


<Bob> OK.  I can see why you were confused, and do not worry, you are not an idiot.  It looks like the author of the User Guide has unwittingly used some very confusing and ambiguous terminology here.  So can you talk me through what you have to do to use this magic box?

<Leslie> First we have to enter some of our historical data; the number of new referrals per week for a year; and the referral and appointment dates for all patients for the most recent three months.

<Bob> OK. That sounds very reasonable.  A run chart of historical demand and the raw event data for a Vitals Chart® is where I would start the measurement phase too – so long as the data creates a valid 3 month reporting window.

<Leslie> Yes, I though so too … but that is not how the black box model seems to work. The weekly demand is used to draw an SPC chart, but the event data seems to disappear into the innards of the black box, and recommendations pop out of it.

<Bob> Ah ha!  And let me guess the relationship between the term ‘percentile’ and the SPC chart of weekly new demand was not explained?

<Leslie> Spot on.  What does percentile mean?


<Bob> It is statistics jargon. Remember that we have talked about the distribution of the data around the average on a BaseLine chart; and how we use the histogram feature of BaseLine to show it visually.  Like this example.

IST_DandC_Model_03<Leslie> Yes. I recognise that. This chart shows a stable system of demand with an average of around 150 new referrals per week and the variation distributed above and below the average in a symmetrical pattern, falling off to zero around the upper and lower process limits.  I believe that you said that over 99% will fall within the limits.

<Bob> Good.  The blue histogram on this chart is called a probability distribution function, to use the terminology of a statistician.

<Leslie> OK.

<Bob> So, what would happen if we created a Pareto chart of demand using the number of patients per week as the categories and ignoring the time aspect? We are allowed to do that if the behaviour is stable, as this chart suggests.

<Leslie> Give me a minute, I will need to do a rough sketch. Does this look right?

IST_DandC_Model_04

<Bob> Perfect!  So if you now convert the Y-axis to a percentage scale so that 52 weeks is 100% then where does the average weekly demand of about 150 fall? Read up from the X-axis to the line then across to the Y-axis.

<Leslie> At about 26 weeks or 50% of 52 weeks.  Ah ha!  So that is what a percentile means!  The 50th percentile is the average, the zeroth percentile is around the lower process limit and the 100th percentile is around the upper process limit!

<Bob> In this case the 50th percentile is the average, it is not always the case though.  So where is the 85th percentile line?

<Leslie> Um, 52 times 0.85 is 44.2 which, reading across from the Y-axis then down to the X-axis gives a weekly demand of about 170 per week.  That is about the same as the average plus one sigma according to the run chart.

<Bob> Excellent. The Pareto chart that you have drawn is called a cumulative probability distribution function … and that is usually what percentiles refer to. Comparative Statisticians love these but often omit to explain their rationale to non-statisticians!


<Leslie> Phew!  So, now I can see that the 65th percentile is just above average demand, and 85th percentile is above that.  But in the confusing paragraph how does that relate to the phrase “65% and 85% of the time”?

<Bob> It doesn’t. That is the really, really confusing part of  that paragraph. I am not surprised that you looped out at that point!

<Leslie> OK. Let us leave that for another conversation.  If I ignore that bit then does the rest of it make sense?

<Bob> Not yet alas. We need to dig a bit deeper. What would you say are the implications of this message?


<Leslie> Well.  I know that if our flow-capacity is less than our average demand then we will guarantee to create an unstable queue and chaos. That is the Flaw of Averages trap.

<Bob> OK.  The creator of this tool seems to know that.

<Leslie> And my outpatient manager colleagues are always complaining that they do not have enough slots to book into, so I conclude that our current flow-capacity is just above the 50th percentile.

<Bob> A reasonable hypothesis.

<Leslie> So to calm the chaos the message is saying I will need to increase my flow capacity up to the 85th percentile of demand which is from about 150 slots per week to 170 slots per week. An increase of 7% which implies a 7% increase in costs.

<Bob> Good.  I am pleased that you did not fall into the intuitive trap that a increase from the 50th to the 85th percentile implies a 35/50 or 70% increase! Your estimate of 7% is a reasonable one.

<Leslie> Well it may be theoretically reasonable but it is not practically possible. We are exhorted to reduce costs by at least that amount.

<Bob> So we have a finance versus governance bun-fight with the operational managers caught in the middle: FOG. That is not the end of the litany of woes … is there anything about Did Not Attends in the model?


<Leslie> Yes indeed! We are required to enter the percentage of DNAs and what we do with them. Do we discharge them or re-book them.

<Bob> OK. Pragmatic reality is always much more interesting than academic rhetoric and this aspect of the real system rather complicates things, at least for a comparative statistician. This is where the smoke and mirrors will appear and they will be hidden inside the black magic box.  To solve this conundrum we need to understand the relationship between demand, capacity, variation and yield … and it is rather counter-intuitive.  So, how would you approach this problem?

<Leslie> I would use the 6M Design® framework and I would start with a map and not with a model; least of all a magic black box one that I did not design, build and verify myself.

<Bob> And how do you know that will work any better?

<Leslie> Because at the One Day ISP Workshop I saw it work with my own eyes. The queues, waits and chaos just evaporated.  And it cost nothing.  We already had more than enough “capacity”.

<Bob> Indeed you did.  So shall we do this one as an ISP-2 project?

<Leslie> An excellent suggestion.  I already feel my confidence flowing back and I am looking forward to this new challenge. Thank you again Bob.

Hot and Cold

stick_figure_on_cloud_150_wht_9604Last week Bob and Leslie were exploring the data analysis trap called a two-points-in-time comparison: as illustrated by the headline “This winter has not been as bad as last … which proves that our winter action plan has worked.

Actually it doesn’t.

But just saying that is not very helpful. We need to explain the reason why this conclusion is invalid and therefore potentially dangerous.


So here is the continuation of Bob and Leslie’s conversation.

<Bob> Hi Leslie, have you been reflecting on the two-points-in-time challenge?

<Leslie> Yes indeed, and you were correct, I did know the answer … I just didn’t know I knew if you get my drift.

<Bob> Yes, I do. So, are you willing to share your story?

<Leslie> OK, but before I do that I would like to share what happened when I described what we talked about to some colleagues.  They sort of got the idea but got lost in the unfamiliar language of ‘variance’ and I realized that I needed an example to illustrate.

<Bob> Excellent … what example did you choose?

<Leslie> The UK weather – or more specifically the temperature.  My reasons for choosing this were many: first it is something that everyone can relate to; secondly it has strong seasonal cycle; and thirdly because the data is readily available on the Internet.

<Bob> OK, so what specific question were you trying to answer and what data did you use?

<Leslie> The question was “Are our winters getting warmer?” and my interest in that is because many people assume that the colder the winter the more people suffer from respiratory illness and the more that go to hospital … contributing to the winter A&E and hospital pressures.  The data that I used was the maximum monthly temperature from 1960 to the present recorded at our closest weather station.

<Bob> OK, and what did you do with that data?

<Leslie> Well, what I did not do was to compare this winter with last winter and draw my conclusion from that!  What I did first was just to plot-the-dots … I created a time-series chart … using the BaseLine© software.

MaxMonthTemp1960-2015

And it shows what I expected to see, a strong, regular, 12-month cycle, with peaks in the summer and troughs in the winter.

<Bob> Can you explain what the green and red lines are and why some dots are red?

<Leslie> Sure. The green line is the average for all the data. The red lines are called the upper and lower process limits.  They are calculated from the data and what they say is “if the variation in this data is random then we will expect more than 99% of the points to fall between these two red lines“.

<Bob> So, we have 55 years of monthly data which is nearly 700 points which means we would expect fewer than seven to fall outside these lines … and we clearly have many more than that.  For example, the winter of 1962-63 and the summer of 1976 look exceptional – a run of three consecutive dots outside the red lines. So can we conclude the variation we are seeing is not random?

<Leslie> Yes, and there is more evidence to support that conclusion. First is the reality check … I do not remember either of those exceptionally cold or hot years personally, so I asked Dr Google.

BigFreeze_1963This picture from January 1963 shows copper telephone lines that are so weighed down with ice, and for so long, that they have stretched down to the ground.  In this era of mobile phones we forget this was what telecommunication was like!

 

 

HeatWave_1976

And just look at the young Michal Fish in the Summer of ’76! Did people really wear clothes like that?

And there is more evidence on the chart. The red dots that you mentioned are indicators that BaseLine© has detected other non-random patterns.

So the large number of red dots confirms our Mark I Eyeball conclusion … that there are signals mixed up with the noise.

<Bob> Actually, I do remember the Summer of ’76 – it was the year I did my O Levels!  And your signals-in-the-noise phrase reminds me of SETI – the search for extra-terrestrial intelligence!  I really enjoyed the 1997 film of Carl Sagan’s book Contact with Jodi Foster playing the role of the determined scientist who ends up taking a faster-than-light trip through space in a machine designed by ET and built by humans. And especially the line about 10 minutes from the end when those-in-high-places who had discounted her story as “unbelievable” realized they may have made an error … the line ‘Yes, that is interesting isn’t it’.

<Leslie> Ha ha! Yes. I enjoyed that film too. It had lots of great characters – her glory seeking boss; the hyper-suspicious head of national security who militarized the project; the charismatic anti-hero; the ranting radical who blew up the first alien machine; and John Hurt as her guardian angel. I must watch it again.

Anyway, back to the story. The problem we have here is that this type of time-series chart is not designed to extract the overwhelming cyclical, annual pattern so that we can search for any weaker signals … such as a smaller change in winter temperature over a longer period of time.

<Bob>Yes, that is indeed the problem with these statistical process control charts.  SPC charts were designed over 60 years ago for process quality assurance in manufacturing not as a diagnostic tool in a complex adaptive system such a healthcare. So how did you solve the problem?

<Leslie> I realized that it was the regularity of  the cyclical pattern that was the key.  I realized that I could use that to separate out the annual cycle and to expose the weaker signals.  I did that using the rational grouping feature of BaseLine© with the month-of-the-year as the group.

MaxMonthTemp1960-2015_ByMonth

Now I realize why the designers of the software put this feature in! With just one mouse click the story jumped out of the screen!

<Bob> OK. So can you explain what we are looking at here?

<Leslie> Sure. This chart shows the same data as before except that I asked BaseLine© first to group the data by month and then to create a mini-chart for each month-group independently.  Each group has its own average and process limits.  So if we look at the pattern of the averages, the green lines, we can clearly see the annual cycle.  What is very obvious now is that the process limits for each sub-group are much narrower, and that there are now very few red points  … other than in the groups that are coloured red anyway … a niggle that the designers need to nail in my opinion!

<Bob> I will pass on your improvement suggestion! So are you saying that the regular annual cycle has accounted for the majority of the signal in the previous chart and that now we have extracted that signal we can look for weaker signals by looking for red flags in each monthly group?

<Leslie> Exactly so.  And the groups I am most interested in are the November to March ones.  So, next I filtered out the November data and plotted it as a separate chart; and I then used another cool feature of BaseLine© called limit locking.

MaxTempNov1960-2015_LockedLimits

What that means is that I have used the November maximum temperature data for the first 30 years to get the baseline average and natural process limits … and we can see that there are no red flags in that section, no obvious signals.  Then I locked these limits at 1990 and this tells BaseLine© to compare the subsequent 25 years of data against these projected limits.  That exposed a lot of signal flags, and we can clearly see that most of the points in the later section are above the projected average from the earlier one.  This confirms that there has been a significant increase in November maximum temperature over this 55 year period.

<Bob> Excellent! You have answered part of your question. So what about December onwards?

<Leslie> I was on a roll now! I also noticed from my second chart that the December, January and February groups looked rather similar so I filtered that data out and plotted them as a separate chart.

MaxTempDecJanFeb1960-2015_GroupedThese were indeed almost identical so I lumped them together as a ‘winter’ group and compared the earlier half with the later half using another BaseLine© feature called segmentation.

MaxTempDecJanFeb1960-2015-SplitThis showed that the more recent winter months have a higher maximum temperature … on average. The difference is just over one degree Celsius. But it also shows that that the month-to-month and year-to-year variation still dominates the picture.

<Bob> Which implies?

<Leslie> That, with data like this, a two-points-in-time comparison is meaningless.  If we do that we are just sampling random noise and there is no useful information in noise. Nothing that we can  learn from. Nothing that we can justify a decision with.  This is the reason the ‘this year was better than last year’ statement is meaningless at best; and dangerous at worst.  Dangerous because if we draw an invalid conclusion, then it can lead us to make an unwise decision, then decide a counter-productive action, and then deliver an unintended outcome.

By doing invalid two-point comparisons we can too easily make the problem worse … not better.

<Bob> Yes. This is what W. Edwards Deming, an early guru of improvement science, referred to as ‘tampering‘.  He was a student of Walter A. Shewhart who recognized this problem in manufacturing and, in 1924, invented the first control chart to highlight it, and so prevent it.  My grandmother used the term meddling to describe this same behavior … and I now use that term as one of the eight sources of variation. Well done Leslie!

The Two-Points-In-Time Comparison Trap

comparing_information_anim_5545[Bzzzzzz] Bob’s phone vibrated to remind him it was time for the regular ISP remote coaching session with Leslie. He flipped the lid of his laptop just as Leslie joined the virtual meeting.

<Leslie> Hi Bob, and Happy New Year!

<Bob> Hello Leslie and I wish you well in 2016 too.  So, what shall we talk about today?

<Leslie> Well, given the time of year I suppose it should be the Winter Crisis.  The regularly repeating annual winter crisis. The one that feels more like the perpetual winter crisis.

<Bob> OK. What specifically would you like to explore?

<Leslie> Specifically? The habit of comparing of this year with last year to answer the burning question “Are we doing better, the same or worse?”  Especially given the enormous effort and political attention that has been focused on the hot potato of A&E 4-hour performance.

<Bob> Aaaaah! That old chestnut! Two-Points-In-Time comparison.

<Leslie> Yes. I seem to recall you usually add the word ‘meaningless’ to that phrase.

<Bob> H’mm.  Yes.  It can certainly become that, but there is a perfectly good reason why we do this.

<Leslie> Indeed, it is because we see seasonal cycles in the data so we only want to compare the same parts of the seasonal cycle with each other. The apples and oranges thing.

<Bob> Yes, that is part of it. So what do you feel is the problem?

<Leslie> It feels like a lottery!  It feels like whether we appear to be better or worse is just the outcome of a random toss.

<Bob> Ah!  So we are back to the question “Is the variation I am looking at signal or noise?” 

<Leslie> Yes, exactly.

<Bob> And we need a scientifically robust way to answer it. One that we can all trust.

<Leslie> Yes.

<Bob> So how do you decide that now in your improvement work?  How do you do it when you have data that does not show a seasonal cycle?

<Leslie> I plot-the-dots and use an XmR chart to alert me to the presence of the signals I am interested in – especially a change of the mean.

<Bob> Good.  So why can we not use that approach here?

<Leslie> Because the seasonal cycle is usually a big signal and it can swamp the smaller change I am looking for.

<Bob> Exactly so. Which is why we have to abandon the XmR chart and fall back the two points in time comparison?

<Leslie> That is what I see. That is the argument I am presented with and I have no answer.

<Bob> OK. It is important to appreciate that the XmR chart was not designed for doing this.  It was designed for monitoring the output quality of a stable and capable process. It was designed to look for early warning signs; small but significant signals that suggest future problems. The purpose is to alert us so that we can identify the root causes, correct them and the avoid a future problem.

<Leslie> So we are using the wrong tool for the job. I sort of knew that. But surely there must be a better way than a two-points-in-time comparison!

<Bob> There is, but first we need to understand why a TPIT is a poor design.

<Leslie> Excellent. I’m all ears.

<Bob> A two point comparison is looking at the difference between two values, and that difference can be positive, zero or negative.  In fact, it is very unlikely to be zero because noise is always present.

<Leslie> OK.

<Bob> Now, both of the values we are comparing are single samples from two bigger pools of data.  It is the difference between the pools that we are interested in but we only have single samples of each one … so they are not measurements … they are estimates.

<Leslie> So, when we do a TPIT comparison we are looking at the difference between two samples that come from two pools that have inherent variation and may or may not actually be different.

<Bob> Well put.  We give that inherent variation a name … we call it variance … and we can quantify it.

<Leslie> So if we do many TPIT comparisons then they will show variation as well … for two reasons; first because the pools we are sampling have inherent variation; and second just from the process of sampling itself.  It was the first lesson in the ISP-1 course.

<Bob> Well done!  So the question is: “How does the variance of the TPIT sample compare with the variance of the pools that the samples are taken from?”

<Leslie> My intuition tells me that it will be less because we are subtracting.

<Bob> Your intuition is half-right.  The effect of the variation caused by the signal will be less … that is the rationale for the TPIT after all … but the same does not hold for the noise.

<Leslie> So the noise variation in the TPIT is the same?

<Bob> No. It is increased.

<Leslie> What! But that would imply that when we do this we are less likely to be able to detect a change because a small shift in signal will be swamped by the increase in the noise!

<Bob> Precisely.  And the degree that the variance increases by is mathematically predictable … it is increased by a factor of two.

<Leslie> So as we usually present variation as the square root of the variance, to get it into the same units as the metric, then that will be increased by the square root of two … 1.414

<Bob> Yes.

<Leslie> I need to put this counter-intuitive theory to the test!

<Bob> Excellent. Accept nothing on faith. Always test assumptions. And how will you do that?

<Leslie> I will use Excel to generate a big series of normally distributed random numbers; then I will calculate a series of TPIT differences using a fixed time interval; then I will calculate the means and variations of the two sets of data; and then I will compare them.

<Bob> Excellent.  Let us reconvene in ten minutes when you have done that.


10 minutes later …


<Leslie> Hi Bob, OK I am ready and I would like to present the results as charts. Is that OK?

<Bob> Perfect!

<Leslie> Here is the first one.  I used our A&E performance data to give me some context. We know that on Mondays we have an average of 210 arrivals with an approximately normal distribution and a standard deviation of 44; so I used these values to generate the random numbers. Here is the simulated Monday Arrivals chart for two years.

TPIT_SourceData

<Bob> OK. It looks stable as we would expect and I see that you have plotted the sigma levels which look to be just under 50 wide.

<Leslie> Yes, it shows that my simulation is working. So next is the chart of the comparison of arrivals for each Monday in Year 2 compared with the corresponding week in Year 1.

TPIT_DifferenceData <Bob> Oooookaaaaay. What have we here?  Another stable chart with a mean of about zero. That is what we would expect given that there has not been a change in the average from Year 1 to Year 2. And the variation has increased … sigma looks to be just over 60.

<Leslie> Yes!  Just as the theory predicted.  And this is not a spurious answer. I ran the simulation dozens of times and the effect is consistent!  So, I am forced by reality to accept the conclusion that when we do two-point-in-time comparisons to eliminate a cyclical signal we will reduce the sensitivity of our test and make it harder to detect other signals.

<Bob> Good work Leslie!  Now that you have demonstrated this to yourself using a carefully designed and conducted simulation experiment, you will be better able to explain it to others.

<Leslie> So how do we avoid this problem?

<Bob> An excellent question and one that I will ask you to ponder on until our next chat.  You know the answer to this … you just need to bring it to conscious awareness.


 

Early Adoption

Rogers_CurveThe early phases of a transformation are where most fall by the wayside.

And the failure rate is horrifying – an estimated 80% of improvement initiatives fail to achieve their goals.

The recent history of the NHS is littered with the rusting wreckage of a series of improvement bandwagons.  Many who survived the crashes are too scarred and too scared to try again.


Transformation and improvement imply change which implies innovation … new ways of thinking, new ways of behaving, new techniques, new tools, and new ways of working.

And it has been known for over 50 years that innovation spreads in a very characteristic way. This process was described by Everett Rogers in a book called ‘Diffusion of Innovations‘ and is described visually in the diagram above.

The horizontal axis is a measure of individual receptiveness to the specific innovation … and the labels are behaviours: ‘I exhibit early adopter behaviour‘ (i.e. not ‘I am an early adopter’).

What Roger’s discovered through empirical observation was that in all cases the innovation diffuses from left-to-right; from innovation through early adoption to the ‘silent’ majority.


Complete diffusion is not guaranteed though … there are barriers between the phases.

One barrier is between innovation and early adoption.

There are many innovations that we never hear about and very often the same innovation appears in many places and often around the same time.

This innovation-adoption barrier is caused by two things:
1) most are not even aware of the problem … they are blissfully ignorant;
2) news of the innovation is not shared widely enough.

Innovators are sensitive people.  They sense there is a problem long before others do. They feel the fear and the excitement of need for innovation. They challenge their own assumptions and they actively seek solutions. They swim against the tide of ignorance, disinterest, skepticism and often toxic cynicism.  So when they do discover a way forward they often feel nervous about sharing it. They have learned (the hard way) that the usual reaction is to be dismissed and discounted.  Most people do not like to learn about unknown problems and hazards; and they like it even less to learn that there are solutions that they neither recognise nor understand.


But not everyone.

There is a group called the early adopters who, like the innovators, are aware of the problem. They just do not share the innovator’s passion to find a solution … irrespective of the risks … so they wait … their antennae tuned for news that a solution has been found.

Then they act.

And they act in one of two ways:

1) Talkers … re-transmit the news of the problem and the discovery of a generic solution … which is essential in building awareness.

2) Walkers … try the innovative approach themselves and in so doing learn a lot about their specific problem and the new ways to solving it.

And it is the early adopters that do both of these actions that are the most effective and the most valuable to everyone else.  Those that talk-the-new-walk and walk-the-new-talk.

And we can identify who they are because they will be able to tell stories of how they have applied the innovation in their world; and the results that they have achieved; and how they achieved them; and what worked well; and what did not; and what they learned; and how they evolved and applied the innovation to meet their specific needs.

They are the leaders, the coaches and the teachers of improvement and transformation.

They See One, Do Some, and Teach Many.

The early adopters are the bridge across the Innovation and Transformation Chasm.

Yield

Dr_Bob_ThumbnailA recurring theme this week has been the concept of ‘quality’.

And it became quickly apparent that a clear definition of quality is often elusive.

Which seems to have led to a belief that quality is difficult to measure because it is subjective and has no precise definition.

The science of quality improvement is nearly 100 years old … and it was shown a long time ago, in 1924 in fact, that it is rather easy to measure quality – objectively and scientifically.

The objective measure of quality is called “yield”.

To measure yield we simply ask all our customers this question:

Did your experience meet your expectation?” 

If the answer is ‘Yes’ then we count this as OK; if it is ‘No’ then we count it as Not OK.

Yield is the ratio of the OKs divided by the number of customers who answered.


But this tried-and-tested way of measuring quality has a design flaw:

Where does a customer get their expectation from?

Because if a customer has an unrealistically high expectation then whatever we do will be perceived by them as Not OK.

So to consistently deliver a high quality service (i.e. high yield) we need to be able to influence both the customer experience and the customer expectation.


If we set our sights on a worthwhile and realistic expectation and we broadcast that to our customers, then we also need a way of avoiding their disappointment … that our objective quality outcome audit may reveal.

One way to defuse disappointment is to set a low enough expectation … which is, sadly, the approach adopted by naysayers,  complainers, cynics and doom-mongers. The inept.

That is not the path to either improvement or to excellence. It is the path to apathy.

A better approach is to set ourselves some internal standards of expectation and to check at each step if our work meets our own standard … and if it fails then we know we need have some more work to do.

This commonly used approach to maintaining quality is called a check-and-correct design.

So let us explore the ramifications of this check-and-correct approach to quality.


Suppose the quality of the product or service that we deliver is influenced by many apparently random factors. And when we actually measure our yield we discover that the chance of getting a right-first-time outcome is about 50%.  This amounts to little more than a quality lottery and we could simulate that ‘random’ process by tossing a coin.

So to set a realistic expectation for future customers there are two further questions we need to answer:
1. How long can an typical customer expect to wait for our product or service?
2. How much can an typical customer expect to pay for our product or service?

It is not immediately and intuitively obvious what the answers to these questions are … so we need to perform an experiment to find out.

Suppose we have five customers who require our product or service … we could represent them as Post It Notes; and suppose we have a clock … we could measure how long the process is taking; and suppose we have our coin … we can simulate the yield of the step; … and suppose we do not start the lead time clock until we start the work for each customer.

We now have the necessary and sufficient components to assemble a simple simulation model of our system … a model that will give us realistic answers to our questions.

So let us see what happens … just click the ‘Start Game’ button.

Http iframes are not shown in https pages in many major browsers. Please read this post for details.


It is worth running this exercise about a dozen times and recording the data for each run … then plotting the results on a time-series chart.

The data to plot is the make-time (which is the time displayed on the top left) and the cost (which is display top middle).

The make-time is the time from starting the first game to completing the last task.

The cost is the number of coin tosses we needed to do to deliver all work to the required standard.

And here are the charts from my dozen runs (yours will be different).

PostItNote_MakeTimeChart

PostItNote_CostChart

The variation from run to run is obvious; as is the correlation between a make-time and a high cost.

The charts also answer our two questions … a make time up to 90 would not be exceptional and an average cost of 10 implies that is the minimum price we need to charge in order to stay in business.

Our customers are waiting while we check-and-correct our own errors and we are expecting them to pay for the extra work!

In the NHS we have a name for this low-quality high-cost design: Payment By Results.


The charts also show us what is possible … a make time of 20 and a cost of 5.

That happened when, purely by chance, we tossed five heads in a row in the Quality Lottery.

So with this insight we could consider how we might increase the probability of ‘throwing a head’ i.e. doing the work right-first-time … because we can see from our charts what would happen.

The improved quality and cost of changing ourselves and our system to remove the root causes of our errors.

Quality Improvement-by-Design.

That something worth learning how to do.

And can we honestly justify not doing it?

The “I am Great (and You are Not)” Trap

business_race__PA_150_wht_3222When we start the process of learning to apply the Science of Improvement in practice we need to start within our circle of influence.

It is just easier, quicker and safer to begin there – and to build our capability, experience and confidence in steps.

And when we get the inevitable ‘amazing’ result it is natural and reasonable for us to want to share the good news with others.  We crossed the finish line first and we want to celebrate.   And that is exactly what we need to do.


We just need to be careful how we do it.

We need to be careful not to unintentionally broadcast an “I am Great (and You are Not)” message – because if we do that we will make further change even more difficult.


Competition can be healthy or unhealthy  … just as scepticism can be.

We want to foster healthy competition … and to do that we have to do something that can feel counter-intuitive … we have to listen to our competitors; and we have to learn from them; and we have to share our discoveries with them.

Eh?


Just picture these two scenarios in your mind’s eye:

Scenario One: The competition is a war. There can only be one winner … the strongest, most daring, most cunning, most ruthless, most feared competitor. So secrecy and ingenuity are needed. Information must be hoarded. Untruths and confusion must be spread.

Scenario Two: The competition is a race. There can only be one winner … the strongest, most resilient, hardest working, fastest learning, most innovative, most admired competitor.  So openness and humility are needed. Information must be shared. Truths and clarity must be spread.

Compare the likely outcomes of the two scenarios.

Which one sounds the more productive, more rewarding and more enjoyable?


So the challenge for the champions of improvement is to appreciate and to practice a different version of the “I’m Great … ” mantra …

I’m Great (And So Are You).

V.U.T.

figure_pointing_out_chart_data_150_wht_8005It was the appointed time for the ISP coaching session and both Bob and Leslie were logged on and chatting about their Easter breaks.

<Bob> OK Leslie, I suppose we had better do some actual work, which seems a shame on such a wonderful spring day.

<Leslie> Yes, I suppose so. There is actually something I would like to ask you about because I came across it by accident and it looked very pertinent to flow design … but you have never mentioned it.

<Bob> That sounds interesting. What is it?

<Leslie> V.U.T.

<Bob> Ah ha!  You have stumbled across the Queue Theorists and the Factory Physicists.  So, what was your take on it?

<Leslie> Well it all sounded very impressive. The context is I was having a chat with a colleague who is also getting into the improvement stuff and who had been to a course called “Factory Physics for Managers” – and he came away buzzing about the VUT equation … and claimed that it explained everything!

<Bob> OK. So what did you do next?

<Leslie> I looked it up of course and I have to say the more I read the more confused I got. Maybe I am just a bid dim and not up to understanding this stuff.

<Bob> Well you are certainly not dim so your confusion must be caused by something else. Did your colleague describe how the VUT equation is applied in practice?

<Leslie> Um. No, I do not remember him describing an example – just that it explained why we cannot expect to run resources at 100% utilisation.

<Bob> Well he is correct on that point … though there is a bit more to it than that.  A more accurate statement is “We cannot expect our system to be stable if there is variation and we run flow-resources at 100% utilisation”.

<Leslie> Well that sounds just like the sort of thing we have been talking about, what you call “resilient design”, so what is the problem with the VUT equation?

<Bob> The problem is that it gives an estimate of the average waiting time in a very simple system called a G/G/1 system.

<Leslie> Eh? What is a G/G/1 system?

<Bob> Arrgh … this is the can of queue theory worms that I was hoping to avoid … but as you brought it up let us grasp the nettle.  This is called Kendall’s Notation and it is a short cut notation for describing the system design. The first letter refers to the arrivals or demand and G means a general distribution of arrival times; the second G refers to the size of the jobs or the cycle time and again the distribution is general; and the last number refers to the number of parallel resources pulling from the queue.

<Leslie> OK, so that is a single queue feeding into a single resource … the simplest possible flow system.

<Bob> Yes. But that isn’t the problem.  The problem is that the VUT equation gives an approximation to the average waiting time. It tells us nothing about the variation in the waiting time.

<Leslie> Ah I see. So it tells us nothing about the variation in the size of the queue either … so does not help us plan the required space-capacity to hold the varying queue.

<Bob> Precisely.  There is another problem too.  The ‘U’ term in the VUT equation refers to utilisation of the resource … denoted by the symbol ? or rho.  The actual term is ? / (1-?) … so what happens when rho approaches one … or in practical terms the average utilisation of the resource approaches 100%?

<Leslie> Um … 1 divided by (1-1) is 1 divided by zero which is … infinity!  The average waiting time becomes infinitely long!

<Bob> Yes, but only if we wait forever – in reality we cannot and anyway – reality is always changing … we live in a dynamic, ever-changing, unstable system called Reality. The VUT equation may be academically appealing but in practice it is almost useless.

<Leslie> Ah ha! Now I see why you never mentioned it. So how do we design for resilience in practice? How do we get a handle on the behaviour of even the G/G/1 system over time?

<Bob> We use an Excel spreadsheet to simulate our G/G/1 system and we find a fit-for-purpose design using an empirical, experimental approach. It is actually quite straightforward and does not require any Queue Theory or VUT equations … just a bit of basic Excel know-how.

<Leslie> Phew!  That sounds more up my street. I would like to see an example.

<Bob> Welcome to the first exercise in ISP-2 (Flow).

Walk Confidently before Running

running_walking_150_wht_8351Improvement is not a continuous process. It has starts and stops, and ups and downs.  Improvement implies change, and that is intentionally disruptive. So the context will determine the progress as much as the change.

A commonly observed behaviour is probably at the root of why the majority of improvements initiatives fail to achieve a significant and sustained improvement.  Trying to run before mastering the skill of walking.


An experienced improvement coach will not throw learners into the deep end and watch them sink or swim.  That is not coaching; it is cruelty.

So the first improvement projects must be doable and done with lots of hands-off support, encouragement and praise for progress.

This has the benefit of developing confidence and capability.

It has a danger of leading to over-confidence though.  Confidence that exceeds capability.

There is a risk that the growing learner will take on a future improvement project that is outside their capability zone.


The danger of doing this is that they fall at the second hurdle and their new confidence can be damaged and even smashed. This can leave the learner feeling less motivated and more fearful than before.


There are a number of ways that an improvement coach can  mitigate this risk:

1. Make the learners aware up front that this is a risk.
2. Scope each project to stretch but not scare.
3. Be prepared to stop and reduce scope if necessary.
4. Set the expectation to consolidate the basics by teaching others.

These are not mutually exclusive options.  Seeing, doing and teaching can happen in parallel and that is actually the most productive way to learn.


As children we learned to walk with confidence before we learned to run … because falling flat on our face hurts both physically and emotionally!

This is just the same.