Understanding Capacity and Demand: A resource pack for healthcare professionals

Chapter

time

length

What's the problem?

00:00:00

00:01:16

Is the demand on your system greater than the capacity? Is the link between demand and capacity really understood?

Running a healthcare system

00:01:16

00:04:18

Is it possible to run a healthcare process with no wait and no waste? Using a model of how patients flow through a 5-step pathway, the audience is asked to run the model so that the end result is no waste for the service and no wait for patients.

What's the result?

00:05:34

00:03:48

The health service often works inefficiently because we don't properly understand how it actually works. What information are you using to try and understand the system? Are you basing your decisions on useful information or meaningless data?

Hitting targets

00:09:22

00:00:52

Distorting the system to make sure targets are met is often the result of not understanding the system in the first place.

Consequences of not understanding variations

00:10:14

00:00:46

What happens because we don't understand our systems? Are you looking for trends where none exist, using past activity data to manage your service? How often have you had business cases rejected due to a lack of quality information?

Interpreting data

00:11:00

00:01:50

What data do you need to properly understand your systems and how should you display and interpret it? Are you able to show that a change you have made has definitely made an improvement?

Statistical process control ( SPC)

00:12:50

00:02:33

What is statistical process control ( SPC) and why is it a useful way to analyse data? Monitoring data over time gives you a better picture of the variation in your service.

Understanding variation

00:15:23

00:01:44

What is variation and how does it affect your service? Learn about common and special variation and how you should interpret variation to make the best use of your resources.

How did the NHS get into this mess?

00:17:07

00:02:41

Summarises the discussion so far. How often do silo thinking and targets create pressure and problems? Making sure you understand the system helps you to quantify what really needs to be done.

Demand and capacity

00:19:48

00:02:05

Demand, capacity, activity and queue are defined here, with an explanation of how to measure them. Do you really know the demand for your service? Should you be counting patients or time? What are the constraints on your capacity?

Why do queues form?

00:21:53

00:04:42

Do queues only form when demand is greater than capacity? Here is a demonstration of how queues form due to the mismatch in the variation in demand and the variation in capacity. Are queues really necessary? Are you relying on waiting list initiative work to keep the queue under control? How predictable is your service?

Utilisation

00:26:35

00:05:10

Is working at 100% utilisation efficient? Are your staff working harder with no obvious improvement to the service? This chapter discusses the impact that pressure to utilise resources fully has on the system. What's wrong with batching? What are the costs of having a queue?

How do we traditionally respond?

00:31:45

00:12:28

How does the health service traditionally respond to having a waiting list? Here are some illustrations of typical responses including delaying the patient, forced booking, carving out capacity, using waiting list initiatives and pressurising the system.

The road to ruin

00:44:13

00:01:25

Failing to understand the system leads to a vicious circle of increased variation and falling activity resulting in longer waits and cutbacks. Does this sound familiar?

So what should we do instead?

00:45:38

00:16:22

Q&A with the audience about alternative ways of managing and measuring systems in order to provide quality services. Topics covered include: investing in people; reducing demand; reducing variation; predicting demand; process design.

What should we do instead? (summary of the above chapter)

01:00:30

00:01:30

Top tips include: focussing on the quality; managing the bottleneck; aiming for effective outcomes.

The patients and the process view

01:02:00

00:15:40

How does the patient view your process? Have you considered following the pathway as a patient? By looking at the number of steps in your process you can identify how many of them actually add value for the patient rather than being the result of a work-around and ultimately wasteful. This chapter looks at the probability of performing each step in the process successfully. How often do you get the right result first time? Reducing the number of steps will increase the chances enormously. Watch out for special cause variation but aim to reduce the common cause variation which affects the majority of patients.

Effects of pooling

01:17:40

00:11:53

A computer model demonstrates the effect of pooling on a queue. Do you have several queues for the same service? Reducing the number of queues will reduce the length of wait - compare your queuing system to that used by the Post Office. Have you looked at the profile of your service? What procedures are the most common?

What do patients think?

01:29:33

00:06:41

This chapter discusses how patients feel about waiting, pooling of lists and travelling for treatment. Here are some illustrations of the effects of seeing patients in turn and matching capacity and demand. How often do you increase demand based on prior knowledge - for example, Breast Awareness week?

Emergency and elective admissions

01:36:14

00:02:13

Which do you believe to be more predictable - emergency or elective admissions? The answer may surprise you.

See "today's" demand "today"

01:38:27

00:01:22

Stop the queue getting out of hand by seeing today's demand today. How should you work out what capacity is needed? By looking at the fluctuation in demand you can calculate the capacity required to maintain a certain level of service. Remember no wait = waste, but no waste = wait.

Setting the right capacity

01:39:49

00:07:30

This computer model demonstrates the effect of capacity on a queue. How should you calculate the required capacity for a service? Here is an example of setting the capacity for a CT service using a specific equation (the 80% rule, see equation below) to take into account the variation in demand.

theoretical capacity = minimum demand + (maximum demand - minimum demand) * 0.8

Behaviour change

01:47:19

00:01:40

Have you noticed the number of hospital cancellations rising and activity falling when the pressure is on to reduce waits and maximise available capacity? This chapter discusses the use of available capacity and the effect it has on staff behaviour.

Length of stay

01:48:59

00:06:40

How does the way we run our system affect the patients' length of stay? Here are some ideas for reducing lengths of stay including improved discharge planning, admission on the day, increased ward rounds improving discharge patterns.

The system approach

01:55:39

00:00:16

Using the system approach improves time, cost and quality whereas the focus is usually on either cost or time.

Question & answer session

Q1: How much data do you need to have to know you fully understand your variables in your system?

01:55:55

00:02:38

Q2: Is the 'no queue' principle a sensible thing?

01:58:33

00:05:34

Q3: Is protected time for radiologist reporting helpful and how does it fit into the model?

02:04:07

00:03:48

Understanding the system

02:07:55

00:07:35

This chapter summarises what to do in the short and long term to improve access to services.

Q4: If your unit has a system that is currently working well, do you still continue to collect data?

02:15:30

00:02:17

Q5: How, in practice, can you go about tackling considerable variability?

02:17:47

00:01:36

Q6: How do we influence the approach to tackling waiting lists?

02:19:23

00:02:53

Q7: Should we wait to collect capacity and demand data until after the summer months when activity is different because there are lots of staff away?

02:22:16

00:01:56

Q8: Can you give us simple tips on how to measure the effect holidays have on the capacity/demand equation?

02:24:12

00:03:33

Chapter

time

length

What's the problem?

00:00:00

00:01:16

• "demand is greater than capacity"
• a lack of understanding of demand

Running a healthcare system

00:01:16

00:04:18

• modelling how patients flow through a 5-step pathway
• models available from www.steyn.org.uk
• the audience attempts to run the model with the aim of no wait and no waste
• where is the bottleneck?
• resources are put into the front-end and all perceived waste in the system is removed

What's the result?

00:05:34

00:03:48

• chaos and increased cost
• delays and deteriorating patient conditions
• reduced staff morale, frustration
• over-complicated board reports
• lack of understanding
• incorrect assumptions and decisions made
• emphasis on targets
• no change to processes
• examples of unintelligible analysis

Hitting targets

00:09:22

00:00:52

• to improve performance against targets we distort the system and distort the data, usually without understanding the system in the first place
• examples of methods that have been used to meet the 4-hour A&E target:
• remove the wheels from the trolleys!
• relabel corridor space as ward!
• don't accept the patient from the ambulance!

Consequences of not understanding variations

00:10:14

00:00:46

• non-existent trends are found
• blame or credit non-responsible individuals
• barriers go up, morale goes down, create fear
• cannot use past performance, cannot predict the future, cannot significantly improve the system
• business cases are based on flawed data and presented badly
• resources are consequently not made available
• using the right information and presenting it clearly in a business case will increase the likelihood of success

Interpreting data

00:11:00

00:01:50

• pitfalls of poor data presentation
• statistics for comparison
• is a change really an improvement?
• example:
• average wait reduced from 70 days to 35 days
• demonstrated with different charts
• importance of monitoring over time
• stop monitoring, effect is reversed

Statistical process control ( SPC)

00:12:50

00:02:33

• an explanation of statistical process control and why it is a useful way to analyse data
• plot data over time to examine the variation
• examples:
• recording the temperature of a patient
• a trip to work

Understanding variation

00:15:23

• every process displays variation
• common cause - consistent pattern, by chance
• should not make decisions based on slight changes
• special cause - assignable, pattern changes over time
• if it has a positive effect incorporate it, if not, remove/avoid it
• example: the pumpkins
• SPC rules to detect whether variation is common or special cause
• SPC points out the questions to be answered, not the answers themselves

How did the NHS get into this mess?

00:17:07

00:02:41

• a lack of understanding of the system
• a lack of real information despite lots of data
• huge pressures
• silo thinking - in the model, step 2 cannot solve the problem by themselves
• using inappropriate measures of performance
• with a "hit the target" mindset:
• waiting lists are due to a lack of resources
• so we increase resources
• which demands value for money
• we think 100% utilisation proves maximum efficiency
• the financial deficit increases whilst patients continue to wait
• SPC example:
• urgent GP referrals to colorectal cancer surgery - quantify the real problem if all patients to be seen within 62 days

Demand and capacity

00:19:48

00:02:05

• what really is the demand for your service?
• usually quoted as average activity for the past 3-6 months
• usually wrong
• demand = what we should do
• capacity = what we could do
• activity = what we did do
• queue = what we should have done
• measure each in the same time frame to compare
• example: compare 5 oesophagoscopies with 5 oesophagectomies
• constraint on capacity = kit or staff
• commonest constraint = staff
• commonest request for additional resource = kit
• but kit is sometimes handed back, as staff was the resource required

Why do queues form?

00:21:53

00:04:42

• demand is greater than capacity (sometimes)
• normally due to the mismatch between the variation in demand and the variation in capacity
• having a queue signals high utilisation, it keeps us busy
• the service was designed to have a queue
• demonstration of the queuing model
• explanation of waiting versus waste
• if demand is truly greater than capacity, the queue rises constantly
• "waiting waits, waste is gone" i.e. demand pushes forward but capacity is lost
• effects of waiting list initiatives:
• impact downstream - no extra funding, queue forms
• end of initiative - holidays, reduced efficiency
• higher impact on normal running of the department
• examples:
• breast clinic - demand is predictable, clinic slots are affected by holidays
• hospital admissions and discharges - less variation in admissions
• elective and emergency admissions - emergency admissions are more predictable than electives (note Richard misquotes here but reiterates correctly)

Utilisation

00:26:35

00:05:10

• is a very judgemental measure
• the assumption is that efficiency = 100% utilisation
• and that having a queue means the resource is fully utilised
• this leads to a push for more resources or a redistribution of resources
• which keeps the pressure on to have a queue
• example:
• surgery lists expand to fill time available, but cases per list will fall
• leads to batch logic
• variable demand placed into set clinics
• which leads to surges in demand
• example:
• once a month rheumatology clinic increases demand on radiology
• is NOT efficient; low unit cost BUT patients wait
• there are associated costs
• of managing the queue
• of deterioration in the queue
• of impact on downstream capacity
• examples:
• CT radiologist spending 2-3 hours per week managing requests
• SPC: CT exam to reporting time
• can lead to silo thinking and money wasted
• utilisation is the driver to pressurise the system - do we really want 95-100% utilisation?
• example:
• compare with jet fuel to Australia

How do we traditionally respond?

00:31:45

00:12:28

• we delay the patient
• keep minor injury patients until there are enough to warrant a medical staff member to treat them
• we use forced booking
• book extra patients into a clinic; everyone waits longer, frustration increases, leads to blocking tactics
• we carve out capacity
• model shows the effect of making an 'urgent' queue; routines wait even longer; add an 'urgent urgent' queue; queue goes out of control
• which leads to increased demand
• lung cancer slots in radiology
• orthopaedic referrals
• results in waste
• orthopaedic patients no longer fit leading to cancelled surgery
• results in churn
• patients still waiting fall ill
• patients start phoning in to be appointed sooner
• patients start to appear the same
• cancer patients - urgents and routines wait the same time
• carve out affects not just lists
• CT scanning schedule - 73 queues!
• dangers of carve out
• we use waiting list initiatives
• perverse incentives, not cost effective, impact downstream
• we pressurise the system
• performance manage with targets - "hit the target, miss the point"
• leads to bullying, reduced quality, lower staff morale, blocking tactics, increased costs
• example: two emergency departments and their 4-hour trolley waits

The road to ruin

00:44:13

00:01:25

• the failure to understand the system leads to increased variation within it - creating a vicious circle
• where capacity plans and contracts are based on average past activity
• leading to the failure to account for the variation in demand and capacity
• which means we fail to deliver the required activity
• so our income is lower than expected
• and our waiting times are longer than guaranteed targets
• leading to increased overtime and initiative work
• and more waiting list validation and breach analysis
• the costs go up
• so cost-cutting begins
• with risk of redundancies and cuts to services
• which reduces the capacity
• and increases the variation in capacity!

So what should we do instead?

00:45:38

00:16:22

• This is a question & answer session with the audience about alternative ways of managing and measuring systems to provide quality services:
• "invest in people":
• invest in the right place and remove waste elsewhere
• change where your money is if it's not working
• usually require skill not kit
• "reduce demand":
• it's not the patient's fault they are waiting
• does the length of wait really matter?
• "right department, right time"
• earlier treatment, easier to treat, shorter length of stay
• self-monitoring, early warning systems, triggers treatment
• "reduce the variation within our control":
• examples: Christmas Eve discharges; ward rounds affecting discharge patterns; weekend ward rounds
• use discharge planning to reduce length of stay rather than discharging patients when needing the bed
• "enhance primary and secondary care relationships"
• maximise the utilisation of primary/secondary care capacity
• "predict demand"
• example: winter pressure planning
• watch out for increased demand due to lower waits, is it for the right reason?
• examples: 64-slice CT; lower threshold for cataract surgery; increased risk of inappropriate surgery and death
• smooth demand
• example: stagger GP home visits which were creating a surge in demand at A&E
• "design the process with the strategic objective in mind"
• compare the design and production of a luxury car with the bolt-on process in healthcare
• "know your strategic objective and how you relate to it"
• otherwise all staff try to achieve their own aim and pull in different directions
• examples: financial versus clinical service; an activity-based service moving to a demand-responsive service

What should we do instead? (summary of the above chapter)

01:00:30

00:01:30

• focus on quality
• manage the bottleneck
• plan for no queue
• focus on flow
• know your service to manage and plan
• aim for an effective outcome

The patients and the process view

01:02:00

00:15:40

• highly recommend following the pathway as a patient
• example: Chairman acting as a lung cancer patient
• aiming for 18 weeks and 62 day pathways
• look at the system as steps in a process
• example: urology clinic - 109 steps!
• get outside opinion
• example: signage, directions
• focus on the value and remove the waste in the process
• are the solutions what the patient would want?
• example: Radio 1 in HDU
• what is the probability of performing each step in the process successfully?
• example: if we want 9 out of 10 patients to be treated perfectly through a 100 step process, the error rate of each step needs to be less than 1 in 1000
• compare with the electronics or aircraft industry with 3.4 defects per million
• we are currently lucky to have 1 in 10 patients without any error
• increase the probability of success by reducing the number of steps before chasing the quality
• examples: patients take their own notes to the appointment; one-stop clinics; lung cancer pathways; direct referral to rapid-access clinic
• investigate special causes but also manage the normal variation down
• don't assume the answer lies within the department
• example:
• time to surgery for colorectal cancer, wanted extra staff, theatre time, beds
• BUT DCAQ showed endoscopy activity was greater than demand, and yet there was still a queue
• so surely they must need more endoscopists!
• BUT activity was less than actual capacity, which was less than the theoretical capacity
• BECAUSE of carve out
• with 73 queues the service was impossible to manage!
• THEREFORE there was no need for extra theatre staff etc.
• example:
• long queues, must need more endoscopists
• actually only 2 toilets in endoscopy unit
• answer - get patients to use bowel preparation at home!

Effects of pooling

01:17:40

00:11:53

• model demonstrates the effect of pooling on a queue
• 8 independent clinics with the same set-up, aiming for a 2-week wait but all go over the 2-week wait
• if all the referrals are pooled the process never goes over a 2-week wait, with the same amount of waste
• reduce carve out
• example: ovarian cancer referrals, pooling reduced waits from 130 days to 30 days
• share diaries to offer patients the first available dates
• discussion:
• patients are educated to prefer certain clinicians by their own doctors
• need to trust your colleagues
• you should not tolerate poor practice
• pareto analysis:
• 20% of procedures account for 80% of the work; 7% for 50%;
• consider whether the rare procedures should be performed at all
• should they be done in one hospital?
• pool the routine procedures
• know what you know - transfer the patient at the appropriate time
• "no service should depend on one individual"
• spread the skills across the department(s) to create a robust and sustainable service
• patients can choose to wait
• "don't just take the solutions and apply them to your service, take the thinking and generate your own solutions"

What do patients think?

01:29:33

00:06:41

• how do patients feel about waiting, pooling of lists and travelling for treatment?
• don't make assumptions about patients - ask them!
• for routine tests, I want to be close to home; for specialist tests, I will travel
• patients will travel some distance for pre-operative assessment, using it as a test run
• patients don't want surprises, they will normally accept being treated by your colleague if necessary if informed in advance
• some patients with dependents need to plan in advance
• would probably not wish to undergo a major procedure 2 days before Christmas
• taking patients in turn leads to a dramatic reduction in waits and smoothes the variation in waits
• example: Scarborough Hospital
• matching variations in demand and capacity can have a big impact
• example: breast clinic - moved from 2 clinics per week with 54 slots to 3 clinics per week with 48 slots; waits reduced from over 2 weeks to 5 days, no vetting required
• use known indicators to plan for increases in demand
• examples: Breast Awareness week; calls to NHS 24; changes in air temperature and pollution
• compare with a supermarket on a hot and sunny local derby day

Emergency and elective admissions

01:36:14

00:02:13

• emergency admissions are more predictable than elective admissions
• examples:
• a North Warwickshire hospital admissions on 14th January
• Manchester Royal Infirmary had "nightmare" Mondays with too many elective admissions
• reducing the variation in elective admissions can reduce the expected number of beds required (but keep the slack)
• example: by reducing the variation in the elective admissions the expected number of beds required fell from 78 to 68 per day

See "today's" demand "today"

01:38:27

00:01:22

• where "today" could mean this hour, this week or this month depending on the type of service you provide
• never let the queue get out of hand
• set your capacity at 80% of the variation in demand and your queue will stay under control
• example:
• if demand varies between 95-105, capacity should be set at 103
• if demand varies between 50-150, capacity should be set at 130
• even though both services have an average demand of 100

Setting the right capacity

01:39:49

00:07:30

• modelling the effect of capacity on the queue
• set your required capacity by calculating 80% of the variation in demand
• = minimum demand + (maximum demand - minimum demand)*0.8
• but is a wait acceptable?
• capacity should depend on the service - for resuscitation you would want 100%
• reduce the impact of holidays using cross-cover, increased capacity beforehand, locum staff
• manage annual leave - don't allow the majority of staff to go on leave at the same time!
• have the right person doing the right job at the right time
• use the right data (not just numbers of patients, but time required to treat the patients)
• example: CT demand, average 79 requests per week, how much capacity is required?
• demand = requests x scan time
• requires 2 extra hours a day - is this cheaper than having to prioritise lists and perform waiting list initiatives?
• know what level of wait you are prepared to accept - balance wait versus waste
• no wait = waste
• no waste = wait

Behaviour change

01:47:19

00:01:40

• discussion of the use of available capacity and the effect this has on staff behaviour
• non-linear relationship between cancellations and occupancy
• once occupancy goes over 85% behaviours change
• the onus becomes one of bed-blocking:
• bring patients in early, keep them in longer, don't accept A&E patients
• no co-operation

Length of stay

01:48:59

00:06:40

• how does the way systems are run affect length of stay?
• examples:
• average length of stay by day of admission - similar patients are treated differently
• reducing the length of stay for a large cohort of patients may increase your average length of stay
• improvements send out the wrong message if interpreted incorrectly
• don't admit if you don't have to - bring the patients back to a clinic in the morning
• admit on the day of surgery not the night before
• aim to reduce the average length of stay of 80% of patients (exclude the long-stayers with social or medical problems) - costs saved can be redistributed elsewhere
• cost analyses of length of stay for
• hip replacement
• mental health bed occupancy
• improve discharge planning but check there is no increase in readmission rates

The system approach

01:55:39

00:00:16

• the system approach improves time, cost and quality whereas the focus is usually on either cost or time

Question & answer session

Q1: How much data do you need to have to know you fully understand your variables in your system?

01:55:55

00:02:38

• use "crude measures of the right thing rather than precise measures of the wrong thing"
• use the data or the quality will remain appalling
• the most common procedure coded in hospitals is 'not otherwise specified'
• ensure that everyone is aware of the need for good quality data and feed it back to them
• statistical process control rules suggest having at least 20 points to detect trends
• use the correct time frame
• find an analyst - maybe this has already been done!

Q2: Is the 'no queue' principle a sensible thing?

01:58:33

00:05:34

• if you remove the waiting list, there will be a surge in demand due to unmet demand
• once this surge passes the level of demand will reset - it may be slightly higher or lower
• example: CT - access times fell so requests dropped - if the test is required it will be available rapidly; unnecessary requests disappeared
• you may need a lead time to organise the patient ( e.g. bowel preparation)
• but why shouldn't a lung cancer patient be seen today?
• on the other hand, the patient may choose to wait (compare GP appointments)
• why not use full booking?
• why worry about loss of control? there's always demand out there!
• no waiting = no worry about taking leave at short notice
• more slack time = better quality service, better motivation, more time for training
• move to day case treatment, even thoracic surgery!
• prove you need a queue!

Q3: Is protected time for radiologist reporting helpful and how does it fit into the model?

02:04:07

00:03:48

• This is a wider discussion of the difference between carve out and segmentation.
• what is the benefit? does it benefit one group of patients or all patients?
• examples: minor injuries "see and treat"; cancer patients
• segmentation:
• reduces the variation in the demand and the process time
• example: uninterrupted reporting allows for more reports to be completed than if the radiologist is interrupted every 30 seconds
• improves the flow for all patients
• increases the return on net assets
• carve out:
• does not control variation, may make it worse
• the flow for one group of patients is improved to the detriment of another group of patients
• capacity is wasted
• to decide if it's right you must monitor the impact in your department
• example: in Heartlands Hospital, simultaneously have uninterrupted reporting and other staff available for questions and support

Understanding the system

02:07:55

00:07:35

• the root cause of delays is variability and high utilisation, not volume
• in the short term, aim to optimise current capacity
• reduce the number of steps and queues
• treat in turn, plan discharges
• maximise skill use, pool capacity
• get it right first time
• in the long term, plan for no queue (or minimal queue)
• measure and shape demand
• plan capacity
• reduce variation in demand and capacity
• back to the healthcare computer model - where is the bottleneck?
• step 5 (discharge) were working with no waste
• reduce the impact of holidays
• reduce the impact of variation
• "balance" the line
• waste will reduce, wait will reduce to a lower level
• gradually feed the backlog into the system
• the demonstration uses one flow whereas radiology, endoscopy and pathology are "hubs" in the hospital with more than one flow which means they are often the centre of attention when actually the problems are caused elsewhere
• needs a global view
• "chain of dependence"

Q4: If your unit has a system that is currently working well, do you still continue to collect data?

02:15:30

00:02:17

• don't measure everything all of the time
• measure what's relevant
• measure before you begin and as you change
• example: waiting times

Q5: How, in practice, can you go about tackling considerable variability?

02:17:47

00:01:36

• don't do waiting list initiatives
• if they must be done, give the initiative to the staff with the shortest waiting list to remove the perverse incentive
• also, fund the support services and staff
• focus attention on getting it done right
• focus attention on giving the highest level of service

Q6: How do we influence the approach to tackling waiting lists?

02:19:23

00:02:53

• challenge senior management
• engage colleagues
• compare with the state of the hospital (take photographs, raise the issue)
• who is providing the advice and guidance to politicians?

Q7: Should we wait to collect capacity and demand data until after the summer months when activity is different because there are lots of staff away?

02:22:16

00:01:56

• start looking, start measuring, start acting!
• sometimes only need small changes, so why wait?
• aim for quick wins
• get the momentum going
• aim to embed a culture of change and service improvement
• use working hours to meet - will improve productivity - make it part of normal work so staff feel engaged and attend

Q8: Can you give us simple tips on how to measure the effect holidays have on the capacity/demand equation?

02:24:12

00:03:33

• no easy solution
• watch out for the wandering bottleneck
• recognise where something isn't working and change it
• or, get the service delivering a reasonable level of work
• don't aim for the ultimate service, aim for simplicity
• want the right resource at the right time and right place
• example: why use nurses to clean theatres, why not have a team of cleaners moving from theatre to theatre at changeover?
• plan ahead - how often is one key member of staff on holiday and nobody knew?