Posts Tagged ‘emergency’
Let’s take a look at how week-by-week profiling can help acute providers with winter pressures. We want to maximise capacity utilisation, and minimise the risk of bed crises, cancellations, and 18-week breaches.
We’ll take it in two stages:
1) Preparing for winter: We will look at how emergency and urgent elective demand are likely to vary, week by week, through the winter; then plan routine elective work around the peaks.
2) During winter: As each winter week goes by, we’ll update this profile with outturn demand and activity, so that our plans for the rest of winter can adapt rapidly and continuously to unfolding events.
Preparing for winter
Nobody knows exactly how winter is going to turn out, so we need to make some reasonable assumptions about how much demand is likely to come in, and how it will vary week by week. A good place to start is by looking at what happened last year or, even better, the last three years, and then adjust it for anything else we know is going to happen.
Armed with this information, we’re ready to start working on our plan. Because we’re focusing on the profiles during winter, let’s assume we have already run our strategic plan for the coming months (based on achieving 18 weeks, or filling the available capacity, or whatever scenario we chose). So we have already worked out the overall demand, activity, and capacity for this future period, as well as the waiting list and waiting times we want to end up with. If our dataset already includes demand and activity profiles then we don’t need any more data and can go straight into the week-by-week profiling.
In this worked example the screenshots are taken from Gooroo Planner, where the Profiling screen looks like this:
The large top chart is the interactive activity profile, and we are going to edit this to reprofile elective surgery around the peaks and troughs in emergency and urgent demand. The large bottom chart is interchangeable by clicking for any of the thumbnails at the bottom, so it can show either activity and urgent/emergency demand, beds, theatres, clinics, or waiting times.
Let’s start by zooming in on the bed profile. We start this analysis using data that is based on last year’s demand profile and last year’s outturn activity profile. We’ve picked a major surgical service, and we’re going to see if we can reprofile it to stay out of trouble over winter.
The blue line shows the the number of beds used by our surgical service, plotted against the left axis, and the straight blue line shows the number of beds notionally allocated to this service. The orange line shows the total beds on our whole hospital site, plotted against the right axis, and again the straight orange line shows the physical on-site bed limit. Clearly, we are heading for trouble in January and February, where the number of beds required is far larger than the number available. Looking at the blue line, we can see that we are making things worse by scheduling so much elective surgery during the winter peak; the “red alerts” we experienced last winter are starting to look disturbingly avoidable.
So let’s start by reducing our plans for elective inpatients during the height of the peak. This is a simple matter of clicking and editing the points on the interactive top chart, to reduce the balance of work profiled during January and February until the editable profile looks like this:
After doing that, we get a bed profile that looks like this:
Much better. But what happens to waiting times as a result of this surgical slow-down? A peek at the waiting times chart reveals this:
The blue line shows waiting times just for the elective inpatient stage of treatment, and the orange line shows the RTT wait for this surgical service: that’s the wait for new outpatients, plus the wait for elective inpatients or daycases (whichever is greater). All waits are on a “90 per cent treated within” basis, so the orange line is comparable with the 18 week target. The bad news is that our waiting list is going to spike over winter, rendering the 18 week target unsustainable for 3 or 4 months.
We don’t want that to happen if we can avoid it. So let’s see if we can front-load some surgery to head off the problem. In real life we would have more than one surgical service to reprofile, but for the sake of this example we’ll try to do it all just with this one. So we’ll crack on with as much elective inpatient surgery as possible over the autumn, then slow down for as short a time as possible to keep beds just nicely full over the winter peak (but not too full – we are working to a target occupancy to allow for in-week fluctuations), and then pick things up again in March to deliver the balance of our planned activity towards the end of the year.
When we’ve finished editing the activity profile, it looks like this:
Now our bed profile looks like this:
That’s fine. Waiting times?
That’s fine too: we’ve front-loaded enough surgery to get the list right down before winter, so that even when it spikes we shouldn’t see any breaches. Then the balance of our planned activity is just right to bring us in on target for year end. (In a real hospital you would have several surgical services to play with, rather than just one, so this example is on the extreme side to illustrate the principle.)
That’s our profile done, then, from the comfort of late summer / early autumn. What are we going to do once the snow starts to fall?
Reacting to events during winter
Fast-forward to late January, and it’s cold. Emergency admissions shot up when the GP surgeries reopened after New Year; nothing unusual in that. But last week it shot up again and we had to cancel surgery. How does this affect our plan?
The first thing to consider is this: does this spike mean that the total amount of demand has gone up, or might this peak be balanced by troughs later on? Frankly, who knows? Overall the external demand for healthcare rises stepwise every few years, and if demand happens to have gone up just in the last week then that may mean something, or nothing. If you want to add an extra chunk of demand to your forecast then that is easily done but, if the end result is forecasts that are more volatile but no more accurate, then what is the benefit? Ultimately it’s your call, but a compromise position might be to update the demand forecast every month, not every week, to smooth the volatility out a bit.
On the basis of a week’s worth of data, then, let’s assume it’s a wobble in the profile not an uptick in total demand. We also have outturn data on the activity we delivered for electives, as well as emergencies. So let’s update both our demand profile and our activity profile with the latest week’s data and see where we stand now.
The loss of surgery means that we are now heading for a 21 week RTT wait at the peak in mid-March, whereas before we were expecting to peak at 18 weeks. Perhaps we should have allowed a bit more margin for error in our original plan. However if our assumption about demand (that this spike is likely to be offset by less demand at other times) is correct, then we should have capacity to bring in the displaced patients over the coming weeks to restore the position, as the revised bed profile shows.
And so it goes, week by week, month by month, until the days start to lengthen again. Forecasting demand is not an exact science, especially at a week-by-week level of detail, so our plans for winter are always going to have a large amount of guesswork mixed in with the logic.
In this worked example, January’s spike in demand caused problems with cancellations and the risk of waiting times breaches. That kind of thing is a risk unless we can provide a more substantial buffer in capacity (e.g. in the form of lower bed occupancy) to absorb the variation. Nevertheless, in this example we were in a much better position than we had been the year before, when we had been galloping merrily towards a severe, prolonged, and utterly predictable bed crisis before the winter had even begun.
This worked example was illustrated using Gooroo Planner with integrated week-by-week profiling; you can see a slideshow version of it here. If you are already using Gooroo Planner then profiling is available to you now: look for the profiling button at the top of the Reports view page. If you aren’t using Gooroo Planner already, and would like to take a look, then email firstname.lastname@example.org for a free on-site demo.
Previously we looked at how the new week-by-week profiling feature is shaping up in Gooroo Planner. Now we’re going to look at something else we’ve been working on: constant-capacity profiling.
The idea is that routine elective activity can be planned around peaks and troughs in non-elective and urgent elective demand. To show an example, the following chart represents (illustratively) the non-elective and urgent demand in ten fictitious clinical services. The clinical services are shown in colour, and the grey columns at the back show the total across all ten services. The heights of the columns represent whatever kind of capacity you want to keep constant.
(These charts won’t actually feature in the implementation… they would be too complicated with dozens of services and 52 weeks. So we’ve used them here just to illustrate what happens. The actual implementation will crunch the numbers, and then let you tinker with each service separately.)
As well as the non-elective and urgent elective work shown above, each service also has some non-urgent elective work that isn’t shown on the chart yet. We want to work out how to profile this non-urgent work across the weeks, so as to keep our capacity usage (and therefore our staffing of that capacity) constant. After running the algorithm, this is what we end up with:
The algorithm does its best to achieve constant capacity across the weeks for all services added together, and as a secondary goal it also tries to achieve constancy for each service separately. In this example, some services did not have quite enough non-urgent activity to allow perfect constancy, and whenever that happened it carefully rebalanced the profiles across the other services and across time. That way, it was able to achieve perfectly constant capacity in total, whilst minimising variations within each service.
There are easier ways of designing the algorithm (such as the methods I used when developing earlier widely-used planning models). But I think this method was worth the effort, because of the better constancy at service level.
What are the drawbacks of profiling for constant capacity? The main objection is that it does not reflect the constraints that drive real-life activity profiles in today’s NHS. Usually, it is not the number of physical beds or theatres or clinics that determines activity (although this does happen from time to time), but the availability of staff. And the availability of staff varies through the year according to when staff take leave.
So why bother creating a constant-capacity calculation at all? Because when we consulted on how week-by-week profiling should work, we received comments that suggested it might be worth it. Although constant-capacity may not reflect current practice in most hospitals, it is something that people aspire to. At the least, it would be a useful starting point for discussions, even if the profile was later tweaked to reflect staff leave more closely.
So we’re building it in anyway. Constant-capacity profiling is very fiddly to do by hand, because it involves tweaking all the (dozens or scores of) services across all the (52?) weeks in the profile, and that’s a big job that is best left to a computer. When it’s all built in to Gooroo Planner, you’ll be able to do things like: hit the constant-capacity button, see what you think, hit the button that replicates last year’s pattern, see what you think, pick the one you like better, tweak it a bit, and end up with something that you think works operationally.
The objectives are firstly to make Planner more interactive, and secondly to reach across the divide from strategic to operational planning. So we hope you find these features useful and practical, whether you are an information analyst, an operational manager, or a clinician. If you have any comments on what we’re building, please let us know. We want to get this right for you.
Doing surgery? Then you need anaesthetics on-site. Obstetrics? Then you need paediatrics.
Acute care is a tangled web of interdependent services, joined by so-called “clinical linkages”. Pull out something innocuous-looking, such as physiotherapy, and the whole thing collapses.
These clinical linkages were all mapped out in an earlier post, and Roy Lilley picked it up in his discussion about competition and regulation (as did Paul Leake). His argument was that competition in healthcare provision could lead to these clinical linkages being unpicked, with disastrous results; therefore any local service reconfigurations would need to be managed (and not left to the forces of competition) in order to preserve these clinical linkages.
So how real is this threat?
Let’s start by sketching out the scenario we are worried about. A healthcare provider (it doesn’t really matter if they are NHS or private) sets up a new elective facility, which attracts work away from the neighbouring NHS acute hospital. This destabilises the NHS hospital and triggers the closure of its acute services (including A&E), much unhappiness for local people, and a political row.
There are three possible ways in which the new provider could destabilise the old:
- by financial cherry-picking: diverting away from the old provider a lot of highly-profitable elective work that had been subsidising the loss-making clinically-linked acute services;
- by deskilling: diverting away a lot of elective activity, so that clinicians at the old provider are no longer seeing a big enough caseload and become deskilled, so that those clinicians can no longer provide a safe acute service on which other acute services depend;
- by poaching: recruiting clinicians away from the old provider, causing the closure of a service on which other acute services depend.
In the Parliamentary committee debates on the Health and Social Care Bill, there was quite a lot of discussion of the problems that might be caused by “cherry-picking”. Monitor responded to these concerns in two ways: firstly to point out that if the problem is that elective procedures are more profitable than non-elective ones, then the solution is to change the tariff price and remove the distortion; and secondly that if economic destabilisation of acute services is the possible result, then Monitor can designate those services as essential and allow extra funding for them.
I would add a further point: it would be rash to assume that elective care is always profitable and acute care is always loss-making. So much in healthcare is characterised by gross and unexplained variations, and so there are likely to be many highly-profitable acute non-elective services, just as there will be many highly-loss-making elective services.
Deskilling is more pernicious, and would not be solved by flinging money at designated services. If surgeons are twiddling their thumbs all day because their elective workload has disappeared down the road, they are not going to be as practiced when surgical emergencies come in. Recruitment and retention would also go to pot. If you lose acute surgery, then acute medicine is at risk and so is A&E. What can be done?
Well, the first question to ask is: where is the new provider going to get its doctors from? In the middle of London, it is quite possible to run a hernia factory from 9-5, Monday to Friday, keep a whole team of surgeons busy, and still leave plenty of elective work around for the rest of London’s NHS doctors; deskilling would not happen in that scenario. But could you do the same in Northampton or Stoke? In practice, you’d probably be using the same NHS doctors who work at the local NHS hospitals, and so they wouldn’t be deskilled, just maintaining their skills on a different hospital site.
For the sake of argument, though, let’s say you did manage to set up an elective factory in the shires without using doctors from the local NHS. Perhaps your medical staff have been brought back from retirement, or want to work family-friendly hours. Would that not pose a threat to the local NHS hospital? Indeed it might. But how might the NHS hospital respond? They could do nothing, and let their surgeons twiddle their thumbs on full pay, but that would be perverse. A more sensible response would be to make their surgeons available to the new provider on attractive enough terms, which sidesteps the deskilling problem and replaces lost income. So it looks as if the old provider could respond to the deskilling threat, and head it off.
What about the third threat: poaching? Well the short answer is that nobody is irreplaceable. The old hospital can just recruit some new doctors. And if the service is so unattractive that it is impossible to recruit, then the old hospital’s problems run much deeper than the arrival of a new elective provider.
So we have seen how a degree of flexibility by the old provider can help sidestep the threat of destabilisation by the new. But we have tacitly assumed in this scenario that both the new and old providers are traditional monoliths who operate hospital buildings, and employ clinical staff, and contract with the NHS commissioners.
Now let’s imagine a world in which those three functions are unbundled. One possible way of doing this would be for the doctors in the old NHS hospital to establish themselves as Chambers and contract directly with commissioners; then the Chambers pays the hospital for the buildings, nurses, diagnostics and so on. Now we can see how much easier it would be to avoid the deskilling problem, which was the most serious challenge we faced above.
Because each Chambers could work across multiple hospital sites, it could respond much more flexibly than a traditional hospital service that was anchored to its buildings. If deskilling ever became an issue, the Chambers could redeploy clinicians across different hospital sites to head off the problem. It could, for instance, supply the clinicians needed by the new provider, including (where it made sense) retired or family-committed doctors.
So it is far from clear that clinical linkages are necessarily threatened by competition in healthcare provision. And even if they are, a flexible and competitive provider market could respond by unbundling provider functions in a way that unties people from buildings.
Hospitals accepting unselected medical emergencies must have on-site surgery.
There’s a lot of guidance like this, from the Royal Colleges, subspecialist societies, NCEPOD, and the Department of Health, all describing in helpful detail the critical links that exist between different acute services.
But each document describes only a few strands in a complex web of interdependencies. Senior clinicians and managers, however, need a system-wide view, but it is difficult to piece together the whole picture from this mass of detail.
The lack of a big picture can waste a lot of time. When acute reconfigurations are being considered, managers and clinicians may get together in a large group to draw up the reconfiguration options. Much later, after a lot of work, some options have to be struck out when a fatal flaw is discovered (such as not being able to separate paediatrics from obstetrics). At worst the lack of a big picture can be dangerous, when piecemeal changes are made locally, to individual services, without realising that they could destabilise the whole hospital.
So we need an overview of these important clinical linkages. Looking only at those 24-hour services that must be provided on the same hospital site, we think the links look something like this. A solid line means that one service must support the service it points to; a dotted line means that it is possible to run the service without that support but procedures must be in place to ensure safety.
There are caveats of course. It isn’t possible to capture all the nuances of this complex guidance in one diagram: for instance, the distinction between a selected and unselected acute medical take is not fully captured. Also there are cases where older guidance states a requirement that is not mentioned in more recent, overlapping guidance; this leaves it unclear whether the requirement has been softened. In the full version of this work, therefore, the diagram is accompanied by the relevant passages from the guidance (referenced by the numbers beside the arrows).
We think this is the first time that acute clinical linkages have been comprehensively published in this way. At a time when acute hospitals and commissioners are under pressure from the EWTD and the financial squeeze, many are considering whether they could transfer services to an adjacent hospital or stop providing them altogether. This map of acute interdependencies should help to show where this can, and cannot, be done safely.
To take just one example, could you save money by downgrading physiotherapy to daytime only? A novice manager might think so. But the answer is clearly no, because that would put at risk the intensive care unit, acute surgery and medicine, and the A&E department. Not a career-enhancing move.
A recent study has just fallen into my lap (under the Chatham House rule). It is the initial findings of a casenotes review of over 100 short-stay (zero- and one-day) emergency admissions at an English acute hospital.
For me the most interesting highlights of these short-stay admissions were:
- Only 33% were appropriate (and only 22% of those from A&E) under the AEP
- For 80%, the grade of doctor making the decision to admit was “unclear” or “not documented”
- 50% of admissions from A&E were in the last half-hour before the 4 hour target was reached
Given that short-stay emergency admissions are common and rising, this presents a huge opportunity to GP Commissioners. In the short term the combination of unclear records, inappropriate admissions, and the absence of previous GP involvement in the patient all point to opportunities to challenge the hospital’s claims for payment. In the longer term, it presents opportunities for GP triaging of A&E attenders, and the establishment of primary and community alternative pathways.
Isn’t that a bit rough on the hospital? Not really. Not just because the inappropriate admissions are, well, inappropriate. But also because there may be darker things going on underneath these headline figures.
The fact that half of admissions from A&E are made in a scramble, just before the 4 hour target is breached, ties in with national figures and offers a clue about why so many admissions are inappropriate. According to the Information Centre:
Of those patients discharged [from A&E] within the final 10 minutes of the 4 hour wait target, the highest proportion (64.7 per cent) were recorded as ‘Admitted / became a lodged patient’.
So late admissions and inappropriate admissions are linked together. Which raises an intriguing question: is the lack of documentation about the admitting doctor also part of the picture? With only minutes to go before the target is breached, perhaps doctors are in such a rush to admit that the notes are left unclear? Or worse, are some patients being admitted “administratively”, by a non-doctor, just to achieve the target (as is sometimes alleged by NHS staff on comment boards, e.g. here, here, and here)?
The problem at the moment is that hospitals are heavily incentivised to behave like this. Such compromise corrodes the soul. If GP Commissioners challenged payments on inappropriate admissions, so that they became a cost to the hospital instead a benefit, then the world could start to turn the right way up again.