Examination of Witnesses (Questions 960
- 977)
WEDNESDAY 29 APRIL 2009
Mr David Fillingham, Ms Rhona MacDonald, Mr Keith
Derbyshire and Dr Stephen Lorrimer
Q960 Lord Rooker:
Do they show any interest in what you do is really what I am after?
Mr Derbyshire: Ten, 15 years ago there was interest
from the Treasury side on how we allocated money and to make sure
that the allocation delivered value for money on the ground, and
also whether there were any lessons to be learned from the NHS
formula compared with the local government formula. So they did
the kind of thing you would expect the Treasury to do, to take
a wider cross-government look at what we were doing, but there
has been little interest in the last ten years, apparently because
it is a transparent process which effectively has got its own
audit mechanism. Lots of academics pore over the work which has
been done to actually check its validity.
Q961 Lord Rooker:
One last question, which is probably completely out of order.
In terms of formulae for allocating funds, have you looked at
other formulae which are used for such a mechanism to assess whether
or not what you are doing is right and fair and transparent, or
whether there is anybody else within the country using formulae
for other purposes? Have you ever assessed other formulae?
Mr Fillingham: Yes, indeed we have, and in fact
as part of our research last time round we commissioned a review
of international approaches to resource allocation in different
health systems. There are essentially two types, systems which
look at individual capitation, insurance-based system—the
Netherlands is a good example of that and the Medicare system
in the US is another example—and then systems which have
an area capitation basis, a population basis such as ours. Many
of those countries which have an area capitation-based approach
have formulae which have similarities to that of the NHS, and
indeed the Scottish formula what it is known as the Arbuthnott
and the approach in Northern Ireland has a very similar approach.
The Welsh formula is slightly different in that it is less based
on the utilisation—this discussion we were having before—and
much more firmly attempts to allocate on the basis of need. It
is probably true to say that the English formula was one of the
first to be developed on a weighted capitation basis and many
of the other systems have emulated aspects of what we have done.
We try and keep up to date with developments in other systems
as part of ACRA's work.
Q962 Lord Rooker:
You have not thought of adopting the Barnett Formula approach
as a replacement for yours?
Mr Fillingham: It may be something which after
this Committee hearing we will take away and consider.
Q963 Lord Smith of Clifton:
I would like to ask you about turbulence. Presumably year on year
there are slight incremental adjustments for most PCTs, but there
have been reports of one or two PCTs, as a result of a surge in
immigration, and so on, where there has been quite a change in
population. Presumably those PCTs are outriders in terms of turbulence,
they are more affected by this. Do you ever look back on this
and the results of what has happened and do you identify, as it
were, those PCTs which are most subjected to turbulence, because
I can imagine, having served on a District Health Authority (as
it then was), that the kind of brake and accelerator effect makes
it terribly difficult to take any medium term planning view?
Mr Fillingham: There are issues about turbulence
in two ways, I guess, and I will let my colleagues comment on
the technicalities. First of all, there clearly is an issue about
population shifts and the formula does aim to account for that.
We use quite sophisticated ONS projections and we also compensate
for the growth areas and the Department for Communities and Local
Government growth points are built into the population characteristics.
There is clearly an issue as well that sometimes those different
populations will have different demands for healthcare too, so
we clearly attempt to take account of that. The other issue about
turbulence is changes to PCT allocations year on year, and I think
that is a very good practical recommendation, whatever the politics,
for a pace of change policy because if we were to move every PCT
immediately in the next target allocation the ones that would
gain could not possibly spend all the money and the ones that
were losing would be in serious difficulty, so having an element
of stability makes sense from a PCT's point of view.
Dr Lorrimer: The only thing I would add is that
immigration/migration generally is a big issue and particularly
perhaps internal migration is a bigger issue for us in this work
rather than perhaps international migration, but we are very actively
involved with a project you may be aware of which ONS are leading
to look to improve the migration statistics and the projections
they produce and we are very hopeful that that will be a big help
as we go forward.
Q964 Lord Smith of Clifton:
Presumably the population projections are going to be even more
volatile in a time of recession because you are likely to get
greater movements of population both internally and within the
EU? One can think of towns which have suddenly had a splurge of
workers and then there is a change to the pound and they all disappeared.
It is impossible to be that sensitive, but nevertheless they can
be quite dramatic changes, one imagines, over a fairly short period
of time?
Ms MacDonald: I do think, speaking as a PCT
chief executive who has to manage the changes, the pace of change
policy becomes really important and also remembering that our
job is to manage the resources we have got. At the end of the
day many PCTs are not at target, some are under, some are over,
and there are all sorts of turbulence in the system. Our job is
to manage, but I do think over the years the pace of change policy
has been used sensibly because my own PCT, as a result of the
last round of recommendations, has moved above target rather more
than I wanted. But in doing that, because the pace of change is
relatively slow at the moment it has not given us a major problem.
That allows us to look ahead, so now we are saying that over the
next X years we know we now have to make that amount of saving.
Chairman: I want to come on to the pace
of change policy in a minute.
Q965 Lord Lang of Monkton:
As you know, we are looking at the Barnett Formula, which covers
a whole range of expenditure far outside just the Health Service,
but are there any aspects or principles you have identified in
your work which underpin your conclusions which might carry over
and apply on a wider basis, or is it so specialised as to have
no value to that extent?
Mr Fillingham: I can comment on that, I guess,
as a private citizen rather than as Chair of ACRA. I think the
essence of the ACRA approach could be used to allocate funding
to the devolved administrations. I think there are probably five
essential elements. The first is being clear about objectives,
so you do need to know on what basis you are allocating resources,
but I think the model, secondly, of having an independent advisory
committee of experts to advise on populations, on weightings,
could be done. Thirdly, I think it is possible, and we have shown
this, to allocate resources according to need rather than just
to demand. Fourthly, then to develop a weighted capitation formula.
I think all of those could apply to your work. Fifthly I think
the other lesson from our work as well which is reflected in the
debate we were just having, is about pace of change and about
recognising the practical difficulties then of making changes
to large elements of public sector funding. I think were the Committee
to consider that, there are three sets of challenges probably
which occur to me. The first straightforwardly would be administrative.
It needs to be set up, it needs to be administered and you need
to choose the right people to be on the advisory committee. There
is a cost, although the cost of ACRA is relatively small actually,
the overhead is not great, but the process and the governance
arrangements, I guess, would be complicated in the Barnett case
for establishing some kind of arrangement like that. Secondly,
there will undoubtedly be technical difficulties because clearly
you are talking about not just health but quite a wide range of
public expenditure and there is a basic decision there about whether
you try and drill down to quite a degree of granularity and develop
a formula for each area of public spend which you then aggregate,
or do you go for some form of proxy measure which determines need
and demand for public services as a whole? There are also, I think,
some technical challenges in data availability and also the way
data is collected across the UK. The third set of challenges clearly
are practical and political. Would there be support from the key
stakeholders for this? How would you manage pace of change? So
I think our conclusion would be that there are a number of things
you could take from this approach which could be made to work,
but there would be some significant challenges to be dealt with.
Q966 Lord Sewel:
Moving upon that a little, your approach has been to make the
formula more and more sophisticated, building up the formula.
I tend to be in favour of building down and concentrating on a
relatively small number of variables which explain most of the
difference. Have you looked at it and said, "Do we actually
need to go to this degree of sophistication? Don't we get 98 per
cent of it by being a lot simpler?"
Mr Fillingham: I could almost imagine you have
sat in on ACRA discussions actually, having put that question!
It is a constant challenge. It is something which, as Chair of
ACRA, is a constant balance because on the one hand our academic
colleagues are often pushing for technical robustness and they
are often in favour of quite complicated and sophisticated adjustments
to the formula. We are always mindful of balancing that against
comprehensibility and I think it is important that non-specialists
can understand how we come to our conclusions and feel that they
are fair. There is a tension between those two objectives which
we try to manage and, yes, one of the statistical tests we do
ask the researchers to apply is how much materiality, how much
difference would this tweaking of the formula make. If it is not
going to make a great material difference, then we may well not
move ahead with that tweaking. There is a risk that with a Committee
like this the formula becomes more and more sophisticated as you
move forward.
Q967 Lord Moser:
My question actually is on a very similar point. In your section
on need there is a paragraph which comes close to what you have
just been talking about. As I understand the thinking, you start
with population, obviously, and you say, "That's too crude.
We've got to look at age and gender. That's the next step."
Then you say, "Well, that's still not quite enough. We must
deal with need." Then you say that the paper says, "Observing
need directly has not proved possible to date." I understand
that, I think. So then you would turn to the statisticians --always
dangerous!—and you say that statistical modelling is used
and what you do there is you take small areas, as I understand
it, and you collect lots and lots of data—I do not know
whether it is real data or theoretical data—and what links
with what, where are the correlations, and out of that you choose
one or two indicators actually to use for the process. Is that
what you actually do? If so, I am left with the puzzle about these
models, how you actually come to choose one or two indicators,
which is what you say you do, I think. Is that right?
Mr Derbyshire: Yes, I think you are 90 per cent
there in that we do look at the utilisation of health services
by a small area and adjust it for relative supply, to try and
identify legitimate needs drivers of different populations. So
for the whole country of England we know what the hospital utilisation
data is and where people live and we look at the utilisation by
small areas, populations of 10,000, adjust it for relative supply,
adjust it for age, and then try and explain the differences. So
a ward in Hull might have a much higher utilisation of hospital
services than a ward in Guildford and we look for legitimate explanations
of that variation linked to mortality data, morbidity data and
socioeconomic data, and we identify models which explain the variation
in utilisation after adjusting for age and need which appear plausible,
are relatively parsimonious and look intuitively correct. Then
there is an element of judgment which is applied by ACRA and TAG
over which models are the most legitimate for use in a resource
allocation formula.
Q968 Lord Moser:
You then end up with one or two indicators. It is a very interesting
question which Lord Sewel just asked you really: can one end up
with just one or two sort of proxy indicators for everything else?
Mr Derbyshire: Going back to the formula which
was current in the 1990s, there were five indicators. I think
in the current formula David is just telling me it is 12, but
that is partly because we have got so many different age bands.
One of the criteria we apply to formulae is simplicity.
Q969 Lord Moser:
That is what we are looking for.
Mr Derbyshire: If simplicity is important, then
you can certainly have simple models as an objective of the whole
process. The original RAWP formula was very elegant and very simple.
It had one variable, which was all age, all cause, mortality.
Q970 Lord Moser:
It is twofold. First of all, simplicity, just so that it is not
too complicated. Secondly, that everybody in the Health Service
and people like me would actually understand what you are up to.
Would you then just end up with a couple of indicators?
Mr Derbyshire: You possibly would. But there
is perhaps a higher level decision than that, which is that we
favour the empirical approach to identify a formula and that is
where the process has got to over time. It needs to be justified
empirically, but you could say it is judgment actually. We could
make a high level judgment that the ratio of needs for transportation
or education, or health, varies according to this single index,
whatever that might be, and have a judgmental formula. You would
over time find it difficult to defend as people who were losing
out from that formula began to do precisely the statistical analyses
you indicate.
Q971 Lord Moser:
That is why you turn to the statistical models, to make it more
defensible?
Mr Derbyshire: The closer people get to target
allocations, the more scrutiny the actual formula generates.
Ms MacDonald: Could I just add one point? We
have talked a lot about the formula and the differential needs,
but especially when we are talking about public expenditure the
cost of meeting the objective is relevant. So at the end of the
day what are you going to apply the formula to, and if you have
got a simple needs formula but meeting health needs could be totally
different costs from meeting education needs and from meeting
transport needs. I think that is one of the questions. When David
talked about the granularity you would want to consider you have
to consider that actually it is not just about differential need,
it is about what is the cost of meeting that need.
Q972 Lord Forsyth of Drumlean:
May I just follow up on Lord Moser's point? If you abandon additional
model complexity, what sort of spread does that give in the allocation?
How much of a difference does it make? Is it just at the margin?
Mr Fillingham: If we are going for a straight
population count, for example -
Q973 Lord Forsyth of Drumlean:
Well, weighted population. If you did not go in for this model
which adjusted, as has just been explained to us, and it was just
a kind of crude approach and you did not have this additional
layer of complexity, what difference does that make to the spread?
I am sorry, I am back with the minister looking at the allocation.
What kind of difference is there? Is it marginal or is it very
significant?
Mr Fillingham: I will let my colleague come
in as well, but in terms of the weighting my recollection when
we looked at this is that if you just took a straight population
count and did not weight for anything else, then the difference
would be very considerable, so I think the biggest loser would
be around 30 per cent and the biggest gainer 21 per cent. There
is a range in between that, if you just went for age and gender
and if you then added other needs. Keith, have we done those calculations?
Mr Derbyshire: Yes, we have done those calculations.
Sticking with the range of 80—121 and the needs adjustment,
then the complexity of the formula does not actually change that
gradient.[1]
The gradient is independent of the complexity of the formula or
the number of variables in the formula. Certainly previous formulae
which have had fewer variables and have had a similar needs gradient.
Lord Forsyth of Drumlean: I am sorry,
I did not understand that answer.
Chairman: I think he is agreeing with
you.
Q974 Lord Forsyth of Drumlean:
Does that mean it does not make very much difference?
Mr Fillingham: I am not sure I understood it
either, but essentially if you go back to the original RAWP (Resource
Allocation Working Party—a predecessor to ACRA) formula
which just used the one elegant mechanism of standardised mortality
rates, the graph for the RAWP formula overall is not wildly dissimilar
to the current graph of allocations, the 80, 120. However, the
position of the particular PCTs on that graph might well change
and the reason for getting to a more sophisticated formula is
because people want to challenge the evidence and people employ
their own statisticians and economists and do their own analyses.
So although having more variables may not change the overall range
of allocations, there is a greater level of confidence that you
are getting it right for each PCT.
Q975 Lord Forsyth of Drumlean:
I understand that, but my question is really just in terms of
these PCTs. The ones who feel aggrieved will be the ones who asked
for more complexity and the ones who are doing well will say,
"This is a very good, fair and simple formula." I understand
that. It is just the extent to which there is variation if you
go for a simpler system, the degree of unfairness there is in
it. It also strikes me—and I am sure I would not want your
job as Chairman—if you start off with the formula it would
be very difficult to change it because you will create winners
and losers and the losers will complain. If one was going to find
some kind of needs based system for allocating the Barnett funds
you would probably want to make it as simple as you can, but at
the same time recognise that people will come along and say, "We
have worked out that if you add this complexity you will get X."
What is in my mind is, how big is X going to be, based on your
experience? The PCTs and the others will argue, of course, but
it is the extent to which it is a problem.
Mr Fillingham: Changes to the formula do take
place based on further research commissioned by ACRA. For example,
the previous formula—and I apologise for the acronyms—which
was known as the AREA formula has now been superseded by the new
formula which is known as CARAN, and the principal difference
was this business at looking at each age band and identifying
the characteristics of that age band and identified need linked
to those. That has made quite a difference to the allocation.
I think around 10 per cent is probably the kind of order of change
that we might be seeing in the individual PCTs. So there was quite
considerable change and you are right, of course, gainers are
quite happy with that change. People who see themselves as losing
out are less happy, but the issue is that there was a range of
concerns raised, particularly by the academic community, about
the previous formula. There have been developments in the statistical
technique and in the availability of data since the previous formula
was done, so it was right for us to update the formula and to
recommend what we felt was now the best evidence based approach,
even though that was going to create some winners and losers,
because that new statistical modelling was available and because
the new data was available. We needed to make those changes and
to update the formula. The other issue, of course—and I
am now going to mention pace of change—is that that alters
the target allocation for the PCTs. The pace at which they get
there is another matter. It may not hugely influence their actual
allocation in a given year.
Chairman: Thank you very much.
Q976 Lord Lawson of Blaby:
May I just ask one question? I have been much impressed by the
conscientiousness and common sense with which you clearly operate
the system with which you have been entrusted. As Mr Fillingham
pointed out, he and Ms MacDonald, unlike the other two witnesses,
have proper day jobs of a very responsible nature and I wondered
how much of your time is taken in doing this. Also, Ms MacDonald
mentioned, very importantly, how you have to look at the cost
of these and she mentioned education as well as health. It occurred
to me that the big difference is that in this country, unlike
in France where they are both done on a national basis, education
is for the most part a local authority responsibility in England
and in other countries in the United Kingdom, whereas health is
national. I wondered if you would care to tell us which of these
two models you think works better.
Mr Fillingham: I will let Rhona think about
that while I answer the first question. It is time consuming.
It is also very interesting. It is quite different from the day
job and trying to get consensus across such a diverse committee
can be a challenge, but it is a very interesting challenge. We
meet around five or six times a year. That tends to be less at
the beginning of our programme and more towards the end, as we
have got to consider the results of research, and they are full
day meetings. I estimate I spend probably a day preparing for
each meeting, reading papers and talking to officials, so probably
somewhere in the order of ten to 12 days a year of my time, slightly
more when I am invited to a Lords Committee!
Ms MacDonald: I spend a similar amount of time,
but in relation to the education point, (a) when I mentioned education
I was referring to the fact that Barnett would have to take account
of different spending streams. Education per se does not feature
in our world and I do not think I know enough about the difference
between how education is allocated and how we do help to actually
make informed judgments. I rather think that even with the local
authority spend there is a national allocation process to local
authorities which has an impact, so I do not think it is that
different.
Q977 Lord Lawson of Blaby:
There is a national allocation system and the local authorities
take no notice of it, they just spend it as they wish. The public
expenditure White Paper is an exercise in fiction, apart from
the total!
Ms MacDonald: I understand, but I think that
has also been a long debate in the NHS, in that we apply different
formulae to different parts of the NHS spend. When it gets to
a PCT, I can determine locally whether to spend it on acute services,
community services or preventative services, so the same principle
applies.
Chairman: Thank you very much indeed.
I am bound to say, as far as I am concerned I found it extraordinarily
helpful. I confess I did not understand the paper with perhaps
the degree of clarity I should, but you have shone a light on
it and I am very grateful. Thank you very much indeed.
1 Gradient is shorthand for the range between high
need and low need areas, eg, the most deprived PCT in England
might have needs 21 per cent above the English average and the
least deprived 20 per cent below the English average. Back
|
