• Health care equity, health equity and resource allocation ...


  •   
  • FileName: AsthanaGibson96.pdf [read-online]
    • Abstract: health care resources should be geographically distributed to. ensure equal opportunity of access to health care for people at ... the highest absolute burden of ill-health, but at those which have ...

Download the ebook

Health care equity, health equity
and resource allocation: towards a
normative approach to achieving
the core principles of the NHS
Sheena Asthana and Alex Gibson
Introduction
Since the 1976 appointment of the Resource Allocation Working
Party (RAWP), health care equity has been an explicit goal of
resource allocation within the NHS. According to this principle,
health care resources should be geographically distributed to
ensure ‘equal opportunity of access to health care for people at
equal risk’. RAWP was superseded by the Resource Allocation
Group in 1995, which was in turn replaced by the Advisory
Committee on Resource Allocation (ACRA) in 1997. Throughout,
equal opportunity of access for equal needs has remained a key
objective of resource allocation. In 1999, however, ACRA introduced
an additional requirement; that resource allocation should
‘contribute to the reduction of avoidable inequalities in health’. This
interpretation of equity as an ‘equal opportunity to be healthy’ is
otherwise known as health equity.
There has been little explicit debate about the tensions that arise
between the principles of health care equity and health equity or to
the fact that, in practice, these goals are difficult to reconcile. A key
difficulty is that the geographical distribution of population ‘needs’
varies according to the principle adopted. In order to promote
"equal opportunity of access for equal needs", the distribution of
funding should reflect the existing burden of disease. In order to
promote an "equal opportunity to be healthy", funding needs to be
targeted so as to reduce the health gap between the most
advantaged and least advantaged groups. This implies that
resources should not necessarily be directed at populations with
the highest absolute burden of ill-health, but at those which have
the worst health in terms of age-standardised measures. The point
is that a population with a high absolute burden of need (perhaps
because it comprises a large proportion of older people) may well, in
age-standardised terms, be relatively healthy. Thus depending on
whether crude or standardised measures are used, the distribution
of ‘need’ is very different.
The aim of this paper is to explore how the current system of
resource allocation in England resolves this tension. Does the
distribution of health care funding reflect the core principle of the
NHS, that of equal opportunity of access for equal needs, or has
health equity displaced health care equity as the key objective of
resource allocation? If such a shift in emphasis has taken place, is
the targeting of additional NHS resources at deprived areas likely to
promote greater health equity? Is the current method used to
distribute resources ‘getting it right’ or should alternative
approaches be considered? In order to address these questions, the
paper provides a description and critique of the current weighted
capitation system and concludes with a brief outline of how a more
normative approach to setting health capitations could – and
should - be developed.
The Weighted Capitation Formula
The technical approach to allocating health care resources also
owes much to the work of original Resource Allocation Working
Party. RAWP recommended that revenue resources for hospital and
community health services (HCHS) should be distributed on the
basis of population, weighted according to the ‘need for health care’
and the costs of providing services. In the intervening period, much
of the technical effort that has gone into formula determination has
focused on the measurement of the 'need for health care'. This is
conceptualised as comprising two elements; 'age-related need' and
'additional need'. The latter concerns that part of a population's
need for health care which is over and above that due to its
demographic composition. This addresses, in other words, the
effect of socio-economic deprivation on a population's healthcare
needs. Unfortunately, whilst the link between deprivation and
health status is well established, it is difficult to quantify this
relationship in terms of resource needs. The variety of different
approaches used since 1976 to measure (or use a proxy for)
additional need is symptomatic of this difficulty.
Since 2003/04 the method used to distribute resources between
Primary Care Trusts has followed the recommendations of the
'Allocation of Resources to English Areas' (AREA) report published
in 2002 (Sutton et al, 2002). Like the previous ‘York Model’, this
adopted the analytical principle that 'statistical modelling of the
relationship between utilisation, socio-economic variables
(including measures of health) and supply factors can identify
which socio-economic variables are indicators of need because of
their effect on utilisation' (Sutton et al, 2002, p.1).
To this end, a two-step procedure was used to model age-related
and additional needs effects. This involved first establishing
average levels of resource use for each of a series of age bands.
These were then used to create 'indirectly-standardised resource
use ratios for small areas' which were compared against a set of
utilisation and supply indicators at small area level (Sutton et al,
p,69). The goal was to define a plausible set of socio-economic
'additional need' indicators which would best explain, once age and
supply factors had been taken into account, how utilisation varied
at ward-level across England.
Any 'utilisation-based' approach such as this must assume that
health service utilisation, once age and supply factors have been
taken into account, is at least a reasonable proxy for health care
needs. Analysis of Health Survey for England data suggested to the
AREA team that some social groups were under-utilising health
services relative to their needs, and thus certain ‘unmet need’
variables were incorporated within the model. In part, this was also
to address ACRA's new health equity criterion.
This two-stage approach to calculating age-related and additional
needs was undertaken separately for the four elements of the
weighted capitation formula. These covered Hospital and
Community Services (HCHS), Prescribing, Primary Care and
HIV/AIDS, and each was assigned a specific set of socio-economic
variables that was to be used as a measure of 'additional needs'. In
addition, all but the Prescribing element of the formula are further
adjusted to take account of variations in the unavoidable costs of
providing health care. Market Forces Factors (MFFs) are applied to
HCHS, Primary Care and HIV/AIDS. The former was also subject to
a further, very small, adjustment known as the Emergency
Ambulance Cost Adjustment (EACA).
For each PCT, a weighted population is thus calculated for each of
the four elements of the overall formula. These take account of the
target population and, as appropriate, the population's age-related
needs, additional needs, MFF and EACA. The four resulting
weighted populations (normalised as necessary) are then combined
to reflect their relative contribution to the overall health care
budget; with, in 2004/6, the HCHS element constituting 77.36% of
the whole, Prescribing 13.23%, Primary Care 8.81% and HIV/AIDS
0.59%. The outcome is a 'Unified Weighted Population' which is
used, once normalised, to calculate the PCT's resource needs vis-à-
vis other PCTs. For 2006/8, two new elements were added, the
Growth Area Adjustment and the English Language Difficulties
Adjustment. Details are given in Department of Health (2005) and
in Galbraith (2008).
The Allocation of Funding in Practice
The implementation of the weighted capitation formula, no less
than its derivation, can be difficult to follow. An Excel-based tool,
developed by Professor Mervyn Stone of UCL, usefully demonstrates
how the various elements of the current weighted capitation
formula impact upon local allocations (Stone, 2006). The tool plots
the sequential impact of the age-related, additional needs, MFF and
EACA variables as used in the HCHS part of the formula, and then
as composite indices, the HIV/AIDS, Prescribing and Primary Care
(GMSCL) components of the formula. It is designed for use with
pre-2006 PCTs (i.e. prior to amalgamation into larger Trusts); is
based on the 2004/6 formula and uses 2003/4 data. However, the
broad effects of the formula remain unchanged.
Using examples from four very different types of area, Figure 1
plots the cumulative impact of the formula’s elements as they build
towards the Unified Weighted Populations used to allocate
resources to pre-2006 PCTs. It shows how, depending on their
particular demographic and socio-economic characteristics, some
PCTs fare better than others.
In the first two examples, Central Manchester and North Dorset,
the additional needs element of the formula opposes and outweighs
the influence of the age-related needs element of the formula. Thus,
although Central Manchester has a relatively young local
demography (an age-related index of 0.91), its Unified Weighted
Population is pushed up to 1.26 by the additional needs variables,
reflecting the high level of deprivation experienced in this area. A
slightly lower than average MFF means that once the HIV/AIDS
and the Prescribing and Primary Care (GMSCL) components are
included (and these also encompass age-related and additional
needs variables), the final outcome is an index of 1.25. In other
words, the Unified Weighted Population for Central Manchester is
significantly larger than its actual population, giving this pre-2006
PCT one of the highest per capita funding allocations in the
country.
North Dorset, by contrast, has a population with significantly
higher than average age-related needs (an index of 1.08), but
significantly lower than average 'additional needs' which, with the
inclusion of the remaining elements, drives the final index down to
0.87. In other words, although the population of North Dorset is
significantly older than average, it received a much lower than
average per capita funding because of the very low 'additional
needs' attached to its relatively affluent population.
Figure 1: Sequentially Incorporated factors in Capitation
Formula
In the remaining two examples, the age-related and additional
needs indices reinforce rather than oppose each other. Thus, the
Unified Weighted Population of Great Yarmouth PCT, which served
an older, deprived population, is higher than its actual population,
whilst Bracknell Forest received a lower than average per capita
allocation, reflecting the relative youth and affluence of this area.
To put these PCTs into a national context, Figure 2 plots all English
pre-2006 PCTs on a graph illustrating the relationship between
their 2006-07 HCHS age-related and additional needs indices
(Department of Health, 2005b). The dashed curve running from the
top-left to the bottom right of the graph marks the point at which
the age-related and additional needs indices cancel each other out.
Other things being equal, PCTs above the line will receive a greater
than average allocation, whilst those below it will receive a lower
than average allocation.
Our first example of Central Manchester falls into the upper left
hand quadrant of the graph. This comprises PCTs which served
younger, more deprived – and typically urban – populations. As can
be seen by the number of PCTs above the dashed curve, the
'additional needs' weighting outweighs the 'age-related needs'
weighting in the majority of cases, many of which received the
highest per capita funding allocations in the country. In direct
contrast, the lower right quadrant encompasses PCTs which, like
North Dorset, served older and more affluent than average
populations. The majority lie below the ‘break-even’ line. In other
words, the positive effect (on allocations) of relatively old
populations tends to be cancelled out by the negative effect of
affluence.
Figure 2: HCHS Age-related versus Additional Needs Indices
In the remaining two quadrants the age-related and additional
needs indices reinforce rather than oppose each other. As a result,
PCTs which, like Great Yarmouth, served older, deprived
populations always received a higher than average allocation,
whilst PCTs with young affluent populations such as Bracknell
Forest always received lower than average allocations. There are,
however, many fewer PCTs in these two quadrants because, in
England, demography and social deprivation are not independent of
one another. In general, areas with older populations tend to be
more affluent whilst those with younger populations tend to be
more deprived. It is thus often the case that the age-related and
additional-need indices oppose each other and, when they are
incorporated into the weighted capitation model, the additional
needs indices usually ‘win the battle’ (Stone, 2007).
The overall impact of the opposing influence of the age-related and
additional need indices is such that PCTs with more ageing
populations would usually be better off if there were no weightings
at all. According to Stone’s analysis of pre-2006 PCTs, if the
distribution of funding was simply proportional to GP-registered
populations, 67% of the most aged PCTs would gain relative to the
current formula, while 52% of the most youthful PCTs would lose
(Stone, 2007). This is an important finding. With per capita costs
for Hospital and Community Health Services currently ranging
from £269 for a child aged 5-15 to £2,799 for over 85 year olds, it is
often assumed that areas with the oldest populations must receive
the greatest funding allocations. Stone’s figures make it clear that
this is not the case.
The introduction of the AREA formula initiated a shift in resources
towards PCTs serving more deprived populations and resulted in
very large differences in targeted funding for PCTs 1 . Thus, in 2005-
06 the average per capita formula allocation was £1,280 in the 20%
most deprived PCTs compared to £927 in the 20% least deprived
PCTs. When geography is taken into account, the positive targeting
of NHS funds is even more apparent. PCTs serving populations that
1
It should be noted that that these are formula allocations rather than actual funding
settlements. For many PCTs, such was the magnitude of the difference between what
they were currently receiving and what the formula calculated they should receive
that it was impossible to immediately reconcile the two – hence differential funding
increases to close the gap or “distance from target”.
are both in the most deprived and most urban quintiles received
the highest average per capita funding allocations (£1,323) while
PCTs serving the most rural and least deprived populations
received the lowest (£922).
It is important to point out that large differences in funding are not,
in themselves, problematic. The very object of the formula is to
identify whether PCT populations have higher or lower per capita
health care needs than the English average. A HCHS age-related
needs index of 0.9 simply means that, in view of what must be a
younger than average population, a PCT is judged by the formula to
require, other things being equal, only 90% of the average per
capita resource set aside for Hospital and Community Health
Services. Similarly, if a (relatively affluent) PCT has a HCHS
additional needs index of 0.9 it simply means that its population,
other things being equal, requires only 90% of the average HCHS
per capita resource. The issue is whether the formula adequately
captures how costs vary with age and deprivation, and whether the
formula adequately captures the health consequences of the local
interplay of demography and deprivation. This is explored below.
A Critique of the Weighted Capitation
Formula
Notwithstanding the statistical sophistication of the empirical
models that underpin the English health resource allocation
system, the legitimacy of deriving health care needs from an
analysis of health care utilisation is highly questionable. Doubts
have long been raised about the extent to which it is possible to
model utilisation data in such a way as to successfully disentangle
health needs from, in particular, the impact of supply effects on
patterns of health service utilisation (Sheldon and Carr-Hill, 1992).
The complexity, and lack of transparency, of the current AREA
formula is a direct consequence of this challenge. In this respect
the AREA model represents the apotheosis of the tendency for
researchers to "become besotted with the production of ever more
refined empirically based formulas’ (Sheldon, 1997).
A reliance on utilisation data presupposes that historical patterns
of service uptake by different care groups are either appropriate or,
if there is evidence of under- or over- utilisation relative to needs,
can be adjusted to better reflect a population's underlying health
needs. It is questionable, however, whether the current approach to
adjusting for either unmet need or unjustified utilisation is
adequate (see Galbraith, 2008).
A selective approach to conceptualising unmet need
The current resource allocation model is rather selective in its
approach to unmet need. For example, it assumes that past
patterns of utilisation across different age groups appropriately
reflect the underlying need for health care. Older people do make
high use of both primary and secondary care. However, several
studies conclude that there is under-utilisation among this age
group relative to need (Peake and Thompson, 2003; Bond et al,
2003; Holmes et al, 2003; Williams et al, 2004; Beswick et al, 2004;
Crome and Natarajan, 2004; Yong et al, 2004; Grande et al, 2006).
This would suggest that the per capita costs allocated to older age
groups may be conservative.
While evidence of unmet need according to age is ignored by the
formula, the HCHS component does incorporate additional
morbidity variables to ‘capture some of the effect of unmet need
where ethnic minority groups and low income groups do not receive
health care services to the same level to that of others with similar
health care characteristics’ (Department of Health, 2005a, p. 20).
Yet, research evidence of a ‘pro-rich’ bias in health care is
equivocal. Reports that poorer social groups have lower than
predicted rates of utilisation (e.g. Ben-Shlomo and Chaturvedi,
1995; Black et al, 1995; Payne and Saul, 1997; Hippisley-Cox and
Pringle, 2000; Morris, Sutton and Gravelle, 2005; Morris et al,
2005) have been challenged by evidence that suggests that poorer
people receive as much if not more health care than richer people
for equal need (e.g. Macleod et al, 2000; Asthana et al, 2004;
Britton et al, 2004; Jones et al, 2005; Adams and White, 2005;
Strong et al, 2006; Kee et al, 2007). Thus, as a recent large-scale
review of inequalities in health care concluded, “the evidence does
not consistently point to poorer access for socio-economically
disadvantaged people, even when need is accounted for; some
studies even suggest that there is a pro-poor bias in the NHS”
(Dixon Woods et al, 2005, p.97). The picture that emerges with
respect to ethnic minority status is similarly ambiguous (Dixon
Woods et al, 2005, p.130-35).
It is important to acknowledge that people from lower income and
ethnic minority groups may face similar difficulties with respect to
processes of presentation, negotiation, adjudication and acceptance
as do older people. Thus, the point is not to suggest that one group
is more or less disadvantaged than another with respect to access
to health care, but to demonstrate that the evidence on inequalities
is complex and equivocal. Against this background, the selective
and unambiguous approach taken to adjusting for unmet need in
the resource allocation formula is questionable.
There is, of course, a geographical dimension to demography,
deprivation and ethnicity. The youngest, most deprived populations
tend to be urban, while populations in rural areas are often both
demographically older and relatively affluent. Urban areas also
have the highest concentrations of ethnic minority groups. The
implications of social bias in health service utilisation (and
subsequent resource allocation) are thus not geographically
neutral. If the model’s assumptions about unmet need are not met
or if, indeed, there is pro-poor bias and/or ageism in health service
use, then one would expect urban areas to be the prime
beneficiaries.
There is some published empirical evidence which suggests that
this occurs. A comparison of utilisation-based measures against
direct health estimates as a basis for setting capitations for the
inpatient treatment of coronary heart disease found that the former
allocated resources to urban deprived areas to a higher level than
implied by morbidity alone. Rural areas, by contrast, appeared to
be under-resourced (Asthana et al, 2004).
Is it possible to adjust adequately for supply?
The weighted capitation formula is not only likely to reflect
systematic inequalities in existing utilisation. There is a strong
possibility that it will reinforce inequality in a vicious/virtuous
cycle of resource allocation. Thus, a higher supply in areas that are
well-resourced in relation to underlying need stimulates higher
rates of utilisation which are in turn rewarded by higher
allocations. By contrast, relatively low levels of funding influence
lower rates of provision and use.
‘Supply effects’ are, of course, controlled for in the statistical
models that inform the English formula. However, critics doubt
whether ‘… modelling techniques (can) deal adequately with a
system where demand, utilisation and supply are so inextricably
linked’ (Sheldon and Carr-Hill, 1992). The current approach to
adjusting for supply uses physical measures such as access and
capacity. This takes inadequate account of the demographic and
socio-economic biases in health service utilisation which have a
geographical dimension which largely parallels that of supply. Thus
there can be 'no justification for the claim that consideration of
supply effects can convert a formula to predict ‘standardized
utilization’ into a formula to predict appropriate standardized
utilization’ (Stone and Galbriath, 2006).
What are we achieving - health care equity
or health equity?
Assumptions regarding the legitimacy of utilisation, together with
the selective approach taken for conceptualising unmet need and
the failure to adjust adequately for supply, suggest that bias (of
uncertain magnitude) is likely to creep into the system of resource
allocation. However, the pattern of allocation to urban deprived and
rural affluent areas principally reflects the relative importance
accorded to age-related and “additional” needs in the calculation of
funding allocations.
The authors of the AREA report intended that their approach would
yield target allocations which would appropriately reflect the
relative impact of age, additional needs and other factors on local
health care needs. However, this depends upon whether ‘needs’ are
defined according to the health care equity or health equity
criterion.
As noted above, the overall effect of the opposing influence of age-
related and additional need indices is such that PCTs with
demographically older populations would often be better off if there
were no weightings at all. By contrast, highly deprived areas benefit
strongly from the formula, even if their populations are relatively
young. The formula clearly operates on the assumption that socio-
economic deprivation has a greater effect on a population’s health
care needs than demography.
Does resource allocation promote equal opportunity of access
for equal need?
There is no doubt that standardised mortality and morbidity rates
are highest, and average life expectancy lowest, in urban and
declining industrial areas where social deprivation is more extreme
(Dorling, 1997; Shaw et al, 1999; Shaw et al, 2005; Asthana and
Halliday, 2006). Here, funding allocations are highest. By contrast,
the most affluent areas, which receive the lowest target allocations,
enjoy the highest levels of health – in standardised terms.
It does not necessarily follow, however, that areas suffering from
the worst health inequalities also have the highest crude rates of
morbidity and thus the highest health care needs. It has now
become so common to age-standardise measures of disease
prevalence that it is easy to overlook the fact that, for most
conditions (mental health being a notable exception), population
age structure is a far more significant determinant of morbidity and
mortality than deprivation (Gibson et al, 2002). As people get older,
they are more likely to develop conditions such as heart disease
and cancers and this places significant demands on health care
resources. Older people are also far more likely to die than younger
people and, because progressive and fatal illness often requires
high intensity care, this has important cost implications
(Seshamani and Gray, 2004).
It is therefore quite plausible that older populations, even if affluent
and with relatively good health status with respect to health
inequalities, will have higher absolute burdens of ill-health. If this
is the case, then the current distribution of funding is unlikely to
secure equal opportunity of access for equal needs: to fulfil the
principle of health care equity, the distribution of health care
funding should reflect the existing burden of disease.
In the absence of a ‘gold standard’ by which to measure the needs
for health services, difficulties arise in assessing whether or not the
current system does promote equal access of opportunity for equal
needs. However, a number of methods lend themselves to this
important issue. As we discuss below, survey-based morbidity data
provide one means of deriving more direct estimates of resource
needs. In addition to offering an alternative approach to resource
allocation, these provide a benchmark against which to assess the
‘fairness’ of current allocations. In a recent analysis, in which
epidemiological estimates were used to provide a proxy measure of
the overall health resource needs of PCTs in the East of England,
the current utilisation-based model was found to overestimate the
resource needs of deprived areas and to underestimate the resource
needs of older areas compared to the morbidity-based approach
(Asthana et al, 2007).
Another approach has been to analyse indicators of organisational
stress (financial and service-related) which can result when local
needs are not satisfactorily met. The systematic pattern of deficits
in recent years suggests that the current resource allocation
formula has failed to address the health care needs of particular
populations adequately (Asthana and Gibson, 2005; Asthana and
Gibson, 2006; Badrinath et al, 2006). Risk of deficit has been
strongly associated with resource allocation. In 2005-06, for
example, only 13% of the 60 PCTs with the highest per capita
allocations ended the year in deficit, compared with 68% of the 60
PCTs with the lowest per capita allocations. With regard to the
population characteristics of deficit PCTs, no less than 71% of PCTs
serving the most affluent and most rural populations failed to
break even, compared to only 6% of those serving the most
deprived and most urban populations. The fact that such a
systematic relationship exists in the distribution of financial
deficits strongly suggests that the current resource allocation
system is not adequately capturing the health care needs of
particular populations.
Does resource allocation promote an equal opportunity to be
healthy?
As urban deprived areas would be expected to benefit
disproportionately from funding allocations for health inequalities,
the fact that they may also be relatively ‘over-funded’ with respect
to the health care equity criterion may appear to be unproblematic.
However, the goals of health care equity and health equity require
very different policy responses.
The vast proportion of NHS resources is spent on curative and
particularly hospital services. Deprived groups that experience
higher rates of premature disease must of course have access to
high quality treatments and procedures as these can play a
significant role in improving quality of life and reducing risk of
mortality. However, the goal of reducing health inequalities is
essentially about prevention. Targeting additional resources at
hospital treatment rather than public health and primary-level care
in deprived areas is tantamount to shutting the stable door after
the horse has bolted.
The goal of reducing health inequalities should instead rest on
policies designed to narrow or, through public health initiatives,
mitigate the effects of, the unequal distribution of the social and
economic resources that influence health. With respect to public
health initiatives (which account for a very small proportion of the
overall health budget), the existing evidence base tells us very little
about the effectiveness of individualised behavioural interventions
in addressing health inequalities (Asthana and Halliday, 2006).
This casts some doubt upon the potential of the current public
health strategy to promote greater health equity.
With respect to the distribution of social and economic resources, it
appears that a reduction in overall inequality is not an aim of the
current government. Low incomes may have improved in absolute
terms, but overall levels of income inequality have remained fairly
stable since Labour assumed power. At the same time, the
distribution of wealth has become more unequal. Education
continues to be characterised by large social class differences, with
the gap widening in access to university places (Galinda-Ruedo et
al, 2004). Similarly, improvements to housing have been made
through reductions in the numbers of substandard homes, yet the
gulf between the rich and the poor in terms of property wealth is
now wider than at any time since the Victorian era (Thomas and
Dorling, 2004). As the unequal distribution of health reflects the
unequal distribution of the social and economic factors that
influence health (Graham, 2004), it is hardly surprising that health
inequalities continue to widen in the first years of the 21st century.
It would therefore seem that the current approach to resource
allocation is flawed with respect to both equity criteria. The NHS
(and particularly national hospital services) has little to contribute
towards the reduction of health inequalities compared to other
sources of variation such as income distribution, education and
housing. Thus, the targeting of core services to urban deprived
populations over and above levels of underlying morbidity is an
ineffective response to health inequalities. It is one, moreover, that
exacerbates health care inequity by underestimating the needs of
older but less deprived populations.
Towards a normative approach to achieving
the core principles of the NHS
However sophisticated the resource allocation system, and however
wide the range of variables used to model the relationship between
socio-demographic and socio-economic variables and health care
utilisation, the fundamental difficulty of disentangling legitimate
needs factors from other policy and supply influences in utilisation
remains. Against this background, questions have been asked as to
whether resource allocation formulae should shift from the current
empirical approach based on observed variations in health
utilisation to the formulation of more normative capitations
(Schokkaert and Van de Voorde, 2004).
It is arguable that in order to promote a shift from what is to what
ought to be, health care capitations should be built on direct
measures of the health needs of populations. There has been little
practical progress in developing more direct approaches to resource
allocation. Wales is unique in using survey-based data on
morbidity as the basis for distributing a large proportion of its
health care expenditure (Gordon et al, 2001), an approach that has
been explicitly rejected for use in Scotland (Carr-Hill and Dixon,
2006). Objections have been less on principle than in response to
data and methodological considerations. In effect, critics have
questioned whether it is possible to reliably measure the burden of
ill-health in different areas, and whether it is then possible to
establish the resources required to meet local health care needs
(McConnachie and Sutton, 2004).
The particular approach adopted in Wales has its limitations and,
as it requires that survey data are collected for all areas, it cannot
be implemented in England. However, an alternative to the Welsh
“sample-based method” is to generate risk-adjusted resource needs
through statistical modelling. A variety of approaches are available
to this end. The most basic use simple attribution and, like the
Welsh method, assume that all cases have the same level of health
care needs (a problematic assumption). More complex methods
apply suitable models (including multilevel models involving both
individual and area level covariates and both fixed and random
effects) to individual survey data to generate predictive
distributions of morbidity. All uncertainty due to the modelling
process is incorporated into local estimates of morbidity, whilst any
further uncertainty regarding the resource required is captured by
sampling from appropriate historic cost distributions. This can be
minimised by attaching age/sex-specific cost data to the
corresponding morbidity estimates. It is thus possible to combine
local predictive distributions of morbidity relative to defined service
care and/or diagnostic groups with equivalent national-level
predictive distributions of per capita resource needs to generate
local risk-adjusted distributions of predicted resources and needs.
One relatively simple approach to modelling estimates of direct
needs (but not estimates of resource needs) is provided by Gibson
et al (2002). Here, cumulative data from the Health Survey for
England were used to create an age, sex and social class matrix of
the prevalence of ischaemic heart disease. Equivalent matrices
were produced for general practices (n=539) in seven health
authorities in contrasting locations in England. These were based
on both patient registration data (which provide age and sex) and
the patient-weighted attribution of census data. The matrices were
then combined to derive an estimate of the disease burden within
the target populations.
While this study generated estimates of morbidity, it did not attach
these to resource weights in order to produce estimates of health
care resource needs. The potential for this was explored in a later
paper published by the same team (Asthana et al, 2004). Here,
symptom-based estimates were produced of PCT-level prevalence of
severe angina and myocardial infarction (MI). Derived from
responses to Rose Questionnaire items in the Health Survey for
England and adjusted for social class, these were combined with an
age/sex resource matrix for the inpatient treatment of coronary
heart disease (based on HRG Reference Costs for inpatient episodes
with a main diagnosis code between ICD10 I20 and I25). In order to
establish a CHD clinical programme budget for each PCT, the
estimated number of people with symptoms of severe angina
and/or MI in each age/sex cohort were multiplied by that cohort’s
average ‘per capita in need’ resource use, then the totals of all
cohorts were aggregated.
While this demonstrated the feasibility of using morbidity as a
basis for setting health care capitations, it produced a budget for
only one clinical area. In a more recent study, in which a simplified
epidemiological approach was used to develop resource need
estimates for post-2006 PCTs in the East of England (Asthana et al,
2007), allocations were developed for 14 Programme Budget
categories, accounting for the major proportion of NHS expenditure.
Using cumulative data from the Health Survey for England
(n=>75,000 individuals), this modelled the relationship between, on
the one hand, an individual’s age, sex and general health status
and, on the other, whether or not they had self-reported suffering
from one of 14 ICD10-based categories of illness. These categories
did not cover the entire spectrum of health care needs, but rather
just those which could be mapped directly onto Programme Budget
categories. Linking age/sex/general health status-based prevalence
rates for different disease categories (derived from the Health
Survey for England) with census-based count of the number of
people in each age/sex/general health status ‘cell’ in each PCT,
estimates were derived of the number of people in each cell likely to
report a disease. Cell totals were summed and then divided by the
total number of people in each PCT to generate overall and disease-
specific prevalence rates. By attaching cost data drawn from
Programme Budget category data, epidemiological-based estimates
of per capita health care resource needs were then produced.
This was an admittedly small sample (there are 14 post-2006 PCTs
in the East of England), but the results were promising. First, the
study took advantage of the fact that, with the recent emergence of
Quality and Outcomes Framework (QOF) data, there is now at least
one independent source of data against which to compare
estimates. It found that epidemiological estimates of the prevalence
of ‘disorders of the heart and circulatory system’ were very highly
correlated with QOF coronary heart disease prevalence rates
(R=0.972) and QOF hypertension prevalence rates (R=0.960),
supporting the efficacy of the epidemiological approach. Second,
the morbidity-based estimates of overall health care resource needs
were a far better predictor of actual programme budget expenditure
(R=0.861) and per capita prescribing costs (R=0.83) in East of
England PCTs than the weighted capitation formula (R=0.62 and
0.44 respectively).
Conclusion
As befits an approach still in its infancy, a wide range of
methodological issues are being raised in explorations of the
normative approach, not all of which have been satisfactorily
resolved. For example, questions remain about the reliability and
validity of survey-based data as a basis for resource allocation.
Problems such as sampling and reporting bias can be reduced
through good survey design and implementation. It is also possible
to validate key dimensions of self-reported health status objectively
(e.g. where surveys also collect physiological measurements and/or
biological markers). In this respect, the Health Survey for England
has several key advantages. There are, moreover, a growing
number of available surveys which include evidence on the
morbidity of individuals alongside socio-demographic and other
characteristics.
Developing methods of attaching meaningful ‘resource needs’ to
morbidity counts also remains a key challenge. Cost data need to
be disaggregated to reflect legitimate variations in the costs of
treating patients in different circumstances and with different co-
morbidities. Outpatient, community and prescribing data do not
lend themselves well to such analysis and, while inpatient cost data
can be disaggregated by age and sex, Hospital Episodes Statistics
(HES) do not include information on socio-economic status. Future
work on the direct approach could therefore fruitfully explore
methods of introducing socio-economic discrimination into the
analysis of cost data and of incorporating overall, rather than just
inpatient, costs.
Despite these reservations, the studies described above do
demonst


Use: 0.0426