Model Emergency Department, Capacity and Demand

NHS England recently published guidance on the Model Emergency Department (ED). The guidance asks Trusts to start work on implementation plans that should include “demand and capacity modelling,” which is more complicated than it sounds. To explain why I say this, it’s probably best to start with an example.

Let’s start with an example

Introducing co-located mental health EDs aims to reduce ED mental health presentations. This is a win for patients in mental health crisis, with presentations to EDs growing fast, currently making up roughly 5% of all ED attendances. But mental health EDs are unlikely to substitute for all ED presentations.

And even if it did, total system demand for mental health crises will likely rise when eventually introduced. A more appropriate space for the police, ambulance services and GPs to refer to, not to mention self referrals to a less stigmatising setting, will drive new demand, on top of existing demand. And some patients will always remain more appropriate in ED.

Suddenly the demand and capacity calculation seems complicated. Similarly to the example, all Model ED recommendations (reconfiguring your front door, streaming processes, extended emergency medicine ambulatory care (EEMAC) areas and same day emergency care (SDEC) pathways) are subject to what health economists refer to as residual demand.

Residual demand

Residual demand is not a new thing. Health economists have studied it for over three decades, in multiple health settings, globally. More recent research by Wyatt, et al. showed that between 2011 and 2019, ED attendances in England grew by 1.29% annually. They found that population growth accounted for 0.27%, changes in age-sex structure 0.09% and health status changes 0.09% of attendance growth. Which only add up to 0.45% (or 35%) of the overall 1.29% growth observed.

The remaining 0.84% (a whopping 65%) of growth could not be explained by standard demographic factors. These are the residual factors we commonly omit in demand and capacity modelling. The paper describes this residual pattern across all hospital services, not just ED. Outpatient growth was 76% residual, non-elective admissions 61% residual and elective admissions 30% residual.

Residual factors affect growth in every part of the healthcare system. It is folly to perform any demand and capacity modelling without considering these.

What are these residual factors?

Although Wyatt’s paper does not specifically measure these, the bulk of available research suggests that three main components drive residual growth.

1. Technology and medical progress

New treatments expand what can be treated. Better diagnostics create new categories of disease. And more intervention capacity enables more procedures per patient. Newhouse estimated that technology accounts for up to 75% of all healthcare cost growth back in 1992 (which was before the internet was publicly available). Several international studies have similarly found that substantial proportions of cost increases are unexplained by demographics or disease burden, with technology identified as the primary driver.

2. Supply-induced demand

Roemer’s 1961 observation “a bed built is a bed filled” best defines supply-induced demand. The macro evidence strongly supports this: despite decades of policy to shift care to community settings, hospital spending grew from 47% to 58% between 2006 and 2021 whilst community care funding shrank in real terms. And patients followed. The premise is simple: establish a new service and it will drive its own new demand, largely unrelated to standard demographics.

3. Access barriers and system design

Patients vote with their feet, and generally choose the path of least resistance. Research shows that around 62% of non-urgent ED attendances occur outside of standard business hours. My elderly neighbours who don’t own a smart phone can’t use the NHS app. How long does a GP have to wait for a specialty team to answer their page? Services that keep mainly to business hours or are more complex to access cannot compete with the ease of access to ED – in spite of long waiting times.

What this means for Model ED implementation

Model ED guidance rightly focuses on configuration: how to organise RAT, streaming, EEMAC, UTCs and SDEC. These matter. But configuration decisions depend on an expected demand. And Wyatt, et al. suggests that around 65% of that demand will come from factors we don’t typically take into account.

Without considering residual factors, improvements in flow and capacity risk getting consumed by unaccounted, residual demand. This leaves us treading water when we think we’re swimming to shore.

EEMAC, SDEC and UTC capacity planning

The general assumption is that if we build alternatives to traditional attendance/ admission pathways, patients currently using those traditional pathways will use the alternatives instead. This will in turn reduce pressure in the traditional pathway. But that assumption only works if residual factors are accounted for.

If we create alternatives, those alternatives will inevitably drive additional demand and only partially substitute demand. Substitution means the same demand shared to a new setting. Addition means more total demand across both settings. If residual growth is not accounted for, demand will eventually outstrip new capacity.

Model ED guidance suggests EEMAC areas to handle patients requiring 4 to 12 hours to complete their emergency care. It encourages SDEC for specialty pathways. And the UEC Plan commits £250 million to 40 new SDEC and UTC sites.

When sizing EEMAC, SDEC and UTC capacity, the calculation isn’t just how many patients currently need this but more importantly, how many will use it once it exists. These are two very different calculations.

The bed arithmetic

The UEC Plan 2025/26 commits to eliminating 21-day discharge delays, potentially saving 500,000 bed days annually. That’s a lot of bed days saved, isn’t it? Lets have a look:

500,000 bed days divided by 365 days in a year is 1,370 beds freed. Given England has approximately 105,000 general and acute beds, this represents a 1.3% capacity increase. So if non-elective admissions grow by 2.69% annually (from Wyatt’s paper), of which 1.65% is residual growth, even if you free 1.3% capacity in one year, that gain is consumed by annual residual growth in non-elective admissions alone!

Flow improvements matter. But if demand grows faster than you can optimise flow, you’re essentially improving how fast you process a queue that keeps growing.

Mental health crisis centres

As already mentioned, model ED guidance recommends mental health EDs to be co-located with Type 1 EDs. The UEC Plan 2025/26 is investing £26 million in these mental health crisis assessment centres. The case is strong: general EDs provide poor environments for mental health crisis care. The challenge is getting the demand and capacity model right.

English EDs currently manage around 3,250 patients in mental health crisis daily. If around half of these patients are suitable to attend a mental health ED instead, and this is matched by new demand, overall demand has gone up by 50%. These are my conservative estimates for illustration – mental health demand in the community vastly exceeds what EDs currently see.

This is not failure. Early mental health crisis intervention improves outcomes. It’s better care in a more appropriate setting, that now also includes previously unmet need. Build appropriate facilities and unmet need becomes visible demand.

What demand modelling should include

Model ED guidance asks trusts to include demand and capacity modelling in implementation plans. Three elements would make that modelling more evidence-based. I cannot stress the value of a data analyst enough.

Explicit demand projections

State expected ED attendance growth and what you think will reduce it. For example: “We expect 1.3% baseline growth from demographic and residual factors. After introducing EEMAC, discharge improvements should reduce this by 0.5%. The remaining 0.8% growth will require further mitigation.” This creates accountability for both flow and demand management.

Current Model ED implementation plans risk specifying configuration targets without reasonable demand projections. Without stating expected growth (including residual growth) and mitigation effects, we’re betting interventions will work without modelling whether they will.

Component-specific interventions

Match interventions to what drives residual growth. If technology drives demand, assess new diagnostics and treatments before rolling them out everywhere. If building capacity creates demand, plan for that rather than assuming substitution. Here are a few practical examples:

Before introducing new diagnostic tests in ED, model the likely impact on investigations per patient and admission decisions. Improved access to diagnostics will drive increased use by both ED clinicians and referrers. Improved MRI access for suspected Cauda Equina is probably a good recent example.
When building an EEMAC area, assume from the outset it will drive new demand and not simply shift existing caseload from main ED. As a tempting escalation space, some of this new demand may not be from ED at all.
When sizing UTC capacity, remember 62% of non-urgent attendances happen outside business hours, so extended GP access up to 8pm probably won’t fully address the pattern. And of course patients will now also have access to a new service, with markedly better performance than an ED minors area.

Realistic timeframes

System transformation takes years, not months. Building EEMAC and UTC capacity whilst training staff, embedding pathways, and changing referral behaviour represents an ambitious timeline.

Separate your sprint targets (what can we deliver this financial year) from your transformation goals (what system changes can we deliver over three to five years). Conflating these creates unrealistic expectations and demoralises staff when short-term metrics don’t improve despite heroic effort.

Common counter-arguments

Flow improvements are demand management

Some argue better flow prevents re-attendances, reduces ambulance conveyances for handover delays, and improves patient experience, leading to appropriate use. This is only partly true. Flow matters for quality and safety. But the evidence shows it’s insufficient alone. You can have excellent flow and still have to catch up on 0.84% annual growth from unaccounted residual factors.

Technology will eventually reduce costs through prevention

This assumes technology reaches saturation where marginal benefits diminish. Health economics research spanning 30 years strongly suggests otherwise. Each technology breakthrough (antibiotics, imaging, genetic testing, etc.) expanded treatable scope rather than reducing cost. Prevention technologies often add surveillance and intervention capacity rather than substitute for existing care.

Some supply-induced demand meets unmet need

Absolutely correct. Not all supply-induced demand is problematic. Building mental health crisis centres likely reveals previously unmet need, which is good. The question is: do we acknowledge this increases total demand, or pretend it substitutes? Honest capacity planning requires distinguishing genuine substitution from meeting unmet need.

Study limitations worth noting

The Wyatt study analysed 2011 to 2019 data, pre-dating COVID-19. Pandemic effects on healthcare utilisation, workforce, and patient behaviour may have altered residual growth patterns. The study used ecological data at hospital level, limiting the ability to examine individual patient pathways. The residual category captures everything not explained by demographics and health status, making it a heterogeneous mix requiring further decomposition.

These limitations don’t invalidate the findings. They suggest residual factors might be even larger post-pandemic, given COVID’s impact on healthcare access, technology adoption and service organisation.

Before submitting your implementation plan

Model ED guidance provides detailed configuration advice. Every trust developing implementation plans should also consider demand. Before finalising plans, three questions might help:

What residual growth do we expect?
Which interventions address which residual components, and how much will they reduce growth?
How confident are we in that estimate?

If the answers are “we haven’t modelled residual factors,” “we’re focusing on configuration,” or “we hope capacity expansion helps,” then the arithmetic suggests modest expectations should be anticipated.

The evidence exists. Research quantifies the components of demand growth. Decades of health economics literature explain what drives the residual. Model ED guidance acknowledges demand modelling matters. The gap lies between knowing what drives demand and planning capacity that addresses those drivers. Implementation plans offer an opportunity to close that gap, or at least acknowledge it exists.