1. Overview

During 2022, the Advisory Committee on Resource Allocation (ACRA) sought feedback from stakeholders on the methodology for estimating needs-based fair shares targets for specialised services.

Commissioners were overwhelmingly supportive of the move to needs based allocations. Providers’ views were more varied, with many expressing concerns, though the majority were supportive.

Many of the concerns that were expressed relate not to the needs-based methodology but to the move to population-based commissioning, regarding the risks of fragmentation of services and the consequent threat to the viability of research centres. These risks are potentially exacerbated by the further shift from population-based towards needs-based population allocations, which also involves funding flowing to commissioners geographically distant from the main research centres. Mitigations for these risks include the governance of integrated care board (ICB) decision-making for delegated services, and the quality standards and provider eligibility criteria that are set nationally.

Another major focus of comment was the need for careful implementation based on an impact assessment and involving a convergence strategy that avoids destabilisation. Though convergence strategy was not the subject of the engagement, the points are well taken and represent guiding principles constraining pace of change in converging on target allocations.

The most notable concern directly related to the needs-based allocation methodology itself was that targeting allocations using the proposed methodology would lead to divestment from integrated care systems (ICSs) with higher proportions of deprived populations.

Analysis of the impact of moving to target allocations does show that the ICSs with the highest proportions of most-deprived-quintile deprivation communities (lower layer super output areas (LOSAs)) would suffer reductions in relative funding. Two factors account for this: that need for specialised services as modelled is associated with greater levels of age-related morbidity, whereas the ICSs including more areas of deprivation have relatively young demographics; and that tertiary providers tend to be located in cities with young populations – in the same ICSs that contain more deprived communities.

However, whereas the resource allocation methodology suggests that ICS populations that include deprived communities have been using more than their fair share of specialised resources, our spend analysis shows that that excess spending is not generally concentrated in the deprived communities; hence it is not the case that a movement towards fair-share funding would necessarily compromise provision of specialised services to those deprived communities.

Nevertheless, there is some evidence that the modelling does not fully take account of healthcare burden of disadvantage. That is part of the justification of the health inequalities unmet need (HI-UN) adjustment. This has been set at a 5% top-slice of funds to be allocated to commissioners covering the small areas with highest levels of avoidable mortality (as defined by the ONS), relative to 10% for core services and 15% for primary care.

A number of the other issues raised require clarification and are the subject of discussion below supported by a revised set of frequently asked questions (FAQs). These include queries on: variation in coding depth between areas; the sensitivity of the modelling to the needs of patients who have moved to be close to their service providers; the design of a specialised market forces factor; adequacy of the modelling given data limitations; flexibility of deployment of delegated funds to allow funding of preventative interventions; the scope of services included (why it is appropriate to include high cost drugs and devices in estimates of need).

Other issues raised require further work and are incorporated in the proposed forward work programme. Engagement respondents were very supportive of the forward work programme, particularly the further investigation of the extent of undiagnosed need, the healthcare burden of deprivation, including investigation of how such issues affect coastal communities. There was also support for work on the refinement of the diagnostic categories used to identify and quantify need.

The engagement, finally, strongly endorsed the proposal to focus effort on understanding variations, including creating sub-models for major service areas.

2. Introduction

Engagement was conducted in 2 phases: 2 webinars (one general and one technical) took place in January 2022. 390 individuals attended, representing 117 NHS trusts, 50 clinical commissioning groups (CCGs) as well as a number of other organisations and think tanks. Comments were collected during the webinars and answered collectively to those who had registered, along with recordings of the webinars. The nature of this first part of the engagement was explanatory, and the comments received were largely seeking clarification. Clarification was provided in the detailed responses sent out to participants, and also written into a detailed methodology paper with a frequently asked questions section annexed. At the end of May 2022, this detailed methodology paper was published on behalf of ACRA with an invitation to comment during June.

Comments on the formal engagement were received from the following 36 organisations and collaboratives. The “yes” and “no” indications next to each respondent in the table below give the broad tendency of these institutions’ answers to the principal engagement questions:

i. Should the methodology for specialised services allocation follow the established approach for CCG-funded services as far as possible?

ii. Do you agree that the approach described for modelling specialised need, including adjustments for HIV and NCC services, provides a sound basis for setting target allocations at the current time?

These indications are necessarily simplistic: many who are marked as “yes” had concerns regarding timing or particular issues, and many who are marked “no” are in favour in principle (answering “yes” to the first question) but strongly concerned not to add to the NHS’ burden of adjustment at the moment.

The indications nevertheless give a broad sense of the spread of opinion on the question of early initiation of transition towards needs-based allocations. Given that some responses are from individual organisations, some from collaborations representing many organisations, some of whom also respond individually, counting responses is not meaningful.

Responses to questions iii, and iv, on the Forward Work Programme, are summarised later in this document.

Table 1: Summary of positions regarding the principal engagement questions

Commissioners

• Bath and North East Somerset, Swindon, and Wiltshire CCG (Yes)
• Birmingham and Solihull ICS/ICB (Qualified)
• Bristol, North Somerset and South Gloucestershire CCG (Yes)
• Devon ICS (Yes)
• Gloucestershire CCG (Yes)
• Norfolk and Waveney CCG (Yes)
• Nottingham and Nottinghamshire CCG (Yes)
• Somerset CCG (Yes)
• Suffolk and North East Essex ICB (Yes)

Providers

• Blackpool Teaching Hospitals Foundation Trust (Yes)
• he Christie NHS Foundation Trust (No)
• East and North Hertfordshire NHS Trust (Yes)
• East of England Specialised Provider Collaborative (Yes)
• East Suffolk and North Essex Foundation Trust (Yes)
• Gloucestershire Health and Care NHS Foundation Trust (No)
• Gloucestershire Hospitals NHS Foundation Trust (Yes)
• Greater Manchester Providers (Yes)Great Ormond Street Hospital for Children NHS Foundation Trust (No)
• Guy’s and St Thomas’ NHS Foundation Trust (No)
• Norfolk and Norwich University Hospitals NHS Foundation Trust (Yes)
• North Bristol NHS Trust (Yes)
• Sheffield Teaching Hospitals NHS Foundation Trust (Yes)
• The Shelford Group (Ambivalent)
• University College London Hospitals NHS Foundation Trust (Ambivalent)
• University Hospital Southampton NHS Foundation Trust (No)
• The Walton Centre NHS Foundation Trust (Yes)

Joint submissions – (commissioners and providers)

• East of England NHS England region (Yes)
• Healthcare Financial Management Association (HFMA) (Yes)
• London NHS England region (Ambivalent)
• Midlands collective response (Yes)
• North East and Yorkshire NHS England region (Yes)
• North Central London System (including both commissioners and providers) (Ambivalent)
• North East London Health and Care Partnership (No)
• South London Specialised Service Programme(No)
• South West region (Yes)

Other

• RSR Consultants (No)

3. Themes from engagement

There follows a summary of the content of the comments received on the questions of the engagement. NHS England responses are set out alongside each issue. To avoid duplication, responses often refer to a set of frequently asked questions and answers (section 4), and a forward work programme (section 5), both based on those published in the engagement document but now amplified to address matters emphasised in the engagement comments.

The principal themes raised are grouped as follows. To reflect the broad tendency of the comments received, we first quote comments, and then list, summarise and respond to the issues raised.

3.1 Support for proposed methodology

Sample comments:

East of England region:

“We agree that the methodology should follow the established approach for CCG-funded services. A general direction of travel to seek a unified allocation model for all services from general to specialised services is welcomed, in particular, to support the direction of integrated commissioning of services for ICS populations. A unified approach for allocation across entire pathways allows ICSs to consider services across the entirety of the health needs of their populations. This provides opportunities to optimise the allocation of resources holistically, and would support commissioning decisions to focus on upstream, care and prevention, and consider the relative need and use between preventative, general and acute, and specialised services.”

East and North Hertfordshire NHS Trust:

“The allocation model is able to allocate funds based on the major drivers for specialised commissioning use, mainly morbidity, and indirectly age and deprivation. It provides a way of modelling based on demand as a proxy for need and a mechanism to understand the drivers behind any variation seen. An improvement compared to activity-based allocation is that it provides the potential to understand if there was an under-or over-provision based on need and creates capacity in the system to provide upstream, joined-up planning in health care services.”

East of England Region, East and North Hertfordshire NHS Trust, and Norfolk and Waveney CCG:

“Risks to the stability of Specialised Service Providers can be managed as resources can continue to be targeted at the organisations that have the specialised staffing and infrastructure required to provide equitable access to specialised services for local populations.”

Gloucestershire Hospitals Foundation Trust:

“It is worthy of particular comment that the model does exclude the price variation between providers. This is sensible and fair. Although this could cause “local difficulties” the panel rightly recognised that this is arbitrary and historic more than a reflection of actual need. In fact closer examination of these differences would support the view the Panel took.”

East of England Specialised Provider Collaborative:

“Overall, we agree that the approach taken to modelling specialised need provides a sound basis for setting target allocations at the current time. In particular, we agree with the proposals to estimate relative need based on key demographic factors, plus unmet need and adjustments for more sparsely populated areas. We also agree with the approach of neutralising supply variables by setting them at average values for all individuals (p.17); while we understand that this will concern some providers given historical differences in pricing, there is a need to rebalance to ensure that allocations are based on relative need.”

East Sussex and North Essex Foundation Trust:

“Yes. This should promote distribution of resources in a way that promotes equal opportunity of access for equal need and that contributes to the reduction of health inequalities that are impacted by healthcare.”

3.2 Impact on integrated care systems with deprived populations

Some respondents were concerned that the model’s output would remove resources from ICSs with deprived populations, even if it was accurately assessing diagnosed need.

Sample comments:

South London Specialised Service Programme:

“Figure 1 [not included in this summary] shows ICSs ranked by deprivation (proportion of 20% most deprived LSOAs), with the bars capturing the estimated income gain/loss for specialised services. The most deprived ICS (Birmingham) appears to lose the largest amount of funding (c. £60 million), and the ICSs indicated to lose the most funding are all on the left hand (more deprived) side of the chart. Figure 2 divides the 42 ICSs into sextiles, ordered by deprivation (1= most deprived). The total predicted gain or loss for each sextile is shown, which demonstrates a clear difference between the loss of the three most deprived sextiles compared with the gain observed in the three least deprived sextiles. Whilst the impact is to be determined (Noting this is based upon 19/20 data which needs to be updated), the indicative changes are concerning.”

“This pattern is most likely due to ACRA’s methodology using age as a key indicator of specialised services need. At ICS level, age appears to be correlated with deprivation, with more deprived areas often having a younger population. The ACRA model allocates higher amounts to areas with older populations, and therefore often less deprived populations.”

North East London Specialised Service Programme:

“The proposed methodology doesn’t account for existing health inequalities; along with other London ICS, we request the funding to take into account, historical under-funding in areas of high deprivation and the impact this will have on long-term health outcomes for our population.”

North Central London Specialised Programme:

“For North Central London as an example we have two Boroughs (Enfield and Haringey) with significant numbers of people suffering varying forms of deprivation. The historic funding models for these Boroughs have led to an under-investment in prevention and early intervention to support these communities and correspondingly they both have significantly higher rates of Non-elective admissions than other parts of London.

“Specifically both Enfield and Haringey have the lowest spend per head of population in prevention and early intervention in diabetes and No investment in supporting prevention of Chronic Kidney Disease (CKD). Not unsurprisingly therefore the two Boroughs have the highest rates of Non-elective admissions in Nephrology and very high rates of people needing dialysis. This is impacting directly on their life expectancy and the quality of their lives. What is also clear is that deprived communities start to experience health issues substantially earlier than those in more affluent areas and therefore a comparison that prioritises age over deprivation will need to adjust for the earlier onset of incidences of disease and ill-health in deprived communities.

“Whilst we are working as a system to address the historic under-funding and ensure that all of our populations are equally served we cannot ignore the fact that we have a legacy of poor health outcomes that will take many years to rectify and in the mean time these populations will need the support of both general and specialist healthcare services. As such, any proposed methodology that does not take into account the existing health of a community, particularly deprived communities, is inherently unfair and open to challenge.”

Response:

1. It is important to note that it is not the case that the fair-shares allocation model uses “age as a key indicator of … need”. There is no a priori assumption that age drives need; rather the data show that age is associated with the morbidities that are followed by use of specialised services. For a number of reasons, the association of such morbidities with age appears stronger than the association with deprivation. There is a deprivation gradient in the model results: for any given age profile, more deprived areas have higher estimated specialised need. It’s just that when you aggregate it all up, it’s the age gradient that dominates.

2. It is true that the model of need does not fully take account of the impact of deprivation upon need. However, the Health Inequalities Unmet Need (HI-UN) adjustment is designed specifically to add resource to areas suffering worse health outcomes in order to adjust for this, and to support reduction in health inequalities.

3. Whilst there are persistent health inequalities that merit action, which have been exacerbated by COVID-19, the HI-UN adjustment is a reflection of the priority attached to this goal. NHS E has taken the view that reducing health inequalities generally requires focus of resources upstream, as reflected in the HI adjustment percentages, which top-slice 15% of the primary care allocation, 10% of the core allocation and 5% of specialised services. ACRA has advised on the metric of disadvantage to be used in allocating the additional funding, but ACRA has consistently advised that it lacks evidence upon which to judge the scale of adjustment required. Research is under way that may supply some evidence perhaps in time for 2025/6 allocations.

4. Interim analysis was carried out to compare estimated actual spend on specialised services in ‘22/23 with a fair-share allocation of specialised spending in that year. ICSs are ordered according to the proportion of their LSOAs in the most-deprived quintile, from the Black Country ICS, with over half its LSOAs in the most deprived quintile, to Surrey Heartlands with less than 1%. If this ordering is segmented into sextiles (six groups of seven ICSs), as was done in the relevant responses to the engagement, the estimated shift of resources as allocations move to a fair share of specialised spend for the sextile with the greatest proportion of deprived communities is a loss of nearly £100m (around 2½%).

5. This is Not in itself grounds for rejecting the target allocation, however. There should be no presumption that the current balance of spending is correct or that a greater proportion of funding than currently should go to those ICBs with a larger share of deprived LSOAs. The allocations formula is designed to target allocations on need for specialised services including need arising from all sectors of the community; there are sound epidemiological reasons for the finding that that need, though associated with deprivation for any given age cohort, is not associated with deprivation overall. More deprived areas tend to be younger and consequently have less need for specialised services; less deprived areas tend to be older and consequently have more need for specialised services.

6. Several factors will determine the extent to which a movement towards fair-shares allocation of specialised services funding will impede efforts by commissioners to meet need related to deprivation in an ICS. These include the extent to which relatively high spending is concentrated in the deprived communities within an ICS (indicating that it is addressing deprivation-related need), and also the implications of distance from fair-share in the other healthcare allocations.

7. On average, from analysis of the patient level contract monitoring (PLCM) datasets, across all ICSs, spend per head in the most deprived-quintile LSOAs is 13.9% higher than in other LSOAs in a given ICS area. In general, there is some association between spending in an ICS relative to model and relative spending on the most deprived LSOAs within an ICS. (This provides circumstantial justification for the HI-UN adjustment in so far as it relates to the health care burden of disadvantage. This will be explored further in the forward work programme; see section 4. Estimate the healthcare burden of disadvantage.

8. However, this association is not evident in the high-spending ICSs with the highest proportions of deprived communities. In those high-spending ICSs with much deprivation, relatively high spending does not appear to be attributable to high spending on deprived communities. For example, Greater Manchester has 41% of its population in deprived LSOAs. It is also substantially over target. But in this ICS, relative spend on deprived areas is lower than average. This suggests that a movement towards fair-share funding need not compromise provision of specialised services to those deprived communities, as the overspending is mainly occurring in other LSOAs.

9. Of the ICSs with a high proportion of deprived areas, the ICS with the greatest extent of overspending relative to the share of resources implied by the specialised model is Birmingham and Solihull. This ICS’s spending is disproportionately concentrated on the half of its LSOAs in the lowest deprivation quintile: specialised spend per head of population in most-deprived quintile LSOAs (which account for nearly half of all the LSOAs in the ICS) is estimated to be nearly 30% in excess of spend per head in less deprived LSOAs. However, Birmingham and Solihull is also exceptional in that movement towards fair-share in specialised spending is more than offset by movement towards target in core spend. In this case, movement towards a combined target would involve an increase in relative funding, with perhaps some shift in emphasis away from specialised to non-specialised, or at least a recognition that some of the services that are classified as specialised for this population are classified as core elsewhere. Overall spending on deprived populations is therefore not in jeopardy.

10. Jointly, Birmingham and Solihull and Greater Manchester account for the targeted drop in share of the whole of the most-deprived sextile; other ICSs gain collectively.

3.3 Variation in coding depth

Some respondents pointed out that coding depth varies by area and provider and questioned the cut off of 12 diagnoses.

Sample comments:

South West region:

“A number of systems in the South West have operated on block contracts for a number of years. This may have reduced the depth of diagnosis coding compared to other systems and could distort the output of the formula between systems. We would wish to understand how comorbidities have been factored into the diagnosis aspects of the formula and how disability is treated. The additional factor linked to diagnosis in the formula is the impact of early or late diagnosis, the later the diagnosis the greater the impact on tertiary services.”

Great Ormond Street Hospital for Children Foundation Trust:

“Testing of the model established that using 12 secondary diagnoses balanced making the best use of data without introducing bias into the model due to differences in depth of coding by providers. There are concerns that although limiting coding to 12 secondary diagnoses may be appropriate for acute hospitals, that this may understate the need for individuals accessing care in specialist hospitals.”

Response:

1. ACRA considers this issue regularly and reviews investigations regarding the optimum depth of coding, judging the optimum balance between sensitivity to variation in severity of morbidity on the one hand against the risk of spuriously interpreting variation in thoroughness in coding for variation in patient morbidity on the other.

2. Significant variation in the depth of coding between providers was observed which could introduce bias into the model as patients attending providers with greater depth of coding could have more morbidities and co-morbidities identified. About 95% of individuals had 14 diagnoses or less recorded per provider. Consideration of the 95th percentile of diagnoses recording suggested that the vast majority of providers coded to around 10 to 12 positions with all providers having a maximum number of diagnoses of at least 12.

3. The concern expressed that parts of the country using block contracts are not coding as thoroughly as elsewhere is problematic. However, the remedy must lie with more meticulous coding – the primary purpose of which is of course clinical.

4. For the specialised model, there is less concern than for core services models with misinterpreting variation in coding thoroughness given that the use of provider-supply variables should neutralise this effect. In principle, this might allow used of greater coding depth (as the GOSH submission proposes). For the current specialised model, we aimed to match the non-specialised model and used the same set of diagnostic groupings as for the non-specialised model; maintaining the same coding depth was consistent with this decision. As part of the forward work programme (section 5.1. Review diagnostic categories) we hope to explore use of more fine-grained categories. In response to this suggestion, we may also consider the use of additional numbers of diagnoses beyond 12 in assessing morbidity.

3.4 Whether deprivation is adequately captured by diagnostic history

Some respondents questioned whether the model accounts for all the need driven by deprivation.

Sample comments:

The Christie NHS Foundation Trust:

“Our first concern is that deprivation … is less salient in the specialised model than the non-specialised model.

“We believe that deprivation is a significant predictor of need for cancer services.

“Cancer Research UK, on 30 September 2020, published figures that revealed that there are circa 20,000 extra cancer cases each year in the more deprived areas of the UK. The starkest differences between the most and least deprived areas of the UK were in smoking-related cancers, which are 3 times higher for the most deprived populations compared to the least deprived.

“Research also shows that, for some cancer types, people from more deprived communities are more likely to be diagnosed at a later stage, giving them fewer treatment options. Data suggests that people from the most deprived communities are less aware of cancer symptoms and report more barriers to seeking help.

“The above is reflected in the results reported in the NHS Digital Publication: Cancer Survival in England, cancers diagnosed 2015 to 2019, followed up to 2020 (published 3 February 2022), that identified that for all cancers, the age-standardised net survival was higher both for males and females living in the least deprived areas when compared to the most deprived areas.

“The above suggests that, for cancer, the use of individuals’ diagnostic history, where this may not be as rich and timely for deprived communities, may not appropriately reflect need and prevalence associated with deprivation, and won’t address undiagnosed need and inequity of access.”

South London Specialised Delegation Programme:

“Given the issues around undiagnosed need …, and the potential for more deprived populations to have greater proportions of undiagnosed need, we would question whether previous inpatient diagnoses are sufficiently capturing the needs of more deprived populations…

“Younger, more deprived populations are likely to have fewer hospital admissions in the last two years, and therefore their diagnosis data is less likely to be included in the model.”

NHS Bristol, North Somerset and South Gloucestershire CCG/ICB:

“Unmet need: we note the comments in the engagement papers the clinical view that the interaction of age and deprivation deters access at a distance. Unmet need is set at 5% in the specialised model, we would like to better understand the evidence for this scale of assessment and how this compares to the unmet need adjustments in other formulas e.g. core CCG and primary care. There may be other indicators of unmet need such as the use of IS services paid for via insurance / self-funded. We would also like to understand more about the assessment of unmet need for older populations.”

Somerset CCG:

“We have just under 10% of the weighted population of England but 19% of the land mass. We would welcome the opportunity to understand more about the how the supply side variables specifically consider the impact of proximity to specialised services. These concerns are highlighted by the comments in the engagement document that undiagnosed need is not reflected in the morbidity inputs in the model estimation which could disadvantage resource allocation to more geographically remote populations.”

NHS Devon ICB:

“Adjustments for Health Inequalities needs to tested. Only 5% is related to HI, and this compares to 15% for primary care and 10% for core ICB. This feels like it may inadequately recognise the impact of HI on need?”

Response:

1. Need that is diagnosed in secondary care but, for reasons of distance and deprivation, goes untreated, is generally captured in the model. So long as some patients with similar diagnostic histories are treated, all similar need will be captured. The process of sterilising the supply variables (treating every patient as having the same access to every provider) attributes need equally to those near and far. Areas from which deprived patients struggle to get access to specialised services then show up as being under target, and ICBs will in due course be resourced to meet that need.

2. However, neither need for primary prevention nor need for earlier diagnosis, both responsible for deprived communities’ bearing disproportionate burden of disease (include from cancer) are addressed by the specialised services fair-shares methodology. This is quite correct, as the methodology is designed fairly to share specialised resource, and neither public health nor case finding is properly the responsibility of specialised services. Whilst it is a policy priority to drive down health inequalities by earlier diagnosis, funding this is not the job of the specialised model, which seeks to track need for specialised services as presented to the service. The broader process of setting target allocations for all services along the pathway from public health to specialised care combines the various estimate of modelled need with other adjustments – including the UN-HI adjustment – to estimate how overall need for resources varies geographically.

3. To the extent that need for specialised services presents – and is therefore included in the PLCM data sets, the model associates that need with population characteristics such as deprivation even in the absence of a diagnostic history. The failure of the model to pick up deprivation as a separate driver of need suggests that diagnostic history is sufficiently sensitive to variation in need driven by deprivation.

4. It is likely however that there is undiagnosed need for specialised services that remains untreated, with consequentially increased mortality. This need will not be picked up in the model, which relies on utilisation to identify need. And it is also plausible that this need is disproportionately left untreated in deprived communities.

5. Need that is unmet, including the need for diagnosis and treatment amongst those who have undiagnosed conditions, is an appropriate object of the HI-UN adjustment. The current adjustment is focused on areas of high avoidable mortality. The adjustment is the subject of a NIHR funded review that will report in early in 2024 and should inform the next allocation. See FAQs 26-28 relating to the HI-UN adjustment.

3.5 Breadth of the diagnostic groupings used to represent need

Some respondents questioned whether the modelling miss the extra resource for some high need but narrowly defined conditions.

Sample comments:

North Central London ICB:

“We believe that Haematology services, which are particularly costly to provide, are an area where the need is high in London and may not be adequately reflected in the formula.”

Gloucestershire Hospitals NHS Trust:

“The model being national will not pick up very localised health issues in some ethnic minority groups e.g. Gaucher’s disease in Hasidic Jews (Golder’s Green in Camden [sic] is a hot spot) and these can have significant (destabilising) financial implications.”

Response: 

1. Where need of this sort is materially important to the resources required by an ICB, and where the need is persistent and stable, the person-based modelling methodology is actually well designed to capture it. The need may be very local, but that local need will be reflected in a local predominance of associated diagnoses.

2. However, the diagnostic categories used by the model have to be fine grained enough to reflect associated variation in need. It is hoped that the adequacy of the current model to this requirement will be investigated in the forward work programme; see section 5.1. Review diagnostic categories.

3.5 Decision to suppress some ethnic coefficients

The model is run with negative coefficients on ethnicities set to zero, on the assumption that these represent access restraint rather than reduced need; respondents questioned this judgment.

Sample comment:

Gloucestershire Hospitals Foundation Trust:

“The adjustment to the model for assumed under representation of ethnic minority groups where the value in the calculation was below 0 is I believe wrong simply because it is an assumption. There is no corresponding assumptions to adjust if over the 0 value. After all what causes a higher number in one area might be mitigated in another service with a lower score. Age profiles and differences in presentation between different ethnic groups will result is differences that do not reflect under representation. The impact of such an assumed adjustment both undermines the model and could unfairly result in a misdirection of resources. The model is a good one but I hope will be adjusted as we learn more to account for a range of factors including ethnic variation in disease load, socio-economic factors that determine access, age etc.”.

Response:

1. The same approach has been taken in the non-specialised model. The rationale is that whereas there is evidence of ethnicity being associated with higher clinical need in some circumstances, there is no evidential basis for assuming lower need on average for certain minority ethnic groups; it is therefore plausible that what the modelling is uncovering are access issues artificially depressing demand from certain groups.

2. In the meantime, sensitivity analysis of impact of suppression of negative coefficients on ICS needs indices has been carried out to test whether this decision is materially affecting allocations. The results are available on request. Broadly, the impact of suppressing the negative-coefficient ethnic variables is to increase needs weighting for specialised services for the London ICSs relative to less diverse parts of the country by between one and two percentage points.

3.6 Concern that patients sometimes move to be closer to a service provider (the halo effect)

Respondents question whether the modelling allows funding to follow patients who move to be near their provider.

Sample comment:

Midlands region:

“Within the Midlands it is felt that that the impact of patients moving towards providers of specialised services is a significant issue and this is demonstrated by the relative use of specialised services in areas with close proximity to specialised tertiary centres”

Response:

See section 4.5, FAQ 24 halo effect.

3.7 Additional costs associated with treating disadvantaged patients

Sample comment:

Unattributed:

“[T]he cost of delivering HIV services to more vulnerable and deprived communities is higher due to the range of methods used to engage people in care.”

Response:

1. The specialised services model uses expenditure rather than cost weighted activity as its dependent variable, and is therefore sensitive to differences in price, and in principle differences in price may reflect differences in costs associated with disadvantage.

2. Nevertheless, there is reason to think that some of the additional costs associated with treating the disadvantaged are not captured by the specialised need model. Although length of stay variation related to difficulties of discharging disadvantaged patients may be picked up for much specialised activity, effects on need indices may be lost in the sterilisation process, which aims to remove unjustified variation in cost between providers. Sterilisation is necessary – as access and efficiency effects, and indeed differences in price unrelated to service-cost, should be suppressed in setting allocations. Hence, even if the model does pick up some burden effects, if these are then sterilised, we will need to add them back again through the HI-UN adjustment. Deprivation related variables do not prove significant in the specialised model, which may be because effects of these kinds are sterilised.

3. Furthermore, support services for patients in specialised services to help them with adherence or recovery in the context of multiple life challenges, as for non-specialised, may be both more important and more costly to provide for disadvantaged individuals, and may be under-provided, and missing from modelling for that reason. These services may not themselves be specialised — or it may not be clear whether they are the responsibility of specialised or non-specialised commissioners (and they may currently be funded by charities). Even if in theory the costs of such services are included in the utilisation model, if in practice the support services are lacking, they should feature in the HI-UN adjustment.

4. See section 4.6, FAQ 26 Health inequalities and unmet need, and Forward Work Programme section 5.4, Healthcare burden of disadvantage.

3.8 The HIV model adjustment

Sample comments:

North East London Specialised Programme:

“We would agree with the allocation of HIV funding on a pro-rata basis. This seems like the most sensible, transparent and pragmatic solution.”

Gloucestershire Health and Care NHS Foundation Trust:

“…current access to HIV services is based on a flexible approach to ensure treatment and support for everyone regardless of where they live and with anonymity. Any change in focus to a population-based resource consumption could create an artificial local barrier and undermine the principles of accepting vulnerable patients from anywhere in the country. As such, a Notional ICB allocation based on the distribution of the c85k HIV patients is unlikely to accurately reflect the demand and complexity of need at a local level and could put at risk the much-needed access, prevention and promotion for all.”

Response:

1. The fair-shares allocation model adjustment for HIV adjusts the funding available to an ICB to manage delegated services for its population. It is quite compatible with patients having discretion to seek treatment elsewhere, with funding flowing across boundaries if required.

2. Provider contracts and payment can still be designed flexibly, and providers will continue to be able to charge commissioners for cross boundary flows.

3.9 Market forces factor (MFF) and the costs of referrals to high cost centres

Respondents questioned whether application of the Market Forces Factor adjustment adequately resources low-cost areas sending patients to providers in high-cost areas.

Sample comment:

East of England Specialised Provider Collaborative:

“MFF: we question whether it is appropriate to apply the MFF for specialised services, given that most specialised services are not delivered by local providers/DGHs. If the MFF is applied, we would expect the MFF adjustment to differ from the CCG allocation methodology such that it includes only specialised services activity when calculating the weighted local authority district MFF. If same weighted MFF adjustment is used as for CCG allocations, this will reflect DGH flows, and disadvantage populations based on the London fringe where a high proportion of specialised services are delivered by London providers.”

Response:

1. The MFF adjusts for the unavoidable differences in unit input costs between areas due to their geographical location. A separate commissioner-MFF adjustment has been calculated for the specialised services model. The specialised services MFF is derived in the same way as the general and acute MFF adjustment, taking account of the share of services to a given population flowing from providers with different MFF adjustments. As patients are more likely to travel to receive specialised services, the specialised MFF is less differentiated than that for other services.

3.11 Model adequacy and stability

Respondents questioned whether the proportion of variation that the model explains gives assurance that it adequately captures systematic variation in need and that is sufficiently stable to be used for allocations.

Sample comments:

South London Specialised Delegation Programme:

“The overall model fit is relatively poor, with the factors feeding into the model explaining only 50% of the variation in the outcome – i.e. the model is missing an explanation for the remaining half of the variation.

“The specialised model is poor in comparison to the general acute model, which manages to explain 85% of the variation.

“This suggests that the inputs to the specialised model are missing a lot of information on the drivers behind the use of specialised services.”

NHS Bristol, North Somerset and South Gloucestershire CCG/ICB:

“As a general point we are concerned that areas of low volume can destabilise a small geography and we would wish to understand the stability of the formula. We understand that the formula will not be revisited until the 2025/26 allocation round and so picking a particular set of data points may not be representative to set the allocation formula for so many years. Has ACRA looked running the formula using more than one year of data to understand the change at system level and so form a view on the stability of a formula? Is there a correlation between size of system and stability of the formula?”

Response:

1. The final model explains 52% of the variation in specialised services costs at the GP practice level. This compares to 85% of variation in cost-weighted general and acute activity that is explained by the model for those services. It is to be expected that the model for specialised services explains less variation than the general and acute model. Fewer people use specialised services, and – to the extent that this incidence is unpredictable – there is greater scope for random variation in the incidence of specialised conditions at a practice level.

2. ACRA judges this level of explanatory power adequate for the purpose of estimating relative need at ICS level.

3. The stability of the model was tested comparing need estimated on utilisation data for 2019/20 with that estimated on the previous year. ACRA has reviewed the results of this comparison and found adequate stability of the specialised services model to give a positive answer to the question, “Is the level of need attributed to ICSs sufficiently stable from year to year to form the basis for forward allocations at ICS level?”

4. The stability test is discussed in Section 5.6.6 of the “Prescribed specialised services needs-based allocations methodology” published alongside this document on the Allocations website (NHS England » Allocations).

5. To cope with the greater random volatility in need year by year that is not captured by the model, risk management mitigations are available. See section 4.4, FAQ 13. Commissioner risk.

3.12 Neonatal costs and the neonatal critical care (NCC) adjustment

A respondent suggested that other neonatal services be grouped together with neonatal critical care as none of the neonatal patients has diagnostic records to enable the model to predict need.

Sample comment:

Great Ormond Street Hospital for Children Foundation Trust:

“It is recognised that as the model uses GP registered populations at the start of [the model estimation year] it means that neonatal critical care expenditure is not captured for children born during the year and is not included in the 2 previous base years for predicting relative need. This issue would also extend to neonates requiring other paediatric specialist services and we would challenge the assumption that the attribution of some funding to young and healthy individuals will adequately cover the funding required for this cohort of children.”

Response:

1. The point is a good one, and in principle all neonatal services should be treated separately. However, these services for neonates are grouped in the aggregated contract monitoring dataset with other paediatric services – so it is not possible to reconcile the PLCM to the outturn data specifically for this group of patients.

2. In practice only some £50m of specialised spending (less than ½ %) is devoted to these infants below the age of one, and some of those patients will have a care record at the end of the previous year. So a proportionate solution to the challenge is to allow these costs to be allocated to ICBs by the model through the age variables and the constant, enabling ICBs to fund specialised costs as they arise.

3.13 Under-representation of outpatient-based services

The reliance on inpatient diagnoses may understate need for those services that are mostly delivered in outpatient settings.

Sample comment:

South London Specialised Delegation Programme:

“Specialised services which treat patients through multiple hospital admissions are much more likely to have relevant data included in the model than specialised services which treat patients directly through an outpatient route.”

Response:

See section 4.5, FAQ 19 Reliance on inpatient diagnoses.

3.14 Concerns regarding incompleteness of the patient level contract monitoring (PLCM) datasets

Some respondents were concerned that the allocations model may underrepresent some specialised service area, for example paediatric rare and expensive procedures due to the incompleteness of the PLCM datasets and their insensitivity to intervention complexity.

Sample comments:

North East London Specialised Programme:

“The agreed threshold of 40% coverage by PLCM seems very low. It is not made clear how many services hover at the 40% mark and how many have much higher coverage. We would recommend that at least a majority of expenditure is accounted for in PLCM before a needs-based model is applied to a service.”

The Christie NHS Foundation Trust:

“We understand that a proportion of specialised commissioning activity is unallocated to ICSs, and that the amount of unallocated activity varies by region. We suggest that you consider the potential impact of this unallocated activity in your modelling, and how to adjust for these data quality issues.”

Great Ormond Street Hospital for Children Foundation Trust:

“Highly specialised services will continue to be funded under current arrangements however there are services at GOSH, being a paediatric specialist provider, that are within broad service lines that would meet the criteria of a highly specialised service. An example of this is surgery for cloacal anomaly where there is an incidence of 1 in 200,000 liveborn children and GOSH provide this service for most referrals in the UK. This currently falls under the paediatric surgery specialist service line but given the highly specialised nature of this small service it is argued that this along with similar services, should be excluded from the needs-based population allocation.

“As a specialist provider GOSH has multiple local pricing arrangements that differ from most other Trusts and it is unclear from the description of the model how this will be addressed to ensure the population need is robustly calculated. These prices may potentially look expensive when compared to other providers however the unit of activity is not comparable, or the cost of a services may not be allocated to individuals. Examples of these include packages of care for paediatric renal dialysis rather than charging items of care separately and paediatric intensive care, where most of the funding is via a block with marginal rates applied for under or over-performance.”

Response:

1. The threshold that allows a service to be included with 40% PLCM coverage is sufficient to achieve a representative sample of patients in order to estimate drivers of their need.

2. Average coverage is 72%; 86 out of 96 services representing 95% of spend have coverage over 50%; and 66 services, representing 54% of spend have coverage over 75%.

3. Regarding variable levels of completeness, this does not affect modelling so long as we have a representative sample of complete records.

4. It is true that local prices may sometimes reflect need and special arrangements, and there is a risk that the modelling will attribute such pricing to supply rather than to need.

5. It is also true that there is a risk that some services are less well represented in the Patient Level Contract Management datasets because more costs are included in block payments.

6. Bias in allocations would result only if the services that are underweighted in these ways have need drivers with different geographical profiles.

7. All these issues will be mitigated if a greater proportion of modelled utilisation can be derived from activity information flows coupled with cost-weights derived from PLICS. See Forward Work Programme section 6 Improving coverage of patient level data.

3.15 Funding prevention including where pathways cross ICS boundaries

The question is raised whether the allocations methodology can take into account the fact that to pre-empt specialised need funding may have to shift upstream to a different ICB.

Sample comments:

South London Specialised Delegation Programme:

“Targeting deprived and younger populations for investing to improve upstream services (prevention, primary and community services) is a key part of an effective strategy to reduce spending in specialised care. ACRA notes that there will be flexibility for funding allocated in respect of specialised services for an ICS population to be applied upstream, including for prevention, however such opportunities for investing in pathway optimisation will be hindered by a reduction in specialised allocation in these younger communities. There are some pan-ICS population flows associated with age (e.g. people moving out of inner south London towards Kent, Surrey or Sussex as they age), which may result in areas like south London receiving insufficient funding for upstream prevention activity amongst their younger population. This inequitable distribution of funding for preventative upstream services would likely cost the NHS as a whole significantly more in the long run.”

North East London Specialised Service Programme:

“Our growing younger population would also benefit from focused upstream services, which would enable us to reduce spending on specialised services and support already strained services. For example, all our boroughs apart from Redbridge have a large number of unidentified people with CVD, If we don’t build upon existing work to identify patients and invest now in upstream services, this will impact on specialised services, and we anticipate 29.6% growth in total Renal Replacement Therapy demand in next 10 years.”

Response:

1. The model is designed to fund current need for specialised services. Other allocations allocate funding of upstream services to pre-empt need. And 15% of the primary care allocation is focused on areas with greatest avoidable mortality through the HI-UN adjustment, to ensure those upstream budgets can be used to address some of the issues raised, e.g. CKD case-finding.

2. In addition, integration of commissioning should enable ICBs to take a strategic view, deploying more resource upstream. It is true that this investment will be less well focused if populations tend to cross borders. But this source of investment in upstream services (through redeployment of specialised allocations) is merely supplemental to the principal allocations for upstream services.

3.16 The range and scope of services modelled for target allocation

Questions were raised regarding the range of services included in the modelling and the inclusion of high cost drugs and devices.

Sample comments:

NHS Devon ICS:

“There is a range of types/levels of specialist activity, and we are keen to understand whether the needs weighting formula as developed applies equally to all specialised services delegated to ICB’s as well as those held centrally, where essentially those delegated would be of a more locally influenceable and those retained centrally highly specialised (high cost low numbers). Is the formula too generic to cover both or does it need to be more specific to services? We would like to see this evidenced.”

NHS London Shared Service:

“The engagement document states that the scope of the services included in the exercise is all those that will be subject to integrated commissioning. However, excluded specialised High Cost Drugs and Devices will not be delegated to ICBs. Instead, as we have recently been informed, they will be centrally managed on a pass-through basis. Therefore, specialised high cost drugs and devices should be excluded from this modelling. This is a particular concern as this is a very high value area and will be distorting some of the content and calculations. For example, the %s used throughout this document (e.g. HIV 3.5% and NCC 5.5% of total spending) are inaccurate if drugs and devices are included.

“In 2022/23 only those services that are in List 1 Annex A (PAR1440-specialised-commissioning-roadmap-addendum-may-2022.pdf (england.nhs.uk) will be delegated to ICBs. It is confusing to have ACRA including within scope services that are not yet ready to be delegated to ICBs. We risk overcomplication when there are multiple versions of services in and out of scope.”

Response:

1. The model includes only services that are ultimately to be delegated. Need drivers for delegable services are modelled collectively, so all services included are covered equally by the resulting needs indices. The aim is to estimate the overall share of specialised resources required by each ICB to commission the full range of delegable specialised services: in any given year, need for different services in a particular ICS will vary randomly, but the aggregate estimate will be more reliable.

2. The determination of which services should ultimately be retained for central commissioning is outside the scope of the engagement. It is based on a number of criteria relating to optimal commissioning arrangements; modelling considerations are not a constraint on delegation.

3. See section 4.3, FAQ 8 Scope of services modelled and allocated.

3.17 Risk of divestment from specialised services and from research centres

There is concern that convergence towards target allocations will add to the risk associated with the move to population-based commissioning, that commissioners collectively will reduce funding of specialised services, particularly reducing the funding of the major centres that have developed excellence and research capability.

Sample comments:

The Shelford Group:

“Overall loss of funding from specialised services – ‘winning’ ICBs may not choose or be able to invest in increased specialised activity for their populations, while ‘losers’ have no choice but to dis-invest. Specialised services could become more expensive and less productive.

“Destabilisation of existing specialised services – increased resources in other areas further away from specialist providers leads to smaller, less effective services (in terms of outcomes, productivity and access) being set-up, despite the limitation of capital and workforce.

“Undermining research and development – the dilution of services from existing centres of excellence may weaken the links between clinical, academic and industry innovators that are so essential to driving improvements in clinical care and to maintaining the UK’s worldwide status in relation to its biomedical and life sciences sector.”

University College London Hospitals NHS Foundation Trust (UCLH):

“An extension in the ‘tail’ of specialised service provision. Whilst we are supportive of transforming services to be closer to home where appropriate, the natural response to the funding transfer will be for ICBs to invest in local service provision, which may not be clinically appropriate or efficient and is counter to the policy of consolidation of specialised services. We should learn from the recent experience of ICBs holding onto elective recovery funding for their own provider organisations rather than using as intended in the national guidance.”

Guy’s and St Thomas’ NHS Foundation Trust (GSTT):

“The method is not suggesting increased levels of funding to allow for widening provision, instead redistributing existing funding geographically. Implied in this must be an assumption that it is appropriate reduce [sic] current service provision, even though these are currently commissioned flows. There is no evidence to support the notion that these services can be reduced or that existing providers can reduce costs accordingly. In effect this is levelling down current provision and there must be an explicit policy to invest more if access is to be increased.”

UCL Partners and Imperial College Academic Health Science Centre (AHSC):

“Many specialist services at our Academic Health Science Centres (AHSCs) support large, internationally regarded research programmes. This clinical academic focus is facilitated by co-location, close collaboration and joint working between the NHS and university partners. It is well established that this research-intensive environment produces better patient outcomes. In addition, the critical mass and quality of our clinical academic endeavour delivers impacts beyond the local patient population; extending across England and beyond.

“…We therefore urge you to consider the impacts of the new allocation methodology on research and how these might be mitigated in order to preserve the national benefits that these programmes bring.

“…If the new allocation model is to be implemented, we would recommend that it is staged, over several years, in order to allow the impact of the new allocation model on service provision to be assessed and the impact on associated academic programmes to be minimised.”

Responses:

1. The shift of commissioning resources to areas with higher relative need, need not result in reduced patient flow to centres of excellence, which may to the contrary be induced to provide outreach services.

2. However, if it happens that delegation results in greater investment upstream, reducing flows of complex patients, then that success should draw research focus with it: research should focus on where care can most cost-effectively be delivered.

3. The movement towards target fair shares will not be immediate so as not to destabilise health economies. ‘Implied in this must be an assumption that it is appropriate reduce current service provision, even though these are currently commissioned flows.’ No such assumption is implied: convergence can take place over a number of years to allow levelling up.

4. The risk of divestment is addressed in section 4.4, FAQ 15. Disinvestment risk.

3.18 Domino impact on highly specialised services

Concern was expressed that specialised centres’ viability might be jeopardised by destabilisingly rapid withdrawal of business.

Sample comment:

North East London Specialised Collaborative:

“Highly specialised services are not standalone; they sit within a service and are delivered by the extended team. If an ICB delegated service is destabilised through an overly speedy re-allocation of resource, then the highly specialised component of that service will also be put at risk.”

Response:

1. Specialised commissioning is being delegated but not devolved to ICBs. NHS England will remain accountable for delivery of specialised services; this means that it has oversight of all aspects of the quality of specialised services delivered to ICS populations. This will include oversight of change proposals to avoid “overly speedy” reallocation of resources.

3.19 Implementation issues

A number of comments related to the timing and risks associated with implementation: from the perspective of avoiding destabilising respectively providers and commissioners.

Sample comments:

The Shelford Group:

“It is, however, hard to judge the reasonableness of the model until we see the outputs it leads to and are able to compare those outputs with our knowledge and experience of these services. As such, whilst we welcome the intention to refine and iterate the model over time, we would suggest an additional step, which would be to test the initial outputs with providers. This would allow us to identify any impacts or consequences which would be worth further investigation and discussion ahead of full implementation.”

Midlands region:

“There is concern across the Midlands that application of a separate Needs Based formula for specialised services without a consideration of ICB target formulas does not recognize overall need of the population. Early data would indicate that in many cases there is a correlation between relative performance against general allocations and specialised allocations which [it] is felt should be considered in the application of these formulas.”

Gloucestershire CCG:

“Some services that are exactly the same are designated as specialised at one provider because the organisation is a provider of specialised services however the same services provided by a non-specialised provider will be charged to CCG commissioners and not considered specialised. Bringing specialised and non-specialised together in one population-based approach is important in getting to a fair allocation of resources. The approach only works if there is alignment in the pace of convergence between all parts of the formula.”

Midlands region:

“Management of fluctuation needs to have robust risk share arrangements between providers and between ICBs. There is a responsibility to develop these locally but there also needs to be a consistent approach to the management of variation between regions that reflects out of area activity and the migration of patients towards specialised centres.”

Response:

1. Impacts on providers are one stage removed and depend upon decisions to be made by commissioners, and upon responses by providers to the challenge to reach out to more distant communities of need. This makes it difficult to anticipate provider impacts. Even in retrospect it will be difficult to attribute change to introduction of Target allocations in the presence of a bigger and more immediate shift from provider based to population-based allocations.

2. Methodology is best judged against the question: does it use the best evidence to assess relative need?

3. Nevertheless, it is intended that convergence policy will also take account of provider impact, constructed on the basis of plausible assumptions regarding the likely shift in demand for services as access equity is achieved.

4. See also: section 4.3, FAQ 12 Geographical variation in the scope of specialised services; section 4.4, FAQ 13 Commissioner risk; and section 4.4, FAQ 14 Impact assessment and convergence strategy.

3.20 Specialised services share of total allocation

Some questions related to how the fair shares allocation methodology should be affected by variation in the balance of service provision: between specialised and non-specialised, and between drugs and devices and other modalities of service.

Sample comments:

Gloucestershire CCG:

“A general point is how does the formula take into account the expected growth in specialised services for the time that the formula will be used? This could be a distortion if the expected growth in services is not uniform and could therefore impact on some populations more than others. For example, the expected growth in devices that the private sector are developing could be a distorting issue over time.”

NHS Devon ICS:

“Key to the success of this process will be ensuring that correct growth rates are applied to this activity in the future. If this includes Drugs and Devices significant growth will need to be provided.

“Also, any changes to specialist service specifications should be properly funded through the ICS allocation process. Although this is a current risk with non-specialist services this is more significant within the specialist service arena.”

Response:

1. The respective allocation quanta for specialised and non-specialised services are determined independently of the fair-shares methodology. The fair shares methodology is applied to carve up each quantum between ICSs.

2. The funding of drugs and devices within the specialised quantum, year by year, is also determined independently of the fair-share methodology.

3. The distribution of need that is being met by drugs and devices and by other services is expected to be relatively stable through the course of an allocation irrespective of any change in the balance of service provision between these different modalities. Therefore, where the share of drugs and devices within the total quantum available to specialised services increases, the share of other services funding towards ICB target allocations for other services will be adjusted accordingly.

4. Frequently asked questions (FAQs)

4.1 Governance

Questions in this section relate to the technical oversight of the development and execution of the methodology for setting needs-based target allocations and to its evaluation.

FAQ 1. Target allocations governance

Who oversees the development of target allocations?

The NHS England Board is advised on target allocations by the Advisory Committee on Resource Allocation (ACRA). The committee is made up of clinicians, academics, including health economists, NHS managers and finance experts, and representatives of central and local government. The committee, and its predecessors, originally the Resource Allocation Working Party (RAWP), has been providing advice since 1974.

FAQ 2. Quality assurance

How do you quality-assure the targets?

Each component of the target, as well as the final targets and actual allocations, goes through a range of quality assurance processes consistent with standards set out in the Macpherson Review of Quality Assurance of Business-Critical Models and the associated Aqua book. This includes detailed investigation of apparently anomalous results. The quality assurance includes:

• peer review
• independent internal review
• clinical review via representatives from the specialised commissioning clinical reference groups (CRGs)
• independent methodological review (ACRA)

FAQ 3. Evaluation

How will you evaluate the shift to needs-based target allocations for specialised services?

It is difficult to separate the specific impact of the allocations from the impact of other management actions to improve services. However academic studies have shown that this approach can improve the equity of outcomes for individuals. See, for example, The impact of NHS resource allocation policy on health inequalities in England 2001-11: longitudinal ecological study | The BMJ. That work confirms the theory that allocating according to need will reduce health inequalities.

For the proposed shift towards needs-based allocation specifically for specialised services, given that it is taking place in the context of and to support integration of commissioning of specialised services with non-specialised, evaluation should assess the impact of driving system allocations overall towards target. There is academic research underway to achieve better understanding and quantification of unmet need; once that’s completed we should then be able to track the degree of unmet need over time and see if it reduces.

4.2 Rationale

These questions relate to the purpose of the introduction of needs-based target allocations for specialised services and to the financial reward structures that underpin realisation of that purpose.

FAQ 4. Rationale

Why is this change being undertaken?

There are two principal aims in estimating relative need for specialised services: equity and efficiency.

• Government’s mandate to NHS England sets the expectation of basing allocations on the principle of equal access for equal need. Estimating relative need for different populations in respect of specialised services, coupled to the advent of population-based allocations for these services converging over time towards needs-based target allocation, is a means to achieving equal access.
• Comparing existing population-based allocations to needs-based targets enables ICBs, initially working with NHS England, to improve services by considering the best balance of services along a patient’s pathway (from prevention to treatment of severe disease). The extent of opportunities for such pathway optimisation, and the scope for other efficiency gains can be revealed by benchmarking utilisation of services against need.

FAQ 5. Prevention incentives

Will there be scope and reward for ICBs to commission preventative services that pre-empt need for more specialist services?

ICBs with delegated responsibility for commissioning specialised services for a given population will have flexibility to optimise resource use along the service pathway, spending on preventative and other upstream services if that is the best way to secure better health outcomes for patients.

Regarding whether financial reward will accrue to the ICB, needs-based target allocations, to which actual allocations will converge over time, will be determined by estimated need. If actual need falls short of this due to effective investment upstream and NHS England is satisfied that delegated responsibility for these services is being discharged properly, the ICB will still receive the same allocation and be free to reinvest the surplus in other services.

FAQ 6. ICB financial efficiency motivation

If NHS England delegates funding but retains accountability for specialised services, how do ICBs benefit from efficiency delivered?

Delegated allocation for specialised services will not be adjusted period on period to take account of variations in actual spend, so ICBs will retain the benefit of any efficiencies that they deliver.

FAQ 7. Levelling down?

Will movement of funding within a limited budget from well-funded and well-served areas to less well-funded and less well-served areas mean that the better served areas will come to suffer worse service and worse outcomes?

The development of specialised services over the last decade has often involved important investment in improved service quality. Rollout of improved services is usually uneven, leading to higher costs in some areas than others. However, needs-based allocations need not compromise the quality of service of those areas that are currently over target for the following reasons:

• To avoid sharp shocks to budgets, actual allocations will be derived from target allocations through convergence policy (described in chapter 8 of the Technical guide to allocations formulae and convergence for 2023/24 to 2024/25). This policy will moderate the speed of movement towards target, ensuring that the minimum growth for those furthest over target is set at a level that allows stability of services and creates confidence for medium term planning.
• In parallel, efforts (outlined below in Forward Work Programme section 7 Variations analysis to support benchmarking of services) to understand what underlies distance from target – the difference between actual spend on specialised services and the target allocation – will enable identification of areas of over/underspend. In some cases, need may be avoided by improving poor-quality primary or secondary care services. In other cases, benchmarking may point to scope for improvements in technical efficiency or economy.
• Where overspend is attributable to relatively high quality of service or to broader access to services, delivered cost-effectively, that would be reason to restrict convergence toward target allocation to what can be achieved through allocation of real increases in funding, so that movement towards target can be achieved by levelling up.
• Where benchmarking reveals lower eligibility thresholds for treatment in some areas that are overfunded relative to target, it is possible that in some cases this represents cost-ineffective care.

4.3 Scope of services included

These are questions relating to the range of services and service-modalities included in the estimation of target fair-share allocations for specialised services, including the geographically-variable impact of the identification rules on the definition of ‘specialised’.

FAQ 8. Scope of services modelled and allocated

What is the range of services the need for which is modelled, and which is included in population-based allocations that converge over time on needs-based target allocations?

Needs-based target allocations are set for the basket of specialised services that is subject to integrated commissioning on a population basis (whether through delegation or other joint commissioning arrangements). However, the scope of services to be delegated will (for the time being) exclude high-cost drugs and devices.

• All services except highly specialised services and the Cancer Drug Fund (CDF) and the Innovative Medicine Fund (IMF) are to be subject to integrated commissioning for defined ICS populations.
• Highly specialised services will remain centrally commissioned as services with few patients at high cost per patient carry a naturally higher level of volatility that can more easily be managed nationally.
• The CDF and IMF will remain centrally managed in order to manage the risk and oversight of the rollout of new drugs.
• High-cost drugs and high-cost devices are included in the modelling of relative need notwithstanding that commercial arrangements for high-cost drugs and devices remain centralised.

The principled arguments for using a broad scope of services including drugs and devices relate both to the technical robustness of the modelling and to the use of its outputs in benchmarking resource use:

• The formula estimates need for services on the basis of individuals’ health status, and that need is not defined in terms of the modality of treatment to be received. So, the need model is likely to be more robust if a broader range of treatments is in scope.
• Similarly, one can better benchmark the utilisation of healthcare resources relative to need using the broadest basket of treatment modalities, so that different trade-offs across the country between modalities of treatment do not distort judgements of efficiency. For drugs that balance is important for example for cancer, given variation in the use of chemotherapy. For devices, that is probably most important for cardiac services: we want to compare ICS populations’ utilisation of resources for cardiac services relative to the need for those services without a bias in favour of those ICSs making more intensive use of high-cost devices.

This also creates an argument not only for including drugs and devices in the assessment of fair shares but also in the actual and target allocations. It would be advantageous for example were the financial reward for diagnosing cancer early (and the business case for investing in detection) to include cost-saving from avoiding chemotherapy because the cancer has been excised early. Ditto for preventative interventions that pre-empt need for cardiac devices. Ideally, from this perspective, ICBs would be financially accountable if they use more than their fair share of resources due to over-use of drugs or of devices.

There are however strong commercial reasons for maintaining central commissioning. Keeping the scope of allocations broad should facilitate design of financial structures that maintain central commissioning yet give local commissioners a measure of financial accountability for the volume of drugs and of devices consumed by their population.

In the meantime, a method is required for carving out a fair share of drugs and devices from the target allocation. Drug and device spend as a share of the total varies systematically with need drivers (notably age: younger populations use fewer drugs and devices as a share of total specialised spending). To adopt a neutral position regarding the role of drugs and devices in meeting service need, we take the baseline share of drugs and devices spend in total resource use for an ICS population as normative. This implies that if an ICS population is using more than its fair share of resources, it will be using more than its fair share both of drugs and devices and of other resources: distance from target for the allocation excluding drugs and devices will therefore be the same as is estimated regarding its use of resources overall.

The target allocation for an ICB is therefore calculated by estimating a fair share of total resources and then scaling down by that ICS population’s baseline proportion of resource spend on drugs and devices, that is by removing from the notional total fair share allocation that proportion that relates to drugs and devices. Thus if its fair share overall is £500 million, and its drugs and devices baseline proportion is α, then its target allocation will be £500m times (1-α). (A similar approach is taken to partitioning the remaining allocation between services to be delegated immediately and any whose delegation is deferred.)

This method will be adapted if in the forward allocation the national share of drugs and devices in overall specialised spending shifts. If the share is projected to expand we assume that the fastest growth is likely to be in systems with more intense use of drugs and devices (as we assume that the intensity of use and the rate of expansion are both driven by demographic and epidemiological factors). This is done by growing the proportion of high-cost drugs and devices (HCDD) spend by a fixed growth factor, so that the total HCDD budget matches national assumptions. So, if for ICB i the proportion of HCDD spend in 2023/24 is α _23/24 then  where  is the HCDD growth factor.

FAQ 9. Mental health

When will MH services be included in needs-based target allocations for specialised services alongside specialised physical health services?

We hope to add MH models of need in the course of 2024, to inform allocations from April 2025.

FAQ 10. Time-varying range of specialised services

How does the model take account of the fact that the scope of specialised services varies over time due to innovation?

The scope of specialised services can change over time with the introduction of new services. Generally, on introduction, these services will not materially affect the appropriate allocation to ICS populations. This is due to: the small scale of new innovations relative to the overall budget; the fact that common factors (age-related morbidity adjusted for deprivation) drive need for most specialised services; and the limited pace of convergence towards target allocations. However, over time, innovations cumulate. This, together with demographic changes, necessitates periodic remodelling of need (see FAQs 31-32, on model renewal). Exceptionally, there may be need for ad hoc allocations.

FAQ 11. Delegation to ICBs and the definition of ‘specialised’

Won’t the modelling quickly become out of date as the definition of which services are specialised changes over time, particularly following integration of specialised commissioning with that for other services?

The integration of specialised commissioning with commissioning of other services will not end the distinction between specialised and non-specialised services; specialised commissioning of in-scope services is being delegated but not devolved from April 2024. Over time, the relative weight of different specialised services may change as discussed in FAQ 10.

FAQ 12. Geographical variation in scope of specialised services

How does the model take account of the fact that the range of services deemed to be specialised varies between places?

As regional commissioning hubs have set aside the Identification Rules (IR) aside in some areas, and as the IR themselves lead to a broader scope of some specialised services in areas in which Provider Eligibility List (PEL) providers are located, it is indeed the case that the range of services deemed to be specialised varies between places.

The model seeks to assess underlying relative need for specialised services, assuming that the right total sum is dedicated to specialised services across the country. In this it properly mirrors the approach taken in the general and acute model. It is true that an area that has a broader scope of services funded as specialised may for that reason appear to be over-spending relative to modelled need; by the same token, however, it should appear to be under-utilising non-specialised services relative to the target allocation for those services. It is therefore important that Distance from Target (DfT) is reviewed in tandem for the specialised and non-specialised models, in the context of careful ex post examination of reasons for variation in spending.

4.4 Destabilisation risk

These questions explore how the financial stability of commissioners and providers of specialised services may be affected by the delegation to ICBs of population-need-based allocations.

FAQ 13. Commissioner risk

Specialised services commissioning was not delegated to CCGs in 2013 in part because the budgetary risk was thought too great for them to bear; how are ICBs to handle this risk?

Commissioning of specialised services shifting over time towards a needs-based allocation to ICS populations, with commissioning being delegated to ICBs. The first step, delegation to ICBs of population-based allocations, in joint committees with NHS E from April ’22 and with full delegation phased in from April ’24, creates budgetary risk for ICBs arising from year-to-year volatility in demand for specialised services. (The transition towards needs-based allocation will little affect this risk.) However:

• The extent to which ICBs can handle volatility exceeds that of CCGs; their greater scale gives them more resilience to manage volatility in demand. There were around three hundred CCGs when specialised commissioning was centralised in 2013; there are 42 ICBs.
• For those services that are more appropriately commissioned on a multi-ICB footprint, risk-sharing arrangements between ICBs mitigate volatility, with ICBs working together in Joint Committees.
• Providers are contracted on a largely fixed basis within financial years under the aligned payment and incentive (API) approach introduced as part of the NHS Payment Scheme. This mitigates commissioner risk in comparison to the Payment by Results arrangements prevailing in 2013.

FAQ 14. Impact assessment and convergence strategy

How will you assess system and provider impact when setting convergence strategy?

Convergence policy sets the speed of movement towards target. Convergence policy for specialised and non-specialised services will be developed in tandem, so that overall system capacity for change is respected.

Impacts on providers are one stage removed and depend both upon decisions to be made by commissioners and upon responses by providers to the challenge to reach out to more distant communities. This makes it difficult to anticipate provider impacts. Nevertheless, convergence policy will also take account of provider impact, constructed on the basis of assumptions regarding the likely shift in demand for services as equity of access is achieved.

FAQ 15. Disinvestment risk

Is there a risk that money moved to areas that are currently under-utilising specialised services (relative to modelled need) will be used to increase funding for non-specialised services, and that consequently overall demand for specialised services will shrink rendering specialised centres financially and clinically unsafe?

There is such a risk. It is mitigated by the following considerations:

• Specialised commissioning is being delegated but not devolved to ICBs. NHS England will remain accountable for delivery of specialised services; this means that it has oversight of all aspects of the quality of specialised services delivered to ICS populations. NHS England regions will routinely monitor performance to identify and address budgetary risk early.
• ICBs with delegated responsibility for commissioning specialised services will be accountable for providing access to these services in line with need and with the service specification. Quality standards and provider eligibility criteria will continue to be set nationally to regulate the entrance of new providers to the market. Guidance will apply to the decision making and application of commissioning changes to ensure they are appropriate and managed.
• Variation analyses and benchmarking can highlight where populations appear to be underserved by specialised services, which should provide ICBs with the information they need to use additional funding to make good shortfall.
• Ultimately, the purpose of integration and delegation is to enable pathway optimisation. If an ICB determines that it is in patients’ interests for example to direct extra funding upstream to pre-empt need for specialised services, and if that leads to a drop-in specialised activity, that would be a successful outcome, notwithstanding that the specialised service would have to adapt to a gradual decline in scale of operation.

4.5 Model specification

The following questions concern the technicalities of estimation of relative need for specialised services.

FAQ 16. Methodology to estimate relative need

What approach do you take to setting target allocations?

There are 4 separate steps involved in all the allocations models, including that for specialised services:

1. We look at the share of the national population in each area, based on GP registered lists, as the starting point for each area’s target allocation or ‘fair share’.
2. We then adjust each area’s share of total resources according to our estimates of whether their relative need for healthcare is higher or lower than the average.

To do this, we use a set of statistical formulae to estimate local healthcare needs. These are built up from pseudonymised data on individuals’ utilisation of services linked to a range of data on the characteristics of those individuals that might plausibly be associated with their need for those services (the need variables). We estimate and set aside factors irrelevant to need that might nonetheless be affecting access and costs (the supply variables).

That means we start with a major assumption that in general, services reach people who need them, even if levels of access and quality and cost of services vary from place to place. So, we can use information about the people who get services to gauge what predicts need.

In our case, information about individuals’ use of specialised services, together with everything else we know about them, including in particular their previous two years’ health records, and what we can deduce about their circumstances from where they live, can be brought together to help us understand what factors are associated with recourse to specialised services. We can then apply that standard model of need up and down the country, smoothing out unfairness to create our target fair-shares allocation.

3. That set of target needs indices is adjusted for unavoidable differences in unit input costs between areas due to their geographical location using the Market Forces Factor (MFF). The specialised services MFF is derived in the same way as the general and acute MFF adjustment, taking account of the share of services to a given population flowing from providers with different MFF adjustments. As patients are more likely to travel to receive specialised services, the specialised MFF is less differentiated than that for other services. The specialised MFF reflects the pattern of service use in the year to February 2020 (the last twelvemonth before data were distorted by COVID-19 effects).
4. The MFF-adjusted fair-share estimate is subject to a further adjustment for health inequalities and unmet need that is missed by the utilisation approach: see FAQs 26-28.

FAQ 17. Utilisation vs prevalence methodologies

Why is the model based on data on the utilisation of hospital services and not the prevalence of disease?

In a nutshell, we lack the information to base allocation on prevalence directly, so instead we exploit utilisation data linked to patient-level information to deduce what characteristics generally lead to specialised service use, and estimate relative need on the basis of the geographical distribution of those characteristics.

Ideally, we would use a bottom up approach to build fair shares, by estimating what care ought to be provided at what cost in different parts of the country. Unfortunately, this is not currently attainable. We lack detailed data on: the prevalence and incidence of disease in England and its severity; and evidence of what is the most efficient and effective treatment that ought to be provided and at what cost.

We therefore make the most of the data available to us at the patient level, but we are careful to use information that represents health needs rather than simply counting activity levels. The patient-level data we use allows us to build a model that can predict costs for each individual, and it draws on information most closely linked to a patient’s health needs, mainly diagnostic information, rather than, for example, number or type of interactions with clinical teams.

FAQ 18. Specialised need arising from poor-quality care upstream

Need for a specialised service might be a consequence of having to compensate for inadequate primary/secondary services. How does the model manage that dynamic?

The personal diagnostic histories that drive the model estimates of need should pick up underlying need for specialised services. As a consequence, those areas where relatively poor primary and secondary services are leading to a larger proportion of morbid patients having recourse to specialised services may find themselves using more specialised resources than their target and will be challenged over time to invest more in secondary prevention to address that issue. It is a virtue of introducing needs-based target allocation alongside integration of commissioning between specialised and non-specialised services that the integrated commissioners will be in a position to optimise services along the pathway in response to such a challenge. (Admittedly, these favourable incentives are somewhat weakened by any inter-ICS migration.)

The model’s use of diagnostic history does not distinguish between early and late diagnosis: late diagnosis will often require more specialised resource. The insensitivity of the model to geographical variation in the propensity to diagnose late creates a helpful reward to areas that focus effort on earlier diagnosis.

Note however that poor-quality primary and secondary services can also reduce demand for specialised services: need may go unidentified and consequently unmet (leading in some cases to premature mortality). Such need may be undiagnosed and therefore missing from the model. Separate work to investigate this is under way. See Forward Work Programme section 5.5 Assess extent of undiagnosed need.

There will also be variation in the quality of upstream services that more distally affects the prevalence of the morbidities that the specialised model does pick up, given that it is recent diagnostic history that it uses to predict need. The specialised model is designed to track such need and fund care notwithstanding that in principle some such need could have been avoided. An ICB that invests in public health and prevention services will eventually have its specialised target allocation revised downwards as its population becomes healthier.

FAQ 19. Reliance on inpatient diagnoses

The historical diagnoses used as need variables are derived from inpatient data; how will need for services primarily provided in outpatient settings be captured?

For services we can test, it appears that patients using outpatient-based services will nonetheless have relevant diagnoses on their inpatient records (perhaps relating to admissions for other issues – given that the model is also sensitive to co-morbidities).

A separate concern is that that areas with a tendency to treat patients in outpatient settings will appear to have less need. Similarly, areas with privately funded health care that is not submitting to the Secondary User Service (SUS) may have their need underrepresented by the model. That concern is somewhat mitigated by the likelihood that such patients may make fewer calls on NHS services.

On the forward work plan we propose to explore linking clinical registry data that will include diagnoses registered in outpatient settings. See Forward Work Programme section 5.1 Review diagnostics categories.

FAQ 20. Expenditure vs. costs

Why is the model based on commissioner expenditure (in the PLCM) rather than provider-costs? Will it be invalidated as expenditure moves away from a price x activity (PbR) basis?

Ideally the model would be built using cost-weighted activity as the measure of persons’ specialised resource utilisation, with the costs-weights derived from estimates of the efficient cost of providing each service. This is the approach taken in estimating the General and Acute allocations model (with most services cost-weighted using Tariff, which in turn is derived from reference costs). In due course, such cost-weights might be derivable for specialised services from the Patient Level Information Costings System (PLICS). However, the coverage and reliability of PLICS is not yet adequate for this purpose (and few specialised services by value are Tariffed based on reference costs).

Instead, we use commissioner expenditure on different services as a proxy for their relative efficient cost, on the assumption that the ratio of aggregate expenditure to costs is unlikely to vary much from service to service. It is true that for many services the amount spent on a given service depends on the local pricing and counting arrangements agreed between NHS England regional teams and service providers, the generosity of which may vary from place to place. However, the modelling takes this into account through the introduction of supply variables for each CCG and for each provider, so that systematic over- or under-pricing will be recognised and excluded from the estimation of relative needs indices.

Given that PLCM data is being used in the model only as a proxy for assessing the relative resources used in providing different services, the shift away from activity-based payment of providers will not in any way invalidate it. How commissioners choose to pay their service providers is totally independent of the determination of the funding allocation within which they conduct their commissioning, and it is only the latter that is estimated by the allocations model.

Nevertheless, we hope in due course to shift estimation of the specialised model to the use of PLICS. See Forward Work Programme section 5.6 Improving coverage of patient level data.

FAQ 21. Research evidence

What independent academic research is used in the target allocations formula?

The allocations methodology has long been based on independent academic advice. The key methodological approach that now underpins our allocations formulae is person-based resource allocation (PBRA) developed during the early 2000s by academics at the Nuffield Trust, University of York, University of Manchester, New York University, Health Dialog and the London School of Hygiene and Tropical Medicine, and developed for mental health services by a team from the University of Manchester.

FAQ 22. Further reading

Where can I find out more detail about the methodology?

Details of the specialised allocations methodology is available on the NHS England allocations webpages.

FAQ 23. Calculations

Can I see the underlying calculations? Is there an equivalent of the exposition book from PCT allocations?

The inputs to target allocations, including the population estimates and the need estimates resulting from the formulae, are set out in the spreadsheets accompanying the technical guidance to allocations. Those spreadsheets include notes and, where feasible, formulae and Stata code to show how the targets are constructed. In addition, the econometric modelling that underpins the formulae is described in detail in the research reports published in the same place (https://www.england.nhs.uk/allocations/).

FAQ 24. Halo Effect

Does the model take account of people who move close to care providers?

We lack evidence of such a halo effect, and it is plausible that it is rare: for acute illnesses people are unlikely to move closer to care; and for longer term conditions, with few exceptions, people are more likely to move near family support. However, we would be interested if there is any evidence of the extent of this phenomenon for different conditions. We would take the following approach to addressing it:

• The modelling accounts for the resource needs of those who have moved close to a hospital so long as their diagnostic record reveals that need.
• In the research programme, we intend to explore whether the diagnostic categories can be refined better to capture those specific groups for whom this is more likely, in particular those with long term conditions on clinical registries. (See Forward Work Programme section 1 Review diagnostics categories.)
• Where evidence emerges of a halo effect, we can ensure that relevant diagnoses are salient in the model inputs. (However, for some services it may be more appropriate for funds to be allocated to places where need arises originally, so to obviate the need for migration of patients to be near provision.)

4.6 Missing from the model

This section of FAQs address concerns that a utilisation-based model estimated using data that is now several years old may miss various categories of specialised need.

FAQ 25. COVID-19

How does the model take account of the impact of COVID-19 and its aftermath (including elective recovery and long COVID) on relative need for different specialised services and on their relative cost?

The model is based on pre-COVID-19 data; no adjustment is made for COVID-19 related costs. This is correct for the time being, as COVID-19 costs, and recovery-funding to address backlog demand, are met by separate ad hoc allocations.

Future models will take into account long COVID and any enduring impact on relative costs related to infection prevention etc. as these will have fed through into the model’s dependent variable.

FAQ 26. Health inequalities

How are you addressing health inequalities?

An adjustment to allow for investment in reducing health inequalities is made to the target allocations derived from the modelling results. This reflects our objectives to tackle health inequalities. The health inequalities and unmet need (HI-UN) adjustment combines this adjustment (for reduction in health inequalities) with an adjustment to recognise that not all need is captured by a utilisation-based model, particularly need that is undiagnosed or that is unusually costly to address.

The HI-UN adjustment thus addresses three aspects of need not captured in the utilisation-based model through which need for services is modelled econometrically: (a) undiagnosed unmet need; (b) the healthcare burden of treating disadvantaged people, where not otherwise recognised; (c) additional resource to reduce inequalities.

The HI-UN adjustment targets a greater share of available resources at communities estimated to be most affected by these gaps, with an expectation that the additional resource will help to meet unmet need and to reduce health inequalities.

The HI-UN adjustment is currently based on data on avoidable mortality at a small area level, aggregated up to system level, and thus takes account of pockets of disadvantage within otherwise better-off systems. ACRA has advised on the metric of disadvantage to be used in allocating the additional funding, but ACRA has consistently advised that it lacks evidence upon which to judge the scale of adjustment required. Research is under way that may supply some evidence perhaps in time for 2025/6 allocations.

In the meantime, NHSE has chosen to apply an HI-UN adjustment of 15% of the primary care allocation, 10% of the core CCG allocation, and 5% of the specialised allocation. This is based on the view that Health Inequalities are best addressed upstream and so the relative weighting should be greater upstream.

Technical details of the derivation of the HI-UN adjustment for all commissioning models are set out in section 4.8 of the Technical guide to allocation formulae and convergence for 2023/24 to 2024/25 revenue allocations.

Work to assess the appropriate scale of the adjustment for specialised services is in included in the forward work programme, see sections 5.4 Estimate the healthcare burden of disadvantage and 5.5 Assess extent of undiagnosed need.

FAQ 27. Multidimensionality of health inequalities

Why are only mortality data used in the health inequalities adjustment when health inequalities are multi-dimensional?

We recognise that health inequalities are a complex and multi-dimensional issue. ACRA’s view is that avoidable mortality is a good proxy for a range of issues relating to health inequalities including deprivation and morbidity, particularly in relation to physical health services, as well as being a health outcome measure in its own right. The forward work programme includes internal analysis and external academic work to review this approach and refresh it should better data and methodologies be available, particularly for mental health.

FAQ 28. Undiagnosed need

The model is driven largely by individuals’ diagnoses received in secondary care. How then will you account for the many people who miss being diagnosed either due to ignoring invitations or because they are waiting for diagnostics?

The model should identify likely need on the basis of historical association of use with morbidity, so it should reveal where specialist need is not currently being diagnosed as expected. However, there may be groups of patients whose need is systematically missing from the record. In some cases, failure to diagnose and refer for treatment will result in avoidable mortality – and the HI-UN adjustment will direct resources to help to address this shortfall (see FAQ 26). This issue is the subject of the work described in Forward Work Programme section 5.5 Assess extent of undiagnosed need.

FAQ 29. Remote communities

Will the modelling of need take into account the additional costs incurred in providing specialised services to remote communities?

This is part of the forward work programme; see section 5.2 Allow for diseconomies of small scale.

4.7 Timing of implementation and renewal

These questions relate to the inevitable delay between modelling of relative need and the implementation of the derived estimates to inform allocations.

FAQ 30. Implementation timing

Wouldn’t it be better to await the refinement of the model promised in the forward work programme?

The agenda for refinement of the modelling is likely to stretch indefinitely forward. ACRA views the outputs of the model as the best currently available estimate of true need. It is better to use roughly right estimates than to leave allocations completely unanchored by estimates of need.

FAQ 31. Model updating

When will the model be updated?

The model will be updated as required to take account of data and modelling improvements and service changes. Although modelling is continually refined, target allocations have typically tended to be set every 2-3 years.

FAQ 32. Reopening allocations

Under what circumstances will allocations be reopened/reviewed?

NHS England reserves the right to change allocations in a number of specific circumstances where the financial stability of the commissioning system is challenged or it is clear that the allocations are no longer fair in their distribution to health economies. Examples of the circumstances under which the allocations will be reviewed include:

• a disproportionate financial imbalance in any part of the commissioning system
• a new government policy with additional funding creating an additional pressure in one area
• a disproportionate increase or decrease in the share of the national population caused by a change to underlying population statistics or changes in the pattern of GP registration
• a disproportionate increase or decrease in the need-weighted share of the total need-weighted population caused by a change to underlying age structures or populations or relative levels of deprivation
• a new national contract or pay award established by the government that requires additional funding or redistribution of resources
• the impact of public sector pensions revaluation and the need to distribute this funding to providers
• any other change in mandate funding.

5. Forward work programme

Separate needs-based allocations for specialised services will continue to be required for so long as accountability for achieving value for these services is separated from that for other services. Work to improve estimation of specialised need takes its place in the forward work programme across the allocations programme as a whole, under the guidance of ACRA and its Technical Advisory Group.

Possible elements in the work to refine estimates of relative need were included in the engagement document, and priorities were sought as follows:

“iii. Do you agree with the proposed forward work programme to refine the model over time set out in section 3.1? Please comment on their relative priority.”

Feedback received on this question and more generally regarding the specialised services allocation methodology suggested the following elements of investigation should be prioritised:

5.1 Review diagnostic categories

Task:

To review the diagnostic categories used to classify individuals’ diagnostic histories. This review should include the option of designing them more closely around rules identifying specialised spending, taking a view service by service of the trade-off between increased precision in identifying enduring specialised need, and avoiding inclusion of supply effects where specific diagnoses are more likely to be given by specialised providers. This will include consideration of use of clinical registry data to capture information about patients, such as those with congenital heart conditions or hepatitis C, whose care is predominantly provided out of hospital, and whose diagnostic history may therefore be missing from the inpatient dataset relied on in the model. We will in this exercise also review the appropriate number of diagnoses to include in modelling comorbidities.

In engagement comments, this was given priority by:

• Great Ormond Street Hospital for Children Foundation Trust
• Suffolk and North East Essex ICB (top priority)
• East and North Hertfordshire NHS Trust
• NHS Norfolk and Waveney CCG

5.2 Allow for diseconomies of small scale

Task:

To investigate diseconomies of small scale in the provision of specialised services where there is a clinical necessity for providing services close to patients’ homes in remote areas, considering in particular neonatal services, radiotherapy; and rural and coastal areas (as referenced by CMO’s 2021 report on health in coastal communities). Work would be required to apply the approach developed in ACRA(2021)16 to specialised services; this found economies’ of scale in department size (mitigated by diseconomies of scale by site and provider). In the first instance, we can investigate whether coastal patients appear to incur systematically higher costs of treatment.

Sample comments:

East of England region:

“… it will be important to explore and understand how to allow and account for diseconomies of small scale, in particular related to providing specialised services close to patients’ homes in the rural and coastal communities (as referenced by CMO’s report on Coastal Communities). Rural and coastal communities are over-represented by populations that are likely to utilise specialised services.”

NHS Bristol, North Somerset and South Gloucestershire CCG/ICB:

“As part of the work on supply side variables we would be particularly interested in understanding the impact of coastal deprivation as set out in the recent CMO’s report on the supply side variables for specialised services. …

“Adjustment for diseconomies of scale: we agree with the need to make adjustments for higher costs in sparsely populated areas and note the link to coastal populations in the future work programme. However, we consider it a key part of the process to set a needs-based formula for 2023/24 to understand any unintended consequences for coastal populations.”

This task was assigned priority also by:

• North East and Yorkshire Region
• Suffolk and North East Essex ICB (top priority)
• Blackpool Teaching Hospital
• East and North Hertfordshire NHS Trust
• NHS Bristol, North Somerset and South Gloucestershire CCG/ICB
• NHS Devon ICS
• NHS Norfolk and Waveney CCG
• Somerset CCG
• East Suffolk and North Essex Foundation Trust
• South West Region
• Gloucestershire CCG

5.3 Explore alternative model structure

Task:

To investigate alternative model structures, in particular models that distinguish estimation of likelihood of requiring specialised services from estimation of the expected cost of such services if required. This should enable model predictions of the numbers of patients separately from predictions of the cost per patient – which would support better understanding of divergence between actual and modelled resource use (‘distance from target’).

5.4 Estimate the healthcare burden of disadvantage

Task:

It is thought that the resources required to serve individuals with similar conditions is greater for those suffering from disadvantages: such individuals may for example require language services, or more follow up to help them with adherence or recovery in the context of multiple life challenges. Engagement responses suggest this is true of HIV services, for example. If the measures of utilisation and cost of provision used to estimate the need model are insensitive to such costs (as is the case for the HIV adjustment), or if these extra challenges systematically result in poor access or poorer quality of service, then modelled need may be insufficiently responsive to deprivation.

There is some support in the data on relative spend on most deprived LSOAs referenced above for the supposition that the healthcare burden of disadvantage is not fully captured by the model. ICS populations using more than predicted specialised resources do tend (with a 0.45 correlation) to have higher relative age-standardised spending specifically upon their most deprived populations (most-deprived quintile LSOAs), suggesting that overspending may be occasioned by the burden of treating disadvantaged individuals.

Work would be undertaken to detail and quantify the extent to which service costs are greater in serving disadvantaged communities. The review should use PLCM and PLICS data to explore patient level costs associated with deprivation; and also, engagement with clinicians and commissioners of services to explore which services (like HIV) and types of costs should be investigated. The question then is whether such costs flow through into the PLCM and therefore are likely to be captured by deprivation variables in the model, and, if not, what adjustment is required.

We would also wish to investigate whether some such need is systematically unmet and missing from modelled need for that reason.

Sample comment:

HFMA (top priority):

“… as the methodology is based upon patient level data, the approach will already include some aspect of the burden of disadvantage, where people are accessing services. However, work is needed to understand the profile of need for specialised services across the country and its relationship with inequality, which is likely to highlight areas of undiagnosed need”

This task was assigned priority also by:

• Sheffield Teaching Hospital
• Blackpool Teaching Hospital (top priority)

5.5 Assess extent of undiagnosed need

Task:

To build on the NIHR unmet need project to assess the extent of need for specific specialised services that is not captured by the diagnosis variables in the model. This work-strand is linked to the review of diagnostic categories, mentioned above. This work necessarily focuses on services for which a methodology is available: for example, undiagnosed need for cancer services is revealed by late-stage presentation; for cardiac services by identifying geographical variation in mortality where the primary cause of death is given as an avoidable cardiac condition, and where there is no recorded diagnosis of a cardiac condition. Renal and HIV undiagnosed need may be suspected in areas in which referral for specialised services is tardy. All such focused assessments can be corroborated for plausibility using prevalence estimates for the conditions underlying specialised need.

Sample comments:

Sheffield Teaching Hospitals:

“[I]f the model could include variables relating to prevalence (even if this were modelled prevalence data with all their shortcomings), then this could allow exploration of unmet need. Historical activity as a proxy for need means there will be some inequality baked into the model particularly where there is variation in diagnosis and referral rates.

“We welcome that the model attempts to adjust for deprivation, and while this will be a significant improvement on previous allocation models, but the use of avoidable mortality as a proxy for this only goes so far. Combining this with an activity-based approach rather than a prevalence one, means that there won’t be a significant shift towards a proportionate universalism that really addresses health inequality, particularly inequality in healthy life expectancy for example.”

Birmingham and Solihull ICS/ICB (top priority):

“Refining the methodology for the health inequalities adjustment as using preventable mortality doesn’t take into account deprivation and ethnicity, unmet need, long term chronic disease and quality of life. No prevalence of disease has been taken into account.”

This task was assigned priority also by:

• HFMA
• NHS Bath and North East Somerset, Swindon and Wiltshire
• Suffolk and North East Essex ICB (top priority)
• Blackpool Teaching Hospital
• East and North Hertfordshire NHS Trust
• NHS Bristol, North Somerset and South Gloucestershire CCG/ICB
• NHS Devon ICS
• NHS Norfolk and Waveney CCG
• Somerset CCG
• East Suffolk and North Essex Foundation Trust
• South West region
• Gloucestershire CCG

5.6 Improve coverage of patient level data

Task:

The patient level information used for contract monitoring in the Patient Level Contract Monitoring (PLCM) datasets used for modelling specialised need is incomplete, and the shortfall in value relative to total spending on specialised services varies between specialities. This could introduce a bias in the estimation of need if for example a service with relatively low coverage is more prevalent in a particular age group than other services, in which case we may be underweighting that age group in the estimate of need. Assessing this is challenging given uncertainty as to whether underrepresentation in the PLCM also relates to patient numbers. We hope to address this problem by improving the comprehensive use of Secondary User Services to cover specialised as well as non-specialised hospital activity, using cost weights derived from the Patient Level Information and Costing System (PLICS). The PLCM datasets would then be needed only to supplement SUS, in particular for drugs and devices. We will keep this issue under review at the next round of modelling; we anticipate that such improvements in coverage will reduce the risk of bias and obviate the need for adjustments.

5.7 Variations analysis

Task:

To undertake detailed work to understand, ICS by ICS, and if possible, place by place and service by service, what is responsible for the distance between actual expenditure on specialised services and target allocations (reflecting modelled need) for specialised services.

The analysis aims to support local systems seeking to bring resource utilisation in line with modelled need through benchmarking of their service provision against others with similar need but different levels of resource use. The aim would be to encourage ICBs to recognise that pace of convergence will be slow enough to enable them to adjust services in good time; to think strategically rather than transactionally about the introduction of a gradual transition towards needs-based allocations.

Analyses could be conducted along the following dimensions:

• by smaller populations: focusing on place – however that is defined (so long as it can be built up from GP-practices)
• separating services: initially focusing on major service areas – cancer, cardiac and renal, neuroscience, critical care services, for each an indicative needs-based allocation has been produced. (A paediatric services model is a plausible addition.)

For each service and for the aggregate model (at ICS and place), the aim would be to distinguish as far as the data allow between the following components of variation:

• in scope of services (where, due to differences of implementation of the identification rules [the IR] determining what services are classified as PSSs, the scope of services funded by specialised commissioning varies). See section 4.3, FAQ 12. Geographical variation in scope of specialised services.
• in the extent to which need is met, distinguishing between variation in access to services – the number of patients treated per head of population (which may be attributable for example to differences in access related to referral-routes or to differences in eligibility thresholds or to capacity constraints including staffing), and variation in spend per patient (reflecting the quality or efficiency with which need is addressed). See above, 3 Explore alternative model structure, for a possible methodology for assessing need-reflective affordable access rates for specialised services.
• in the extent of presenting need relative to modelled need, considering in particular differences in utilisation of specialised services attributable to differential effectiveness of upstream services, for example in primary and community services. Improving these upstream services might be an appropriate strategy to reduce spending on specialised care in some areas.
• in outcomes: at least as a control analysis, to see whether variation in overall resource use relative to the combined allocation models is associated with variation in outcome measures such as standardised avoidable mortality rates (after controlling for variation in wider determinants of health).

In the engagement document, the proposal to explore why ICS populations’ use of specialised resources varies from that predicted by the need-based fair-share estimates (“variations analysis”) was presented as the second part of a possible forward work programme, in section 3.2, and accompanied by the following question:

• “iv. Do you agree with the proposed forward work programme to undertake variations analysis to support benchmarking of services set out in section 3.2? Please comment.”

Sample comments:

Midlands region:

“There is a view from the Midlands that there is differential access to specialised services specifically linked to supply of services. Whilst we recognise that this is not a factor in developing needs-based formulas, we do feel that the alternative routes of treatment that result from the geographical supply of provision needs to ultimately [be] considered in the calculation of future allocations.”

South West region:

“We would also like to propose adding to the work programme to include a more outcomes-based approach to resource allocation.” [Linking variation analysis to outcomes was a major theme in the comments.]

East of England Specialised Provider Collaborative:

“Children’s: the document states that four indicative sub-models are being created for cancer, cardiac, renal and neurosurgery services, and that initial variations analysis will focus on these areas. We suggest that you should also consider creating an indicative sub-model for children’s, particularly given that utilisation of children’s services may display different patterns versus the overall utilisation of physical health services.”

Publication reference: PRN01046i